COMPARATIVE EFFECTIVENESS OF QUALITY IMPROVEMENT INTERVENTIONS FOR PRESSURE ULCER PREVENTION IN U.S. ACADEMIC HOSPITALS WILLIAM VINCENT PADULA III

Transcription

1 COMPARATIVE EFFECTIVENESS OF QUALITY IMPROVEMENT INTERVENTIONS FOR PRESSURE ULCER PREVENTION IN U.S. ACADEMIC HOSPITALS by WILLIAM VINCENT PADULA III B.S., Northwestern University, 2006 M.S., Dartmouth College, 2008 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Pharmaceutical Sciences Program 2013

2 2013 WILLIAM VINCENT PADULA III ALL RIGHTS RESERVED

3 This thesis for the Doctor of Philosophy degree by William Vincent Padula III has been approved for the Pharmaceutical Sciences Program by Kavita V. Nair, Chair Robert J. Valuck, Advisor Jonathan D. Campbell Mary Beth F. Makic Heidi L. Wald Date 7/22/13 ii

4 Padula, William, Vincent (Ph.D., Pharmaceutical Sciences) Comparative Effectiveness of Quality Improvement Interventions for Pressure Ulcer Prevention in U.S. Academic Hospitals Thesis directed by Professor Robert J. Valuck ABSTRACT Objective: To compare the effectiveness of quality improvement (QI) interventions for hospital-acquired pressure ulcer (HAPU) prevention among U.S. academic medical centers of the University HealthSystem Consortium (UHC). Methods: We surveyed UHC hospitals to longitudinally characterize adoption patterns of QI interventions for HAPU prevention between in response to CMS nonpayment policy for HAPUs. Characterization was based on the QI best-practice framework for QI strategy which includes 25 QI interventions organized into four domains: Leadership; Staff; Information Technology; and Performance & Improvement. Survey data was merged to quarterly hospital-level HAPU incidence rates to measure the effect of each QI intervention at reducing HAPU incidence. Utilizing an effect size analysis, we calculated derivatives of overall HAPU reduction for each QI intervention. A t-test compared marginal effect size in the first three quarters following adoption to remaining periods of adoption for each hospital. An analysis of covariance (ANCOVA) tested the correlation between of QI interventions and HAPU incidence variability while controlling for Medicare policy, age, gender, length-of-stay, case-mix index, and intensive-care unit (ICU) admission. iii

5 Results: A representative sample of fifty-five UHC hospitals responded to the survey, of which 53 (96%) indicated use of QI interventions in HAPU prevention. All QI interventions fit within the QI best-practice framework, thereby validating its structure. The effect size analysis identified five QI interventions with clinically meaningful effectiveness by reducing HAPU incidence greater than 1 case per 1,000 patient admissions between , including: Leadership Initiatives; Visual Tools; HAPU Staging; Skin Care; and Patient Nutrition. The t- test returned that the greatest reductions in HAPU incidence occur earlier in the adoption process (p<0.05). The ANCOVA model found that Medicare policy and ICU admission are the primary indicators of variability in HAPU incidence. Conclusion: The effect size analysis identified five QI interventions that have a meaningful effect on HAPU prevention in UHC hospitals. These QI interventions can be used in support of an evidence-based protocol for HAPU prevention. Many hospitals that began implementing QI interventions in response to CMS nonpayment policy experienced significant reductions in HAPU incidence immediately following initiation. publication. The form and content of this abstract are approved. I recommend its Approved: Robert J. Valuck iv

6 DEDICATION I dedicate this work to my family: William, Judith and Lauren. This work was achieved because of your love and unending support. v

7 ACKNOWLEDGEMENTS I would like to begin by thanking my primary mentor, Prof. Robert Valuck. Prof. Valuck and I spent countless hours together over the past four years conceptualizing a study of quality improvement within the framework of comparative effectiveness research. What began as an idea was made real into this thesis with his vast understanding of methods in health outcomes and dedication to improving the evidence base of research that improves the entire system of patient care. I am grateful to each of my thesis committee members who dedicated much of their professional time to ensure diversity in contributions to this research as well as maintain its integrity in the tradition of doctoral research. Prof. Kavita Nair played an integral role as chairperson for the committee, and steered the project toward success from its early conception. Prof. Jonathan Campbell offered continuous and valuable analytical input that filled voids in the research that few others might anticipate when focused so closely on the task at hand. Prof. Mary Beth Makic brought passion to the issue of HAPU prevention based on her front-line experiences in nursing; her consistent belief that We Can Do It! with her arms flexed gave reassurance to the committee that this research had an important purpose. Dr. Heidi Wald contributed essential policy insight to our understanding of why HAPU prevention is still an issue despite promising signs of improvement. Dr. Manish Mishra was the spark that lit a wildfire in my interest on this topic of pressure ulcer prevention. Since my earliest exposure to working with vi

8 him at Dartmouth, his passion for quality improvement and patient safety became my passion as well. I am grateful for Manish s many late-night reassurances that this research will not go overlooked, nor will it be the end of an enlightening partnership to fix the system of health care delivery from the ground up. Thanks to each of the faculty mentors at the Center for Pharmaceutical Outcomes Research, University of Colorado. Prof. Anne Libby, Prof. Heather Anderson, Dr. Vahram Ghushchyan, and Prof. Patrick Sullivan each kept an open door to my research concerns and offered crucial input to this study s success through the process for doctoral candidacy. I am indebted to my mentors at The Dartmouth Institute who inspired me with their understanding of quality improvement theory. Thanks to Dr. Mark Splaine, Dr. Paul Batalden, Prof. Gene Nelson and Dr. Greg Ogrinc for their input throughout my graduate studies. I hope that I have appropriately built research upon the theoretical foundations that you each instilled. This research would not have happened without the voluntary contributions of such good people. Carol Ruscin and Connie Chambers were indispensible for their access to UHC data. Their reassurance early-on that this research was possible saved it from certain destruction! Thanks also to Paula Gipp, our CWOCN in the trenches at University of Colorado Hospital who related her experiences to me, uncensored. The pilot test participants at University of Colorado and Dartmouth College provided such an important step forward in this study by validating the framework for pressure ulcer prevention and building the survey instrument. From vii

9 University of Colorado, thanks to Dr. Daniel Matlock, Prof. Paul Cook, Dr. Lisa Price, and Dr. Ethan Cumbler. From Dartmouth College, thanks to Mary Catherine Rawls, Mary Jo Slattery, and Carmeleta Beidler. The project was successful because of all the nurses who responded to my request for participation in the survey. Special thanks to the following individuals who gave permission to be acknowledged: Jill Bick, Sarah Lebovits, Carol Hall, Donna Truland, Myra Varnado, Katherine Constable, Molly Holt, Donella Doctor, Mary Montague, Pat Pezzella, Christine Berke, Teri Garin, Susan Klaus, Sharon Lykins-Brown, Felicia Jones, Sunniva Zaratkiewicz, Lisa Nowicki, Robert Maurer, Dawn Carson, Aaron Clousing, Donna Thomas, Carolyn Watts, Cindy Walker, Michael Kingan, Barbara Koruda, Jill Trelease, Melissa Sisko, Christine Baker, Jocelyn Goffney, Tod Brindle, Bill Falone, Mary Anne Lewis, Penny Jones, Cheryl Garnica, Heather McEntarffer, Barbara Provo, Katherine Mehaffey, David Mercer, Shirley Alltop, Christa Heinsler, Teresa Gegax-Martinez, Melissa Pickett, Nancy Bryant, Destinee Eakle, Shirley Sheppard, Merrill Fraser, and Donna Geiger. Thanks also to those respondents who will remain anonymous. Your dedication to patient safety was enough, but your additional time to complete this survey will not be taken in vain to improve your field. I am beholden to my friends that kept me sane throughout the past five years. To Adele Wilhelm, thanks for keeping me upright throughout the emotional and physical distress of a doctorate. To Dr. Brett McQueen and Dr. viii

10 Julia Slejko, thank goodness Colorado brews good craft beers, and even more thanks that you both were up to drinking a few with me. And most importantly, words cannot express my full gratitude to my family for their undying support and inspiration. My grandparents William, Florence, Normand and Genevieve made ultimate sacrifices which in turn gave me this opportunity that I have tried not to take for granted. To my parents, Bill and Judith, and to my sister, Lauren, I am forever grateful. You always told me I could, and never said I couldn t. You listened to everything I had to say, and you always responded in a caring, loving and supportive manner. For the greatest family in the world, this dissertation belongs to you. ix

11 TABLE OF CONTENTS CHAPTER I. INTRODUCTION.. 1 Background 1 Problem Statement... 6 Conceptual Framework...7 Specific Aims. 9 Research Design.12 II. LITERATURE REVIEW AND CONCEPTUAL FRAMEWORK 16 Overview Review of Related Literature 18 Conceptual Framework. 76 III. RESEARCH DESIGN Specific Aims Aim Aim Aim Considerations for Study Design 120 IV. RESULTS Study Population Aim 1 Results: Survey of HAPU Prevention Aim 2 Results: QI Intervention Effectiveness. 147 Aim 3 Results: QI Strategy. 167 x

12 V. DISCUSSION 176 Aim 1 Discussion Aim 2 Discussion Aim 3 Discussion Conclusions REFERENCES APPENDIX A. KEY TERMS AND DEFINITIONS. 211 B. SURVEY INSTRUMENT. 218 xi

13 LIST OF TABLES Table 2.1 Theoretical QI elements of the best-practice framework organized by domain, taken from Nelson et al. (2007) Modified best-practice framework to reflect elements of QI for HAPU prevention Data sources and parameters utilized for the proposed study Mock data of number of hospitals representing QI interventions in each hospital quarter Power of samples for an effect size analysis Patient demographics among all UHC Hospitals ( ) Hospital-level characteristics among responding UHC hospitals a Patient-level characteristics among responding UHC hospitals b Patient-level characteristics among non-responding hospitals Characteristics of HAPU prevention for UHC Hospitals (N=55) Hospital-level characteristics of QI adoption Overall adoption of QI interventions in hospitals pre- and post-cms policy intervention Elements of the QI best-practice framework by domain Scope of QI strategies by year for UHC hospitals (N=55) Overall changes in scope of QI interventions for hospitals pre- and post-cms policy intervention Overall trends in scale of QI strategies by UHC hospital (N=55) Results of t-test for overall changes in QI scale at UHC hospitals pre- and post-cms policy intervention Self-reported influences of HAPU prevention, Unadjusted results of effect size analysis by QI intervention 153 xii

14 4.14 Statistical test of effect sizes for early adopters compared to late adopters of QI interventions for HAPU prevention across UHC hospitals Statistical test of effect sizes for immediate adoption of QI interventions compared to lag periods for HAPU prevention within UHC hospitals Analysis of Covariance (ANCOVA) results for overall QI adoption relative to HAPU incidence Analysis of Covariance (ANCOVA) results for QI adoption by bestpractice domain relative to HAPU incidence Analysis of Covariance (ANCOVA) results for QI adoption by elements of the best-practice framework relative to HAPU incidence Analysis of Covariance (ANCOVA) results for QI adoption by elements of the best-practice framework relative to HAPU incidence, without controlling for CMS nonpayment policy Sensitivity analysis of effect size results by quarterly adjustments in QI adoption periods Effect sizes of combinations of effective QI interventions QI strategies of hospitals with the greatest reductions in HAPU incidence rates QI strategies of hospitals with the lowest average HAPU incidence rates 174 xiii

15 LIST OF FIGURES Figure 2.1 Inverted pyramid of multi-tier health care system with evidencebased and quality measures, modified from Quinn (1992) Four levels for organization theory pertaining to HAPU prevention Structured improvement of hospital systems with PDSA cycles, modified from Nelson et al. (2007) Time-dependent effectiveness of QI Intervention, modified from Mohr (2000) Reflexivity of QI interventions on HAPU prevention Strengths and weaknesses of research designs for effective QI strategies, modified from Berwick and Goldmann (2008) Illustrated design of separate-sample pretest-posttest, modified from Campbell and Stanley (1963) Illustrated design of separate-sample pretest-posttest with control group, modified from Campbell and Stanley (1963) Multiple time-series design, modified from Campbell and Stanley (1963) A conceptual framework for implementation science through the translation of evidence-based practice into quality improvement for pressure ulcer care, modified from Gonzalez et al. (2012) Box diagram of inpatient populations and related data sources Scope of QI strategies by quarter for UHC hospitals Increases in scale by combination of leadership QI interventions Increases in scale by combination of staff QI interventions Increases in scale by combination of IT QI interventions Increases in scale by combination of P&I QI interventions 140 xiv

16 4.6 Increases in scale by leadership QI intervention Increases in scale by staff QI intervention Increases in scale by IT QI intervention Increases in scale by P&I QI intervention Graphical depiction of time-dependent effect sizes by leadership QI intervention Graphical depiction of time-dependent effect sizes by staff QI intervention Graphical depiction of time-dependent effect sizes by IT QI intervention Graphical depiction of time-dependent effect sizes by performance and improvement (P&I) QI intervention Graphical depiction of time-dependent effect sizes by combination of QI interventions 171 xv

17 CHAPTER I INTRODUCTION Background Hospital-acquire pressure ulcers (HAPUs) are a leading issue of patient safety in the U.S. HAPUs are a classification of hospital-acquired condition (HAC) that result in extremely high hospital costs as well as patient morbidity and mortality. Despite physicians and nurses greatest strides to prevent conditions such as HAPUs, the Institute of Medicine (IOM) landmark report First Do No Harm made clear that in many cases, just the opposite was occurring from their efforts. 1 Occasionally patients left in a clinician s care can endure more harm than good. HACs are identifiable, preventable errors in medical care that involve serious consequences for patients. 2 HACs represent a diverse set of complications such as HAPUs, central line infections, catheter-associated urinary tract infections, air embolisms, and falls. 3 HAPUs, like other HACs, can be avoided by following evidence-based prevention guidelines. 4 In the past decade medical literature has incorporated a growing evidence base for instituting techniques associated with HAPU prevention. 5 Furthermore, recent policy interventions at state and federal levels incentivize hospitals to prevent HAPUs. Pressure Ulcer Prevalence, Incidence, and Economic Factors On August 1, 2007, the U.S. Centers for Medicare and Medicaid Services (CMS) announced that they be implementing a nonpayment policy that would 1

18 cease to cover the cost for certain HACs, including pressure ulcers. 6 Previously, CMS included payment for HACs as part of the prospective payment for a primary diagnosis, which in essence functioned as a system of monetary reward for poor care. 7 This new CMS nonpayment policy was not enacted until October 1, 2008; however, the announcement of reimbursement changes immediately incentivized hospitals to do more to reduce the incidence of these complications. 8 Since this policy was implemented, hospitals have not received payment for treating pressure ulcers unless the condition was noted as present on admission (POA). Pressure ulcers are localized injuries to the skin and underlying tissue, which usually occur over a bony prominence as a result of pressure in combination with shear or friction. 9 Presenting pressure ulcers range from a nonblanchable erythema of intact skin to deep ulcers reaching bone which can lead 10, 11 to bacteremia, sepsis, and death. This range lends to pressure ulcer diagnosis by stage, varying from stage I (mild) to IV (severe) in terms of progressive morbidity. 12 Pressure ulcers can also be classified as unstageable or as a suspected deep tissue injury (DTI), which appears as a blood-filled blister. 13 HAPU incidence grew at an alarming rate in the 1990s and early 2000s. The Agency for Healthcare Research and Quality (AHRQ) recently determined that HAPU incidence increased by approximately 62% from 1993 to 2003, and increased by 80% for pressure ulcers in all patient care settings from 1993 to , 15 A national cross-sectional study by Whittington and Briones 16 (2004) 2

19 measured HAPU incidence near 7%, and a report by the National Pressure Ulcer Advisory Panel (NPUAP) suggested that HAPUs incidence ranges anywhere from 0.4% to 38% across various acute care settings in the U.S. 17 In 2009, HAPUs affected approximately 3 million adults at a national economic burden estimated at $11 billion for the year. HAPUs range in cost by individual patient 5, from $500 to $70,000 per case. In addition to being expensive, advancedstage pressure ulcers can be fatal. Overall HAPU mortality is estimated at 7.2%. 21 The overwhelming epidemiologic and economic burden that HAPUs present in the U.S., together with the patient suffering they cause, has prompted clinician advocate groups to begin campaigning for a universal goal of 0% HAPU incidence. 22 Pressure Ulcer Prevention Practices The NPUAP and Wound, Ostomy and Continence Nurses (WOCN) Society have researched and established evidence-based guidelines that clinicians should follow as a HAPU prevention protocol. 23 Both the NPUAP and WOCN have published an effective HAPU prevention protocol which includes modern support surfaces (e.g. pressure-relieving hospital mattresses), frequent patient repositioning, a focus on improving patient nutrition, and managing moisture and incontinence. 23 A risk-assessment instrument such as the Braden Scale should be used on patients as part of a prevention protocol upon admission and at regular intervals thereafter (e.g. daily or with changes in patient condition) to assess individual need for the extent of HAPU prevention 5, 24 modalities. 3

20 The HAPU prevention protocol is often organized by hospital staff in the form of a checklist or other intervention, and studies have found the prevention protocol is cost-saving compared to treating a pressure ulcer The prevention protocol is effective when implemented consistently. Nonetheless, without customizing the design of the prevention protocol in a way that simplifies implementation for intended hospital clinicians, prevention protocol initiation is subject to variability through each clinician s subjective interpretation. Such variability can jeopardize successful efforts of preventive care. 28 Examples of variability include inconsistent intervals between patient repositioning, or varied interpretations of how certain patient factors (e.g. age, obesity) impact Braden scoring. 29 Evidence suggests that clinical deviation from consistently initiating the entire prevention protocol places patients at greater risk for pressure ulcer 5, development. Quality Improvement Efforts for Prevent Pressure Ulcer Prevention Given the complexity of HAPU prevention and protocol initiation, 33 clinicians could benefit from a dynamic strategy to incorporate the prevention protocol into the clinical process with consistent initiation. According to the Gonzalez et al. 34 (2012) framework of implementation science, evidence-based practice does not translate directly into clinical practice. A sustainable strategy that organizes health care delivery systems with greater efficiency assures improved patient outcomes. Quality improvement (QI) strategies could effectively increase adoption of the prevention protocol by organizing evidence- 4

21 based practices for improved outcomes. In addition, QI strategies have potential to reduce practice variation through subtle integration into the protocol. 35 Berwick 36 (2007) describes QI as a construct that can facilitate clinicians to better understand the health care environment and nurture effective action on care. Preventive QI interventions take many forms including checklists, data sharing, team building, and other media that are customized for particular hospital cultures with the intent of increasing prevention protocol initiation. 37 QI interventions are often targeted for use in clinical microsystems since these are the smallest replicable units of inpatient care and represent the multitude of 38, 39 different cultures and services within a single hospital. Robust QI interventions should span multiple clinical microsystems while still achieving global aims. 40 A clinical microsystem can be defined as the combination of a small group of clinicians who work together in a defined setting to provide care to individuals (i.e. a discrete subpopulation of patients). 36 As a functioning unit, microsystems have clinical and business aims, linked processes, a shared environment of information and technology, and delivers care that can be measured as performance outcomes. 41 Microsystems evolve over time with the inclusion of new practices and QI initiatives to continously improve care and other available services. Comparative effectiveness research (CER) of QI interventions in clinical microsystems is possible with proper resources. The University HealthSystem Consortium (UHC) represents over 180 academic medical centers in the U.S. 5

22 that pool data on multiple patient safety and quality metrics including HAPU outcomes. UHC provides a valuable data source for studying HAPU prevention since these hospitals study evidence-based practices including the HAPU prevention protocol. As research institutions, UHC hospitals often explore the effects of QI intiatives. Though UHC does not offer magnification of QI approaches in individual clinical microsystems, it is possible to view each hospital as a separate clinical entity utilizing QI with evidence-based practices. The establishment of UHC data pooling prior to CMS nonpayment policy for HACs permits studying the effectiveness of QI interventions for HAPU prevention while controlling for changes in policy. Problem Statement Limited empirical evidence on the use of QI interventions indicates a lacking certainty about whether QI interventions are effective at improving patient 7, 42 outcomes. Furthermore, CER lacks information on direct comparisons of QI strategies, which are combinations of different QI interventions. 35 These uncertainties represent a gap that is pronounced for HAPUs since CMS nonpayment policy was instated for HACs; hospitals have been focused on novel approaches to improving HAPU prevention. Researching the effectiveness of QI interventions that, when coupled with the evidence-based prevention protocol, reduce HAPU incidence would add to the body of literature about QI effectiveness. In addition, such research could have direct implications on how clinicians initiate the prevention protocol with patients who are at-risk for HAPUs. 6

23 Coupling the prevention protocol with a recommended QI strategy using a CER framework can reduce practice variations and standardize hospital care, thereby improving HAPU prevention. Problem Significance The IOM recently published 100 top priority areas for CER, including: compare the effectiveness of different quality improvement strategies in disease prevention, acute care, chronic disease care, and rehabilitation services for diverse populations of children and adults. 43 By developing a list of recommended QI interventions for HAPU prevention, hospitals can direct their current efforts in HAPU prevention towards such QI interventions which are evidently effective. Following time-dependent adoption of QI interventions in conjunction with CMS nonpayment policy also offers insight as to whether such reimbursement policies effectively trigger reductions in HAPU incidence at hospitals and if such policies can transform practice at the level of the clinical microsystem. 41 Conceptual Framework The conceptual framework of this proposal is based on Gonzalez et al. 34 (2012) which suggests that in order to achieve desirable patient outcomes with evidence-based practice, the system of health care delivery requires extensive restructuring. QI theory represents a mechanism for organizing delivery systems to properly implement evidence-based practices, such as the HAPU prevention protocol. QI theory is organized into actionable items referred to as 7

24 interventions, and multiple effective, complementary interventions that bolster evidence-base practice contribute to an overall QI strategy. Because of the influence that CMS non-payment policy has on improving HAC outcomes, organization theory according to Trisolini (2011) provides an understanding of how QI strategies are effective in the hospital setting (2011). 44 First, general trends suggest that adoption begins with support from hospital leadership and active participation from clinicians who initiate the HAPU 37, 45 prevention protocol. Second, QI interventions should bolster use of the current evidence-base practice (i.e. the prevention protocol), rather than alter or eliminate proven practices. Third, QI interventions should be validated as effective components of practice in terms of clinician buy-in and meeting global aims (i.e. reduced HAPU incidence). 22 Measuring these two aspects of validity often require clinician surveillance for consistent initiation, and closely monitoring quality metrics such as HAPU incidence rates. Using a quasi-experimental approach, this conceptual framework allows for different hospital sites to naturally test various QI interventions. Hospitals with statistically significant reductions in HAPU incidence combined with QI adoption can contribute data on effective QI interventions. Modelling this approach as an interrupted time-series to control for CMS nonpayment policy provides a basis for direct comparative effectiveness of QI interventions. This framework culminates as CER since QI strategies used in conjunction with the HAPU prevention protocol are based on consideration for all possible combinations of QI interventions in use by hospitals. 8

25 Best-practice Framework of Quality Improvement Interventions Nelson et al. developed a best-practice framework for QI interventions by suggesting that QI approaches should encompass four domains of clinical practice: (1) Leadership; (2) Staff; (3) Information and Information Technology; and (4) Performance and Improvement. 28 These four domains are scaled by multiple QI interventions (e.g. checklists, team huddles, etc.), which hospitals can consider for adoption with respect to their financial, personnel, and time dependent resources. QI strategies consist of multiple QI interventions for a targeted outcome, such as HAPU prevention. A QI strategy with interventions from multiple practice domains is believed to have good scope. Given multiple QI interventions within each domain, a strategy that utilizes more than one element from a domain is also referred to having broad scale. Dynamic QI strategies are broad in both scope and scale. According to Berwick 46 (2009), an effective QI strategy should support patient-centered care, integrate with diverse clinical microsystems, and is dynamic with regards to the best-practice framework. 47 Specific Aims Current evidence lacks information on the effectiveness of QI interventions for improving patient safety and quality care. Specific to HAPU outcomes, there may be multiple QI interventions that contribute to improved HAPU prevention, but there is no CER on specific QI interventions or strategies identifying an effective approach. Three aims in this proposal will test whether QI 9

26 interventions are effective vehicles for implementing evidence-based practices with regards to HAPU prevention, and determine which QI interventions are most effective. These aims will utilize HAPU incidence data from a retrospective cohort of UHC hospitals as well as survey data of QI interventions by hospital. The quasi-experimental study is designed as an interrupted time-series of changes in HAPU incidence associated with adoption of hospital-level QI interventions, overlapping with CMS nonpayment policy. Aim 1 The first aim is to describe characteristics of inpatient QI interventions for HAPU prevention and measure hospital-level changes in QI adoption before and after CMS nonpayment policy. A survey will yield information on QI interventions utilized by UHC hospitals for HAPU prevention. There are multiple sub-aims of this objective: a. Using key informant interviews with multiple QI experts at two hospitals i. Develop a best-practice framework for QI interventions based on Nelson et al. 28 (2007) that pertains specifically to HAPU prevention. ii. Develop, refine and pilot test a survey instrument that could be used to collect longitudinal data on adoption of QI interventions. b. Administer a web-based survey to collect data on hospital-level patterns of QI adoption for HAPU prevention at UHC hospitals. 10

27 c. Measure changes in hospital-level QI adoption patterns before and after CMS nonpayment policy at UHC hospitals. d. Describe national patterns of QI intervention typologies in terms of scope and scale before and after CMS nonpayment policy at UHC hospitals. Aim 2 The second aim is to measure the effectiveness of QI interventions at reducing HAPU incidence before and after CMS policy intervention. This aim will yield data on the most effective QI interventions involved in HAPU prevention while controlling for environmental interruptions caused by CMS nonpayment policy. Aim 2 consists of two sub-aims based on concerns for overall effectiveness and variability of HAPU prevention: a. Measure the effectiveness of QI interventions at reducing HAPU incidence. b. Measure changes in variance of HAPU incidence with QI and policy. Aim 3 The third aim is to characterize hospital-level QI strategies and compare the effectiveness of these strategies according to HAPU incidence rates. This aim will yield information about the comparative effectiveness of combinations of QI interventions practiced by UHC hospitals that are most successful at preventing HAPUs. There are two sub-aims of this objective: 11

28 a. Identify effective combinations of QI interventions for HAPU prevention. b. Characterize QI strategies of hospitals that successfully prevent HAPUs. Research Design This proposed study will provide hospitals with a catalogue of the most effective QI interventions in conjunction with the HAPU prevention protocol. The study design is a retrospective cohort of academic medical centers to perform an interrupted time-series quasi-experiment of QI intervention effectiveness. Empirical methods will also determine if any QI interventions are more effective when practiced in combination, or whether high-performing hospitals have specific QI strategies worth emulating. Moreover, the study design will illustrate the impact of CMS nonpayment policy on motivating hospitals to improve preventive care through the adoption of novel clinical approaches. Purpose This study s purpose is to identify, characterize and compare the effectiveness of QI interventions for HAPU prevention at the hospital-level. These objectives will be achieved with HAPU incidence data from a retrospective cohort of UHC hospitals as well as survey data of QI interventions by hospital. An effect size analysis will measure changes in HAPU incidence associated with adoption of hospital-level QI interventions while incorporating the presence of CMS 12

29 nonpayment policy. QI interventions are derived from the best-practice framework developed by Nelson et al. 28 Once characterized in the survey, hospital-level QI interventions will be ranked by effectivess in terms of reducing HAPU incidence. QI strategies from hospitals with the greatest improvements during observation will also be included among recommended QI interventions. Other hospitals without effective QI strategies can emulate the results of this study to improve consistent implementation of the HAPU prevention protocol. Nature The proposed study will fill a gap by identifying QI interventions that support the HAPU prevention protocol by reducing HAPU incidence. Furthermore, this study is a contribution to CER evidence with a focus on recommending effective QI strategies for HAPU prevention based on hospitallevel performance. The study is structured into three specific aims that rely on analysis of HAPU outcomes among UHC hospitals. As a quasi-experimental design, the study design directly compares multiple hospital QI interventions in a natural setting. Since this design utilizes time-series and mixed methods to achieve predictive knowledge, there is greater certainty about the generalizability of the findings as opposed to traditional controlled trials. 48 Epidemiology. Quarterly HAPU incidence in each UHC hospital will be calculated as a function of hospital admissions for all admitted inpatients. Patients diagnosed with a stage III or IV pressure ulcer not POA will be included in the calculation based on AHRQ s patient safety indicator for HAPUs. These 13

30 incidence rates will then be linked to each QI intervention practiced by a hospital during a specified quarter for an effect size analysis. This calculation of incidence will update previous literature on the epidemiology of HAPUs specifically for academic medical centers. Furthermore, this information will provide the field of nursing with a benchmark for HAPU incidence based on findings from leading hospitals in the U.S. Survey. The survey in Aim 1 uniquely follows the best-practice domains developed by Nelson et al. to characterize QI interventions. This survey will capture information on how UHC hospitals follow the HAPU prevention protocol. The survey will also identify external influences on HAPU prevention and QI adoption, such as CMS nonpayment policy. Overall, this survey identifies QI interventions in practice and provides a foundation for forming QI strategies within hospitals. Observational Data. The study aims to analyze nationally-representative UHC data of academic medical centers throughout the country. UHC hospitals are diverse geographically and in terms of the populations each one serves. HAPU outcomes are closely monitored in UHC data, thus providing a reliable data source. Many of the UHC affiliates include schools of nursing, or have obtained Magnet recognition for nursing excellence, and therefore emphasize the importance of novel methods in HAPU prevention such as use of the prevention protocol and investigation of QI interventions for HAPU prevention. 14

31 Significance The proposed study follows the timeline of changes in HAPU care for analysis. Changes to CMS nonpayment policy created new incentive for hospitals to prevent HAPUs. 6 However, the period following adoption of a QI intervention is likely to be the period where care actually improved. Results of this proposed study will be used to drive the adoption of preventive QI in the hospital setting. This study will also identify QI strategy bundles that can protect patients from HAPU incidence, especially in hospitals lacking resources to carry out their own trials on QI interventions. Implications of Research Findings This study will track changes in HAPU incidence and inform the field of outcomes research how CMS policy can affect hospital practice. Contributions to QI literature will be made in the form of a catalogue of QI interventions that are best utilized for purposes of preventing HAPUs as well as potentially other HACs or preventable acute illnesses. In summary, applying a quasi-experimental design with effect size calculation will be a novel use of mixed methods, thus securing the robustness of the findings. 15

32 CHAPTER II REVIEW OF LITERATURE AND CONCEPTUAL FRAMEWORK Overview When the Centers for Medicare and Medicaid Services (CMS) modified its Inpatient Prospective Payment System in 2007, its intention was no longer rewarding hospitals with payment for preventable harms. 7 This policy intervention was a pay-for-performance (P4P) based initiative, that is, a prospective system of value-based payment that CMS implemented following demands for provider accountability to high quality and cost-effective health care. 49 A U.S. congressional mandate required that CMS focus its attention on reimbursement for hospital-acquired conditions that were considered avoidable 27, 50 and their prevention cost-effective. Since October of 2008 hospitals no longer receive payment for a number of hospital-acquired conditions (HACs) such as hospital-acquired pressure ulcers (HAPUs). 3 Although such conditions are not always avoidable, the CMS nonpayment policy has made HAPU prevention an internal issue for hospitals to decipher. 51 The HAPU prevention protocol was institutionalized by CMS nonpayment policy since it was the primary evidence-based guideline that could prevent HAPUs. Organizations prioritizing HAPU prevention such as the Agency for Healthcare Research and Quality (AHRQ) and the National Pressure Ulcer Advisory Panel (NPUAP) have researched and endorsed the prevention protocol as an effective measure in HAPU prevention among all patients. 23 Hospital 16

33 clinicians who have been willing to incorporate the prevention protocol into standard patient care may relieve hospitals from an elevated financial burden, as well as contribute to improved quality care and patient safety. Physicians and nurses exist at the interface of caring for the patients needs and preventing HAPUs. Prevention is a complex issue because the highrisk patients for HAPUs are of older age, higher acuity, and malnourished when compared to other hospitalized patients, but any patient may still be at risk. 52 Given all these tasks for the prevention of one acute hospital condition, clinicians could benefit from a dynamic strategy of quality improvement (QI) interventions that reengineers the delivery system of HAPU prevention for consistent initiation of the prevention protocol. Nelson et al. propose a best-practice framework for QI interventions, which suggests that QI should be diverse in scope, intervening in four domains of clinical practice: (1) Leadership; (2) Staff; (3) Information Technology; and (4) Performance and Improvement. These four domains are scaled by 20 elemental QI interventions, which hospitals can consider for adoption based on their financial, personnel, and time dependent resources. To this point there has not been an investigation of which QI interventions work most effectively in support of the HAPU prevention protocol. It could be that there is one or multiple interventions that, when adopted into health care delivery, improve implementation of the prevention protocol. Given that there are many types of QI interventions and multiple iterations of a QI strategy within the bestpractice framework, the field would benefit from comparative effectiveness research on the QI intervention(s) that best support initiation of the HAPU 17

34 prevention protocol. 35 Though it is difficult to measure frequent implementation of the prevention protocol at the interface between the patient and clinician, reductions in HAPU incidence rates could be correlated with reduced variation of protocol initation. 53 Review of Related Literature Before discussing the conceptual framework and study design that are addressed in this proposal, I present an overview of the philosophical rationale for this study and review the related empirical literature. Additional definitions to key terms are located in Appendix A. Pressure Ulcer Care Pressure ulcers are localized injuries to the skin and underlying tissue. The NPUAP defines pressure ulcers as occurring over a bony prominence as a result of pressure in combination with shear or friction. 9 A number of contributing factors exist that are associated with pressure ulcer risk, such as age and length of stay (LOS). Allman s 54 (1989) seminal review of pressure ulcer outcomes suggested that as many as 50% occur in patients over the age of 70. Among geriatric patients, pressure ulcers are associated with a fourfold increase in the risk of death. Bergstrom and Braden 55 (1992) made observations about patients in multiple skilled nursing facilities who were mostly seniors (i.e. 65 years or older). The observational study found that pressure ulcers developed in over 30% of seniors, in contrast to 0% of patients younger than 65 who had an anticipated 18

35 LOS greater than 10 days. Given that the average LOS in the Braden and Bergstrom study was 28 days, they concluded that high LOS was associated with increased pressure ulcer risk. Interestingly, randomized controlled trials (RCTs) of pressure relieving surfaces such as Inman et al. 56 (1993) do not include inpatients at-risk for a pressure ulcer unless they have stayed greater than three days. More recently, Magahalles (2007) described being hospitalized and elderly as the primary risk factors for pressure ulcers. Shea 12 (1975) was the first to classify pressure ulcers by four stages. Black (2005) reported on deep tissue injuries (DTIs) as a type of wound being distinct unique from the system of pressure ulcer staging, but suggested that DTIs should not be overlooked or treated differently. 13 In 2007, NPUAP updated the definitions for pressure ulcer stages and included in the definition unstageable pressure ulcers and suspected DTIs. 9 Stage I is defined by intact skin with non-blanchable redness in a localized area. In stage II, there is partial thickness loss of the dermis which can result in a shallow open ulcer. Subcutaneous fat may be exposed in stage III as all top layers of skin are loss, but bone muscle and tendons still remain protected. Stage IV pressure ulcers can include exposed bone, tendon or muscle, and undermining or tunneling due to infection can appear. Suspected DTIs can evolve quickly into worsening stages such as III or IV since their detection is at first difficult in some patients. 9 According to AHRQ 57, a pressure ulcer can be classified as hospitalacquired if the patient has an LOS of greater than five days and no pressure ulcers were present on admission (POA). AHRQ only includes stage III and IV 19

36 pressure ulcers in its definition of HAPUs. Certain patients are exempt from the definition of a HAPU, including those who are pregnant, paralyzed (i.e. hemiplegia, paraplegia, or quadriplegia), diagnosed with spina bifida, had a skin graft, or transferred from another health care facility. In exempt HAPU cases, the hospital is not directly responsible for the cost associated with pressure ulcer care among patients covered by CMS. 58 Prevention Protocol. The most current version of the HAPU prevention protocol was published by Ratliff and Bryant 23 (2003) following a meeting of the Wound, Ostomy, and Continence Nurses (WOCN) Society. The protocol stated that clinicians should consistently practice five key components: (1) pressure ulcer risk-assessment; (2) frequent patient repositioning; (3) transferring patients onto proper support surfaces; (4) managing moisture and incontinence; and (5) maintaining patient nutrition. The NPUAP endorsed this protocol with the caveat that not all pressure ulcers are avoidable, 51 so clinicians should still regularly reevaluate patient risk after admission. Pressure ulcer risk-assessment refers to examining a patient s condition with an instrument such as the Braden Scale. The Braden Scale was introduced by Bergstrom and Braden 24 (1987) as a tool to risk-stratify patients for pressure ulcers, thereby allowing nurses to focus their preventive efforts on patients who are most susceptible. The scale consists of six domains by which clinicians should score the patient. First is sensory perception, under which it is expected that the patient should be able to respond to pressure-related discomfort. The second domain is related to patients skin exposure to moisture; patient skin that 20

37 is in direct contact with sources of moisture such as perspiration or urine is of greater risk for pressure ulcer development. Third is the patient s measure of physical activity. Patients who can remain mobilized either by walking or moving out of the bed occasionally are more likely to avoid pressure ulcers. Fourth is the patient s measure of mobility while in bed. For patients who remain bedfast, their ability to still turn side-to-side under their own strength is a meaningful advantage. The fifth domain pertains to nutrition; patients who can maintain an appetite and regular eating schedule are more likely to maintain strength, mobility and physical activity. The final domain of the Braden Scale is friction and shear, which relates to pressure applied directly on different areas of skin. Scoring on the Braden Scale can range from a low of 6 points to a high of 23 points. Braden scores of 19 or higher indicate that the patient is not at great risk for a pressure ulcer. Scores of range indicate that the patient is at risk for a pressure ulcer. A patient develops moderate risk in the range of 13-14, high risk from 10-12, and very high risk of 9 points or less. As the patient s risk increases within these classifications, components of the prevention protocol such as patient repositioning should be repeated more frequently. 59 Magahalles 29 (2007) declared that the scoring system of the Braden Scale was accurate, but adjustable for patients with risk factors not explicit to the instrument. Magahalles suggested that patients with other present risk factors should be advanced to the next level of risk, such as from moderate risk to high risk. These risk factors include advanced age (e.g. seniors), increased LOS, chronic illness, diastolic blood pressure below 60 mmhg, and uncontrolled pain. 21

38 Furthermore, bariatric patients with a body-mass index (BMI) greater than 40 should be placed on pressure relieving support surfaces. Braden and Bergstrom reaffirmed the use of the Braden Scale as a predicatively valid risk-assessment instrument in multiple populations including the institutionalized elderly 55 (1992) and nursing home inhabitants 60 (1994). Lyder et al. 61 (1998) investigated the use of the Braden Scale in elderly blacks and Hispanics, thus expanding the validity of the instrument to minority populations. Use of the Braden Scale as a part of the prevention protocol is commonplace in adult patient populations. An alternative to the Braden Scale for adult patient risk-assessment is the Norton Scale 62 which is more commonly used in European hospital settings, and clinicians utilize the Braden Q Scale 63 for pediatric populations. Keast et al. 64 (2008) described the other components of the prevention protocol in greater detail than Ratliff and Bryant did previously. Frequent patient repositioning should be fulfilled once every 4-6 hours, however greater frequency such as 1-2 hours would increase patient safety. Patients that are indicated atrisk for a pressure ulcer based on their risk-assessment should be transferred onto a proper preventive support surface such as an air-fluidized mattress or have an overlay placed on top of their current bed mattress. The patient s skin should be moisturized daily using skin cream, and those with incontinence issues should have bed sheets replaced regularly. Finally, patients should be fed on schedule or administered intravenous nutritional supplements if oral intake is not possible. 22

39 A systematic review by Reddy et al. 65 (2006) included randomizedcontrolled trials (RCTs) that tested the efficacy of different components of the prevention protocol. Reddy s findings validated that using specialized support surfaces, patient repositioning, optimizing nutritional status, and managing moisture and incontinence were all appropriate strategies that contribute to pressure ulcer prevention. Treatment Options. When prevention efforts fail and a pressure ulcer develops, clinicians should follow a standard wound care protocol. Brem and Lyder 66 (2004) outline wound care as beginning with a physical examination, followed by debridement and wound moistening. The wound should be tested for bacterial culture, which can determine if the protocol includes the use of antiseptics. Surgery to remove the infected area is a final option if the previous steps fail to suffice in the healing process. The physical examination consists of a comprehensive evaluation for effective management of the wound. Lyder 5 (2003) outlines the evaluation steps. First, the wound should be cleansed with saline solution, avoiding antiseptics, and then measure the depth of the wound and note the amount of exposed tissue. This initial evaluation allows the clinician to declare a stage for the pressure ulcer and determine next steps. There are two commonly used tools to aid clinicians in staging a pressure ulcer that can be used during wound care initiation. Bates-Jensen 67 (1997) developed the Pressure Sore Status Tool (PSST), and Stotts et al. 68 (2001) developed the Pressure Ulcer Scale for Healing (PUSH). The PSST is very 23

40 sensitive to staging the wound with 13 separate items, but it takes a long time to complete. The 3-item PUSH tool is simpler and quicker to use, but less sensitive to subtle variations in patient wounds. Once the pressure ulcer is staged, the clinician should commence wound care. Cannon and Cannon 69 (2004) emphasize that pressure on the wound site must be relieved immediately. Following pressure relief, debridement with a topical dressing is necessary. Lyder (2003) differentiates between dressings applied with gauze and non-gauze dressings, of which non-gauze dressings are more expensive but also more cost-effective. 5 Flack et al. 70 (2008) also argues for wound healing with negative-pressure Vacuum Assisted Closure (VAC) therapy since it surpassed both traditional and advanced wound dressings in a comprehensive cost-effectiveness analysis. Lyder 5 (2003) describes surgical intervention as a viable alternative for full-thickness pressure ulcers, although recurrence rates are higher. Surgical interventions include direct closure, skin grafts, skin flaps, musculocutaneous flaps, and free flaps. Singer and Clark 71 (1999) warn against the use of allografts, a more common type of skin graft which can grow quickly with surrounding skin. Allografts are not appropriate for permanent coverage of fullthickness pressure ulcers. Autologous grafts are recommended in cases of advanced-stage pressure ulcers. In addition to wound care, Breslow et al. 72 (1993) demonstrated the importance of managing nutrition with increased dietary protein for improved healing. In this study, malnourished patients who received nutrients consisting of 24

41 24% protein had greater shrinkage in the surface area of a pressure ulcer than patients in a control group with a 14% protein diet. Pressure Ulcer Outcomes Epidemiology. A seminal article by Braden and Bergstrom 73 (1989) described stage I and II pressure ulcers as the most prevalent among high-risk patients. This study screened 2,000 nursing home patients and tracked 200 of them who scored 17 or below on the Braden Scale. Of those patients, 75% developed a pressure ulcer and 95 patients were classified as stage I or II. Few patients developed stage III or IV pressure ulcers in the 2½ year study. Though Braden and Bergstrom did not investigate HAPUs in an acute care setting, their methodology and predictive validation of the Braden Scale provided a framework for further research in HAPU incidence, prevalence, and risk. Since the Braden and Bergstrom study, national bodies have directed efforts towards characterizing HAPU epidemiology throughout the 1990s and early 2000s. Bergstrom et al. 18 (1992) reported for the U.S. Department of Health and Human Services that as many as 3 million adult Americans are affected by pressure ulcers. AHRQ 14 (2006) measured that HAPU incidence increased by approximately 62% from 1993 to 2003, and increased by 80% for pressure ulcers in all patient care settings from 1993 to The NPUAP estimated that HAPU incidence ranges anywhere from 0.4% to 38% across 16, 17 various acute care settings in the U.S. In addition to their widespread incidence and prevalence, advanced-stage pressure ulcers can be fatal. Reddy et al. 21 (2008) cited a national HAPU- 25

42 attributed mortality rate of 7.2%. Makic 74 (2007) reported that this mortality rate can total as many as 60,000 deaths per year. Padula et al. 35 (2012) reported a higher mortality rate of 16.6% among stage III and IV HAPUs in a national query of academic medical centers listed in the University HealthSystem Consortium (UHC) between Lindenauer et al. 75 (2012) discussed that for a related condition, hospitalacquired pneumonia, infection mortality rates have dropped between However, this fact is deceiving due to recent changes in diagnosis coding that are not captured by UHC outcomes reporting. Mortality rates have actually remained relatively constant among pneumonia patients at UHC hospitals, as has secondary diagnosis of pneumonia. Lindenauer et al. is indicative of inconsistent coding procedures, which becomes a limitation of any epidemiologic outcome for hospital-acquired conditions. An epidemiologic study of HAPUs by Whittington and Briones 16 (2004) reported the incidence of all staged pressure ulcers (I through IV) at a national level. The study by Whittington and Briones is recognized as one of the strongest available epidemiologic studies of all staged pressure ulcers reported in the past decade. 35 The study tracked over 200,000 patients in acute care settings across the U.S. Pressure ulcer prevalence ranged from 14-17% over a four year period ( ). Pressure ulcer incidence fell from 9% to 7% from 1999 through Whittington and Briones only reported specific incidence rates of the top two pressure ulcer stages, stages I and II; thus this study confirmed earlier results of Braden and Bergstrom. Based on these results, 26

43 incidence of stages III and IV can be calculated as a remainder. Unfortunately there is little clarity about the incidence of individual stages in the analysis by Whittington and Briones. Overall, the field is lacking nationally representative studies of HAPU incidence and prevalence by each pressure ulcer stage. The epidemiology of DTIs is even less quantifiable. However, there is general consensus that stage I and II pressure ulcers are the most prevalent and incident. Since CMS policy intervention in 2008, the field has focused on the relative reduction in HAPUs in the acute care setting with use of the HAPU prevention protocol, though evidence lacks national representativeness. A systematic review by Reddy et al. 65 (2006) investigated pressure ulcer prevention with the complete HAPU prevention protocol. Reddy et al. established that there was a need for well-designed experimental studies to evaluate the epidemiology of HAPUs and concurrently quantify HAPU prevention with the prevention protocol and other techniques. Reddy et al. reviewed RCTs in multiple settings including long-term care, rehabilitation, acute care, and combinations. This systematic review qualitatively evaluated which components of a standard HAPU prevention protocol are efficacious by virtue of the fact that the review was limited 35, 76, 77 to RCTs. It found that all five components of the prevention protocol contributed to reduced HAPU incidence and prevalence, though the information of the study was not quantifiable since the review s qualitative results were limited to Yes or No regarding whether preventive interventions reduced HAPU incidence. 27

44 A meta-analysis of observational data by Comfort 53 (2008) quantified HAPU incidence reduction in the acute care setting. 53 Comfort reported HAPU incidence reduction in terms of an odds ratio, for which the 95% confidence interval was This study included nine tertiary medical centers including academic and non-academic medical centers actively practicing the prevention protocol. Following the Comfort study, Padula et al. 27 (2011) recalculated the confidence interval of odds ratios into a single, weighted arithmetic mean of 0.335, suggesting that the prevention protocol has a central tendency to reduce HAPU incidence. Other epidemiologic literature on HAPUs has refocused from the national impact of HAPU incidence and prevalence over to HAPUs with data specific to the hospital-level (i.e. the clinical macrosystem) 40 and a hospital s clinical subunits (i.e. the clinical microsystem). 38 The process to improve prevention of HACs can begin with an analysis of the current epidemiology in the hospital setting, including incidence, prevalence, as well as quality measures. 78 Yet, even with all the analytical methods available to evaluate the epidemiology and quality measures of HACs in the hospital, it often remains unclear where a clinician or health professional team should begin in data collection. Combining hospitallevel data with data specific to clinical units for HAPUs can provide complementary information about quality care and patient safety that is actionable by clinicians. 79 Lemaster 80 (2007) reported on quarterly HAPU prevalence in two microsystems at Genesis Hospital in Davenport, IA over a 3½ year period. 28

45 Lemaster observed that even though patient care at Genesis included the HAPU prevention protocol, there were still HAPUs observed. HAPU cases ranged by quarter from 0-17 HAPUs in one clinical microsystem, and 0-21 HAPUs in another microsystem. Lemaster presented significant variation in HAPU prevalence between quarters, suggesting that there was more to HAPU prevention than the prevention protocol. Random noise may account for some HAPU incidences in clinical units. Nonetheless, Genesis could have benefited from additional interventions to control external forces on clinical microsystems that counteracted HAPU prevention efforts. A number of epidemiologic studies have since focused less on the central tendency of HAPU outcomes, and more on the variation in HAPU incidence and prevalence. The points of variation for HAPU outcomes may be indicative of periods when initiation of the prevention protocol deviated from consistency as defined by the AHRQ/NPUAP guidelines. Makic 74 (2007) discuss that barriers in acute care practice such as a lack of continuity in skin care interventions inhibit consistent implementation of the prevention protocol, thereby supporting this theory of variation. By measuring time-dependent variation in HAPU incidence and prevalence, hospitals may be able to pinpoint the root of the issue and intervene accordingly. Morton et al. 81 (2010) has established that the use of aggregate hospitallevel data to evaluate HAPU outcomes is necessary to evaluate variation of key measures due to the fact that HAPUs are relatively rare events. Kottner et al. 82 (2010) utilized statistical process control (SPC; also referred to as Shewhart 29

46 charts) charts for aggregate data to determine the level of quality with which clinical staff prevent HAPUs by utilizing recommended components of the prevention protocol (e.g. consistent patient evaluation with the Braden scale). 83 SPC chart analyses are statistical metrics that illustrate control in health care delivery at the microsystem level to study and more deeply understand improvement efforts related to HAPU prevention. Padula et al. 79 (2012) provided examples of macrosystem analysis of HAPU outcomes using regression methods with aggregate data. In addition, Padula et al. analyzed HAPU prevention from the level of the clinical microsystem simultaneously with SPC charts. The study observed 337 incident HAPUs among 43,844 inpatients in four microsystems of a single hospital. A probit regression model predicted the correlation of age, gender, and length-ofstay on HAPU incidence (pseudo-r 2 = 0.096). Panel data analysis determined that for each additional day in the hospital, there was a 0.28% increase in the likelihood of HAPU incidence. An SPC chart of HAPU incidence rates (i.e. p- chart) showed a mean incidence rate of 1.17% remaining in statistical control. An SPC chart of time between HAPU incidences (i.e. t-chart) showed the average time between events for the last 25 HAPUs was days. In one special cause signal, there was a 57-day period between two incidences during the observation period. This particular signal would be indicative of how variation could inform improvement if hospitals could recreate the system of delivery as it existed during that time period. Finally, a third SPC chart (i.e. p-chart) addressing Braden Scale assessment showed that 40.5% of all patients were 30

47 risk-stratified for HAPUs upon admission. Based on this analysis Padula et al. correlated HAPU incidence and risk factors with a target for improvement: more consistent use of the Braden Scale. Economics. Pressure ulcers represent a tremendous economic burden in the U.S. National estimates approach $11 billion among 3 million adults per year. 18, 61 19, 20 HAPUs can range in cost from $500 to $70,000 per patient. These estimates of cost are scaled by pressure ulcer stage since common stage I and II pressure ulcers are less complicated to treat. The extended inpatient stay to treat a stage III or IV pressure ulcers with medical and surgical interventions can escalate costs by thousands of dollars. Beckrich and Aronovitch 84 (1999) calculated exact figures for the cost of HAPUs by stage. For each stage, the following results were published in terms of cost per case: (1) stage I HAPUs cost $125.54; (2) stage II HAPUs cost $125.55; (3) stage III HAPUs cost $13,937.38; (4) stage IV HAPUs cost $14,008.85; and (4) unstageable HAPUs cost $14, These values represent add-on costs for medical and surgical patients who developed a HAPU. Xakellis and Frantz 85 (1996) studied pressure ulcer costs in a managed care organization across multiple health care settings. The mean cost to treat a pressure ulcer, including hospitalization and long-term care, was $2,731 per pressure ulcer. This cost, excluding hospital costs was $489. Considering that some patients in their study had multiple pressure ulcers, the mean cost per 31

48 patient was $4,647. Xakellis and Frantz study separated out pressure ulcer costs in multiple care settings inpatient hospitalization and long-term care. A cost analysis of HAPUs by Pappas 86 (2008) differentiated pressure ulcer costs for medical and surgical patients. Medical inpatients diagnosed with a pressure ulcer saw a cost increase of $2,384 per case. Surgical inpatients experienced a cost increase of $25. Pappas explanation for the cost of treating pressure ulcers is mostly explained by nursing time, which accounts for 50% of wound care costs. Pappas discussion neglected to explain the high cost differential between medical and surgical HAPUs. It could be that the more costly surgical cases (e.g. neurosurgical, cardiovascular, etc.) that do not develop a HAPU could offset the cost increase relative to patients observed with a HAPU. Inpatients who underwent certain less costly procedures could represent those at greater risk. Furthermore, as a caveat to Pappas conclusions, nursing time spent on wound care could be seen as an opportunity cost, since nurses would likely remain occupied with other obligations during a normal shift. A number of studies have also focused on the cost and cost-effectiveness of components of the prevention protocol for HAPU prevention. Besides nursing time, hospital beds represent the most expensive component of the HAPU prevention protocol. Mackey 87 (2005) outlined the costs of pressure-relieving support surfaces (i.e. hospital mattresses and overlays) designed for HAPU prevention through the Medicare reimbursement pay structure. Since the Medicare Part B Support Surface Policy implemented in 1996, hospitals can seek 32

49 partial reimbursement for transitioning inpatients from standard beds to specialized pressure-relieving surfaces for HAPU prevention such as an airfluidized bed. Hospitals can receive reimbursements ranging from $17-32 for overlay/mattress rentals or $ for new overlay/mattress purchases. These beds would typically go into place for patients considered at high-risk for a HAPU. For patients with staged HAPUs, Medicare will reimburse hospitals for more complex bed systems ranging from $377-3,200 for rentals. Iglesias et al. 88 (2006) reported on the Pressure Relieving Support Surfaces (PRESSURE) Trial which examined the cost-effectiveness between utilizing overlays or mattresses for the prevention of HAPUs. With mattresses as a comparator, the incremental cost-effectiveness ratio (ICER) of this analysis was per hospital days avoided. This ICER suggests that mattresses dominated overlays as the preferred approach to HAPU prevention. Xakellis and Frantz 89 (1998) studied the costs of varying degrees of intensity in initiation of the prevention protocol in long-term care. Prior to protocol initiation, the mean cost for prevention and treatment of a HAPU was $113 (SD = $345). Following the prevention protocol, the add-on cost for prevention and treatment was $100 (SD = $157). This early cost study suggested that the prevention protocol reduced overall costs for a hospital population. A cost-utility analysis by Padula et al. 27 (2011) compared use of the prevention protocol to a standard care approach in preventing HAPUs from a societal perspective. Padula et al. estimated the cost of the prevention protocol as $54.66 per day per inpatient. The lifetime model, which included states for all 33

50 staged pressure ulcers and DTIs, as well as medical and surgical patients, concluded that consistent initiation of the prevention protocol was actually costsaving. The prevention protocol also improved health-related quality of life as seen by an increase of quality-adjusted life years (QALYs) over standard care. Other cost-effectiveness analyses by Pham et al. 90 (2011), Inman et al. 56 (1993), and Iglesias et al. 88 (2006) reported qualitatively similar results; the overall cost increase due to improving HAPU prevention was minimal compared to the cost of treating pressure ulcers. Organizational Initiatives in Pressure Ulcer Prevention Surgical operations occur in the U.S. at an astounding 30 million procedures per year. 91 Unfortunately, there are numerous adverse events associated with each procedure. Results from the Harvard Medical Practice Study by Leape et al. 92 (1991) reported 1,133 preventable adverse events from about 30,000 patient records, of which 48% were related to surgical cases. Gawande 93 (1999) found similar rates of adverse events in surgical patients in a comprehensive observational study of Colorado and Utah hospitals. The vast number of patients with adverse events is due in part to the nature of modern surgery. As Griffin 91 (2007) points out, efforts to improve surgical safety, postoperative care and critical care are complicated by the diversity of cases. Clinical teams entrusted with initiating safety measures such as the HAPU prevention protocol in the surgical setting must make constant adjustments to the needs of the individual depending on length of the procedure and patient characteristics. According to Bosk 94 (1979), when a HAPU occurs in surgical sectors, it is a 34

51 technical error of the clinical team since proper initiation of evidence-based protocol is not carried out in its entirety The model of system categorization developed by Zimmerman et al. 95 (2001) and Glouberman and Zimmerman 33 (2002) can be used to establish a linkage between preventable conditions such as HAPUs with surgical outcomes. System categorization can be broken down into simple, complicated, and complex tasks: simple tasks can be answered yes/no ; complicated tasks utilize if/then algorithms; and complex tasks are relationally dependent events with unpredictable outcomes. Although the HAPU prevention protocol is based on a list of simple tasks, its completion is conditional upon more intricate tasks. Hader 45 (2007) points out that thorough initiation of the prevention protocol requires financial and material resources necessitating support from hospital administration gaining administrative buy-in is a complicated task. Each component of the prevention protocol is a complex task since the hospitalized patients are unique. Clinical teams must respond differently to a range of Braden scores, nutritional demands, repositioning regimens that alter by weight and mobility, as well as varying degrees of incontinence. According to Berwick 46 (2009), the way in which clinical teams respond to the needs of individual patients is termed patient-centered care. Patientcentered care can include the complex tasks associated with customizing the HAPU prevention protocol. Berwick claims that providing patient-centered care supersedes evidence-based practice, in this case the HAPU prevention protocol itself. Patient-centered care can be seen as in conflict with a systematic drive for 35

52 standardization of health care delivery with protocols and technology that are effective and economical. 96 Clinicians who provide patient-centered care using treatments that fit the needs and autonomy of the individual patient achieve 47, 97 greater outcomes. Multiple organizations, including the Institute of Medicine (IOM), the Institute for Healthcare Improvement (IHI), and the Joint Commission have initiated campaigns and regulations to specifically promote HAPU prevention with the use of the prevention protocol in a manner that is patientcentered. Institute of Medicine. In 1999 the IOM released the report To Err is Human on the basis of Leap et al. 92 (1991), Gawande et al. 93 (1999), and related findings. To Err is Human stated that while human error is natural and can lead to adverse events in the medical setting, systematic flaws within the healthcare environment are the likely source of medical error. 98 Systematic improvement could reduce the likelihood of compound human error that exposes patients to harm. 98 With this understanding, the IOM initiated a landmark campaign in health care referred to as First Do No Harm. First Do No Harm was a compilation of evidence-based directives that strived to motivate various constituents of health care. 1 First Do No Harm classified a broad range of medical errors. While there was nothing specific to HAPU prevention or other hospital-acquired conditions, there was consideration for multiple related issues: failure to employ indicated tests (e.g. the Braden Scale); error in the performance of an operation, procedure 36

53 or test (e.g. prevention protocol initiation); inadequate monitoring or follow-up of preventive treatment; and other system failures. 99 The IOM put research into action by making four key recommendations from First Do No Harm. 1 First it suggested creating a Center for Patient Safety within AHRQ to set national goals for patient safety and track such progress. This center in AHRQ would also become the primary source for research funding related to patient safety. Second, there should be a national system of mandatory and voluntary reporting for errors within hospitals as a way to address issues that clinicians encounter without punitive consequences. Third, regulators and accreditors such as the Food and Drug Administration (FDA) and the Joint Commission should raise standards and expectations for improvement in safety through the actions of oversight organizations, group purchasers, and professional groups. Fourth, healthcare organizations should create safety systems through the implementation of safe practices at the delivery level. Visible attention to safety and training in evidence-based practices could hopefully improve care at all levels. Following First Do No Harm, the IOM released a second report titled Crossing the Quality Chasm. Unlike the first report, Crossing the Quality Chasm called for a complete system redesign to health care delivery by claiming that medical errors could not be patched up with straightforward recommendations. Berwick 100 (2002) explained six dimensions of system redesign in the second report: (1) safety; (2) effectiveness; (3) patient-centeredness; (4) timeliness; (5) efficiency; and (6) equity. Berwick also noted that these dimensions would need 37

54 to improve at all levels of the health care system: the patient; the clinical microsystem; the macrosystem; and the environment and laws that govern action in health care. Considering the message of Crossing the Quality Chasm as it relates to HAPU prevention, achieving better outcomes means not only adjusting how the prevention protocol reaches the patient, but reorganizing clinical teams to ensure that the prevention protocol is consistently initiated without harmful variation. Institute for Healthcare Improvement. In December of 2006, IHI began a campaign with the goal of reducing unintended physical injury resulting from or contributed to by medical care (including the absence of indicated medical treatment that requires additional monitoring, treatment, or hospitalization, or that results in death. 101 McCannon et al. 101 (2007) explained that this initiative, titled The 5 Million Lives Campaign, was an extension of IHI s previously successful The 100,000 Lives Campaign, and aimed to prevent at least 5 million adverse events in a two year period ( ). Berwick et al. 102 (2006) stated that the purpose behind The 100,000 Lives Campaign was to encourage hospital adoption of preset goals in a pre-specified time frame by IHI in order to garner a personal (hospital) sense of responsibility towards improving quality of patient care. The 100,000 Lives Campaign included preventing several hospitalacquired conditions including surgical site infections, central-line infections, and adverse drug events. The 5 Million Lives Campaign continued the mission of the former campaign, but as Lewis 103 (2007) pointed out, IHI specifically added HAPUs and five other preventable conditions to the prioritized list. 38

55 The 5 Million Lives Campaign concluded with results that were informative with respect to HACs, especially HAPU prevention. Despite Berwick 46 (2009) message to move past evidence-based practice into the realm of patientcentered care, many hospitals were not prepared for this progression by A descriptive study by Leasure et al. 104 (2006) found that of nurses involved in The 5 Million Lives Campaign, few had consistently navigated evidence-based practice: 64% of nurses read one or more specialty journals; 53% read a nursing journal; 20% did not regularly read any professional journals; and 0% read a journal dedicated to the publication of original research. Makic et al. 105 (2011) note that although nurses knowledge of evidence-based practice is not the only factor involved in improving care, it is an essential prerequisite to avoid further harm. In fact, carrying on established traditions in preventive care may actually do more harm than benefit to patients, and certainly does not help hospitals work toward a patient-centered delivery system. Despite these barriers to improvement in patient care, IHI declared multiple successes from The 5 Million Lives Campaign by December, IHI 106 (2008) reported that over 4,000 hospitals in the U.S. committed to adopting the goals set by The 5 Million Lives Campaign. Among many accomplishments, HAPU incidence dropped nationwide in many hospital networks, such as New Jersey hospitals which reported a 72% overall reduction in HAPUs during the campaign. Berwick et al. 107 (2006) discuss that measurable success is an indicator that hospital clinicians became more self-educated on evidence-based 39

56 practices and reached a new standard with which to initiate elements of patientcentered care, including QI, to further improve adverse event outcomes. 102 The Joint Commission. As a regulating body of hospital reporting, the Joint Commission has deemed status from CMS to accredit hospitals and nursing homes. Among its missions, the Joint Commission monitors hospital outcomes associated with patient safety and quality care. In addition to CMS policy intervention to discontinue reimbursement for HAPUs, Golladay et al. 108 (2010) state that CMS mandates that hospitals report all incidences of HAPUs and other HACs. Misreporting can affect a hospital s accreditation status with the Joint Commission, therefore hospitals have no incentive to skew HAPU incidence. By recognizing HAPU occurrences within the hospital, clinicians may review the context of incidence and take further steps to prevent future cases in the diverse set of hospitalized patients. 31 Since the Joint Commission established this policy on reporting, subsequent studies have shown hospital self-reporting of adverse event outcomes to be quite reliable. Williams et al. 109 (2006) reported a 94.2% agreement between hospitals and the Joint Commission reviews on patient diagnoses at discharge. These data suggest that reported HAPU incidences should be reliable. CMS Policy and Organizational Change In June of 2008, Kurtzman and Buerhaus 8 (2008) reported that by October of that same year CMS would discontinue reimbursement for a number of HACs, including HAPUs. 110 According to Gray 111 (2008) the CMS nonpayment policy 40

57 could be viewed as a P4P measure. Stone et al. 112 (2010), reporting for a group of HAC experts, discuss the primary intent of CMS policy is as an external motivator to hospital QI efforts for preventing HACs including HAPUs. Since this nonpayment policy was introduced, CMS has monitored HAPU incidence as a measure of hospital performance. 113 According to Schaffer 114 (2008), HAPU incidence is a hospital-wide indicator of effective patient safety, but does not evaluate provider-level performance since preventing HAPUs exemplifies a team approach; it is a joint effort of multiple nurses and physicians. As Trisolini 115 (2011) suggests, the theoretical underpinnings of value-based policy interventions should engage individual providers to improve their practice. However, Golden and Sloan 116 (2008) and Shortell and Kaluzny 117 (2006) both expressed concern that hospital-level financial incentives may be poor extrinsic motivators for improved performance at the level of the provider. A system of interventions that improve the delivery process of the HAPU prevention protocol could more effectively fulfill global aims, such as reducing HAPU incidence to zero as shared by Duncan 22 (2007). QI interventions have not been ruled out as a source of systematic change to lower HAPU incidence since QI establishes a link between provider engagement and policy intervention. 35 Overall, only certain motivators can incentivize clinicians to improve their practice, and an organizational framework that includes process improvement in systematic delivery appears most appropriate to describe the theoretical foundation of the value-based policy agenda by CMS. In line with Trisolini 44 (2011) claims on theory of organizational change, successful HAPU prevention 41

58 stems from a multi-level organizational construct between CMS policy and hospitals, as well as multidisciplinary interactions between nurses and physicians. 44 A disconnect between these levels exists due to the dichotomous measurements of success, with regards to HAPU prevention: (1) CMS measures hospital performance by HAPU incidence; and (2) clinicians measure success incrementally with frequent completion of patient Braden Scales and other components of the prevention protocol. The overall relationship is reflected in 28, 39 the inverted pyramid scheme of health care (Figure 2.1). Organization Theory. Nonpayment policy targeted for HAPU prevention is unique since it focuses only on provider organizations such as hospitals, and not on individual clinicians. Hospitals are incentivized to prevent HAPUs even without reward because a pressure ulcer care can cost $500-70,000 per case. 65 Since this cost does not directly affect physicians and nurses at the interface of patient care, hospitals must work through the multi-level construct, transforming nonpayment policy into a meaningful motivator for clinicians. Kimberly and Minvielle 118 (2003) describe this construct in four levels: (1) ownership; (2) institutional layers; (3) organizational cultures; and (4) change management and QI. The relationship of these levels to HAPU prevention is illustrated in Figure 2.2. Hospitals and other health care institutions represent ownership that employs multiple physicians and nurses. CMS measures HAPU outcomes data at the hospital-level, so improved prevention is a measure of hospital-wide response to value-based policy. 42

59 Figure 2.1. Inverted pyramid of multi-tier health care system with evidencebased and quality measures, modified from Quinn 39 (1992). 43

60 Figure 2.2. Four levels for organization theory pertaining to HAPU prevention. 44

61 Hospitals restructure external payments into intrinsic incentives for physicians and nurses, such as P4P, thereby translating policy into something clinically meaningful. Benefits transferred from ownership could be in the form of a reorganized payment structure such as salary incentives or other arrangements considering that there is no reward for HAPU incidence. Such benefits could also be in some form of reward based on effective care and recognition for improved clinical performance, such as reduced HAPU incidence. According to Town et al. 119 (2004), the institutional layers of health care represent divided roles of hospital administration, medical and surgical services, and the lower hierarchies of physicians and nurses responsible for ensuring HAPU prevention. Godfrey et al. 40 (2008) also present these layers similar in style to the inverted pyramid scheme in Figure 2.1 as the macrosystem (hospital), mesosystem (department), and microsystem (clinical unit). The lower layers are where clinicians investigate methods to prevent disease in response to demands from above, which should align with value-based policy. Recognizing the limitations and variation of a prevention protocol, Landon et al. 120 (1998) suggest that clinicians can investigate alternatives to improve performance with other clinician groups as resources. As clinicians physicians and nurses work jointly to reduce HAPUs, they develop a culture of prevention with precise aims. The global aim is represented by hospital- and policy-level efforts to eliminate avoidable pressure ulcers. 22 The specific aim is more unique to the culture of the hospital, which asks, How can we intervene in current practice to further reduce HAPU 45

62 incidence. Smalarz 121 (2006) predicts that clinician cultures prone to brainstorming ideas and producing actionable items that help answer this question can achieve higher quality outcomes. An important step towards a solution is that hospital culture should work in parallel with the intent of nonpayment policy and retain evidence-based practices such as the HAPU prevention protocol. 44 This aligns with the findings of Padula et al. 79 (2012) which hypothesize that increases in consistent risk-assessment with the Braden Scale at the microsystem level could lead to reduced HAPU incidence as measured at the macrosystem level. Finally, through organization theory, clinicians identify change management and QI intervention as the solution to high HAPU incidence. Mohr et al. 41 (2002) claim that QI interventions developed in the microsystem by the clinicians who intend to use it are more likely to adopt it for prevention. Godfrey et al. 40 (2008) state that this approach to development assures clinician buy-in of QI, as opposed to hospital- or policy-level interventions. Fisher 122 (2006) describes high-level interventions as entities that are applied externally upon a microsystem commonly in the form of a value-based payment (e.g. P4P). Organizational Change with Quality Improvement. Trisolini 44 (2011) identifies six characteristics of the most influential QI interventions for change. These characteristics include: (1) leadership; (2a) working in teams and (2b) a culture of learning; (3) information technology; and (4a) care coordination and (4b) patient-centered medicine. Structural approaches to QI development that integrate these characteristics must capture the iterative design process: an 46

63 intervention is rarely ideal after primary implementation. 123 Trisolini 44 (2011) identifies development of the above characteristics in four stages. An advanced structure reflecting Trisolini s stages onto QI development was coined by Demming 124 (1983) as the Plan-Do-Study-Act (PDSA) paradigm. Nelson and Batalden 125 (1993) describe PDSA cycles as able to achieve higher performance levels with continued reassessment and modification of QI interventions (Figure 2.3). Lyder et al. 126 (2004) approach to mapping improvement of HAPU riskassessment (Braden Scale completion) with PDSA cycles in multiple hospitals illustrates significant improvement in HAPU reduction. The organizational approach has several consequences as a result of its complex structure. Organization theory intends to link the influence of valuebased policies and hospitals with providers in the clinical microsystem. 41 This linkage produces a cultural awareness of HAPU prevention at the point-of-care, thus leading to changes in clinical processes that reduce incidence. However, organization theory unintentionally produces multiple levels of accountability with different aims in mind. 122 Hospitals are accountable for overall performance in terms of HAPU incidence, whereas clinicians are accountable for completing discrete steps in the prevention protocol to achieve the hospital s higher aims. The clinical microsystem is recognized as the basic, most replicable unit in hospital setting; therefore, fostering accountability in the microsystem leads to multiple iterations of higher performance hospital-wide and achievement of benchmarks set by CMS nonpayment policy

64 Figure 2.3. Structured improvement of hospital systems with PDSA cycles, modified from Nelson et al. 28 (2008). 48

65 In conclusion, organization theory appropriately describes the use of policy interventions such as P4P to improve HAPU prevention. Hospitals are the targets of value-based payments for reduced HAPU incidence, which must then motivate clinicians to improve their individual practices. Such practices include the use of the HAPU prevention protocol, but often require additional interventions to institutionalize the prevention protocol within a hospital setting. Organization theory provides structured support for the development and use of QI interventions, which can aid clinicians in improving practices by reducing variation in evidence-based practice such as inconsistent Braden Scale completion. With the entire spectrum of health care reorganized in this perspective, hospitals should undergo improved performance that will lower operating costs and meet the expectations of governing bodies. Quality Improvement Nelson et al. 28 (2007) conceptually define QI as the study of intervention or system redesign for better patient outcomes, better systematic performance, and better professional service. QI is considered a strategic system organized by interventions that can institutionalize an evidence-based practice among clinical staff. More closely related to this topic, the goal of QI is to improve HAPU outcomes through consistent implementation of the prevention protocol. QI interventions are devices, concepts, or organizational tools adopted by clinicians that change their approach to caring for patients with respect to HAPU prevention. Multiple QI interventions are grouped into a QI strategy, creating a unique combination of interventions practiced by a hospital to improve HAPU 49

66 prevention. In combination with the HAPU prevention protocol, QI strategies may lead to further reductions in HAPU incidence. Organization theory is a useful construct to describe the necessity for QI interventions in health care. Wheeler et al. 127 (2007) view value-based payments as a medium that gives hospitals new justification to test QI interventions throughout clinical microsystems, and evaluate these tools success at responding to high-level benchmarks. 127 The six characteristics in organization theory described above by Trisolini (2011) properly reflect the design needs of QI interventions as well. Best-practice Framework for Quality Improvement. QI interventions can be quite diverse in design. A system that classifies QI interventions would be useful in their study to ensure that full scope of organization theory is covered. Nelson et al. 28 (2007) organize QI interventions into what is called a best-practice framework. This framework categorizes QI interventions within a simplified scope of organization theory: (1) Leadership; (2) Staff; (3) Information and Information Technology; and (4) Performance and Improvement. Given the theoretical foundations of this framework, clinicians pressured by external motivators such as CMS nonpayment policy can utilize QI interventions from the best-practice framework to implement change at the level of the clinical microsystem. The four domains of Nelson et al. 28 (2007) best-practice framework are scaled by 20 elemental QI interventions (Table 2.1). The elements of the bestpractice framework are general to all fields of health care. These QI 50

67 interventions can be coupled with components of the prevention protocol such as risk-assessment with the Braden Scale or patient repositioning. Table 2.1. Theoretical QI elements of the best-practice framework organized by domain, taken from Nelson et al. 28 (2007). Domain of QI Practice Description of Noteworthy Practice Leading Organizations Annual retreat to promote mission, vision, planning, and deployment throughout microsystem Open-door policy among microsystem leaders Shared leadership within the clinic (for example, among physician, nurse and manager) Use of storytelling to highlight improvements needed and improvements made Promotion of culture to value reflective practice and learning Intentional discussions related to mission, vision, values Staff Daily huddles to enhance communication among staff Daily case conferences to focus on patient status and treatment plans Monthly all-staff (town hall) meetings Continuing education designed into staff plans for professional growth Screening of potential hires for attitude, values, and skill alignment Training and orientation of new staff into work of clinic Information & Information Technology Performance & Improvement Tracking of data over time at microsystem level Use of feed forward data to match care plan with changing patient needs Information systems linked to care processes Inclusion of information technology (IT) staff on microsystem team Use of benchmarking information on processes and outcomes Use of data walls and displays of key measures for staff to view and use to assess microsystem performance (e.g. checklists) Extensive use of protocols and guidelines for core processes Encouragement of innovative thinking and tests of change The elements across the four domains of the best-practice framework are stand-alone QI interventions. Any one of the elements could be applied independently for clinical practice, such as in support of the HAPU prevention protocol. In doing so, a clinical microsystem could observe improvements in the quality of care provided. Given that, it is also likely that practicing multiple 51

68 elements of the best-practice framework simultaneously could result in a greater improvement in quality care. Multiple elements that are practiced in combination, are referred to as a QI strategy. QI strategies can vary by hospital considering the multitude of QI interventions listed in the best-practice framework that can go into one strategy. Nelson et al. 28 (2007) suggest that the best QI strategy is one that is dynamic in both scale and scope. The scope of a QI strategy refers to having elements across many of the domains of the best-practice framework; a QI strategy with elements from each of the four domains would be considered most dynamic in terms of scope. On the other hand, scale refers to a strategy with many elements from a single domain. For instance, a strategy aiming to improve its transfer of information using highly technical devices might have dynamic scale by including all four elements of the Information & Information Technology domain. Thus, the best QI strategies will have multiple QI interventions representing all aspects of the best-practice framework. In terms of efficiency, some QI strategies may not need every single element of the framework in order to improve practice in an area such as HAPU prevention there may be a select set of QI interventions from the entire best-pracitce framework that works most effectively. Padula et al. 27 (2011) warn about additional investment in HAPU prevention that may come with diminishing returns. Assuming an estimated hospital cost of $54.66 per patient per day to initiate the prevention protocol 52

69 significantly reduces HAPU incidence, doubling the investment in prevention (i.e. $109.32) with add-on QI interventions may observe a further reduction in HAPU incidence. The reduction in incidence may not be proportional to the investment. However, considering Padula et al. (2011) threshold cost estimate of $ per patient per day where HAPU prevention is no longer cost-effective, there are likely many QI strategies that would prevail as cost-effective in the hospital setting. Standardizing Quality Improvement. Organization theory is limited for lacking forethought in the iterative design process, which culminates by integrating QI interventions into regular practice. Following PDSA cycles, Nelson et al. (2007) suggest that QI interventions enter a prolonged period of standardization and further evaluation, Standardize-Do-Study-Act (SDSA). 28 Assuming this is correct, Trisolini s discussion of organization theory illuminates its only clear shortcoming. Once QI interventions are put into place and working, do they continue to meet the expectations set through policy of governing bodies of health care? A caveat of Nelson et al. about hospitals short-sightedness of QI interventions is a potential pitfall to steady, long-term improvement in quality and patient safety. 28 Hospital efforts to improve quality and cut costs are based on carrying out projects. Projects often succeed in the short-term, but fail to hold gains in the long-term or lead to collateral improvement in other areas of patient care. According to Nelson et al. 128 (2004), few organizations properly utilize QI theory to continually measure, improve, and adapt to changing demands in quality and 53

70 high-valued care. This issue can lead to early conclusions about a hospital s success in meeting global aims, while in the long-term performance dwindles. A report by Lindenauer 129 (2008) described the experience of a large QI collaborative of hospitals in the U.S. that was committed to reducing hospitalacquired infections among post-operative patients. Although the collaborative successfully initiated the use of several QI interventions with the PDSA paradigm, the effort failed to show improved results by lowering infection rates. An explanation for this outcome is open to speculation, however it may have been the result of inconsistent SDSA cycles. Lindenauer even suggest that the methods for measurement were insufficient compared to more novel methods recently applied in studies of economic and epidemiologic burden of infections, and related improvement efforts. Clinical Context of Adoption. Hospitals that succeed in overcoming the short-sightedness of project-based QI initiatives still have several issues to confront. Given that QI strategies are meant to focus a microsystem s cultural awareness on a clinical issue such as HAPU prevention, 118 varying cultures could have different contexts for adoption of QI interventions. Damschroder et al. 130 (2009) claim that the type of QI strategy chosen for a microsystem is dependent on the clinical context for adopting such a strategy. Thus, two microsystems focused on the same goal (e.g. HAPU prevention) may actually have different contexts for reaching that goal. According to Fink et al. 131 (2012) and Wald et al. 132 (2012), these contexts can be quite diverse, and fall under 54

71 both internal and external classifications of influence on prevention of a hospitalacquired condition. For examples of these contexts, take two clinical microsystems. One clinical microsystem measuring its own HAPU incidence might reach a selfimposed level of unacceptable incidence (e.g. 10% incidence); whereas, another microsystem at a different hospital is pressured by hospital administration to prevent any HAPU cases because of CMS nonpayment policy. The two examples here have different contexts for HAPU prevention, the first based internally on high incidence, and the other based externally on reimbursement policies imposed regardless of microsystem performance. Both microsystems have the same endpoint, which is to prevent HAPUs. By Damschroder et al. 130 (2009) logic, if each case becomes a high-performing microsystem then both likely have unique QI strategies in place that improve HAPU prevention. Batalden et al. 133 (2011) explain that these are traditional QI strategies which enhance performance, but do not necessarily achieve optimized system performance. There may still be one or several QI strategies reflected by one microsystem s QI interventions or a combination of the two microsystems that is most effective for HAPU prevention. Long-term Effectiveness of Quality Improvement. Determining the effectiveness of QI interventions, such as for HAPU prevention, requires precise measurements. Typically, a microsystem will begin with baseline measures prior to intervention, so that when QI adoption takes place there are outcomes for comparison. Following a baseline calculation, QI intervention starts and can 55

72 potentially improve HAPU outcomes at great margins with time. Mohr 134 (2000) illustrated the effect of QI on an outcome as an S-curve (Figure 2.4). 134 QI interventions that are adopted early in the observation period can be tracked to the point of improvement at steady-state. In reality, start-up periods can vary depending on clinicians learning curve to integrate the QI intervention(s) into their current practices. The direction of effect can be positive or negative, and there s evidence to suggest both. The study of sepsis prevention by Castellanos-Ortega et al. 135 (2010) evaluated the effects of a staff-based QI intervention on sepsis prevention. Their report indicated immediate reductions in sepsis cases following QI adoption. However, QI interventions in information technology, such as electronic health record (EHR) systems, indicate initial setbacks in effectiveness due to a higher learning curve. Spetz and Keane 136 (2009) report that a new EHR system at a small rural hospital initiated greater patient errors before there was any indication of improved quality care. This result was unfortunate considering that Spetz and Keane described the hospital system as high-performing prior to EHR implementation. Reflexivity. The theory of reflexivity can be used to explain why Spetz and Keane (2009) observations yielded negative outcomes following adoption of QI interventions. Van Loon and Zuiderent-Jerak 137 (2011) describe high-performing clinical microsystems as a combination of patients and social change. Social change exists in the microsystem when clinicians take on new initiatives to improve patient care, such as reducing HAPU incidence through implementation 56

73 Figure 2.4. Time-dependent effectiveness of QI Intervention, modified from Mohr 134 (2000). 57

74 of the prevention protocol. Under this delivery system, avoidable HAPUs are prevented as the device of social change is successfully integrated into the microsystem. When the system is impacted by an external factor such as a QI intervention, the theory of reflexivity suggests that new factor can reduce microsystem performance. As opposed to the previous situation where social change (Δ 1 ) was used consistently in an efficient microsystem with patients to achieve a desirable effect, the unnecessary addition of a QI intervention (Δ 2 ) yields an unintentional effect on social change (Δ 1 ). The result is a less efficient clinical microsystem than prior to the addition of QI interventions (Figure 2.5). Comparative Effectiveness Research The American Recovery and Reinvestment Act of 2009 authorized $1.1 billion in research grants to the National Institutes of Health (NIH) and AHRQ for clinical outcomes, effectiveness, and appropriateness of items, services, and procedures that are used to prevent, diagnose or treat diseases, disease disorders, and other health conditions. 96 These research initiatives have been grouped together as comparative effectiveness research (CER) in an effort to guide U.S. health care policy. The IOM defines CER as the generation and synthesis of evidence that compares the benefits and harms of alternative methods to prevent, diagnose, treat, and monitor a clinical condition or improve the delivery of care. 138 Following the American Recovery and Reinvestment Act, the IOM released a report of 100 priority areas of CER that would guide funding from NIH 58

75 Figure 2.5. Reflexivity of QI interventions on HAPU prevention. 59

76 and AHRQ. 139 Of these priority areas, 50% represented research to compare aspects of the health care delivery system, while the other half consisted of specific clinical issues in fields such as cardiology, neurology, and women s health. Among the priority areas of health care delivery, the IOM included compare the effectiveness of different quality improvement strategies in disease prevention, acute care, chronic disease care, and rehabilitation services for diverse populations of children and adults. 43 There is great potential for QI to improve HAPU prevention, but based on the best-practice framework it is difficult to discern which QI interventions, or QI strategies, are most effective. CER of QI interventions for HAPU prevention would be a meaningful response to the IOM s priority areas since it has the potential to improve patient safety and reduce economic burden nationally. Comparative Effectiveness Research of HAPU prevention. Chou et al. (2013) 140 developed a systematic review of all previous trials and cohort studies that had measured the effectiveness of elements of the HAPU prevention protocol. Their study reviewed protocol domains including risk assessment, support surfaces, skin care (e.g. patient repositioning; managing moisture and incontinence), and nutrition. Chou et al. questioned the validity of risk assessment tools such as the Braden Scale for accurately predicting HAPU risk or properly risk-stratifying patients. The review did reaffirm air-fluidized beds and other novel support surfaces as more effective than standard care. Though the systematic review by Chou et al. was a comprehensive evaluation of the effectiveness of the prevention protocol, it hand-picked studies 60

77 of varied quality pre- and post-cms nonpayment policy. It also evaluated each element of the prevention protocol in isolation rather than reviewing the effectiveness of the protocol as a single entity. Considering recommendations of the NPUAP that all elements of the prevention protocol need to be implemented consistently and in entirety, Chou et al. fell short in their attempt to establish a line of CER for HAPU prevention. Comparative Effectiveness Research of Quality Improvement. Over the past decade some evidence has reported on CER of different QI interventions in the acute care setting. Although, the generalizability of such QI research is limited due to narrow case study designs. According to Mohlenbrock and Kish 141 (2011), there is a need for more generalizable CER of QI interventions directed towards physicians and other clinical staff in the hospital setting. Several CER studies of QI interventions in acute care settings are described below, which leaves room for improvement in the field regarding study designs. Boesch et al. 142 (2012) compared the effectiveness of clinical bundles (i.e. QI strategies) for the prevention of tracheostomy-related pressure ulcers preand post-implementation in a pediatric ventilator unit. There were standard and extended clinical bundles, where the extended clinical bundles reflected more dynamic QI strategies in terms of scope and scale. Boesch et al. utilized PDSA cycles to develop the clinical bundles and measured pressure ulcer incidence rates associated with each experimental group. The study reported significant reductions in pressure ulcer incidence for patients on the extended bundles. Boesch et al. findings are restricted to pediatric inpatients and do not account for 61

78 the impact of CMS nonpayment policy. The potential success of preventive clinical bundles in other clinical microsystems, including ones in other parts of the case study s hospital is questionable. On another topic of HAC prevention, Castellanos-Ortega et al. 135 (2010) compared the effectiveness of different QI interventions for sepsis prevention in a hospital-wide case study, thus adding to research on the robustness of QI. This study also evaluated the effectiveness of multiple combinations of QI interventions combined into QI strategies. Ferrer et al. 143 (2008) evaluated the effectiveness of multiple QI interventions and two QI strategies at preventing sepsis and avoiding sepsis mortality across 59 hospital intensive care units (ICUs) nationwide. Ferrer et al. used an experimental pretest-posttest design to evaluate the effectiveness of sepsis prevention associated with QI adoption at all ICUs. Ferrer et al. found that a sepsis management bundle (i.e. QI strategy) produced the highest levels of compliance with sepsis prevention protocols and improved outcomes in the ICU. The combined study designs of Castellanos- Ortega et al. and Ferrer et al. offer some insight into characteristics of an improved study of QI interventions across multiple microsystems and hospitals before and after QI adoption. However, the evidence base lacks studies that are designed with consideration of extrinsic factors such as policy intervention. CER of QI interventions pertaining to HAPU prevention is complicated by the impact of CMS nonpayment policy. On the one hand, CMS policy could be the spark to ignite hospital initiatives for improved quality and safety by coupling QI strategy with evidence-based practice. However, CMS policy could also 62

79 directly affect HAPU outcomes. Controlling for CMS nonpayment policy is complex since it is non-existent at first, then in October, 2008 it affects every hospital. Several levels of study design address this issue, with more complex designs better suited for control. In the process of developing a CER study design of QI interventions, it is critical to consider validity. A hospital-based study is usually internally valid since it is a study of actual HAPU cases. Other hospitals should be able to utilize the results of a study to develop a prevention protocol coupled with QI interventions that effectively prevent HAPUs. In order to achieve generalizable (externally valid) results with a thorough exploration of QI interventions, Berwick and Goldmann 48 (2008) recommend utilizing a CER study design that validly predicts effective QI interventions without confounding by exogenous factors, such as CMS nonpayment policy (Figure 2.6). This is difficult since CMS policy affects every U.S. hospital following However, such approaches would benefit the evaluative effectiveness of QI without the pitfalls described by Lindenauer 129 (2008). The space below explores the choices for study design as well as the pros and cons for each. Experimental Pretest-Posttest. An effective study design could emulate a pretest-posttest design in its most basic form. Each QI elements from the bestpractice domain represents an independent study group. Each group that appears in the period spanning an observation period has some measureable outcome O 1. If outcome is HAPU incidence, then incidence is expected to reduce with exposure to QI. When CMS nonpayment policy (X) is implemented 63

80 Figure 2.6. Strengths and weaknesses of research designs for effective QI strategies, modified from Berwick and Goldmann 48 (2008). 64

81 in 2008, it confounds the effect of QI exposure on improved HAPU outcomes. 144 Thus, all outcomes following policy intervention (O 2 ) are potentially confounded. It is difficult to quantify direct effects of a QI intervention on HAPU incidence in the O 1 -X-O 2 framework. CER of QI interventions for HAPU prevention in the pretest-posttest framework could falter, firstly, because differences in effect between observation periods that cross X are confounded. Campbell and Stanley 144 (1963) discuss that unbiased transition across X is only plausible if the event is rare and shortlived. However, CMS nonpayment policy affects all hospitals in the U.S. at once and is still in place. Second, the lack of multiple observations before or after X weakens a hypothesis of maturation. 144 QI interventions potentially improve HAPU outcomes at great margins with time. Consider the effect of QI on an outcome as the S-curve in Figure QI interventions that are adopted early in the observation period (e.g. prior to CMS nonpayment policy) are tracked to the point of maturation, or steady-state. Whereas, QI interventions that do not appear in hospitals until late in observed time (e.g. following CMS nonpayment policy), cannot be evaluated with control for the effect of CMS policy. Difference-in-differences Estimator. According to Wooldridge 145 (2009) the difference-in-differences estimator improves upon basic pretest-posttest construct because it controls for the exposure effect on outcomes surrounding event X. Consider a scenario where during multiple observation periods from before and after CMS nonpayment policy, some hospitals may not declare the use of any QI interventions, whereas the rest of the hospitals contribute HAPU 65

82 outcomes data to some of the best-practice elements. If hospitals that elect to do-nothing still experience some improvement in HAPU prevention, especially after CMS policy intervention, then these hospitals illustrate the direct effect of policy on HAPU prevention. By controlling for this effect in experimental groups by subtracting the difference in HAPU incidence following policy (Equation 2.1), the difference-in-differences estimator, δ, isolates the effect of QI interventions on HAPU incidence and mitigates bias. δ = (Outcome pre,qi Outcome pre,do-nothing) (Outcome post,qi Outcome post,do-nothing ) (2.1) Difference-in-differences would be straightforward to calculate in a comparative effectiveness study of QI interventions given the unit of analysis: cases of HAPU incidence. If a clinically meaningful improvement in HAPU incidence (i.e. 1 case per 1,000) was observed immediately following CMS nonpayment policy across all do-nothing hospitals, then this amount could be subtracted as an absolute amount from the detectable difference due to a QI intervention. 113 Quasi-experimental Comparative Effectiveness Research. When financial and time-limited resources are not an issue, one should ideally use a goldstandard in research design. RCTs are viewed in many fields of science as an ideal design, 146 but as Berwick and Goldmann 48 (2008) suggest, RCTs have flaws in exploration of effective QI strategy. Considering the theoretical 66

83 foundation for developing QI interventions, to assign QI interventions at different hospitals randomly could result in failure given various contexts. Any indication of a forced concept or practice upon clinicians receives push-back, according to Pronovost and Vohr 37 (2010). On the other hand, even if a QI intervention shows marginal improvement, Rothwell 76 (2005) argue that the motivational circumstances for adopting a QI intervention (e.g. a policy intervention or clinicians reaction to high HAPU rates) lack real-world context in an RCT. A study design falling more so in the upper right quadrant of the plane in Figure 2.6 would provide more meaningful results for clinicians. The study design of the RAND Health Insurance Experiment reported by Levy and Meltzer 147 (2008) as a quasi-experimental approach is praised as a gold-standard for measuring the effects of health care interventions in a natural setting. Such a design would be practical for the study of effective QI interventions for certain reasons. First, it has no pre-assigned control group. A control group to QI would likely be a do-nothing approach, which is counterintuitive to patient safety and shown not to be cost-effective. 27 Second, the design would allow for each hospital to form its own QI strategy of meaningful interventions, leading to a higher likelihood of improvement in HAPU prevention. A quasi-experimental design of QI interventions would have several features. The design would include a diverse set of many hospital sites, such as a combination of UHC hospitals as well as groups of rural, non-academic, and small/medium-sized hospitals. This study population would allow for testing of QI 67

84 interventions in multiple hospital settings (e.g hospital sites), as opposed to only academic medical centers. The quasi-experimental approach would begin at present and collect data on HAPU outcomes prospectively through three phases of QI adoption. From start-up, it could take as long as 8-12 months before QI interventions are put in place at each hospital. A benefit here is that QI interventions could be put into place simultaneously at each hospital so there is no time-variance, though this is difficult to achieve. Then PDSA cycles would result in another 8-12 months testing and revising components of the QI strategy. Finally, hospitals reach a global aim such as Duncan 22 (2007) goal of 0% HAPU incidence or a steadystate in HAPU prevention, and the QI intervention becomes a part of the hospital s standard practice. Prospective data would likely be collected in equal time intervals such as quarterly since hospitals report quarterly to CMS on quality measures such as HAPU incidence. 108 Ultimately, the data would appear longitudinally by hospital for approximately 12 quarters. Quasi-experimental CER studies would likely focus on two outcomes for good measure. Based on the methodology of Padula et al. 79 (2012) one outcome would be a macrosystem measure such as hospital-wide HAPU incidence since it is the standard measurement used by CMS for hospital performance. The second measure would be of microsystem performance, such as tracking Braden Scale completion upon admission in each patient record. This second measure is unique because it explores the effectiveness of QI interventions to improve care inline with Berwick 100 (2002) interpretation of 68

85 delivery system redesign from Crossing the Quality Chasm: patients at-risk for a HAPU; Braden Scale use in the microsystem; HAPU incidence in the macrosystem; and CMS nonpayment policy for HACs. QI interventions aimed at HAPU prevention are most likely to focus on improving implementation of the Braden Scale and other prevention guidelines to achieve reductions in HAPU incidence. Measuring Braden scale initiation tests the predictive validity of the scale to risk-stratify patients and also reveal circumstances where HAPUs are unavoidable, such as in an extremely comorbid 29, 51 population. Measuring Braden Scale initiation would require additional personnel to scan patient charts for a sampling or all study participants depending on volume. Both outcome measurements would ultimately achieve a gold-standard in QI research by monitoring both a clinical evidence measure (i.e. risk-assessment initiation with the Braden Scale) and a clinical quality measure (i.e. HAPU incidence) for QI intervention success. 28 The difference-in-differences estimator applies well to a quasiexperimental framework, which by definition allows for exogenous factors to occur naturally in the study setting. 145 The example in Equation 2.1 is plausible because a study group of hospitals are always practicing at least one QI intervention during the observation period. Under such circumstances, a donothing control group may not be accessible. Quasi-experimental designs such as the RAND Health Insurance Experiment, which studied four cost-sharing models of insurance, utilized hypothesis testing with multiple experimental groups, but no controls

86 According to Manning et al. 148 (1998) applying a quasi-experimental separatesample pretest-posttest design work around the lack of a true control group in CER of QI interventions. Under this framework, Campbell and Stanley 144 (1963) compare outcomes of two subgroups of a QI intervention directly before and after policy intervention. Such a comparison relies on the fact that each subgroup is different (Figure 2.7). In the separate-sample pretest-posttest, each subgroup (R) is randomly assembled for practicing the same QI intervention. The hospitals that make up each subgroup are different since certain hospitals add or drop the QI intervention from their overall strategy with time. HAPU incidence rates (O) of the two subgroups are directly comparable. While CMS policy intervention (X) is noted and controlled for in subgroup 2, X is irrelevant to subgroup 1. The quasi-experimental design has greater strengths than a basic pretestposttest experiment, but has some weaknesses. Including outcomes from a control group (e.g. hospitals that do-nothing) in contrast to the separate-sample pretest-posttest further strengthens the validity of the findings (Figure 2.8). Internal validity of a separate-sample pretest-posttest control group quasiexperimental design is quite exceptional. 144 The temporal effectiveness of a QI intervention, as well as its point of maturation is identifiable. Results are reliable (repeatable) and account for random error HAPU incidence due to variation in QI interventions and prevention protocols. It reduces sources of bias inherent in the study design, such as the use of the un-validated survey instrument. In addition, results of this design are widely considered generalizable since each subgroup 70

87 exists unaffected or fully immersed in CMS nonpayment policy, and the donothing alternative provides a difference-of-differences. By hybridizing all previous CER design considerations, the quasiexperimental design takes the shape of what Campbell and Stanley 144 (1963) refer to as a multiple time-series design (Figure 2.9). Consider the three phases of QI adoption based on the PDSA illustration in Figure 2.3. There is an initial observation period of HAPU incidence outcomes (O) where no interventions are in place. Following exposure of a uniform variable (X), such as QI interventions, observations are tracked for further outcomes. Although a control group as referred to in RCTs would not be part of this experiment, there may still be the opportunity to add a non-equivalent control arm to the ideal study design. 144 Rather than a classic do-nothing approach, hospitals deemed to already have a robust QI strategy in place during the entire observation period could act as controls. By collecting HAPU outcomes data for such hospitals without intervening in their current QI strategies, they could act as controls. The interest of holding a control group in quasi-experimental CER would be to note any exogenous factors that may present during observation, such as further CMS policy intervention or a revised HAPU prevention protocol. Variations due to exogenous factors in the control group could then be adjusted for in the experimental group. Overall, a quasi-experimental study design would reflect Figure 2.9. The multiple time-series design consists of a number of other measurements. The primary exposure variables of the study remain QI 71

88 Subgroup 1: R O (X) Subgroup 2: R X O Figure 2.7. Illustrated design of separate-sample pretest-posttest, modified from Campbell and Stanley 144 (1963). Subgroup 1: R O (X) Subgroup 2: R X O Subgroup 3: R O Subgroup 4: R O Figure 2.8. Illustrated design of separate-sample pretest-posttest with control group, modified from Campbell and Stanley (1963). O O O O X O O O O O O O O O O O O Figure 2.9. Multiple time-series design, modified from Campbell and Stanley (1963). 72

89 interventions. The QI interventions would likely be categorized into each element and four domains from the best-practice framework. 28 Since multiple combinations of QI interventions at different hospitals are probable, each hospital or set of hospitals with unique QI strategies would act as an independent study group. HAPU incidence and Braden Scale utilization are the outcome measures of interest. Hospitals that have the greatest absolute improvement in HAPU incidence and Braden Scale use would be deemed to have to most effective QI strategies. The endpoint would be a list of those particular hospitals QI interventions in place for HAPU prevention. Given that these interventions are compared head-to-head by hospital from the same starting point, the resulting QI interventions might become a new standard of care in HAPU prevention, such as an addition to the prevention protocol. Observational Comparative Effectiveness Research. Berwick and Goldmann (2008) discussion of need for multiple robust designs in comparing effectiveness of QI interventions still leaves the possibility of several approaches to conducting CER. According to Dreyer 149 (2011), robust observational studies provide a useful construct for CER since effectiveness is measured with realworld data. Thus, observational CER holds with it meaningful external validity in a unique construct. 77 Observational CER takes on three time periods during the course of a study: baseline, treatment start, and follow-up. This study design allows researchers to use retrospective data rather than restricting CER to prospective studies, and follows the same development process as in Figure 2.3. For the 73

90 study of QI interventions in HAPU prevention, the period prior to CMS policy intervention is where most baselines are measured. This is the period where hospitals are least likely to practice QI interventions for HAPU prevention. At some point during this period it is expected that at least some U.S. hospitals will contribute data on HAPU outcomes to at least one element of the best-practice framework. For example, assume that by the first quarter of 2006 no hospitals have adopted a QI intervention team huddles. However, in the second quarter of 2006, several hospitals claim to start using team huddle interventions as a QI intervention for HAPU prevention. The average HAPU incidence data for these hospitals will contribute to the treatment start effect of team huddles for the second quarter of Since a baseline is still needed to calculate a difference in effect between the quarter of treatment start and a prior quarter, the HAPU incidence data from UHC for the same three hospitals in the first quarter of 2006 will represent the baseline data. This example represents the same scenario as following a patient on medication, whose baseline data prior to intervention contributes to that medication. Following a baseline calculation, QI interventions are incorporated into practice and can potentially improve HAPU outcomes at great margins with time. Consider the effect of QI on an outcome the same as Mohr 134 (2002) S-curve (Figure 2.4). 134 QI interventions that are adopted early in the observation period can be tracked to the point of steady-state. Further Comparative Effectiveness Research Design Considerations. A novel application of quasi-experimental design to evaluate the effectiveness of an 74

91 macrosystem-level intervention was presented by Castellanos-Ortega et al. 135 (2010). The study prospectively followed multiple units of one hospital (i.e. emergency department, critical care, intensive care, etc.), at baseline for 12 months. Following the initial observation period, the hospital implemented a cultural intervention called the Surviving Sepsis Campaign which included a global aim to reduce sepsis by 25% in the study departments, and a specific aim to re-educate all personnel in sepsis prevention. With the cultural intervention in place, hospital physicians and nurses initiated two sepsis prevention bundles (QI strategies): (1) a 6-hour resuscitation bundle for early-onset sepsis: and (2) a 24-hour management bundle for additional follow-up. Upon completion of the study, the authors noted a reduction in sepsis incidence and claimed success for their intervention. Castellanos-Ortega et al. 135 (2010) utilized a design equivalent to Campbell and Stanley 144 (1963) separate-sample pretest-posttest design since the subgroups of patients before and after their cultural intervention were separate groups. In fact, the study indicates that there were significant changes in those with sepsis across the subgroups, including prior risk factors. This study design is a useful framework to represent in a comparison of QI interventions for HAPU prevention since the units of observation (e.g. hospital-level groups in this case) pre- and post- policy intervention are unlikely to be the identical even though multiple hospitals could exist in both subgroups. As a caveat to using quasi-experimental design, hospitals with experienced QI practice, theoretically, should not be affected at all by the 75

92 looming presence of CMS nonpayment policy. A preventive QI intervention should be developed at the level of the clinical microsystem. 38 An intervention developed in the microsystem by the clinicians that intend to use it are more likely to adopt it for prevention. 41 This approach to development assures clinician buy-in of QI, as opposed to a macrosystem intervention that is forced down upon 40, 150 a microsystem in the form of a mandate or value-based payment. Therefore, robust QI interventions made for the clinical microsystem are, in theory, not susceptible to the ripples of changes from above. 28 If there is strong indication that QI interventions developed by U.S. hospitals are innovative and robust, then changes in HAPU incidence should only be a product of QI adoption. Yet, the grounds for this assumption are less than substantial without a more extensive survey and interview design to validate standardization in each hospital s approach for developing QI interventions. Conceptual Framework The conceptual framework for this study is adapted from implementation and dissemination science as well as quality improvement theory, which describe the translation of medical evidence into practice policy, and public health improvements according to Gonzales et al. 34 (2012). Under this framework, certain QI interventions coupled with the evidence-based prevention protocol can more effectively prevent HAPUs, thereby improving overall patient safety (Figure 2.10). Evidence-based practices including the prevention protocol are implementable, but only after a health care delivery system is properly designed 76

93 to enable such practices. QI interventions are a set of resources that cultivate advanced health care delivery systems in order to incorporate evidence-based practices as part of a clinical routine. Assuming that the evidence-based prevention protocol is part of standard practice in hospitals, providers screen all admitted hospital-inpatients for HAPUs and initiate the protocol in regular intervals. However, subtle variations in how providers initiate the protocol can place the hospital inpatient at greater risk for developing a pressure ulcer. These variations could stem from a number of issues such as different interpretations of Braden scoring as discussed by Magahalles 29 (2007), or inconsistent protocol implementation suggested by Makic 74 (2007). Random noise may also account for some pressure ulcer incidences. The mechanisms that govern this variation depend on the influence of regulating bodies (e.g. CMS, IOM, IHI and the Joint Commission). Quality improvement interventions are another mechanism to control variation of the prevention protocol. The best-practice domains further organize the conceptual framework. Quality Improvement Typology From the information gleaned by surveying hospitals adoption of QI interventions, QI interventions will be classified by the four domains of the best- 77

94 Stakeholder Organizations CMS Policy IOM IHI JCAHO Health Care Delivery Systems Nelson et al. Best-practice Framework for Quality Improvement Interventions Individuals Hospitals Patients Providers Figure A conceptual framework for implementation science through the translation of evidence-based practice into quality improvement for pressure ulcer care, modified from Gonzalez et al. (2012). 78

95 practice framework. 28 For instance, one UHC hospital, Dartmouth-Hitchcock Medical Center (DHMC) uses data walls and a checklist as components of its QI strategy for HAPU prevention, which are performance & improvement elements. By surveying all hospitals that participate in UHC data pooling, hospitals claiming to have QI interventions in place that complement the prevention protocol will act as the cases. UHC hospitals that do not already have adopted QI interventions will act as a comparison group for analysis of trends in HAPU incidence. In summary, all hospitals may experience some reduction in HAPU incidence over time because of CMS policy intervention. However, the case hospitals may experience a further decrease in incidence because of increasingly dynamic QI strategies. With the best-practice domains as a framework for QI in this study, QI interventions as they pertain specifically to HAPU prevention are modified relative to the list presented by Nelson et al. (2007) in Table 2.1. The modifications to QI interventions are based on feedback from experts in the field of HAPU prevention (Table 2.2). In addition, several more elements have been included that pertain specifically to HAPU prevention. CMS Policy Intervention This study not only monitors the effect that QI interventions have on HAPU outcomes, but also the effect that policy has on hospitals and indirectly on preventive care. By studying hospitals before and after policy intervention, it is possible to observe what QI interventions, if any, were in practice prior to CMS 79

96 Table 2.2. Modified best-practice framework to reflect elements of QI for HAPU prevention. Domain Intervention Description Leadership 1 Program Mission Annual programs to promote mission of pressure ulcer prevention; highlight results of prevention efforts 2 Prevention Awareness A cooperative ongoing effort among a multidisciplinary committee (e.g. nurses and physicians) in prevention awareness 3 Leadership Initiatives Promotion of routine leadership initiatives (e.g. unit poster displays, informational flyers, outcomes boards) 4 Admin Support Improved administrative support to staff for participation in prevention programs with QI 5 Prevention Protocol Incorporate pressure ulcer prevention protocol into institutional policies and procedures 6 Benchmarking Participation in a benchmarking project to assess baseline rates of pressure ulcer outcomes 7 Wound Team Establish multidisciplinary wound care team for pressure ulcer prevention Staff 1 Performance Measures Regular discussions among staff about pressure ulcer performance measures 2 Team Huddles Consult-driven huddles daily or weekly to enhance communication between staff for at-risk pressure ulcer patients 3 All-Staff Meetings Frequent all-staff (town hall) meetings to discuss prevention guidelines 4 Wound/QI Team Established wound care clinician approach/teamwork with QI people (e.g. review of epidemiology data; quality reporting; shared leadership) 5 Prevention Education Continuing education about prevention 6 Staff Training Training and orientation for new or unfamiliar staff Information & Information 1 Data Tracking Data tracking, analysis and dissemination of pressure ulcer rates within units and hospitals (e.g. data wall displays) 2 EHR Risk Assess A standardized risk assessment tool built into an Electronic Health Record (EHR) system 3 Electronic Alarm Establish an electronic trigger/alarm system related to pressure ulcer risk 4 EHR Implementation Implementation of an Electronic Health Record (EHR) system Performance & Improvement (P&I) 1 Braden Scale New or increased frequency of a risk assessment scale for admitted patients (e.g. Braden, Braden Q, Norton) 2 Visual Tools Visual reinforcement tools (e.g. checklists, posters, or bundled interventions) for follow-through with prevention protocols 3 Beds New beds or surfaces (including low airloss or bariatric beds) 4 HAPU Staging Adjustment in pressure ulcer stage reporting to clarify documentation of staging in patient records 5 Skin Care New skin care products or creams 6 Incontinence New incontinence wicking underpads 7 Repositioning Establish a formal repositioning regimen, or made changes to an existing one 8 Nutrition Establish a formal nutrition regimen HAPU indicates Hospital-acquired Pressure Ulcer; QI, Quality Improvement 80

97 nonpayment policy intervention. A significant increase in the use of QI following policy intervention and a subsequent reduction in HAPU incidence is worth noting. Organization theory suggests that value-based policy motivates increased initiation of prevention protocols and the exploration of QI adoption in order to prevent negative outcomes. 44 Thus, inclusion of the CMS policy intervention in the timeline of observation is a strength of the study. Despite the nearly universal impact of CMS nonpayment policy, there is still the ability to control for this policy intervention during analysis. By controlling for policy, this research can focus on the effect of QI interventions at reducing HAPU incidence rates and variation in HAPU prevention. These conclusions are based on the premise in this conceptual framework that QI interventions are what directly contribute to HAPU prevention, as well as extrinsic policy motivators. 81

98 CHAPTER III RESEARCH DESIGN Specific Aims Evidence supporting the effectiveness of quality improvement (QI) interventions for improving patient safety and quality care is lacking. 35 Specific to hospital-acquired pressure ulcer (HAPU) outcomes, there may be multiple QI interventions that contribute to improved HAPU prevention, but there is no comparative effectiveness research (CER) on specific QI strategies. Three aims in this proposal will test whether QI interventions are an effective vehicle for implementing the evidence-based HAPU prevention protocol, and determine which QI interventions are most effective. This work will utilize HAPU incidence data from a retrospective cohort of University HealthSystem Consortium (UHC) hospitals as well as survey data of QI interventions by hospital. Methodology The study is designed as an interrupted time-series quasi-experiment 151 of changes in HAPU incidence associated with adoption of hospital-level QI interventions, overlapping with nonpayment policy instated by the Centers for Medicare and Medicaid Services (CMS). The following sections of this chapter discuss the study design, analytic plan, and other methodological considerations. This study design requires several sources of hospital-based data. There are 82

99 multiple steps in this analysis in order to measure HAPU incidence correlated with the implementation of QI interventions. Data Sources This study aims to analyze nationally-representative HAPU outcomes at the hospital-level (Figure 3.1). The data sources chosen are based on parameters available for achieving each study aim (Table 3.1). UHC s Clinical Database and Resource Manager (CDB/RM) will provide nationwide, comparative data and analytic tools on HAPU incidence and patient characteristics from select tertiary care academic medical centers. CDB/RM data is reported quarterly by UHC hospitals to a central data manager, so all HAPU outcomes will be available quarterly from 2007 through 2012 for individually participating hospitals. As of 2011, there were 181 hospitals contributing information to the UHC CDB/RM; in 2008 at the time of CMS policy intervention there were 170 UHC hospitals reporting inpatient data. In addition to aggregate UHC data, a survey instrument will be designed and sent out to all UHC hospitals to collect corresponding data on exposure to QI interventions adopted by each hospital. To test the survey instrument s content validity, it will be pilot tested at University of Colorado Hospital (UCH) in Aurora, CO and Dartmouth-Hitchcock Medical Center (DHMC) in Lebanon, NH. 83

100 All US Hospitals Academic Medical Centers Survey Respondents Data Source University Hospital Consortium (UHC): HAPU outcomes with diverse QI Interventions UHC hospitals that respond to the survey: Specified QI interventions Figure 3.1. Box diagram of inpatient populations and related data sources. 84

101 Table 3.1. Data sources and parameters utilized for the proposed study. Parameter UHC Clinical Data Survey Age x Gender x Hospital LOS x Mortality x ICU Admission x Case-Mix Index x Medical/Surgical Procedures x HAPU Incidence x Hospital Staffing x HAPU Prevention Protocol x Contextual Influences x Quality Improvement Strategies x Aim 1 Evidence is currently lacking on QI interventions in use for a multi-site study, or the time periods of QI adoption. The first aim is to describe characteristics of inpatient QI interventions for HAPU prevention and measure hospital-level changes in QI adoption before and after CMS nonpayment policy for hospital-acquire conditions (HACs). A survey will yield information on QI interventions utilized by UHC hospitals for HAPU prevention. There are multiple sub-aims of this objective: a. Using key informant interviews with multiple QI experts at two hospitals i. Develop a best-practice framework for QI interventions based on Nelson et al. 28 (2007) that pertains specifically to HAPU prevention. 85

102 ii. Develop, refine and pilot test a survey instrument that could be used to collect longitudinal data on adoption of QI interventions. b. Administer a web-based survey to collect data on hospital-level patterns of QI adoption for HAPU prevention at UHC hospitals. c. Measure changes in hospital-level QI adoption patterns before and after CMS nonpayment policy at UHC hospitals. d. Describe national patterns of QI intervention typologies in terms of scope and scale before and after CMS nonpayment intervention at UHC hospitals. Hypotheses Multiple hypotheses stem from a comparison of hospital trends of QI adoption before and after the intervention of CMS nonpayment policy: Null Hypothesis (H1 0 ): QI adoption patterns do not change at the hospital-level. Alternative Hypotheses (H 1 ): Adoption of QI interventions at the hospital-level a. Increases in overall quantity among UHC hospitals following CMS policy intervention (post-2008). b. Covers a dynamic typology from the best-practice framework: i. The scope (i.e. more best-practice domains) of QI interventions increases following CMS policy intervention. 86

103 ii. The scale (i.e. more elements of the best-practice framework) of QI interventions increases per domain in practice following CMS policy intervention. Study Design The survey instrument (Appendix B) characterizes QI interventions at the hospital level before and after CMS nonpayment policy. The survey instrument is designed for tracking hospital exposure to QI interventions based on the Nelson et al. 28 (2007) best-practice framework by quarter from The bestpractice framework has been modified to reflect QI interventions that are most applicable to HAPU prevention (Table 2.2). To avoid missing a potential hospital-level QI intervention, respondents also have the option to fill-in their own QI interventions. Participation in the survey will be requested of all Certified Wound, Ostomy and Continence Nurses (CWOCNs) at UHC hospitals in the U.S. CWOCNs are best suited for this task since they are familiar with preventive practices throughout their own hospital for HAPU prevention. UHC hospitals typically have one or multiple CWOCNs per hospital. Each of these CWOCNs is in charge of implementing new QI strategies for various wound prevention initiatives. These CWOCNs can either respond directly to questions about applicable QI interventions in use for HAPU prevention, or forward the instrument to other hospital personnel best suited for responding to requested information. The survey instrument will contain three domains of questioning: (1) general approach to HAPU prevention; (2) QI intervention typology; and (3) 87

104 influencing factors on QI adoption and HAPU prevention. First, the instrument evaluates hospitals general approach to HAPU prevention. There is an interest in validating the extent to which each hospital initiates the evidence-based prevention protocol. A line of questioning in this first domain also concerns with the type of nursing staff that initiate the prevention protocol. A CWOCN is more likely to initiate the prevention protocol frequently with minimal practice variation, as opposed to a less experienced nurse with other experience in bedside care. The second domain of questioning is an organized checklist of QI interventions that could potentially be in use by each hospital to support initiation of the evidence-based prevention protocol. Survey respondents are expected to indicate which QI interventions are in place or undergoing testing within the hospital based on the best-practice framework. Each QI intervention should relate to HAPU prevention in some manner. QI interventions in the questionnaire are organized by best-practice domain: (1) leadership; (2) staff; (3) information and information technology; and (4) performance and improvement. Performance and improvement elements also include equipment adoption or upgrades to the prevention protocol. Respondents should also indicate the dates for adoption periods of each QI intervention, or whether the QI intervention is in ongoing usage. These questions will be used to later evaluate the level of dynamic scope and scale each hospital QI strategy involves. The indicated dates of adoption will be used for the interrupted time-series analysis to determine if any QI interventions are correlated with changes in HAPU incidence. 88

105 The third domain of the survey instrument includes questioning to determine the context of QI adoption. Context can include hospital-level initiatives as a result of CMS nonpayment policy or other influences. Context for QI adoption could also relate to issues such as noticeably high HAPU rates within a hospital, excessive job turnover in nursing staff within a certain timeframe, or integration of complex systems such as an electronic health record (EHR) system. Procedures Aim 1 utilizes a survey instrument that is not previously validated. Developing the appropriate questionnaire for the survey will require content validation through a pilot test. Following pilot testing, dissemination of the survey instrument to a national group of QI experts will require several specific steps to ensure a sufficient response rate. Pilot Test Protocol. The two proposed pilot test site hospitals for the current form of the questionnaire are UCH and DHMC, which both contribute patient outcomes data to UHC. Each of these hospitals has developed extensive protocols in HAPU prevention, as well as possesses experts in the field who frequently investigate QI strategies. The following protocol for the pilot test will ensure that the questionnaire is properly evaluated and edited before it is administered nationally: 89

106 1. Identify at least five people at each hospital who would be considered experts in QI strategy or HAPU prevention. At least one member of each site should also be familiar with UHC data reporting. 2. Provide the questionnaire to each individual in paper form. Allow each respondent two weeks to review the questionnaire for face validity. Encourage that they edit components of the instrument that they would prefer to see change in order to ensure validity. 3. Arrange a 30-minute interview with each pilot participant to receive feedback on instrument design. 4. Following review of survey instrument by all pilot test constituents at UCH and DHMC, upload electronically to an online version. 5. Re-administer the instrument to pilot test participants a second time in electronic format to ensure instrument validity has not been lost during transition. 6. Make final edits to the questionnaire based on feedback from the electronic version. Contact CWOCN Directors. Following completion of the pilot test, the questionnaire will be disseminated for completion by CWOCN directors at each UHC hospital via . There are several steps in this protocol in order to identify the proper point of contact for each hospital to avoid biased or unreliable responses. 90

107 1. Utilize the Wound, Ostomy and Continence Nursing (WOCN) Society central directory of CWOCNs to forward the cover letter and link to the questionnaire to experts at each hospital. 2. a cover letter and link to the questionnaire to each individual on the list. Typically, those connected through the UHC system are familiar with other personnel in their department or medical center who are familiar with HAPU incidence data and QI strategy. In the event that the person contacted can complete the questionnaire themselves, they will likely do so. Those who are not familiar with the survey request will forward it through their hospital network to those who are capable of completing all sections. 3. In the event that a hospital provides no response: i. Contact CWOCNs individually by phone and to identify a person best fit to complete the questionnaire. ii. Contact the hospital s office of Chief Nursing Officer and request information on personnel who would be capable of participating in the survey. 4. Access to the web-based survey will be through a link at the hospital-level. Therefore, as responses are saved, it will avoid the complication where multiple different responses can be given to the same question. 5. If a person completes the questionnaire, but leaves key information out of the responses, request permission to contact this person 91

108 directly and ask to re-administer the questionnaire orally or to clarify certain unanswered or misanswered questions. If it becomes certain that the respondent is not a topic expert, determine if there is another colleague at their hospital who would be better suited for the study. Survey Protocol. To begin the survey, the list of contacts will be notified that their name is available to us through the WOCN Society directory. Relevant contacts will receive a cover letter and link to the online questionnaire. The link will direct them to an electronic version of the questionnaire that they can complete at their convenience. Follow-up cover letters will be sent to individuals that do not respond or complete the questionnaire in 4-week intervals. The cover letter will also contain a request for permission to follow-up with respondents by phone or for any reason. Those who wish not to be contacted have the right to respond to us and request privacy. Those who respond to the questionnaire offering to share their prevention protocol will be contacted directly to obtain such information. The online survey will be available for five months, or until an acceptable number of participants respond to ensure that the study is well powered. If the response rate is low, UHC contacts will be reached directly by phone for request to participate. Study Population The population of the study design for Aim 1 includes UHC hospitals and pertinent CWOCN directors within each hospital. 92

109 Pilot Test Participants. The two proposed pilot test sites to evaluate the content validity of the survey instrument are UCH and DHMC. QI content experts at UCH include physicians, nurses, and research faculty. Physicians participating in the pilot test will offer face validity specifically to QI interventions in place at UCH. Nurses will offer input on the domains regarding the HAPU prevention protocol, QI strategy, and motivational incentives. Research faculty will offer multiple dimensions of input. Faculty within the Center for Pharmaceutical Outcomes Research (CePOR) at the University of Colorado will review the overall design of the survey instrument and feasibility with the research protocol. Faculty within the College of Nursing at University of Colorado will review the psychometric qualities of the survey instrument for points of improvement. DHMC will offer overlapping input from physicians and nurses to UCH with similar experiences regarding HAPU prevention and QI strategy. Research and medical directors at the Center for Leadership and Improvement within The Dartmouth Institute for Health Policy and Clinical Practice will be able to validate the specificity of QI interventions for HAPU prevention as each relates to the best-practices framework. UHC Hospitals. There are 122 academic medical centers in the U.S. that reported inpatient outcomes to UHC quarterly as of Some of these medical centers have multiple sites, such as Mayo Clinic which has three sites in Minnesota, Florida, and Arizona. Each hospital site will act as a separate entity for evaluation in this study, thus bringing the total number of hospitals to 93

110 approximately 180. Separating these academic medical centers into each constituent accounts for the assumption that each hospital may have a separate strategy and motivational contexts for HAPU prevention. These hospitals are diverse geographically and vary by qualifying characteristics such as Nursing Magnet Recognition. CWOCN directors at each hospital will receive a questionnaire to participate in the survey. A target response rate of 30% (54 hospitals out of 180) is predetermined to obtain a representative sample of prevention protocols and QI strategies among these hospitals. UHC CWOCN Directors. There is at least one CWOCN for most of the 122 academic medical centers and affiliates who are in direct contact with the WOCN Society. The list includes each individual s address, phone number, and hospital location. Unfortunately, there is not enough information on the list to immediately narrow the names to those who are directly involved in reporting HAPU incidence, nor is there the suggestion that any of these personnel are familiar with HAPU prevention protocols and QI strategy. All CWOCN directors with an understanding of wound care nursing will be contacted with the intention of identifying a HAPU prevention expert. Survey Biases and Assumptions The study design has a number of potential biases that, according to Aday and Cornelius 152 (2006), could affect reliability of the survey. The following section lists these biases and discusses some study assumptions that address each bias. First, it is assumed that all hospitals initiate the prevention protocol as prescribed by the National Pressure Ulcer Advisory Panel (NPUAP) in order to 94

111 limit HAPU incidence. 23 The survey instrument requests that each hospital shares a copy of their HAPU prevention protocol in their response to validate this assumption. Second, there is potential recall bias from respondents who fail to accurately describe start and end dates of adoption periods for a QI intervention, or fail to recall all QI interventions in use during the observation period. There is also potential channeling bias related to QI interventions used at a hospital that are not specifically for HAPU prevention but reported as such. For certain QI interventions such as adoption of novel underpads or transition to an electronic Braden scale, this is not a concerning issue since both types of interventions are specific to HAPU prevention. The mention of more general QI interventions such as team huddles lack specificity towards HAPU prevention. If hospitals experience clinically meaningful effects concurrent with the use of unrelated QI interventions, results of this study could incur false-positive recommendations. Thus, the study is dependent on the knowledge of the survey respondent. It will be difficult to determine if the survey respondent is the proper authority on QI to represent their hospital. By discussing the survey by phone with potential respondents prior to completion, the chances of selection bias are minimized. According to Dillman 153 (2006), response bias could affect the study design if hospitals with less-touted QI strategies choose not to respond. Those hospitals who utilize QI interventions in preventive practice will be more likely to report their actions in the survey. 95

112 Survey Validity and Reliability Since the survey instrument has not been previously tested, there is concern about the degree to which particular components of the survey measures achieve intended outcomes, otherwise known as survey validity. 152 The survey pilot test is intended to introduce content validity to the questionnaire. Aday and Cornelius 152 (2006) define content validity as the process of expert judges giving their opinions on whether the survey captures the domains of interest on the face of it. Given that that the pilot test participants are all experts in the field of wound care, HAPU prevention, or QI theory, each member would be able to validate some component of the survey domains to ask the proper line of questioning. According to Aday and Cornelius, 152 construct validity is captured in the design of the survey as well since the line of questioning is based on the evidence-based HAPU prevention protocol and the best-practice framework of QI 23, 28 interventions. By using two evidence-based resources, the questionnaire is ensured that the construct of the survey is appropriately framed. Pilot testing will further verify the survey s construct since experts will either confirm or update the list of QI interventions in the best-practice framework to better reflect best practices of HAPU prevention. Other forms of survey validity and reliability discussed by Aday and Cornelius, such as criterion validity, are not captured in the survey. Criterion validity in survey research is considered the most substantial form of validity for reducing the subjective interpretation of responses. Typically, survey design 96

113 includes a gold standard line of questioning (e.g. a previously validated questionnaire, the SF-6D) or an empirical measure (e.g. heart rate or blood pressure). However, it is difficult to capture criterion validity without a previous gold standard for monitoring adoption of QI interventions. The novelty of this study design also yields limitations. Reliability, which is most easily captured through test-retest surveying, will not be verified based on this study design. Test-retest monitors the ability of respondents to provide the same answers repeatedly, such as verification of QI interventions in place during certain time periods at the hospital. Test-retest usually requires sending out the questionnaire multiple times in spaces of two weeks or greater. Since the study population of UHC QI directors is remote, requesting multiple iterations of the survey is not feasible. Analytic Plan Aim 1 proposes multiple hypotheses about the adoption patterns of QI interventions in relation to CMS policy intervention. The null hypothesis is that the quantity of QI interventions remains constant across the period of observation ( ). However, it is quite realistic that adoption patterns of QI interventions increase over time, and most likely following CMS policy intervention. Welch s T-test. A t-test will be utilized to test part A of the Aim 1 hypothesis that the overall quantity of QI interventions increases following CMS policy intervention. Specifically, Welch s t-test is selected since the samples preand post- policy intervention will likely have unequal sample sizes and unequal 97

114 variance as hospital participation in UHC data pooling has increased in more recent years, as well as increases in hospital with QI strategies in place. This calculation will be done by counting the number of QI interventions per hospital in each time period (pre- and post-cms policy intervention). These counts will then be averaged among all hospitals reporting QI interventions in the survey for preand post- time periods. The t-test will generate a test statistic following a normal distribution to indicate whether the change in QI interventions following policy intervention is both positive and statistically significant at the 95% confidencelevel. Equation 3.1 below shows this calculation: 0Txx1 0 s2s2(3.1) 10N N1In Equation 3.1, the subscript values represent time-dependent samples pre-policy intervention (t=0) and post-policy intervention (t=1). X represents the mean count of QI interventions among all hospitals in each time-dependent sample. S stands for the variance among hospital-level QI intervention counts for each time, and N is the hospital sample size in each group. T is the test statistic that will measure changes of statistical significance at the 95% confidence-level. Degrees of freedom will be calculated based on variance and sample size in order to related T to a level of significance. 98

115 Testing Dynamics of QI Adoption Patterns. Part B of the Aim 1 hypotheses predicts that UHC hospitals will broaden the scope and scale of QI interventions in place for HAPU prevention following CMS policy. Using a t-test for each sub-hypothesis in part B, the increase in dynamics of QI interventions is measureable. The first sub-hypothesis suggests that scope (i.e. more best-practice domains) will increase overall following CMS policy intervention for each UHC hospital. To test this hypothesis, counts will be made of each hospital s total number of domains applied in QI initiatives before and after policy intervention. These counts will be averaged for t 0 and t 1 among all hospitals. The difference in means for these to time periods will be measured for overall effect and statistical significance. The mean difference should be great enough to suggest significant changes at the 95% confidence-level. Welch s t-test will be applied to test the statistical significance of the mean difference since the sample sizes and variance pre- and post-policy intervention will most likely change. In the second sub-hypothesis, a test will be developed for an increase in scale (i.e. more elements of the best-practice framework). This test will only measure increases in QI interventions within domains already in place at a hospital prior to CMS policy intervention. For each hospital utilizing QI interventions prior to CMS policy intervention, the domain(s) and number of interventions within each domain will be monitored. Following policy intervention, the number of interventions remaining in each initial domain will be counted, and the difference taken compared to the pre-policy time period. With the differences 99

116 calculated for each hospital s QI domains, the mean will be taken across differences of all domains. The hospital-level means will be averaged across all hospital samples. Finally, using a one-sample student t-test (Equation 3.2), the overall mean will be compared to an expected change of 0.0 (i.e. the null hypothesis) to determine if there is statistically significant growth in scale. Tx0.0 S(3.2) NxIn Equation 3.2, represents the overall mean increase in QI interventions taken from hospital-level increases in scale. S represents variation in the overall mean, and N represents the number of hospitals sampled for the calculation in growth of scale. The comparator 0.0 reflects no change in scale post-policy intervention. Aim 2 Gaps in evidence exist about whether QI interventions are an effective component of patient safety, and more specifically if certain QI interventions are effective components of a HAPU prevention protocol. The second aim is to measure the effectiveness of QI interventions at reducing HAPU incidence before and after CMS nonpayment policy. This aim will yield data on the most effective QI interventions involved in HAPU prevention while controlling for environmental interruptions caused by CMS policy. Aim 2 consists of two sub- 100

117 aims based on concerns for overall effectiveness and variability of HAPU prevention: a. Measure the effectiveness of QI interventions at reducing HAPU incidence. b. Measure changes in variance of HAPU incidence with QI and policy. Hypotheses There are multiple hypotheses to test of Aim 2: Null Hypothesis (H2 0 ): HAPU incidence rates remain constant following the adoption of QI interventions before and after CMS policy intervention. Alternative Hypothesis (H2 1 ): Among effective QI interventions a. The greatest changes in HAPU incidence rates across hospitals occur early in the adoption process of QI interventions. b. The greatest changes in HAPU incidence rates within hospitals occur early in the adoption process of QI interventions. c. Variation in HAPU incidence will be reduced following CMS policy intervention. Study Design The study design of the second aim follows the form of an interrupted time-series quasi-experiment with a hospital cohort. Following completion of the survey in Aim 1, survey response data for QI interventions will be merged with HAPU incidence data by each UHC hospital for a type of discriminant analysis 101

118 that Tabachnick and Fidell 154 (2007) refer to as an effect size analysis. The UHC CDB/RM will provide synchronized outcomes data on quarterly HAPU incidence at the hospital-level. HAPU incidence rates will be calculated as a function of quarterly HAPU cases relative to all at-risk inpatient admissions. Quarterly HAPU incidence rates will be linked to each QI intervention practiced by a hospital during a specified quarter for the effect size analysis. This study makes hypotheses about the differences in effect for each QI intervention presented in the best-practices framework. Effect size analysis can utilize survey response data for classifying structure into separate factors. Each factor represents the average change in HAPU incidence between quarters for all hospitals practicing a particular QI intervention. At the endpoint of Aim 2, QI interventions with the greatest effects on improved HAPU incidence will become recommended components of QI strategies for existing hospitals. Based on CMS s criteria for improved HAPU prevention, only QI interventions that show a clinically meaningful reduction of at least 1 HAPU case per 1,000 inpatients per quarter will be considered for recommendation. Overall, this component of the study will present a list of the most effective QI interventions that hospitals can consider to adopt based on limited resources. The selection of an effect-size analysis with outcomes data from a retrospective cohort is based on a balance of issues. First, it is a practical approach in CER. An RCT or quasi-experimental alternative could be costly and take years to complete. This observational study allows access to years of 102

119 retrospective data at no cost, and the web-based survey will only take several months to complete. Second, UHC data is readily available and has multiple benefits. UHC data was designed with the intention of analyzing patient quality and safety outcomes. Combining it with exposure data on QI interventions illustrates a unique application of UHC data that will broadly impact prevention of hospital-acquired acute conditions. Third, the retrospective nature of this research allows simultaneous observation of the effects of not only QI interventions, but policy intervention. Results of this study will impact how hospitals practice prevention, and affect the way payers such as CMS influence patient quality and safety with value-based initiatives. Finally, by performing an effect size analysis of changes in HAPU incidence between all hospital-quarters, we achieve in calculating a derivative in the trend of overall prevention. This derivative is valuable insight into the actually effectiveness of a QI intervention over an extended time series as opposed to a simple cross-section of time. Following effect size analysis, an analysis of covariance (ANCOVA) will determine if QI interventions reduce variation in HAPU incidence as much as mean reduction in HAPU incidence. 155 The expectation is that the adoption of clinically meaningful QI interventions will reduce overall variation in HAPU 79, 82 incidence. This analysis of variance will be done while controlling for covariates such as and length of stay (LOS) since both are significantly 54, 55, 156 correlated with HAPU incidence. Specific QI interventions recommended following the effect size analysis could be empirically related to variance as well. 103

120 Controlling for CMS Policy Intervention Since CMS policy changes occur during the period of observation for this analysis, there are concerns that this policy intervention could skew results. CMS nonpayment policy could be viewed as an effect modifier because while there are likely QI interventions in place at some UHC hospitals prior to the policy intervention, performance may not have been optimized yet. The occurrence of CMS policy further improves HAPU prevention and motivates clinicians to consistently initiate the prevention protocol without variation. Adoption of novel or additional QI interventions could support this effort. In contrast, the CMS policy intervention is not a confounder, nor does it exhibit direct causality to HAPU prevention. The CMS policy does not conflict with the adoption of QI interventions, and the correlation of QI interventions with HAPU prevention. CMS policy would not alone prevent HAPUs either, based on organization theory by Trisolini 44 (2011) that there is a division between policy impacting the clinical macrosystem, and action that leads to HAPU prevention in the microsystem. 40 Therefore, it is a combination of CMS policy and effective QI interventions that bolster consistent use of the prevention protocol that reduce HAPU incidence. Controlling for CMS policy when measuring variation with the ANCOVA model is crucial to emphasizing the effects of QI interventions. Analytic Plan There are several phases to analysis for Aim 2, which are described in the space below. 104

121 HAPU Incidence. Quarterly HAPU incidence in each hospital will be calculated as a function of hospital admissions (Equation 3.3) for all at-risk inpatients. The most basic form of this approach is presented by Rothman 157 et al. (2008). Patients with a secondary diagnosis for a stage III or IV pressure ulcer (ICD-9 code 707.0*) not POA will be included in the numerator, in accordance with the Agency for Healthcare Research and Quality (AHRQ) patient safety indicator (PSI) #3 for HAPUs. 57 All admitted inpatients patients considered at-risk by the AHRQ indicator will be included in the denominator. Incidenc#ofnewHAPUcasese #ofhospitaladmisions(3.3) This study makes hypotheses about changes in HAPU incidence for each effective QI intervention presented in the best-practices framework. These incidence rates will then be linked to each QI intervention practiced by a hospital during a specified quarter for an effect size analysis. Only stage III and IV HAPUs will be counted in the incidence calculation since these are well-documented in patient records. While all HACs must be closely tracked, hospitals are not required to report stage I or II HAPUs to CMS, and would not be reimbursed for such incidences otherwise. This fact provides justification for focusing on stage III and IV HAPUs which are consistently found in patient records. 105

122 Effect Size Analysis. The first part of the hypothesis for Aim 2 suggests that the greatest changes in HAPU incidence occur early in the QI adoption process. The changes in incidence associated with each effective QI intervention will be calculated in an effect size analysis. According to Tabachnik and Fidell (2008), this approach analyzes variance associated with individual predictors. In the case of this study, changes in the dependent variable HAPU incidence are predictable with adoption of QI interventions. QI interventions are believed to effect HAPU incidence indirectly by reducing variation in initiation of the evidence-based prevention protocol. The purpose of this analysis is to identify the most effective QI interventions at reducing HAPU incidence between quarters or over a period of multiple quarters. As hospitals indicate in the survey that they are using QI interventions, then more hospital-level HAPU incidence rates from UHC data will contribute to each QI intervention s effect in a quarter. Differences in HAPU incidence between quarters are calculated to then determine and effect size for each QI intervention. The mathematical construct for this approach appears in Equation 3.4: nefecsizeincidence-incidenc t,qt1,qn (3.4) ein the above equation, effect size is calculated as a function of HAPU incidence rates for two different quarters (t and t-1), in which a QI intervention (Q n ) appears sequentially. When a QI intervention first appears in the study, the 106

123 immediate past quarter (t 0 ) will act as the first quarter for the effect size calculation. The average effect size is taken across all hospitals utilizing a QI intervention for each quarter to represent the overall effect of a QI intervention. This effect size analysis assumes that there is no lag time between a hospital s claim to start a QI intervention and the effect it has on HAPU outcomes. The quarter after treatment begins is considered the follow-up period. Hospitals that began in the starting quarter of an intervention will most likely continue to contribute effect size measures to the QI intervention s effectiveness. Hospitals indicating in the survey that they no longer use a QI intervention are dropped from further contributions to effectiveness of the QI intervention. Hospitals that add the QI intervention to their strategy later in the study after others have already started using the intervention will begin contributing HAPU outcomes data to that intervention. Although, by this point the intervention has already passed start-up, hospital latecomers are assumed to contribute only to the follow-up period. This assumption is based on the latent effects of the CMS policy intervention. The study endpoint will result in a list of effective QI interventions. All three CER observational periods as defined in Chapter II contribute to the effectiveness of an intervention. The follow-up period will contribute the most units of time since follow-up can begin as early as the fourth quarter of 2007 for a QI intervention and continue through the second quarter of Effective QI interventions will only be listed if they have a clinically meaningful effectiveness 107

124 by reducing HAPU incidence by 1 case per 1000 patient admissions between quarters. Test Across Hospitals. A two-sample test of proportions (t-test) will be used to test Hypothesis 2a that reductions in HAPU incidence across hospitals occur early in the QI adoption process. More specifically, the greatest effect sizes are observed in the first two quarters of QI intervention at the first hospitals that adopt the QI interventions. Therefore, hospitals that responded early to CMS nonpayment policy experience the greatest improvements in HAPU prevention. This period could also be referred to as the time between initiation and the first follow-up quarter, and between the first and second follow-up quarter as described in Chapter II. Therefore, at least three consecutive quarters of observation for a QI intervention are needed following initiation. The exact time of initiation for a QI intervention is complicated by the time span of a quarter. For example, a survey respondent who indicates that their hospital began a QI intervention in the first quarter of a year could be implying that the intervention started sometime in January, February or March. Since the survey is not sensitive to exact start and end times within each quarter, the exact quarter of initiation is ignored for this analysis and assumed to be a pre-initiation quarter. Instead, the first quarter after initiation as indicated on the survey is where observation begins for testing significance in the effect size analysis. The test of proportions t-test will compare the mean effect sizes of the first two quarters to the mean annual effect sizes of all remaining quarters for each QI intervention. The test of proportions will reflect the same mathematical construct 108

125 at the Welch s t-test since here it applies potentially unequal samples and variation for each time period. The QI interventions with the highest, statistically significant effects on reducing HAPU incidence will be listed as recommended QI interventions for their immediate improvement effects. Test Within Hospitals. A two-sample test of proportions (t-test) will be used to test Hypothesis 2b that reductions in HAPU incidence within hospitals occur early in the QI adoption process. More specifically, the greatest effect sizes are observed in the first two quarters of QI intervention at each hospital that adopts the QI interventions, regardless of when the adoption period takes place during observation. The analysis will proceed exactly as the across-hospital test in Hypothesis 2A, except the initiation period for each hospital s QI interventions will be clarified during data collection. The test of proportions t-test will compare the mean effect sizes of the first two quarters to the mean annual effect sizes of all remaining quarters for each hospital s QI interventions. The test of proportions will reflect the same mathematical construct at the Welch s t-test, but with equal sample sizes since the same number of hospitals must contribute to initiation and lag periods of adoption. Example Effect Size Analysis. For an example of the effect size analysis, consider the mock data presented in Table 3.2. The survey will reveal information on how many hospitals are practicing each QI intervention, such as intervention 6 (e.g. frequent town hall meetings according to the best-practice framework) which is adopted by 16 hospitals in each of quarters 1 and 2 of 109

126 observation. The average of HAPU incidence data will be gathered for each of 16 hospitals for both quarters, and the difference taken. If the difference is positive showing a reduction in HAPU incidence between quarters then intervention 6 has an improvement effect on HAPU incidence. If the difference shows no change or negative, then intervention 6 does not show an improvement effect. Table 3.2. Mock data of number of hospitals representing QI interventions in each hospital quarter. QI Hospital-Quarter Intervention Q 0 Q 1 Q 2 Q 3 Q Once QI intervention and HAPU incidence data are recorded quarterly for all hospitals between , the average effect as a function of differences in incidence between quarters (Equation 3.4) of each QI intervention will be calculated across all quarters. QI interventions with the greatest improvement effects will become recommended components of QI strategies for existing hospitals. Based on CMS s recommendations, only QI interventions that show an improvement effect, as well as a clinically meaningful reduction of 1 HAPU case 110

127 per 1000 inpatients per quarter will be considered for recommendation. Overall, this component of the study will present a list of the most effective QI interventions that hospitals can consider to adopt based on their financial, time, and personnel limited resources. Power Analysis. For the study of effect sizes to be well-powered, at least several hospitals should be contributing incidence data sequentially to a QI intervention order to observe statistically significant changes in effect. If the minimal meaningful change in effect between two quarters is 1 HAPU case per 1,000 inpatients, then at least eight hospitals need to contribute incidence data to a particular QI intervention. Likewise, if only one hospital is contributing data to a particular QI intervention between quarters, then that hospital s effect on incidence should be greater, near 8 cases per 1,000 inpatients. Table 3.3 below illustrates a range of possible sample sizes (N) and the associated changes (power = 0.80; alpha = 0.05). The targeted survey response rate of 30% is expected to attain enough hospitals with variable QI strategies to fulfill this power. Table 3.3. Power of samples for an effect size analysis. Hospitals Power Alpha HAPU Incidence Standard Deviation N Cases per 1,000 Cases per 1,

128 The results of this power analysis simplify the effect size analysis because of certain conditions. When an effect size is too small for a given sample of hospitals contributing to a QI intervention, it will be censored from the analysis. This study is only concerned with reporting on clinically meaningful and statistically significant effect sizes. Variation By obtaining mean effectiveness measures for a set of clinically meaningful QI interventions related to HAPU incidence, there will be variance between mean effectiveness measures for each quarter. There is an interest in analyzing this variance in HAPU incidence before and after interruptions (i.e. CMS policy intervention, adoption of new QI interventions) to determine if with each interruption there is a reduction in variance. Typically, an analysis of variance (ANOVA) would be used to compare the difference in means of quarterly HAPU incidence for each QI intervention in order to observe impact over time. However, given that there are a number of covariates related to HAPU incidence, an ANCOVA model is preferable. An ANCOVA model in this study would consider several covariates that impact HAPU incidence, including CMS policy, age, LOS, hospital magnet status, and case-mix index (CMI). According to Berlowitz et al. 158 (1996), variance of these covariates is likely, especially age and LOS which are not constant between hospital quarters and associated with greater HAPU risk. Though magnet status is relatively stable for individual hospitals, there could be fluctuations in status over the observation period of multiple years which would 112

129 impact HAPU prevention due to an increase in nursing quality. Furthermore, CMS policy, which acts independently of QI adoption patterns, is at one time nonexistent, and then omnipresent to all US hospitals after October, By adjusting for these covariates, a direct impact of QI interventions on HAPU incidence reduction is measurable. Analysis of Covariance. Hypothesis 2c suggests that variation in HAPU incidence is reduced after CMS policy intervention with the addition of QI interventions. This hypothesis will be tested with an analysis of covariance (ANCOVA), which will determine if QI interventions reduce variation in HAPU incidence. The analysis will control for the following covariates: CMS Policy; CMI; hospital magnet status; gender; ICU admission; mortality; medical or surgical procedures; and hospital size (bed count). Some Evidence suggests that many of these characteristics can significantly alter HAPU incidence. Other covariates will be tested for their overall fit in the model, and remain in the final ANCOVA model if applicable. Restricted models will test this fit. The expectation is that the adoption of clinically meaningful QI interventions reduce overall variation in HAPU incidence, so that hospitals are in control of special cause signals. 82 The two-way ANCOVA model appears in Equation 3.5, which longitudinally correlates the dependent variable HAPU incidence to QI interventions for each hospital (i) and quarter (t) while adjusting for pertinent covariates. The model is a random effects model since it incorporates a diverse population of UHC patients from multiple academic medical centers and various 113

130 114 patient settings to evaluate the correlation between QI interventions and HAPU incidence. it11109it8it7it6it5it4it3it2it10itεbedsβmortalityβrgicalmedical/suβicuβgenderβcmsβcmβmagneβlosβageβqβαincidence (3.5.) Statistically significant reductions in HAPU incidence variation following CMS policy intervention at the 95% confidence-level will be indicative of an improvement in incidence variation. An F-test will produce a test statistic for this change associated with policy intervention. The test statistic is computed by dividing the explained variance between groups (i.e. hospitals) for a QI intervention by the unexplained variations within groups (i.e. error). The level of significance will determine if there is a change in variation over time. The significance of the effect of CMS policy on variation will also be test in the model to determine if incidence variation reduces following policy intervention using an F-test. Sensitivity Analysis A sensitivity analysis of the effect size measures will be conducted to assess the robustness of results by varying durations of the case hospitals QI intervention adoption periods. By varying adoption periods of QI interventions at each hospital by one quarter before and after the claimed adoption period, the sensitivity analysis improves results affected by potential recall bias. These adjustments also test the null of the assumption that QI interventions take effect immediately; there may be lag times between QI adoption and effectiveness that

131 the sensitivity analysis would better capture. Upon adjustment of each hospital s adoption periods by one quarter, the effect size analysis will be re-evaluated to determine if there is a reordering of the most effective QI interventions that yield HAPU reductions of greater than 1 case per 1,000 admissions. Study Population The second aim utilizes data on hospital inpatients in order to calculate quarterly HAPU incidence by hospital in the UHC system. Data on HAPU incidence will be combined with information from Aim 1 for UHC hospitals. Hospital inpatients in this study will be separated into two groups for this study: those with diagnosed HAPUs, and those who are at-risk for development of a HAPU. HAPU patients must be diagnosed with a HAPU not POA according to AHRQ PSI #3. 57 Any discharges with an ICD-9 code of a pressure ulcer (707.0*) in any secondary diagnosis field among cases meeting the inclusion and exclusion rules. Patients can be excluded from a HAPU diagnoses if they meet any of the following criteria, according to AHRQ: LOS less than 5 days Principal diagnosis of pressure ulcer (ICD *) upon admission MDC 9 (skin) and MDC 14 (pregnancy) Diagnosis of hemiplegia (342*), paraplegia or quadriplegia (344*), or quadriplegia Diagnosis of spina bifida (741*) or anoxic brain damage (348.1) 115

132 ICD-9 procedure code for debridement or pedicle graft before/on same day as surgical procedure (834.5, 862*, 867*) Diagnosis of a stage I or II pressure ulcer or unstageable (707.2*) Transfer from a different hospital, SNF, or other health care facility The hospital inpatient population utilized for inclusion in the denominator for a HAPU incidence calculation will include all hospital inpatients who are atrisk for developing a pressure ulcer not POA. This includes all hospital inpatients that are not excluded by the above AHRQ quality and safety indicators, since such patients are not considered at true risk for a HAPU. Data Collection The UHC CDB/RM is set up to enter direct data queries about patient populations. Each query will be written up as a separate report and formatted for direct use by CDB/RM coders in order to download data for use in Microsoft Excel. Queries will be organized into HAPU and general at-risk inpatient populations so that resulting data can be utilized for HAPU incidence calculations and effect size analysis. The exposure variable for this study is the QI intervention, of which there are multiple classifications. The survey data will collect hospital-level exposure to each QI intervention for analysis in Aim 2. Upon adoption of a QI intervention, hospital exposure is continuous by quarter until the end of UHC data in 2011, or a hospital indicates that the use of a QI intervention has ceased. The outcome variable is HAPU incidence. HAPU incidence will be determined by direct queries of HAPU cases and total patient admissions per 116

133 hospital. Incidence is not given, but calculated as a function of the two measures. Other parameters collected in this study include hospital- and patient-level indicators of HAPUs. Hospital-level indicators include magnet status and the number of hospital beds. Patient-level indicators include age, LOS, case-mix index, gender, mortality, and patient status as a medical or surgical discharge. Aim 3 Some QI interventions may have positive, synergistic effects on HAPU prevention as a QI strategy, as found by hospitals currently practicing multiple QI interventions simultaneously. The third aim is to characterize hospital-level QI strategies and compare the effectiveness of these strategies according to HAPU incidence rates. This aim will yield information about the comparative effectiveness of QI strategies practiced by UHC hospitals that are most successful at preventing HAPUs. There are two sub-aims of this objective: a. Identify effective combinations of QI interventions and compare the effectiveness of these combinations as hospital-level QI strategies. b. Characterize QI strategies of hospitals that successfully prevent HAPUs. 117

134 Hypotheses By observing hospital trends of QI adoption during the observation period, there are multiple hypotheses: H 0 : Hospitals do not exhibit similar patterns of adoption of QI strategies before or after CMS policy intervention. H 1 : The most effective hospital-level QI strategies a. Contain particular combinations of QI interventions b. Are dynamic by principle of the best-practice framework. c. Have many of the same QI interventions in common. Study Design There is an interest in knowing what QI interventions hospitals use in combination for QI strategies to effectively prevent HAPUs. This aim will review components of the survey to determine which QI interventions that hospitals are consistently practicing in combination with the prevention protocol. Descriptive statistics will report what portion of responding UHC hospitals practice QI interventions. An effect size analysis will show if any recommended QI interventions are practiced as part of a QI strategy by multiple hospitals. There is potential for multiple hospital groups to share similar combinations of QI interventions that result in clinically meaningful reductions of HAPU incidence. QI interventions will be separated into groups based on clinically meaningful effectiveness. This approach will follow the same effect size analysis discussed in Aim 2 for combinations of QI interventions. 118

135 Finally, hospitals with the greatest reductions in HAPU incidence will be reviewed for their patterns of QI strategy. The most successful hospitals may have the most dynamic QI strategies, broad in both scope and scale according to the best-practice framework. Hospitals will be identified in terms of high performance by one of three definitions: (1) hospitals with the lowest mean HAPU rates from ; (2) hospitals with the greatest reductions in HAPU rates from start to finish of the observation period; and (3) hospitals with the lowest HAPU rates at the end of observation (e.g. lowest rates between January- June, 2012). This information could be valuable to other hospitals that do not have the complete resources to investigate QI strategies on their own. Analytic Plan After collection of time-dependent data on QI interventions, effective QI interventions will be grouped together in pairs and tested in an effect size analysis as QI strategies. Groups of effective QI interventions represented for two or more quarters will then have a calculated effect size analysis. QI strategies that are well-powered according to the power analysis in Aim 2. For example, eight hospitals must represent a QI strategy to detect a clinically meaningful effect of 1 case per 1,000. QI strategies that have clinically meaningful, well-powered effects will be reported as effective components of a QI strategy for HAPU prevention. Comparative Hospital Effectiveness. Groups or individual UHC hospitals that are viewed as high-performers of HAPU prevention will be reviewed for their QI strategies. These high performers come in three forms: (1) hospitals with the 119

136 lowest mean HAPU rates from ; (2) hospitals with the greatest reductions in HAPU rates from start to finish of the observation period; and (3) hospitals with the lowest HAPU rates at the end of observation. Among these hospitals, their common trends in QI strategy will become part of recommended QI strategy for other hospitals to note. Considerations for Study Design There are several advantages and disadvantages for aspects of the study design, including the retrospective cohort and interrupted time-series. These issues are discussed below. Retrospective Cohort The first consideration of this study design is use of retrospective data. Looking back at HAPU outcomes and QI interventions over six years has the distinct advantage of crossing over CMS policy intervention for hospital-acquired condition reimbursement in An intervention of this magnitude rarely appears in hospital outcomes, and given federal policy regarding reimbursement, this study fills gaps in literature about the effectiveness of QI interventions as well as policy intervention for HAPU prevention. Prior to 2008, there is the advantage of observing hospital trends in QI adoption without the influence of policy intervention. There are also multiple disadvantages of a retrospective design. Tracking hospital exposure to specific QI interventions for HAPU prevention is difficult with a survey due to recall bias; the start and end dates are not completely reliable. 120

137 Second, CMS policy intervention could bias reporting of HAPU incidences since hospitals receive no reimbursement. However, the second disadvantage of retrospective data is negated by the Joint Commission s stringent regulation of hospital reporting for all stage III and IV HAPUs. 108 The cohort model of the study has its own advantages as well. A cohort study tracks hospitals from the point of exposure to QI interventions. If the desired endpoint is attained, clinicians have expressed satisfaction in results that identify effective QI interventions or reaffirm their current standards of practice. Tracking the cohort from exposure makes the study highly predictive of effective strategies, as previous cohort studies in HAPU prevention also indicate. 60 Interrupted Time-Series The decision to use an interrupted time-series for effect size analysis for in Aim 2 for selecting effective QI interventions is two-fold. First, effect size analysis provides useful results of the most effective QI interventions. An overlapping cluster analysis provides results of packaged QI interventions based on empirical evidence. In reality, such QI strategies may consist of combinations of interventions that only work well at select hospitals, and do not apply externally to other hospital locations. Second, there is concern that a this study is not well powered for a cluster analysis base on preliminary results. Ideally, several hundred hospitals would be needed to perform a cluster analysis. Therefore, an effect size analysis is the primary form of evaluation, while clustering is remains an exploratory third aim. 121

138 Validity and Reliability The overall study design achieves desirable validity and reliability. Results will be useful to many U.S. hospitals since the study setting includes UHC hospitals of many different sizes and demographic populations. The measures occur in a real-world setting as opposed to a controlled setting of a clinical trial. Every hospital in the study faces the dilemma presented by CMS policy intervention, as do U.S. hospitals outside of UHC. Such hospitals can utilize a list of QI interventions presented in the recommendations of the study, especially since many have limited resources. Non-academic hospitals less apt to study QI interventions can translate the results directly into actionable items. Other non-uhc academic hospitals can compare their approach to UHC institutions and adjust their QI strategies to benefit HAPU prevention. The study design is internally valid since its outcomes data reflect standardized measurements of stage III and IV HAPU incidence defined by AHRQ. The survey component of the study is also validated by the lists of QI interventions presented in the seminal text on QI research and endorsed by the IHI. 28 Reliability of the study is based on the fact that the HAPU outcomes are presented uniformly from one source, UHC, and regulated by the Joint Commission. 108 The internal validity of the study is questionable on several levels. Despite hospitals claims to practice QI, certain interventions may only exist in a fraction of patient wards relative to all those inpatients at-risk for HAPUs; therefore, the study would not be based on hospital-level QI interventions. Initiation of the 122

139 HAPU prevention protocol may also vary beyond acceptable levels for certain hospitals despite claims. If a hospital s prevention protocol is improperly initiated, it could lead to poor HAPU outcomes and confound effects of QI on the same outcomes. According to Shadish et al. 159 (2002), based on the philosophy of quasiexperimental design (Chapter II), CMS policy intervention could drastically skew outcomes data on HAPU incidence. This policy could also fundamentally change documentation and coding for HAPU diagnoses. 75 If hospitals are not reimbursed for HAPUs after 2008, then policy intervention could bias the reporting of HAPU incidence in favor of reduced incidence and fundamentally change documentation and coding. Reporting bias would lead to greater perceived effectiveness of QI interventions in place before and after the policy intervention. However, misreporting can jeopardize a hospital s accreditation with the Joint Commission, therefore hospitals are not incentivized to skew HAPU incidence rates. 108 This study follows the assumption set by Padula et al. 35 (2012) that HAPU incidence rates reported by UHC hospitals are accurate. Still, controlling for policy intervention is a valid concern. One approach would be to censor data during the period affected by CMS policy, but doing so would have two consequences. First, it s difficult to determine when CMS policy actually begins affecting clinician behavior. Policy effects could start as early as 2007 when the changes were announced, or following October, 2008 when the reimbursement policy was enforced. Second, censoring loses quarters of data when QI interventions could potentially enter phases of start-up. Missing these 123

140 time periods would bias the results for such interventions since measure of effectiveness is only based on steady-state follow-up. Overall, it is more appropriate to keep outcomes data surrounding CMS policy intervention in the study design, and control for it using a method similar to the one suggestion above under a separate-sample pretest-posttest control group design. The alternative approaches would be to control using a simple difference-of-differences, or at least note the context of QI adoption with incurred biases. Study Assumptions Innate in the design of this study are several assumptions about the effectiveness of QI interventions on improving HAPU outcomes. First, the global aim for HAPU incidence at U.S. hospitals is simple: 0%. 22 Given that national HAPU incidence hovers near 7%, measurable range of QI effectiveness is limited. 16 A second assumption of QI interventions is that they begin to take immediate effect; there is no lag time between adoption and effect on HAPU outcomes. The direction of effect can be positive or negative, and there s evidence to suggest both. Castellanos-Ortega et al. 135 (2010) which evaluated the effects of QI bundles on sepsis prevention indicated immediate reductions in sepsis cases following QI adoption. However, QI interventions in information technology, such as electronic health record (EHR) systems, indicate initial setbacks in effectiveness due to a higher learning curve. However, Spetz and Keane 136 (2009) reported that a new EHR system (i.e. QI intervention falling 124

141 under the domain of information and information technology) at a small rural hospital initiated greater patient errors before there was any indication of improved quality care. 125

142 CHAPTER IV RESULTS Study Population UHC Hospitals Hospitals of the University HealthSystem Consortium (UHC) represent a diverse set of inpatient services and patient populations. Since 2007 the consortium has grown from 147 academic medical centers to more than 200 at present. At the time that this study was conceived approximately 180 medical centers participated in the UHC Clinical Database/Resource Manager (CBD/RM). Fifty of these UHC hospitals are recipients of Magnet recognition for nursing excellence. Over 4.28 million inpatients considered at risk for developing hospital-acquired pressure ulcers (HAPUs) were admitted to these hospitals between These patients were admitted under a number of different circumstances, including admission into an intensive care unit (ICU) as well as medical and surgical procedures. Of these patients, 27,289 reportedly developed a HAPU according to the Agency for Healthcare Research & Quality (AHRQ) patient safety indicator (PSI) #3. The great majority of HAPU cases occurred prior to the Centers for Medicare and Medicaid Services (CMS) policy intervention in 2007 when rates exceeded 80 cases per 1,000 inpatients. Table 4.1 describes the patient demographics of the entire UHC inpatient population. 126

143 Table 4.1. Patient demographics among all UHC Hospitals ( ). Characteristic 2007* ** N UHC Hospitals and affiliates Nursing Magnet Hospitals n/a n/a Total Inpatient Admissions 207, , , ,116 1,026, ,808 4,288,181 Total Avg LOS (Days) Index LOS Mortality (Deaths per 1000) Age (%/Year) , (%/Year) , (%/Year) ,215,218 >=65 (%/Year) ,714,744 Female (%/Year) ,144,895 Case-mix Index Medicare (Cases/Year) 95, , , , , ,341 1,852,914 Medical (Cases/Year) 121, , , , , ,533 2,610,045 Surgical (Cases/Year) 86, , , , , ,274 1,715,369 HAPUs (Cases per Admissions/Year) 16,903 6,753 1,313 1, ,289 HAPU Rate (Cases per 1000) HAPU Avg LOS (Days) HAPU Index LOS HAPU Mortality (Deaths per 100) Age (%/Year) , (%/Year) , (%/Year) ,194 >=65 (%/Year) ,669 Female (%/Year) ,163 Case-mix Index Medicare (Cases/Year) 1,801 3, ,633 Medical (Cases/Year) 1,583 3, ,610 Surgical (Cases/Year) 1,250 3, ,679 * 2007 only represents data from the 4th quarter; **2012 only represents data between 1st and 2nd quarters; HAPU indicates hospitalacquired pressure ulcer; LOS, length of stay; UHC, University HealthSystem Consortium Survey Respondents Of 180 UHC hospitals targeted for the survey, 55 hospitals (30.5%) reported having an established HAPU prevention protocol, QI interventions, and internal or external influential factors to HAPU prevention. This response rate met a target response rate of 30%, based on an expectation that 54 responses would provide a representative sample of UHC hospitals. Between , some of these hospitals had not yet begun sharing information with the UHC 127

144 CBD/RM. Table 4.2 illustrates the trends of how many hospitals participated in UHC data pooling each year and the number of hospitals with Magnet recognition. Table 4.2. Hospital-level characteristics among responding UHC hospitals. Characteristic 2007* ** UHC Hospitals Hospital-quarters Nursing Magnet Hospitals *2007 only represents data from the 4th quarter; **2012 only represents data between 1st and 2nd quarters; UHC indicates University HealthSystem Consortium These 55 UHC hospitals are representative of the entire consortium with diverse patient populations based on trends and variation among demographic and procedural characteristics. Table 4.3a indicates that of the 1.59 million inpatients meeting study inclusion criteria, 5,208 HAPU cases occurred between in the sample population. HAPU rates varied by year from over 14 cases per 1,000 in 2007 to a reduction of less than 1 case per 1,000 in Thus, the majority of these HAPU cases were discharged between Some key trends to note based on the population of surveyed hospitals include that HAPU cases were relatively older than the general inpatient population. A greater proportion of patients over the age of 65 developed HAPUs, and fewer young adults compared with the general population. HAPU patients had much higher length of stay (LOS), and greater rates of ICU 128

145 Table 4.3a. Patient-level characteristics among responding UHC hospitals. Characteristic 2007* ** N Total Inpatient Admissions 76, , , , , ,926 1,590,022 Total Avg LOS (Days) Index LOS Mortality (Deaths per 1000) ICU Admissions (%/Year) Age (%/Year) , (%/Year) , (%/Year) ,510 >=65 (%/Year) ,167 Female (%/Year) ,492 Case-mix Index Medical (Cases/Year) 42, , , , , , ,497 Surgical (Cases/Year) 33, , , , ,351 75, ,496 HAPUs (Cases/Year) 1,034 2, ,208 HAPU Rate (Cases per 1000) HAPU Avg LOS (Days) HAPU Index LOS HAPU Mortality (Deaths per 100) ICU Admissions (%/Year) Age (%/Year) (%/Year) (%/Year) ,597 >=65 (%/Year) ,495 Female (%/Year) ,217 Case-mix Index Medical (Cases/Year) 537 1, ,979 Surgical (Cases/Year) 497 1, ,208 *2007 only represents data from the 4th quarter; **2012 only represents data between 1st and 2nd quarters; HAPU indicates hospitalacquired pressure ulcer; LOS, length of stay; UHC, University HealthSystem Consortium admission, mortality rates and case-mix index (CMI) than the general population. There appeared to be relatively even numbers of males and females at risk for HAPUs, as well as HAPU risk on the basis of medical or surgical procedures. These trends are consistent with the hospitals depicted in Table 4.3b, which were UHC hospitals that did not respond to the survey. In general, the populations of responding and non-responding hospitals are similar, though there is a slightly higher rate of HAPU incidence during in the nonrespondents. This difference in incidence may be indicative of UHC hospitals 129

146 that had fewer interventions in place to prevent HAPUs, and relate to nonreporting bias in the survey of this study. Table 4.3b. Patient-level characteristics among non-responding hospitals. Characteristic 2007* ** N UHC Hospitals and affiliates Nursing Magnet Hospitals n/a n/a Total Inpatient Admissions 130, , , , , ,882 2,698,159 Total Avg LOS (Days) Index LOS Mortality (Deaths per 1000) Age (%/Year) , (%/Year) , (%/Year) ,588 >=65 (%/Year) ,121,231 Female (%/Year) ,362,605 Case-mix Index Medical (Cases/Year) 79, , , , , ,144 1,738,548 Surgical (Cases/Year) 52, , , , , ,957 1,033,873 HAPUs (Cases per Admissions/Year) 15,869 4, ,081 HAPU Rate (Cases per 1000) HAPU Avg LOS (Days) HAPU Index LOS HAPU Mortality (Deaths per 100) Age (%/Year) (%/Year) , (%/Year) ,288 >=65 (%/Year) ,249 Female (%/Year) ,762 Case-mix Index Medical (Cases/Year) 1,046 1, ,631 Surgical (Cases/Year) 753 2, ,471 * 2007 only represents data from the 4th quarter; **2012 only represents data between 1st and 2nd quarters; HAPU indicates hospitalacquired pressure ulcer; LOS, length of stay; UHC, University HealthSystem Consortium Trends in demographic characteristics for the study population in Table 4.3a including LOS, mortality, ICU admission, age, gender, CMI, and medical or surgical procedure are similar to trends of the overall UHC population in Table 4.1. The study population also over-represents hospitals with Magnet 130

147 recognition, thereby assuring a representative sample of hospitals with nursing excellence to address issues including HAPU prevention. A noteworthy trend of UHC hospitals is that despite the overall reduction in rates of HAPU incidence, the rates of mortality and ICU admission did not drop as drastically. This could imply that hospitals improved routine prevention of HAPUs for most inpatients. Those inpatients who developed HAPUs were among the most ill patients at UHC hospitals, especially during the period following CMS policy intervention between Aim 1 Results: Survey of HAPU Prevention Of the 55 responding hospitals, 54 of the surveys were completed by a certified wound, ostomy and continence nurse (CWOCN), and one survey was completed by the directing podiatrist of a wound care center. These 55 hospitals represented a variety of different approaches to HAPU prevention on the bases of prevention protocols and quality improvement (QI) strategies. HAPU Prevention Out of 55 responding hospitals, 51 indicated that they had an evidencebased HAPU prevention protocol in place, and 54 hospitals had CWOCNs on staff (Table 4.4). Most hospital-based prevention protocols were carried out regularly by staff nurses, but some hospitals indicated that the CWOCN remained the primary initiator of the prevention protocol. New beds purchased during the period of interested with the intent of HAPU prevention were in place at 38 hospitals, of which the average age of a new bed was 3.25 years. 131

148 Nineteen of the responding hospitals were willing to share a copy of their prevention protocol along with submission of the survey. A qualitative review of these prevention protocols determined that each follows the five components of the prevention protocol endorsed by the National Pressure Ulcer Advisory Panel (NPUAP), including: (1) Braden Scale risk-assessment; (2) patient repositioning; (3) managing moisture and incontinence; (4) patient nutrition; and (5) improved support surfaces. Table 4.4. Characteristics of HAPU prevention for UHC Hospitals (N=55*). Characteristic N % Mean SD Median Min Max HAPU prevention protocol Hospitals with CWOCNs CWOCN (FTEE) Protocol initiation: Staff Nurse CWOCN No answer Hospitals with new beds Age of new beds (in years) *180 UHC hospitals were surveyed about their HAPU prevention protocols between , of which 55 responded (30.6% response rate). CWOCN indicates Certified Wound, Ostomy, Continence Nurse; FTEE, full-time equivalent employees; HAPU, hospital-acquired pressure ulcer; UHC, University HealthSystem Consortium QI Adoption Over the course of the study period ( ), 53 hospitals (96.4%) indicated adopting QI interventions improve compliance with their evidencebased HAPU prevention protocol (Table 4.5). Only two hospitals reported not having included QI strategies in place. The number of hospitals initiating a QI 132

149 strategy increased during observation from 43 hospitals in 2007 to 53 hospitals in Many of the hospitals with QI strategies included those with Magnet recognition, and the proportion of Magnet hospitals grew during observation. Table 4.5. Hospital-level characteristics of QI adoption. Characteristic * N UHC Hospitals reporting QI Interventions % QI of total respondents Hospital-quarters ,074 Nursing Magnet Hospitals *2012 only represents data between 1st and 2nd quarters; HAPU indicates hospital-acquired pressure ulcer; QI, quality improvement; UHC, University HealthSystem Consortium In order to test Hypothesis 1a that adoption of QI interventions at the hospital-level increases in overall quantity among UHC hospitals following CMS policy intervention, we measured the average number of QI elements in place at hospitals pre- and post-cms policy intervention. Prior to CMS nonpayment policy, 53 hospitals had an average of 7.5 QI interventions in place as part of their overall QI strategy. Following the policy intervention in the 4 th quarter of 2008, the number of QI interventions increased to over 12 on average per hospital. This increase was found to be statistically significant at the 95% confidence-level by a Welch s t-test (Table 4.6). 133

150 Table 4.6. Overall adoption of QI interventions in hospitals pre- and post-cms policy intervention. Total of QI Interventions Time Period Hospitals Hospital-quarters Mean* SD Min Max Pre-CMS Policy Post-CMS Policy *Overall change in mean maximum number of elements in QI strategy across hospitals pre- vs. post- CMS policy intervention is statistically significant at the 95% confidence-level (p = ; 95% CI: to -2.42); the mean includes hospitals that indicated practicing 'other' QI interventions not listed on the best-practice framework. CMS indicates Centers for Medicare and Medicaid Services. Modified QI Best-practice Framework The best-practice framework was modified from the Nelson et al. (2007) 28 framework and pilot tested for validation, as discussed in Chapter 3. The final version of the framework applied during the survey consisted of 25 QI elements, in addition to four other categories that did not overlap with other elements. Based on the number of responses indicating use of each QI intervention over time, all elements were well-powered for the analyses in Aim 2. The modified framework and abbreviations for each element that will be referred to throughout this chapter appear in Table 4.7. QI Scope and Scale Analysis of the survey results by the four QI domains and 29 elements (including Other QI interventions for each domain) indicated increases in both scope and scale over the course of six years. The greatest increases in scope and scaled occurred following CMS policy intervention, and these increases were statistically significant. 134

151 Table 4.7. Elements of the QI best-practice framework by domain. Domain Abbreviation Survey Description Leadership 1 Program Mission Annual programs to promote mission of pressure ulcer prevention; highlight results of prevention efforts 2 Prevention Awareness A cooperative ongoing effort among a multidisciplinary committee (e.g. nurses and physicians) in prevention awareness 3 Leadership Initiatives Promotion of routine leadership initiatives (e.g. unit poster displays, informational flyers, outcomes boards) 4 Admin Support Improved administrative support to staff for participation in prevention programs with QI 5 Prevention Protocol Incorporate pressure ulcer prevention protocol into institutional policies and procedures 6 Benchmarking Participation in a benchmarking project to assess baseline rates of pressure ulcer outcomes 7 Wound Team Establish multidisciplinary wound care team for pressure ulcer prevention 8 Other Leadership Other types of QI interventions that might be classified under leadership roles Staff 1 Performance Measures Regular discussions among staff about pressure ulcer performance measures 2 Team Huddles Consult-driven huddles daily or weekly to enhance communication between staff for at-risk pressure ulcer patients 3 All-Staff Meetings Frequent all-staff (town hall) meetings to discuss prevention guidelines 4 Wound/QI Team Established wound care clinician approach/teamwork with QI people (e.g. review of epidemiology data; quality reporting; shared leadership) 5 Prevention Education Continuing education about prevention 6 Staff Training Training and orientation for new or unfamiliar staff 7 Other Staff Other types of QI interventions that might be classified under staff roles Information & Information Technology (IT) 1 Data Tracking Data tracking, analysis and dissemination of pressure ulcer rates within units and hospitals (e.g. data wall displays) 2 EHR Risk Assess A standardized risk assessment tool built into an Electronic Health Record (EHR) system 3 Electronic Alarm Establish an electronic trigger/alarm system related to pressure ulcer risk 4 EHR Implementation Implementation of an Electronic Health Record (EHR) system 5 Other IT Other types of QI interventions that might be classified under information technology Performance & Improvement (P&I) 1 Braden Scale New or increased frequency of a risk assessment scale for admitted patients (e.g. Braden, Braden Q, Norton) 2 Visual Tools Visual reinforcement tools (e.g. checklists, posters, or bundled interventions) for follow-through with prevention protocols 3 Beds New beds or surfaces (including low airloss or bariatric beds) 4 HAPU Staging Adjustment in pressure ulcer stage reporting to clarify documentation of staging in patient records 5 Skin Care New skin care products or creams 6 Incontinence New incontinence wicking underpads 7 Repositioning Establish a formal repositioning regimen, or made changes to an existing one 8 Nutrition Establish a formal nutrition regimen 9 Other P&I Other types of QI interventions that might be classified under performance and improvement measures HAPU indicates Hospital-acquired Pressure Ulcer; QI, Quality Improvement 135

152 Scope. Scope of QI strategies saw upward trends at all hospitals during observation (Figure 4.1). Many hospitals began with few domains zero or one domain during the period of ; however, following CMS policy intervention the majority of hospitals underwent increases in QI strategies that utilized all four domains. By 2012, 45 out of 55 hospitals had QI strategies of broad scope that included all four domains of the best-practice framework (Table 4.8). Concurrently, the number of hospitals utilizing only one, two or three domains experienced an overall decrease by Hospitals also experienced increases in scope by addition of specific domains to overall QI strategy for each hospital between Leadership interventions were most frequent, increasing by within-hospital scope from 40-63%. Staff interventions increased in overall adoption from 32-53%. Information & Information Technology (IT) interventions increased from 31-55%, and Performance & Improvement (P&I) interventions increased from 18-40%. Table 4.8. Scope of QI strategies by year for UHC hospitals (N=55). Hospital-Level Scope* Hospitals with 0 domains Hospitals with 1 domain Hospitals with 2 domains Hospitals with 3 domains Hospitals with 4 domains *Annual data on QI strategy scope is based on what hospitals reported practicing at the mid-point of each year (i.e. end of second quarter); HAPU indicates hospital-acquired pressure ulcer; QI, quality improvement; UHC, University HealthSystem Consortium 136

153 Under Hypothesis 1b, there is an interest in knowing if QI interventions cover a dyamnic typology from the best-practice framework. The first part of this hypothesis specifically tests whether the scope of QI interventions increases following CMS policy intervention. By measuring the average number of best practice domains represented in each QI strategy among hospitals pre- and post- CMS policy, we tested the difference using a Welch s t-test. Findings from the survey indicated that among 53 hospitals there was an average increase in the scope of QI strategies during observation from an average of 2.5 domains prepolicy to 3.3 domains post policy. This increase was statistically signficant at the 95% confidence-level (Table 4.9). Scale. As with trends of scope, the scale of QI strategies increased during observation by a number of different measures. Within each domain of the bestpractice framework, hospitals typically expanded the scale of their QI strategy to include more elements of each domain between (Figures ). By 2012, most hospitals were utilizing the majority of elements within each domain (Table 4.10). Table 4.9. Overall changes in scope of QI interventions for hospitals pre- and post-cms policy intervention. Total of QI Interventions Time Period Hospitals Hospital-quarters Mean* SD Min Max Pre-CMS Policy Post-CMS Policy *Overall change in mean maximum number of domains in QI strategy across hospitals pre- vs. post- CMS policy intervention statistically significant at the 95% confidence-level (p = ; 95% CI: to -.29). CMS indicates Centers for Medicare and Medicaid Services. 137

154 Trends in Scope of QI strategies (N=55) Proportion of Hospitals Quarter (January, June, 2012) Hospitals with 0 domains Hospitals with 1 domain Hospitals with 2 domains Hospitals with 3 domains Hospitals with 4 domains Figure 4.1. Scope of QI strategies by quarter for UHC hospitals. 138

155 Trends in Adoption of Combined Leadership QI Interventions (N=55) Proportion of Hospitals Quarter (January, June, 2012) Hospitals with 0 Hospitals with 1 Hospitals with 2 Hospitals with 3 Hospitals with 4 Hospitals with 5 Hospitals with 6 Hospitals with 7 Hospitals with 8 Figure 4.2. Increases in scale by combination of leadership QI interventions. Trends in Adoption of Combined Staff QI Interventions (N=55) Proportion of Hospitals Quarter (January, June, 2012) Hospitals with 0 Hospitals with 1 Hospitals with 2 Hospitals with 3 Hospitals with 4 Hospitals with 5 Hospitals with 6 Hospitals with 7 Figure 4.3. Increases in scale by combination of staff QI intervention. 139

156 Trends in Adoption of Combined IT QI Interventions (N=55) Proportion of Hospitals Quarter (January, June, 2012) Hospitals with 0 Hospitals with 1 Hospitals with 2 Hospitals with 3 Hospitals with 4 Hospitals with 5 Figure 4.4. Increases in scale by combination of IT QI interventions. Trends in Adoption of Combined P&I QI Interventions (N=55) Proportion of Hospitals Quarter (January, June, 2012) Hospitals with 0 Hospitals with 1 Hospitals with 2 Hospitals with 3 Hospitals with 4 Hospitals with 5 Hospitals with 6 Hospitals with 7 Hospitals with 8 Figure 4.5. Increases in scale by combination of P&I QI interventions 140

157 Table Overall trends in scale of QI strategies by UHC hospital (N=55). Hospital-Level Scale* * N Leadership Hospital-quarters Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Staff Hospital-quarters Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Information & Information Technology (IT) Hospital-quarters Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Performance & Improvement (P&I) Hospital-quarters Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with Hospitals with *Annual data on QI strategy scope is based on what hospitals reported practicing at the mid-point of each year (i.e. end of second quarter); HAPU indicates hospital-acquired pressure ulcer; QI, quality improvement; UHC, University HealthSystem Consortium 141

158 According to the information above on scale, leadership had large increases, with the majority of hospitals practicing 5, 6 or 7 QI interventions within that domain. Staff interventions increased from a point where 21 hospitals were practicing zero staff interventions, to most adopting 4-5. There was a similar increase under the IT domain since most hospitals were practicing 3-4 interventions by Finally, P&I interventions remained spread out across the spectrum since this domain has the most options for application. P&I interventions still exemplified an upward trend from zero or one intervention to two or more across all hospitals with QI strategies. Hospitals saw increases in scale on an element-by-element basis as well for those elements labeled according to the best-practice framework (Figures ). Under the leadership domain, all elements had significant increases, with Program Mission showing the greatest increase. Six of seven staff interventions increased significantly in scale, with Team Huddles undergoing the greatest increase. All five IT interventions increased significantly. Electronic Alarm accounted for the greatest adoption among IT intervention. Finally, seven of nine P&I interventions had significant increases in scale, with Skin Care increasing the most. A test of proportions t-test was applied to test the signficance of increasing scale within each QI domain. This test aligns with the second part of Hypothesis 1b, which assumes that the scale of QI interventions increases per domain in practice following CMS policy intervention. Four t-tests were performed to find that there were statistically signficant increases in scale through all four domains 142

159 Trends in Adoption of Leadership QI Interventions (N=55) Proportion of Hospitals Program Mission Prevention Awareness Leadership Initiatives Admin Support Prevention Protocol Benchmarking Wound Team Quarter (January, June, 2012) Figure 4.6. Increases in scale by leadership QI intervention. Trends in Adoption of Staff QI Interventions (N=55) Proportion of Hospitals Performance Measures Team Huddles All-Staff Meetings Wound/QI Team Prevention Education Staff Training Quarter (January, June, 2012) Figure 4.7. Increases in scale by staff QI intervention. 143

160 Trends in Adoption of IT QI Interventions (N=55) Proportion of Hospitals Data Tracking EHR Risk Assess Electronic Alarm EHR Implementation Quarter (January, June, 2012) Figure 4.8. Increases in scale by IT QI intervention. Trends in Adoption of P&I QI Interventions (N=55) Proportion of Hospitals Quarter (January, June, 2012) Braden Scale Visual Tools Beds HAPU Staging Skin Care Incontinence Repositioning Nutrition Figure 4.9. Increases in scale by P&I QI intervention. 144

161 of the best-practice framework at the 95% confidence-level (Table 4.11). Of these increases, the greatest increase in adoption was seen in leadership QI interventions, followed in order by staff, P&I, and IT. Table Results of t-test for overall changes in QI scale at UHC hospitals preand post-cms policy intervention. Pre-CMS Policy Post-CMS Policy Change P- 95% CI Hospitals Hospital-quarters Mean SD Hospitals Hospital-quarters Mean SD Δ Value Lower - Upper Lead Staff IT PI Bolded p-values indicate statistically significant increases in scale at the 95% confidence level (p<0.05). Influences of HAPU Prevention Survey respondents self-reported a number of internal and external influential factors that may have led to an increased prioritization of HAPU prevention (Table 4.12). The majority of respondents indicated that among external influential factors, key drivers included: financial concerns; application for Magnet recognition; data sharing among peer institutions; and regulatory issues including oversight by JCAHO. Concerning internal influential factors that were indicated by a majority of respondents included: hospital prevention campaigns; the availability of nursing specialists such as CWOCNs; and the level of preventive knowledge among hospital staff. Summary of Aim 1 Findings Completion of the survey by 55 UHC hospitals indicated that most QI hospital-level QI strategies were broad in terms of scope and scale according to the best-practice framework. The survey also validated the study s interpretation 145

162 of best-practice framework for HAPU prevention since almost all listed QI interventions fit the framework, and few hospitals added other QI interventions in their responses. The effort of Aim 1 to organize QI intervention adoption longitudinally and merge it to quarterly HAPU incidence rates and covariates prepares the analyses of Aim 2. Table Self-reported influences of HAPU prevention, Characteristic N* % External Influential Factors Financial Professional - Magnet Application Professional - Hospital Collaborative Professional - Data Sharing Regulatory Other Internal Influential Factors High HAPU incidence Hospital prevention campaign High nursing job turnover Availability of nurse specialists Level of preventive knowledge Corporate hospital influence Patients with high BMI Other *Out of 55 total respondents. BMI indicates body-mass index; HAPU, hospital-acquired pressure ulcer. 146

163 Aim 2 Results: QI Intervention Effectiveness The 53 hospitals indicating the adoption of QI provided information to measure the effectiveness of QI interventions at reducing HAPU incidence. Aim 2 utilized an effect size analysis to measure quarterly derivatives of QI intervention effectiveness. In addition, an analysis of covariance (ANCOVA) model that included most covariates available from the UHC CDB/RM measured the impact of QI interventions on the variability of HAPU incidence. Several hypothesis tests were performed during these analyses in order to identify significant and meaningfully effective QI interventions of the best-practice framework. Finally, a sensitivity analysis of the effect size measures was performed to detect any uncertainty in the outcomes. Effect Size Analysis Effect sizes were calculated for the periods between all quarters where HAPU incidence rates existed (i.e. from the fourth quarter of 2007 until the second quarter of 2012). Nearly all of the QI interventions were well-powered during each quarter to calculate effect sizes. After quarterly effects were calculated, the average effect sizes across all quarters of observation were calculated. Effect sizes that exceeded 1 HAPU case per 1,000 inpatients were considered clinically meaningful. Quarterly effect sizes for each QI intervention were plotted over time in Figures These plots are organized by domain and depict time periods when the greatest effects take place during observation. Overall trends suggest that the largest effects occur early in the observation period in the time 147

164 leading up to CMS policy intervention. Following the 4 th quarter of 2008, the effect size of each QI intervention appears to decrease and oscillate around a net-zero effect. The sustained stabilized effect indicates a reduction in the variability of HAPU incidence as a result QI adoption, which would improve process control on HAPU prevention within the inpatient setting. The findings of the effect size analysis suggested that five QI interventions were effective to a clinically meaningful extent (Table 4.13). These QI interventions included one leadership QI intervention as well as four P&I QI interventions. Leadership Initiatives which suggested the inclusion of unit poster displays and outcomes boards was the meaningfully effective leadership QI intervention. The four P&I interventions that were meaningfully effective included: (1) Visual Tools ; (2) HAPU Staging ; (3) Skin Care ; and (4) Nutrition. While many other QI interventions had notable effect sizes, none exceeded the predetermined cutoff of clinical meaningfulness. The connection between Leadership Initiatives and Visual Tools suggests a potential handoff in the application of unit-based displays. This handoff begins at the level of administrative support, followed by clinician buy-in for QI initiatives at the level of the clinical microsystem. In addition, there is a likely association between HAPU Staging, Skin Care, and Nutrition as effective QI interventions. These three QI interventions reiterate the validity of certain components of the evidence-based prevention protocol which have undergone extensive investigation and scientific advancement since CMS policy intervention. 148

165 Effect Sizes of Leadership QI Interventions Effect Size (HAPU Cases per Inpatient Admission) Program Mission Prevention Awareness Leadership Initiatives Admin Support Prevention Protocol Benchmarking Wound Team Quarter (4th Quarter, nd Quarter, 2012) Figure Graphical depiction of time-dependent effect sizes by leadership QI intervention. 149

166 Effect Sizes of Staff QI Interventions Effect Size (HAPU Cases per Inpatient Admission) Performance Measures Team Huddles All-staff Meetings Wound/QI Team Prevention Education Staff Training Quarter (4th Quarter, nd Quarter, 2012) Figure Graphical depiction of time-dependent effect sizes by staff QI intervention. 150

167 Effect Sizes of IT QI Interventions Effect Size (HAPU Cases per Inpatient Admission) Data Tracking EHR Risk Assess Electronic Alarm EHR Implementation Quarter (4th Quarter, nd Quarter, 2012) Figure Graphical depiction of time-dependent effect sizes by IT QI intervention. 151

168 Effect Sizes of P&I QI Interventions Effect Size (HAPU Cases per Inpatient Admission) Braden Scale Visual Tools Beds HAPU Staging Skin Care Incontinence Repositioning Nutrition Quarter (4th Quarter, nd Quarter, 2012) Figure Graphical depiction of time-dependent effect sizes by performance and improvement (P&I) QI intervention. 152

169 Table Unadjusted results of effect size analysis by QI intervention. Hospital QI Intervention N Quarters Effect Size Clinically Meaningful Leadership Program Mission No Prevention Awareness No Leadership Initiatives Yes Admin Support No Prevention Protocol No Benchmarking No Wound Team No Staff Performance Measures No Team Huddles No All-Staff Meetings No Wound/QI Team No Prevention Education No Staff Training No Information & Information Technology (IT) Data Tracking No EHR Risk Assess No Electronic Alarm No EHR Implementation No Performance & Improvement (P&I) Overall Effect Braden Scale No Visual Tools Yes Beds No HAPU Staging Yes Skin Care Yes Incontinence No Repositioning No Nutrition Yes BOLDED effect sizes are clinically meaningful. CMS indicates Centers for Medicare & Medicaid Services; HAPU, Hospital-acquired Pressure Ulcer; QI, Quality Improvement 153

170 Across-hospital QI Effectiveness. A t-test was developed to test Hypothesis 2a that the greatest changes in HAPU incidence rates across hospitals occur early in the adoption process of QI interventions. This test was performed with an interest in studying where the greatest effect sizes occurred relative to CMS policy intervention. Thus, hospitals that began adoption of QI interventions early in 2007 had their effect sizes compared to that of hospitals that initiated QI intervention after an initial observational period of 3 quarters. According to a comparative measure of early vs. late adoption of each QI intervention in the framework, 24 out of 25 QI interventions showed greater effect sizes early in the observation period rather than late (Table 4.14). Of these 24 noted QI interventions, seventeen had early effect sizes that were greater than late effect sizes by a statistically significant margin. Among the significant QI interventions from this test were those identified as clinically meaningful from the initial effect size analysis presented in Table Within-hospital QI Effectiveness. The alternative Hypothesis 2b explores whether the greatest changes in HAPU incidence rates within hospitals occur early in the adoption process of QI interventions. To test this hypothesis, data were organized to calculate the effect sizes within each hospital for the first two interval quarters following adoption initiation. The average of these effect sizes were compared to that of all remaining quarters for each hospital s lag period when QI interventions were still in place. The difference between initial and lag periods were taken, and compared by t-test for statistical significance. 154

171 Table Statistical test of effect sizes for early adopters compared to late adopters of QI interventions for HAPU prevention across UHC hospitals. Early vs. Late Adoption Effectiveness P 95% CI QI Intervention Early Adoption Late Adoption Δ Value Lower Bound - Upper Bound Leadership Program Mission < Prevention Awareness < Leadership Initiatives < Admin Support < Prevention Protocol < Benchmarking < Wound Team Staff Performance Measures < Team Huddles All-Staff Meetings Wound/QI Team Prevention Education < Staff Training < Information & Information Technology (IT) Data Tracking EHR Risk Assess < Electronic Alarm EHR Implementation Performance & Improvement (P&I) Braden Scale Visual Tools Beds HAPU Staging Skin Care < Incontinence Repositioning Nutrition < BOLDED P-values indicate statistical significance at the 95% confidence-level. N/A indicates that an effect size was not calculable due to an under-powered sample for the QI intervention. CMS indicates Centers for Medicare & Medicaid Services; HAPU, Hospital-acquired Pressure Ulcer; QI, Quality Improvement This within-group comparison found that nineteen out of 25 QI interventions had greater effect sizes during the initial period of adoption which were statistically significant (Table 4.15). These QI interventions included the five clinically meaningful QI interventions from the effect size analysis. Two QI interventions presented a significant reverse effect, in which the greater effect 155

172 was seen following the initiation period. Four remaining QI interventions had no statistically significant effect when comparing the periods of initiation to lag. Analysis of Covariance The second part of Aim 2 proposed the development of a regression model to evaluate the impact of QI interventions and related covariates on the variability of HAPU incidence. In particular, an ANCOVA model was applied to detect reductions in variance of HAPU incidence following CMS policy intervention through the application of QI interventions. While the impact of CMS policy on HAPU incidence is indirect in theory, the changes in policy should influence adoption of QI interventions that in turn lead to greater process control for implementation of the prevention protocol. Multiple regression models were developed and tested in order to indentify a best-fit ANCOVA model. In general, each analysis modeled QI interventions as the main explanatory variables regarding HAPU incidence. The organization of data permitted a longitudinal analysis of hospitals with QI interventions in place relative to control hospitals that had no QI interventions in place quarter-toquarter. By controlling for additional covariates related to HAPU incidence, each model was tested for improved overall fit. The model with the greatest fit became the referent model of covariance for the study. 156

173 Table Statistical test of effect sizes for immediate adoption of QI interventions compared to lag periods for HAPU prevention within UHC hospitals. Initial vs. Lag Adoption Effectiveness QI Intervention Initial Effect Lag Effect Δ Value Lower Bound - Upper Bound Leadership Program Mission < Prevention Awareness < Leadership Initiatives < Admin Support < Prevention Protocol < Benchmarking < Wound Team < Staff Performance Measures Team Huddles All-Staff Meetings Wound/QI Team < Prevention Education < Staff Training < Information & Information Technology (IT) Data Tracking < EHR Risk Assess < Electronic Alarm EHR Implementation < Performance & Improvement (P&I) Braden Scale Visual Tools < Beds HAPU Staging Skin Care < Incontinence Repositioning Nutrition BOLDED P-values indicate statistical significance at the 95% confidence-level. CMS indicates Centers for Medicare & Medicaid Services; HAPU, Hospital-acquired Pressure Ulcer; QI, Quality Improvement P 95% CI Covariates. The covariates proposed in Chapter 3 included: age; LOS; CMS Policy; CMI; hospital magnet status; gender; ICU admission; mortality; medical or surgical procedures; and hospital size (bed counts). These covariates were each included in restricted models with overall presence of a hospital-level QI strategy. Compared to restricted models, the regression model that controlled for nine of the ten listed covariates simultaneously was the best-fit model based on an optimal R 2 value, even though some covariates were not statistically 157

174 significant. Hospital size (bed counts) was not a covariate that fit the model well based on restricted models, so left out of the final model. Based on this finding, the analysis proceeded with consideration for nine covariates while observing the direct correlation between QI interventions and HAPU incidence. A Hausman test found that a random effect model was the best form of the model to proceed under. Variance of QI. An initial ANCOVA model evaluated the correlation between overall QI intervention adoption and HAPU incidence (Table 4.16). The overall model had good fit measured by an R 2 of The correlation between QI interventions and HAPU incidence was statistically significant and explained a moderate proportion of variability according to the sum of squares. However, the marginal effect of QI on HAPU incidence was positive; this is a counter-intuitive correlation since a positive marginal effect suggests that QI interventions would increase HAPU incidence. The resulting marginal effect of QI is due primarily to the number of controls in place on the model, as opposed to the method of direct assessment between QI interventions and HAPU incidence in the effect size analysis. By including other covariates that account for a majority of the variability in the model (e.g. ICU admission, Magnet recognition, surgical procedures, and CMS Policy), the broad QI term may not be well differentiated for its impact on HAPU incidence. 158

175 Table Analysis of Covariance (ANCOVA) results for overall QI adoption relative to HAPU incidence. Covariate Beta Coefficient Marginal Effect Partial SS % of Total SS 95% CI QI E E-05 CMS CMI 4.11E E E E Magnet E Age (31-50) 1.63E E E Age (51-64) E Age (>64) E LOS -6.26E E E E E-06 Female 4.05E E E ICU E Death -5.10E E E Surgical E Residual Total Adjusted R-squared = BOLDED values indicate statistically significant coefficients at the 95% confidence-level; CI indicates confidence-interval; CMI, Case-mix Index; CMS, Centers for Medicare & Medicaid Services; ICU, intensive care unit; LOS, length-of-stay; QI, quality improvement; SS, sum of squares. The adjustment for CMS policy intervention accounted for the greatest proportion of variability in HAPU incidence among covariates at over 34%, as well as having a negative marginal effect. These findings suggest that CMS policy intervention was a primary correlate to HAPU prevention. The residual sum of squares accounted for the remaining 54% of variability in the model, which implies that other factors not explicit in the model or this study could account for some constituents of HAPU incidence. Variance of Domains. Compared to the initial ANCOVA model, a second model tested the impact of QI interventions organized by best-practice domain on HAPU incidence variability. This model had a slightly improved overall fit according to an R 2 of (Table 4.17). Although the partial sum of squares 159

176 for each domain suggested some impact on variability of HAPU incidence, the correlates of each domain were not statistically significant. Table Analysis of Covariance (ANCOVA) results for QI adoption by bestpractice domain relative to HAPU incidence. Covariate Beta Coefficient Marginal Effect Partial SS % of Total SS 95% CI Leadership E Staff E IT E P&I CMS CMI E E Magnet E Age (31-50) E Age (51-64) E Age (>64) E LOS -7.02E E E E-06 Female E ICU E Death E Surgical E Residual Total Adjusted R-squared = BOLDED values indicate statistically significant coefficients at the 95% confidence-level; CI indicates confidenceinterval; CMI, Case-mix Index; CMS, Centers for Medicare & Medicaid Services; ICU, intensive care unit; IT, information & information technology; LOS, length-of-stay; P&I, performance & improvement; QI, quality improvement; SS, sum of squares. As in the most general ANCOVA model, CMS policy intervention accounted for the most significant variability among covariates. Other significant covariates included the same three as previously: ICU admission, Magnet recognition, and surgical procedures. The residual sum of squares accounted for the remaining 54% of variability as in the previous model. Variance of QI Interventions. The third and most comprehensive ANCOVA model evaluated the correlation between each QI element separately and HAPU incidence, while controlling for all remaining covariates. This model 160

177 yielded an improved fit according to an increased R 2 of (Table 4.18). This high goodness of fit implies that subtle variation in HAPU incidence is wellexplained by adoption patterns of individual QI interventions rather than groupings by domain or overall adoption. On an elemental basis, 9 of the 25 QI elements had statistically significant correlations to HAPU incidence. However, only four of these elements had negative marginal effects. The alternate signs of marginal effects offer inconsistent conclusions about the directional impact of QI interventions on HAPU incidence. Tabulating all partial sum of squares for QI interventions accounts for 5.38% of the total variability in the model. Thus, controlling for QI interventions explain a significant amount of variability in the model and improve the overall predictability of HAPU incidence. CMS policy intervention and ICU admission were the only statistically significant covariates, and both had negative marginal effects. Again, CMS policy explained most of the variability in HAPU incidence through 31% of the partial sum of squares. The presence of individual QI interventions reduced the residual sum of squares to only 46% of the total compared with previous models. An additional model was developed in Table 4.19 which was identical to the model above, but did not control for CMS nonpayment policy. The purpose of this model was to review the effects of QI interventions on HAPU incidence independent of CMS policy. The results identified three leadership interventions and one P&I intervention that significantly impacted variance of HAPU incidence and had negative marginal effects. These variates represented a greater 161

178 Table Analysis of Covariance (ANCOVA) results for QI adoption by elements of the best-practice framework relative to HAPU incidence. Covariate Beta Coefficient Marginal Effect Partial SS % of Total SS Leadership Program Mission E Prevention Awareness E Leadership Initiatives E Admin Support E Prevention Protocol E Benchmarking E Wound Team E Staff Performance Measures E Team Huddles E All-Staff Meetings E Wound/QI Team E Prevention Education E Staff Training E Information Technology (IT) Data Tracking E EHR Risk Assess E Electronic Alarm E EHR Implementation E Performance & Improvement (P&I) Braden Scale E E Visual Tools E Beds E HAPU Staging E Skin Care E Incontinence E Repositioning E Nutrition E CMS E CMI E Magnet E Age (31-50) E Age (51-64) E Age (>64) E LOS -9.99E E E E E-06 Female E ICU E Death E Surgical E Residual Total % CI Adjusted R-squared = BOLDED values indicate statistically significant coefficients at the 95% confidence-level; CMI indicates Case-mix Index; CMS, Centers for Medicare & Medicaid Services; ICU, intensive care unit; LOS, length-of-stay; QI, quality improvement; SS, sum of squares. 162

179 Table Analysis of Covariance (ANCOVA) results for QI adoption by elements of the best-practice framework relative to HAPU incidence, without controlling for CMS nonpayment policy. Covariate Beta Coefficient Marginal Effect Partial SS % of Total SS 95% CI Leadership Program Mission E Prevention Awareness E Leadership Initiatives E Admin Support E Prevention Protocol E Benchmarking E Wound Team E Staff Performance Measures E Team Huddles E All-Staff Meetings E Wound/QI Team E Prevention Education E Staff Training E Information Technology (IT) Data Tracking E E EHR Risk Assess E Electronic Alarm E EHR Implementation E Performance & Improvement (P&I) Braden Scale E Visual Tools E Beds E HAPU Staging E Skin Care E Incontinence E Repositioning E Nutrition E CMI E Magnet E Age (31-50) E Age (51-64) E Age (>64) E LOS E E E-06 Female E ICU E Death E Surgical E Residual Total Adjusted R-squared = BOLDED values indicate statistically significant coefficients at the 95% confidence-level; CMI indicates Case-mix Index; ICU, intensive care unit; LOS, length-of-stay; QI, quality improvement; SS, sum of squa 163

180 proportion of the partial sum of squares compared to the previous models, but the fit of the equation was much lower considering its r-squared of Based on fit, an ANCOVA model controlling for CMS nonpayment policy better explains changes in variance of HAPU incidence than QI interventions alone. Sensitivity Analysis The sensitivity analysis evaluated uncertainty in the results of the effect size analysis. Potential uncertainty existed in the results of the survey and effect size analysis by two sources. First, recall bias by survey respondents could result in misreporting of start and end dates for QI interventions. Second, the study assumption that no lag time existed between QI initiation and a clinically meaningful effect could be overestimated the time between initiation and a period of effective change could extend past one quarter. To control for these uncertainties, the sensitivity analysis altered the start and end quarters of each QI intervention by one quarter in each direction. These changes in adoption were done at the hospital-level. As a result, two analyses tested uncertainty through evaluations of both an expanded period of QI adoption, and a reduced period of QI adoption. The sensitivity analysis identified only five circumstances where the results were uncertain (Table 4.20). Under the circumstances when the period of QI adoption was reduced by one quarter at the start and end dates, two P&I interventions no longer had clinically meaningful effects compared to the basecase analysis. These two QI interventions were Visual Tools and HAPU Staging. Likewise, the addition of one quarter to the start and end dates noted 164

181 Table Sensitivity analysis of effect size results by quarterly adjustments in QI adoption periods. Unadjusted Effect Clinically Effect Size Meaningful Quarter Reduction Quarter Addition Clinically Clinically Effect Size Meaningful Δ Effect Size Meaningful QI Intervention Δ Leadership Program Mission No No No Prevention Awareness No No No Leadership Initiatives Yes Yes Yes Admin Support No No No Prevention Protocol No No No Benchmarking No No No Wound Team No No No Staff Performance Measures No No No Team Huddles No No Yes All-Staff Meetings No No Yes Wound/QI Team No No No Prevention Education No No No Staff Training No No No Information Technology (IT) Data Tracking No No No EHR Risk Assess No No No Electronic Alarm No No No EHR Implementation No No No Performance & Improvement (P&I) Braden Scale No No No Visual Tools Yes No Yes Beds No No No HAPU Staging Yes No No Skin Care Yes Yes Yes Incontinence No No No Repositioning No No No Nutrition Yes Yes Yes BOLDED values indicate that the sensitivity analysis identified an alternative interpretation of clinically meaningful results than the base-case analysis. CMS indicates Centers for Medicare & Medicaid Services; HAPU, Hospital-acquired Pressure Ulcer; QI, Quality Improvement uncertainty in the clinically effectiveness of three QI interventions. Two staff interventions, Team Huddles and All-staff Meetings, were not clinically effective in the base-case analysis, but identified as effective in the sensitivity analysis. The opposite was the case for HAPU staging, which had a reduction in clinically meaningful effectiveness following the quarterly addition in hospital adoption periods. 165

182 The sensitivity analysis suggests that for the great majority of QI interventions, the results of the effect size analysis are robust to adjustments controlling for recall bias and lag time in QI intervention effectiveness. However, two QI interventions that are key components of the effect size analysis results, Visual Tools and HAPU Staging, present uncertainty and should be accepted as effective parts of a QI strategy for HAPU prevention with caution. The findings for Team Huddles and All-staff Meetings offer potential insight into staff interventions that are effective components of a QI strategy which base-case findings did not detect. Summary of Aim 2 Findings By characterizing effect size as a derivative of HAPU incidence for all observed time, certain QI interventions reduce HAPU incidence by a clinically meaningful measure. These effective interventions include one leadership intervention (Leadership Initiatives) and four P&I interventions (Visual Tools, HAPU Staging, Skin Care, and Nutrition). Elements of the framework were wellpowered to perform this analysis. Although 19 other QI interventions did not show clinically meaningful effectiveness, each had significant positive effects according to several statistical measures. The ANCOVA model evaluated the impact of QI interventions on variability of HAPU incidence. This model provided contrast to the effect size analysis by controlling for covariates that may have measurable impact on HAPU incidence, as opposed to QI interventions alone The model including adjustments for experimental controls, including hospitals that did not adoption QI strategy during 166

183 periods of observation. CMS policy intervention had a significant impact on HAPU incidence with great magnitude. Because of this impact on variability by CMS policy, it is difficult to differentiate the unique impact of certain QI interventions on HAPU incidence especially considering that QI adoption time periods are simultaneous to the period of impact by CMS policy intervention. Nonetheless, certain leadership as well as performance and improvement QI interventions had significant marginal effects on reduction in HAPU incidence. Aim 3 Results: QI Strategy The purpose of Aim 3 was to investigate the effectiveness of combinations of QI interventions as part of a robust QI strategy, broad in both scope and scale. To address this aim, we evaluated combinations of QI interventions in an effect size analysis, as well as examined QI strategies of high performing hospitals in the study for overlapping QI interventions. Ultimately, this aim was carried out to identify patterns of QI interventions among UHC hospitals following CMS policy intervention. Aim 3 developed several exploratory sub-hypotheses to address the comparative effectiveness of hospital-level QI strategies. First, effective QI strategies are dynamic by principle of the best-practice framework. Second, effective QI strategies have many of the same QI interventions in common. Third, dynamic QI strategies are comparatively more effective at reducing HAPU incidence than other forms of strategy following CMS policy intervention. 167

184 Effect Size Analysis of QI Strategies In follow-up to the effect size analysis of Aim 2, effective QI interventions that were found clinically meaningful were re-evaluated for effectiveness in combinations. For example, since Leadership Initiatives and Visual Tools were each identified as effective individually, select hospitals that included both QI interventions in their QI strategies were evaluated for effect size. This evaluation simulated the comparative effectiveness of QI strategies based on observational data. In all, five QI interventions were tested together totaling ten possible oneway combinations. Each combination was found to have a clinically meaningful effect by reducing HAPU incidence by at least 1 case per 1,000 inpatients (Table 4.21). Given the selective sample sizes for each combination evaluated, only three combinations were well-powered to draw conclusions from the effect size analyses. These combinations of QI interventions that were statistically and clinically meaningfully effective included the following: (1) Leadership Initiatives and Visual Tools; (2) Leadership Initiatives and Skin Care; and (3) Visual Tools and Nutrition. The remaining seven combinations had sample sizes under a minimum of eight hospitals practicing each combination of QI interventions in order to assume a well-powered sample. 168

185 Table Effect sizes of combinations of effective QI interventions. Leadership Visual Tools HAPU Skin Care Nutrition QI Intervention Initiatives Staging Leadership Initiatives x x x x x Visual Tools x x x x HAPU Staging x x x Skin Care x x Nutrition x BOLDED effect sizes indicate statistical significance at the 95% confidence-level. HAPU indicates Hospital-acquired Pressure Ulcer; QI, Quality Improvement. The effect sizes in Table 4.21 indicate a synergistic effect of QI intervention combinations on HAPU incidence. These effect sizes are greater on average than the effect sizes of individual QI interventions evaluated in Aim 2. Two of the three notable combinations include a leadership intervention and P&I intervention, trending towards a dynamic QI strategy in terms of scope. A time-dependent plot of effect sizes for these combinations of QI interventions offers similar insight to plots of individual QI interventions (Figure 4.14). Greatest effect sizes are observed around the time that the announcement for CMS policy changes through the actual policy intervention took place (4 th quarter, th quarter, 2008). Following adoption of the common QI intervention combinations, HAPU prevention appears to stabilize, thereby reducing the effect size toward net-zero. High-performing Hospital QI Strategies There were multiple definitions of UHC hospitals considered highperformers of HAPU prevention based on reduced or consistently low HAPU incidence rates. Studying these hospitals QI strategies may offer some insight 169

186 into effective forms of QI strategy that would be worthwhile for other hospitals to emulate. The three definitions of high-performers, along with the QI strategies of high-performing hospitals be each definition are reviewed here. Definition 1: Lowest Overall HAPU Rates. The first definition of highperforming hospitals consists of those which had consistently low HAPU rates in recent observation, regardless of performance during prior periods. This definition was derived based on the assumption that QI strategies continuously improve in terms of scope and scale following CMS policy intervention. Therefore, HAPU prevention performance improves with adoption of sustainable QI strategies, which are likely observable during the latest periods since CMS policy intervention. During the span of January, 2012 through June, 2012, sixteen hospitals in the study sustained 0% HAPU rates. These hospitals shared a number of QI interventions in common, as well as exhibited dynamic scope and scale. Over 80% of these hospitals shared the following QI interventions by domain: Leadership: Prevention Protocol; Wound Team Staff: Prevention Education; Staff Training IT: Data Tracking; EHR Risk Assessment Nothing was adopted consistently by these hospitals based on the P&I domain. Additionally, all sixteen of these hospitals had at least one leadership intervention in their QI strategies. 170

187 Effect Sizes of QI Intervention Combinations 0.01 Effect Size (HAPU Cases per Inpatient Admission) Leadership Initiatives & Visual Tools Leadership Initiatives & HAPU Staging Leadership Initiatives & Skin Care Leadership Initatives & Nutrition Visual Tools & HAPU Staging Visual Tools & Skin Care Visual Tools & Nutrition HAPU Staging & Skin Care HAPU Staging & Nutrition Skin Care & Nutrition Quarter (4th Quarter, nd Quarter, 2012) Figure Graphical depiction of time-dependent effect sizes by combination of QI interventions. 171

188 Definition 2: Greatest HAPU Incidence Reductions. The second definition consists of those hospitals that exhibited significant marginal improvements in HAPU incidence rates during the period of observation. These hospitals QI strategies could offer validation of the best-practice framework since following adoption, HAPU rates were reduced. Two hospitals in the study exemplified great reductions in HAPU incidence over the five-year period by driving down HAPU incidence over 40 cases per 1,000 inpatients. In the 4 th quarter of 2007, Hospital A observed a HAPU rate of HAPU cases per 1,000, and Hospital B s rate was HAPU cases per 1,000. By 2 nd quarter of 2012, Hospital A s rate dropped to 1.06 cases per 1,000, and Hospital B had a reduced rate of 1.00 cases per 1,000. Both hospitals exhibited QI strategies that were dynamic in both scope and scale, and included most QI interventions from the best practice framework (Table 4.22). Definition 3: Lowest Average HAPU Rates. The third definition of highperforming hospitals is based on those hospitals that maintained the lowest average HAPU incidence rates across every quarter of observation. Although these hospitals may not have experienced as great of effect sizes as hospitals by the other two definitions, these hospitals have sustained process control in the realm of HAPU prevention over several years. Three such hospitals were found to have the lowest average HAPU incidence rates: Hospital X maintained a rate of cases per 1,000; Hospital Y maintained a rate of cases per 1,000; and Hospital Z maintained a rate of cases per 1,000. Each of these hospitals had broad scope and scale 172

189 dynamics, as indicated in Table The overlapping QI interventions of these hospitals included Benchmarking, Wound Team, Prevention Education, Staff Training, EHR Risk Assessment, and Visual Tools. Table QI strategies of hospitals with the greatest reductions in HAPU incidence rates. QI Intervention Hospital A Hospital B Shared QI Interventions Leadership Program Mission x x x Prevention Awareness x Leadership Initiatives x x x Admin Support x Prevention Protocol x Benchmarking x x x Wound Team x x x Staff Performance Measures x x x Team Huddles x All-Staff Meetings x Wound/QI Team x x x Prevention Education x x x Staff Training x x x Information Technology (IT) Data Tracking x x x EHR Risk Assess x x x Electronic Alarm x x x EHR Implementation x x x Performance & Improvement (P&I) Braden Scale x x x Visual Tools x x x Beds x x x HAPU Staging x x x Skin Care x x x Incontinence x Repositioning x x x Nutrition x x x 173

190 Table QI strategies of hospitals with the lowest average HAPU incidence rates. QI Intervention Hospital X Hospital Y Hospital Z Shared QI Interventions Leadership Program Mission x x Prevention Awareness x Leadership Initiatives x x Admin Support x Prevention Protocol x x Benchmarking x x x x Wound Team x x x x Staff Performance Measures x x Team Huddles x x All-Staff Meetings Wound/QI Team x x Prevention Education x x x x Staff Training x x x x Information Technology (IT) Data Tracking x x EHR Risk Assess x x x x Electronic Alarm x x EHR Implementation x x Performance & Improvement (P&I) Braden Scale x x Visual Tools x x x x Beds x x HAPU Staging x Skin Care x x Incontinence x x Repositioning x Nutrition 174

191 Summary of Aim 3 Findings The effect size analysis of QI intervention combinations may be most informative to hospitals of the 3-4 QI interventions to base a QI strategy upon. Leadership Initiatives, Visual Tools, Skin Care, and Nutrition all had clinically meaningful effects on the reduction of HAPU incidence. In addition to these findings, the QI strategies of high-performing hospitals all appeared to be dynamic in terms of scope and scale despite diversity between strategies. Therefore, hospitals studying adoption of QI interventions that support HAPU prevention can approach the issue with a base of effective QI interventions. With these results in mind, hospitals can consider which staff and IT interventions may support successful follow-through for consistent implementation of the prevention protocol. 175

192 CHAPTER V DISCUSSION Aim 1 Discussion This study provided seminal data regarding longitudinal adoption patterns of quality improvement (QI) interventions among academic medical centers in the University HealthSystem Consortium (UHC). The survey responses from 55 hospitals reflected a representative sample of all UHC hospitals based on similar trends of inpatient demographics and hospital-acquired pressure ulcer (HAPU) outcomes. These responding hospitals set an example for all US hospitals based on their common implementation of the evidence-based HAPU prevention protocol and goals for clinical excellence as exhibited by Magnet recognition amongst a majority of respondents. Fifty-one hospitals (92.7%) reported implementing an evidence-based prevention protocol, and 53 respondents (96.4%) indicated use of QI interventions for HAPU prevention. These data exemplify efforts to further improve patient safety in the inpatient setting. UHC hospitals made significant progress to develop effective QI strategies for HAPU prevention following the announcement of a non-payment policy for hospital-acquired conditions (HACs) by the Centers for Medicare and Medicaid Services (CMS). 112 The study found dynamic structure in these QI strategies based on scope and scale according to the best-practice framework developed by Nelson et al. (2007). 28 QI strategies with broad scope and scale are indicative 176

193 of improvements in HAPU prevention through consistent implementation of the evidence-based HAPU prevention protocol. The HAPU incidence rates calculated for Aim 1 indicate significant reduction of HAPU incidence since the announcement of CMS policy intervention between HAPU incidence in the past three years has stabilized at less than 1.0 cases per 1,000 inpatients across UHC hospitals. This reduction in HAPU incidence across UHC hospitals has occurred despite the fact that there are few noticeable differences in the patient population in terms of age, gender, or procedural status. Stabilization of complications for the incident HAPU population such as mortality and intensive care unit (ICU) admission may suggest that patients who still developed HAPUs were the most morbid inpatient cases. Thus, hospital clinicians successfully improved HAPU prevention for the majority of inpatients. Many of those who still developed a HAPU since CMS policy intervention could be categorized as unavoidable cases according to the National Pressure Ulcer Advisory Panel (NPUAP). 51 Survey responses indicated that most UHC hospitals support hiring multiple Certified Wound, Ostomy and Continence Nurses (CWOCNs) to manage HAPU prevention. CWOCNs are responsible for developing QI strategy and gaining support from both hospital administration and other clinical staff. Most hospital CWOCNs delineate implementation of the HAPU prevention protocol to staff nurses and focus on special cases. Although, some hospitals reported that CWOCNs implemented the prevention protocol themselves. 177

194 This study validates the best-practice framework as a structure for QI strategies pertaining to HAPU prevention. After working closely with clinical experts to modify the framework for HAPU prevention, the survey responses did not suggest any alterations to the framework. All responses fit within the four domains of the framework, and few QI interventions fell outside of the 25 specified elements. The time-dependent trends in QI intervention adoption show significant overall increases, as well as significant increases in scope and scale during observation. The greatest increases took place when comparing QI adoption trends before and after CMS policy intervention. Hospitals prioritizing HAPU prevention should emulate the practice habits of respondents by broadening the dynamics of their QI strategies. Given the reductions in HAPU incidence coinciding with adoption patterns of QI interventions, hospitals that focus on enhancing all components of their QI strategy with regards to leadership, staff, information and information technology (IT), and performance and improvement (P&I) interventions gain the best likelihood of improving HAPU outcomes. While data from Aim 1 identify specific QI interventions based on frequency, the analyses in Aim 2 and Aim 3 provide quantifiable recommendations for QI interventions. Action to expand QI interventions should be based on those results which follow in this chapter. The influences that respondents selected as part of their efforts to enhance HAPU prevention provide meaningful insight into the pressures that hospitals face. The greatest external and internal influential factors were 178

195 financial constraints and availability of nurse specialists, respectively. The influence of hospital finances based on the fact that CMS no longer reimburses for HACs suggests that HAPUs particularly burden hospital budgets. Given the cost-effective nature of the HAPU prevention protocol over treatment, hospitals have recognized the importance of investing in prevention. 27 The issue of nurse specialist availability complements the financial investment in HAPU prevention since CWOCNs and other specialists take on leading roles in protocol implementation and developing QI strategies. These specialists come at an increased cost compared to staff nurses. Still, most hospitals appear to be investing in multiple CWOCNs to address the issue of HAPU prevention, along with many other HACs. Aim 1 Limitations Several limitations exist in Aim 1 based on the study design and certain findings. Aim 1 results were based almost entirely on the survey, which has certain inherent validity and reliability issues. Since the survey received 54 responses from 180 targeted hospitals, there is some question about the applicability of these findings outside of the sample population or UHC hospitals in general. Results should be considered with caution to other types of U.S. hospitals such as small, rural, or community medical centers. The most concerning issues in the survey design are recall bias and reporting bias. Recall bias stems from the fact that respondents are expected to indicate start and end dates of QI interventions over the course of the six years between to the nearest quarter. Some respondents had not worked 179

196 for their current hospital for this entire duration, therefore limiting the knowledge of QI interventions adopted prior to starting their current position. Other respondents acknowledged that their awareness of start dates for QI interventions were only estimates, especially dating past several years. While these estimates do not impede a descriptive analysis of QI adoption trends in Aim 1, it does limit validity of the results in Aim 2 prior to the sensitivity analysis. A second administration of the survey to all respondents could have impeded recall bias by ensuring that start and end dates are consistent between both administrations. This approach was evaded since respondents were volunteering their time away from primary duties to ensuring patient safety. Reporting bias is also concerning since the 53 hospitals indicating QI intervention adoption could represent the majority of UHC hospitals focusing on QI for HAPU prevention. Hospitals that didn t respond may have not been so inclined since the survey instrument did not pertain to their preventive efforts. It remains an assumption that the 55 hospitals responding to the survey are a representative sample of UHC hospitals. On that note, this study only represents nonprofit academic medical centers of the UHC, and does not necessarily extend to other hospital classifications. Validation of the survey instrument was limited to the pilot test of experts in the fields of QI theory and HAPU prevention at two UHC sites. A larger pilot with participation from additional sites would have been ideal, but was difficult based on time and resource constraints. Fortunately, survey responses added 180

197 few elements to the best-practice framework, and open-ended comments did not imply that the survey instrument was impertinent to the topic of HAPU prevention. Aim 1 Implications The survey successfully characterized longitudinal adoption patterns of QI interventions for HAPU prevention at the hospital-level. Three hypotheses for this survey related to increased adoption of QI interventions were tested and confirmed. Overall QI intervention adoption increased significantly, as well increases in scope and scale, following CMS policy intervention in the 4 th quarter of CMS policies with financial implications are effective motivators for hospitals to explore novel methods for improving patient safety, especially with respect to HAC prevention. Given that the NPUAP has devised an evidence-based protocol for HAPU prevention, hospitals have a point of reference for improving patient safety. The best-practice framework offers structure for successful implementation of the prevention protocol. Hospitals can devise QI strategies based on a variety of different combinations of QI interventions within each domain to promote consistent protocol implementation. While this study offers insight into the QI intervention adoption trends by respondents, other hospitals should consider developing a QI strategy based on individual needs of its staff at the level of the clinical microsystem. Nonetheless, a QI strategy that contains the four domains of the best-practice framework leadership, staff, IT, and P&I ensures the greatest likelihood of successful HAPU prevention. 181

198 Aim 2 Discussion UCH medical centers responded to CMS nonpayment policy for HAPUs with significant changes in practice. All hospitals took great strides to incorporate evidence-based prevention to eliminate HAPUs. These efforts appeared to be relatively successful, since HAPU incidence rates are not as great or alarming as once existed during the early 2000s. The effect size analysis offered comparative data on effective QI interventions that have direct application to pressure ulcer prevention in the inpatient setting. Responses from 53 hospitals indicating QI adoption for HAPU prevention provided a well-powered sample to test the effectiveness of QI interventions. Results of the effect size analysis indicate that one leadership QI intervention, Leadership Initiatives, and four P&I interventions Visual Tools, HAPU Staging, Skin Care, and Nutrition - offer clinically meaningful reductions in HAPU incidence. These results provide a useful starting point for hospitals considering adoption of QI interventions to improve consistent implementation of the HAPU prevention protocol. The connection between Leadership Initiatives and Visual Tools is noteworthy since these two QI interventions are related. Leadership Initiatives take place in upper tiers of hospitals, where the administrative and financial support are necessary in order to promote utilization of unit poster displays, information flyers, outcomes boards, checklists, etc. With that support, clinical staff such as CWOCNs can carry out the demonstration of visual reinforcement tools at the level of the clinical microsystem. 41 Such tools have been used 182

199 previously to target successful follow-through of evidence-based prevention protocols, thereby leading to improved prevention. 37 Recommendations of this study are based on a robust methodology, the effect size analysis, in which the effectiveness is characterized as a derivative of HAPU incidence reduction for each QI intervention over a time series. While other QI interventions were not found clinically meaningful by a predetermined threshold of HAPU prevention (i.e. 1 HAPU case per 1,000 inpatients), many QI interventions did present preventive potential. The sensitivity analysis found that 96% of iterations from the effect size results were robust to change. Some variance from the sensitivity analysis highlighted multiple QI interventions including staff interventions that could be effective under different circumstances than reported in the survey. On the other hand, some QI interventions including recommended results were found less effective. It could be that given confirmation of Hypotheses 2a and 2b which describe the greatest effect sizes occurring early in adoption, adding or removing periods of crucial effectiveness early in observation significantly alter the results. Overall, the changes in effectiveness from the sensitivity analysis were only small shifts in effect size. Hospitals should refer to the base-case results of this effect size analysis for structuring a QI strategy in HAPU prevention, and consider other QI interventions to support implementation of the prevention protocol based on individual clinician needs and the availability of resources. Testing the within- and across-hospital changes in QI intervention effectiveness over time found that the greatest reductions in HAPU incidence 183

200 occurred early in the adoption process. Across all hospitals, those that started applying QI strategy to HAPU prevention prior to CMS policy saw the greatest effectiveness in HAPU prevention. Likewise, within-hospital trends found that the greatest reductions in HAPU incidence occur in the first three quarters following QI intervention adoption. In either case, QI interventions are indicative of reductions in HAPU incidence as well as reduced incidence variability soon after adoption takes place. The sustainability of QI interventions in clinical processes maintains improved HAPU prevention for an extended duration. Therefore, hospitals should target HAPU prevention with QI interventions that are highly supported by clinicians for long-term adoption rather than on a temporary project basis. 128 There was a challenge in differentiating the impact of CMS policy from adoption of QI interventions on changes in HAPU incidence. The analysis of covariance (ANCOVA) model, which controlled for both QI adoption and CMS policy, did not identify the same level of positive correlation between QI interventions and HAPU prevention as did the effect size analysis. CMS policy intervention in the 4 th quarter of 2008 was highly correlated to reductions in HAPU incidence, and accounted for the greatest changes in HAPU variability. However, the nonpayment policy alone does not explain reductions in HAPU incidence. According to Gonzales et al (2012), 34 the conceptual framework for implementation science best explains the series of events that occurs between policy, evidence-based practice, and QI interventions to improve prevention. Since the announcement for CMS policy changes occurred shortly before the 184

201 actually policy intervention, it is difficult to sort out of this study the effectiveness of QI interventions independent of policy. Compared to the ANCOVA model, the effect size analysis offers the impact of QI interventions on a relative scale, and provides the best actionable information for hospital clinicians. Although the ANCOVA models did not offer much insight into the risk of patient demographic factors such as age, gender, length of stay (LOS), and procedural status on HAPU outcomes, previous evidence clarifies these risk factors. Allman (1989) 54 previously identified elderly patients at greater risk for HAPUs. Tescher et al. (2012) 160 found that among a large patient populations, many subscale scores of the Braden Scale and subscales for medical/surgical status and patient diagnostics improved HAPU risk assessment. In contrast, this study adds to existing literature by quantifying the impacts of nonpayment policy and QI intervention adoption to improving HAPU prevention. Limitations The effect size analysis and ANCOVA model were limited by several factors. First, the study relied upon recollection of start and end dates for QI interventions at hospitals through a survey of HAPU prevention experts. Recall bias in this survey could effect the measurements in both analyses of this study since periods of QI adoption are not completely reliable. Second, the period when QI interventions begin to take effect after adoption is not certain. We assumed immediate effect following the first quarter of initiation, but there could be a lag between initiation and effectiveness. The sensitivity analysis was meant 185

202 to address the inherent issues of recall bias and lag effect by measuring effect sizes for altered start and end dates. Reporting bias could have also limited the results of this study based on the survey data. Hospitals that observed positive effects following QI adoption are more likely to report their findings than hospitals with no effect or negative outcomes. The effect size analysis should be accepted with precaution that not all hospitals might experience the same level of effectiveness as the hospitals in this study. There may be some issues with the reliability of UHC data due to transcription errors in coding at the level of data entry or variability between patient and billing records. These issues are beyond the scope of correction in this project. We assume that as UHC becomes a more established source of data, these issues are less prevalent. HAPU incidence rates following CMS policy intervention are confounded by national changes in diagnostic coding through electronic health record (EHR) systems. Since HAPUs are not coded as present-on-admission (POA) or as a primary diagnosis code, patients are first coded for their admitted health state. According to Lindenauer et al. (2012), 75 the lack of incentive to code secondary diagnoses, especially if space for diagnosis coding in an EHR is full, leaves HAPUs uncoded in EHRs. Thus, HAPU incidence rates could possibly be greater than found in this study. This concern is mitigated by the fact that hospitals are mandated to report all incidences of HACs, and that misreporting 186

203 can jeopardize a hospital s accreditation status with the Joint Commission on 35, 108 Accreditation of Healthcare Organizations (JCAHO). The issue of coding neglect does not supersede the possibility that HAPUs are occasionally miscoded as well. The process for HAPU staging is complicated by measurements which occupy nurses time. Variation in the approach to staging could result in a stage II pressure ulcer being diagnoses under stage II, or vice versa. The results of this inconsistency could certainly skew HAPU incidence rates of this analysis which only considers stage III and IV HAPUs. HAPU incidence rates should be accepted in this study with some caution; however, the largely significant reduction in HAPUs following CMS nonpayment policy is unlikely a factor of miscoding alone. Findings of the ANCOVA model were limited by difficulty in defining the interaction between covariates such as the effects of CMS policy intervention versus adoption of QI interventions on HAPU incidence. Data suggests that CMS policy was a crucial driver in QI adoption patterns. Since the greatest effects in HAPU incidence reduction took place overlapping with the period of hospital awareness for CMS policy changes, it remains unclear what the effectiveness of QI interventions are alone. In addition, the ANCOVA model lacked other covariates that may have developed a better-fit model, such as Braden scores for HAPU patients, patient diagnoses and comorbid conditions, laboratory data, as well as prescribed pharmacotherapy. These additional covariates may have been shown to accurately predict risk of certain patient populations. For example, laboratory 187

204 data on patient BMI would offer useful information since overweight patients are at greater risk for HAPUs than patients of normal weight. 160 Finally, the ANCOVA model only studied the additive correlation of covariates on the dependent variable, HAPU incidence. Further models in this area could consider other types of correlations between HAPU risk factors including age and LOS. A possible approach to testing these correlations would be to see how multiplicative, exponential, and inverse correlations of certain risk factors or QI interventions may improve the fit of the overall model. Implications This study offers clear directive on five QI interventions that are effective based on a robust methodology. Hospitals investigating novel approaches to HAPU prevention should consider adoption of the following QI interventions to improve the effectiveness of their QI strategy. First, hospitals should gain administrative support to promote routine leadership initiatives (e.g. unit poster displays, informational flyers, outcomes boards). By developing a campaign of visual tools that promote HAPU prevention awareness, hospitals clinicians are more likely to take notice and interest in support of a cause. The follows closely with the second QI intervention, Visual Tools. Visual Tools represents the leadership initiatives adopted and carried out on the hospital floor by clinicians following support of administration. Additional scientific tools such as checklists, posters, or bundled interventions help to ensure follow-through of the prevention protocol. 188

205 The remaining three QI interventions identified as meaningfully effective in this study are HAPU Staging; Skin Care; and Nutrition. These three components of a QI strategy are taken directly from the HAPU prevention protocol as improvements to existing protocol. A protocol with improvement updates to the staging process are more likely to identify early-stage HAPUs that could otherwise lead to deleterious cases. Acquisition of new skin care products and creams ensures that hospitals are constantly updating their supply of products to the most current and effective. Finally, an emphasis on a nutrition regiment for patients rich in protein, vitamins and other essential nutrients deters the advancement of HAPUs. Using these elements in the context of a dynamic QI strategy in terms of broad scope and scale improves the overall chances of reducing HAPU incidence. Based on QI theory, these QI interventions should be directed towards the support of an evidence-based HAPU prevention protocol. According to the NPUAP, the prevention protocol is the only valid guideline for preventing HAPUs. QI interventions that supersede this conceptual framework of implementation science may not improve HAPU prevention. As CMS implemented its nonpayment policy for HACs in 2008, there were significant reductions in HAPU incidence. These incidence reductions directly coincided with adoption patterns of QI interventions at UHC hospitals. Therefore, CMS policy is a key driver in improving patient safety in hospital medicine. While policy does not directly affect HAPU incidence rates, the external influence of 189

206 financial regulations incentivized hospitals to respond efficiently by improving HAPU prevention through adoption of novel QI interventions. The field would benefit from ongoing research of different types of QI interventions and combinations as strategies that improve HAPU prevention. Information on other HAPU outcomes including Braden scoring, comorbid diagnoses, laboratory data, and pharmacotherapy at the level of clinical microsystems could contribute to greater HAPU risk-stratification among hospital inpatients. By identifying new high-risk patient groups, additional QI interventions may become available that assist in implementation of the prevention protocol to these patients. Another study on HAPU outcomes that would add to the current evidence base would include analyses of the interactions between covariates such as CMS policy and certain QI interventions. The interaction between CMS policy and QI adoption on HAPU incidence is difficult to parse out. The direct interactions between different QI interventions such as Leadership Initiatives and Visual Tools would provide helpful insight into the actual effect of this combination on HAPU incidence while controlling for other covariates. More comprehensive ANCOVA models evaluating these interactions would offer useful information to stakeholders in HAPU prevention. Aim 3 Discussion The study is a novel approach evaluating the comparative effectiveness of QI strategies using real-world data for HAPU outcomes. The results of this study 190

207 suggest that certain combinations of QI interventions have an improvement effect on HAPU prevention. Applying an effect size analysis to study the effectiveness of QI intervention combinations identified an enhancement in HAPU incidence reduction over individual QI interventions. This finding supports the theory that effective QI strategies are broad in scope and scale, and based on elements of the best-practice framework. Three combinations of QI interventions are recommended for adoption to support an evidence-based HAPU prevention protocol: (1) Leadership Initiatives and Visual Tools; (2) Leadership Initiatives and Skin Care; and (3) Visual Tools and Nutrition. The first two of these combinations cover issues of scope since both include domains of Leadership and P&I interventions. The third combination focuses more on issues scale since Visual Tools and Nutrition are each P&I interventions. These combinations of QI interventions apparently have a synergistic effect on reducing HAPU incidence. Previous analyses to evaluate the effectiveness of individual QI interventions found lesser effect sizes than these combinations. Furthermore, these QI intervention combinations showed a positive impact on the reduction of HAPU incidence variability. These improvements in effect and variability coincided with implementation of the CMS nonpayment policy. The study was based on a survey of 53 hospitals reporting use of QI interventions to support evidence-based HAPU prevention. The survey responses were combined with HAPU incidence data to analyze the effect sizes of QI interventions over a six-year period. While the study was well-powered to 191

208 evaluate individual effect sizes, it was not designed to study effect sizes of withingroup combinations of QI interventions. Consequently, there are seven other combinations of QI interventions highlighted in this study as potentially beneficial combinations for a QI strategy. However, evidence in inconclusive as to whether the clinically meaningful effect sizes of additional QI intervention combinations are statistically significant. In addition to the robust quantitative analysis of QI intervention combinations, this study reviewed the qualities of QI strategies among highperforming UHC hospitals. Hospitals were selected based on consistent efforts to reduce HAPU incidence by one of three definitions. First, hospitals were evaluated based on consistently low HAPU rates between January and June of Second, hospitals with the greatest reductions in HAPU incidence since CMS policy intervention were evaluated for overlapping qualities in QI strategy. Third, the study reviewed hospitals with the lowest mean rates of HAPU incidence during the entire period of observation ( ). By all three definitions, the best performing UHC hospitals in the sample had broad scope and scale dynamics to their QI strategies. Wound Team was consistently utilized by hospitals under each definition as a leadership element. The same applied to staff QI interventions Prevention Education and Staff Training, as well as IT QI intervention EHR Risk Assessment. Visual Tools appear most often among P&I interventions. While this study did not identify some of these QI interventions as effective through empirical methods, scope and scale dynamics imply that these QI interventions could provide a good 192

209 foundation upon which to develop a QI strategy for HAPU prevention. By gaining support from all stakeholders for evidence-based practice at the level of the clinical microsystem, including hospital administration and clinical staff, consistent implementation of the evidence-based prevention protocol stands the best chance of success. Limitations The evaluation of effect sizes for QI intervention combinations was limited by imperfect control for CMS policy intervention, as well as possibly other covariates that affect HAPU risk. The greatest effect sizes for each QI intervention combination occurred during overlapping time spans with CMS policy intervention in As a result, it is difficult to define the amount of HAPU incidence reduction attributed solely to QI adoption. However, QI theory suggests that the policy has little effect on actual reduction of HAPU incidence. The only valid device that leads to HAPU prevention is the evidence-based protocol endorsed by the NPUAP. Policy may have incentivized hospitals to explore novel methods for improving consistent implementation of the prevention protocol, such as through adoption of QI interventions. The sample size which was powered for effectiveness of individual QI interventions limits the interpretation of these results with respect to QI intervention combinations. Although the effect sizes of these QI interventions are clinically meaningful, none are well-powered. Therefore, evidence on the comparative effectiveness of QI strategies remains inconclusive. Investigating QI 193

210 strategy from the standpoint of the best-practice framework with broad scope and scale dynamics is an appropriate starting point. The exploratory evaluation of QI strategy effectiveness for high-performing UHC hospitals is limited by inherent recall and reporting biases of the respondents. The survey implemented to collect information on QI strategies depends on recollection of QI adoption patterns over a six-year period. Hospitals may have overstated the adoption of certain QI interventions, or neglected to mention other components of their QI strategies for HAPU prevention that no longer exist. Additionally, reporting bias would suggest that hospitals with the most comprehensive QI strategies responded to the survey. There may be other UHC hospitals that manage HAPU prevention more efficiently than those in the sample and still achieve significant reductions in HAPU incidence. The caveat of diminishing returns discussed in Chapter 2 arises again in Aim 3. Since this investigation lacks the economic impact of QI strategy expansion, hospitals should approach expansion of existing QI strategies with careful consideration for the return on investment in terms of HAPU prevention. Bundled payment systems from CMS and other insurance companies limit hospitals resources to implement new strategies for HAPU prevention despite the potential cost-effectiveness. Implications This study explores the comparative effectiveness of QI strategies using a novel empirical approach. Certain combinations of leadership and P&I QI 194

211 interventions could significantly reduce HAPU incidence rates. By coupling these effective combinations of QI interventions with other elements that enhance the scope and scale of an overall QI strategy, hospitals may be able to reduce HAPU rates to a point where only unavoidable incidences remain. The field would benefit from further research on the comparative effectiveness of QI strategies in a prospective quasi-experimental approach. This study has narrowed the potential combinations of QI interventions from the best-practice framework to several that optimize implementation of the evidencebased prevention protocol. By testing these combinations in separate clinical microsystems within variable hospital settings, clinicians can make conclusive decisions about a final set of QI interventions for HAPU prevention. Taking a prospective approach to this experiment counters the issues of recall bias and lag effect of policy on QI adoption since the study can track QI adoption from the moment it occurs. Conclusions In conclusion, this study successfully tests QI interventions as effective vehicles for implementing the evidence-based HAPU prevention protocol. This test stems from the theoretical foundation that QI interventions effectively alter preventive processes at the level of the clinical microsystem. Following the establishment of CMS nonpayment policy for HACs, hospitals are left to address issues that circumvent successful prevention efforts. Based on the conceptual framework for implementation science by Gonzalez et al. (2012), 34 effective 195

212 prevention begins by validating an evidence-based protocol to improve patient safety. Following the adoption of a QI strategy that structures the preventive process to support needs of hospital administration and staff, consistent implementation of the HAPU prevention protocol will ensue. Considering the conceptual framework of implementation science, this study emphasizes the development of QI strategies to support evidence-based practice in response to policy. There is less of a concern in this study about comparative effectiveness of the HAPU prevention protocol since previous literature extensively evaluates the effectiveness of each protocol component. If there was more reliable data about usage of specific prevention protocol components such as frequency of risk assessment with the Braden Scale or patient repositioning, it would be possible to control for changes in HAPU prevention directly correlated with the prevention protocol. This study design would be difficult to address without a more intensive patient chart review. Due to the emphasis of this study on evaluating the effectiveness of QI interventions, it utilizes a conceptual framework of implementation science while moving toward a great understanding of specifically QI scientific theory. To differentiate these two sciences, the framework makes clear that evidence-based practices such as the prevention protocol are what need to be implemented. Such practices cannot be introduced into existing health care delivery systems without reengineering such systems. QI theory devises strategies that lead to system redesign so that evidence-based practices can be implemented consistently for intended purposes. 196

213 The QI best-practice framework developed by Nelson et al. (2007) 28 offers an effective structure on which to model a QI strategy for HAPU prevention. QI strategies that are dynamic in terms of scope and scale, thereby including effective elements of leadership, staff, IT, and P&I domains, are most likely to lead to successful prevention. The survey of UHC hospitals indicated that the modified framework for this study was an accurate depiction of pertinent QI interventions to HAPU prevention. Following CMS policy intervention, data indicate a significant increase in adoption of QI interventions for HAPU intervention. In most cases, these QI interventions were adopted to support evidence-based prevention, reaffirming established theory. Empirical evidence points to specific QI interventions that are significant factors in HAPU prevention. Leadership Initiatives, Visual Tools, HAPU Staging, Skin Care, and Nutrition are recommended QI interventions on which to develop a dynamic QI strategy. These QI interventions yield clinically meaningful improvements in HAPU prevention both individually, and more so in combination. Many UHC hospitals that observed the most successful improvements in HAPU prevention utilized these QI interventions, reiterating the importance of each in a QI strategy. This research provides a model approach to the comparative effectiveness of QI interventions for HAPUs as well as many other HACs that place patients at risk and burden hospital staff. Moving forward, the field would benefit from identifying QI interventions that support evidence-based practices for all HACs. By focusing on patient-centered outcomes at the level of the clinical 197

214 microsystem rather than hospital-wide outcomes, further research could improve upon this model and discern greater clarity. The observational nature of this research provides real-world applicability to the findings on QI intervention effectiveness. Hospitals in the US which are not in a position to conduct research on HAPU prevention to the extent that academic medical centers are able can utilize the findings of this study as a foundation for initiating or improving a thorough strategy to effectively reduce HAPU incidence. 198

215 REFERENCES 1. The Learning Healthcare System: Workshop Summary. In: Olsen LA, Aisner D, McGinnish JM, ed. Institute of Medicine (US) Roundtable on Evidencebased Medicine. Washington, D.C.: National Academies Press; Serious Reportable Events in Healthcare. National Quality Forum, 2008 Oct. (Accessed February 28, 2012, at 3. Wald HL, Kramer AM. Nonpayment for Harms Resulting From Medical Care: Catheter-Associated Urinary Tract Infections. JAMA 2007 Dec 19;298: Ayello E, Lyder CH. A New Era of Pressure Ulcer Accountability in Acute Care. Adv Skin Wound Care 2008 Mar;21: Lyder C. Pressure ulcer prevention and management. JAMA 2003;289: Welton JM. Implications of Medicare reimbursement changes related to inpatient nursing care quality. J Nurs Adm 2008 Jul-Aug;38: Leatherman S, Berwick D, Iles D, Lewin LS, Davidoff F, Nolan T, Bisognano M. The business case for quality: case studies and an analysis. Health Aff (Millwood) 2003 Mar-Apr;22: Kurtzman E, Buerhaus PI. New Medicare Payment Rules: Danger or Opportunity for Nursing? Am J Nurs 2008;108: National Pressure Ulcer Advisory Panel. Pressure Ulcer Stages Revised by NPUAP. Washington, DC: NPUAP; 2007 Feb. 10. Galpin JE, Chow AW, Bayer AS, Guze LB. Sepsis associated with decubitus ulcers. Am J Med 1976; Wall BM, Mangold T, Huch KM, et al. Bacteremia in the chronic spinal cord injury population: risk factors for mortality. J Spinal Cord Med 2003;25: Shea JD. Pressure Sores: Classification and Management. Clin Orthop Rel Res 1975;112: Black JM. National Pressure Ulcer Advisory Panel: Moving toward consensus on deep tissue injury and pressure ulcer staging. Adv Skin Wound Care 2005;18: Russo CA, Elixhauser A. Hospitalizations related to pressure sores, Healthcare Cost and Utilization Project (HCUP) Statistical Brief #3. Rockville, MD: Agency for Healthcare Research and Quality; 2006 Apr. 199

216 15. Russo CA, Steiner C, Spector W. Hospitalizations Related to Pressure Ulcers among Adults 18 Years and Older, Healthcare Cost and Utilization Project (HCUP). Rockville, MD: Agency for Healthcare Research and Quality; 2008 Dec. 16. Whittington K, Briones R. National Prevalence and Incidence Study: 6- year sequential acute care data. Adv Skin Wound Care 2004;17: Cuddigan J, Ayello EA, Sussman C, Baranoski S. Pressure Ulcers in America: Prevalence, Incidence, and Implications for the Future. Reston, VA: National Pressure Ulcer Advisory Panel; Bergstrom N, Allman RM, Carlson CE. Clinical Practice Guideline Number 3: Pressure Ulcers in Adults: Prediction and Prevention. Rockville, MD: Public Health Service, US Dept of Health and Human Services; Gordon M, Gottschlich MM, Helvig EI. Review of evidence-based practice for the prevention of pressure sores in burn patients. J Burn Care Rehabil 2004 Sep-Oct;25: Kuhn BA, Coulter SJ. Balancing the pressure ulcer cost and quality equation. Nurs Econ 1992 Sep-Oct;10: Reddy M, Gill SS, Kalkar SR, Wu W, Anderson PJ, Ronchon PA. Treatment of pressure ulcers: a systematic review. JAMA 2008;300: Duncan K. Preventing Pressure Ulcers: The Goal is Zero. The Joint Commission Journal on Quality and Patient Safety 2007; Ratliff CR, Bryant DE. Guideline for prevention and management of pressure ulcers. In: WOCN clinical practice guideline; no 2. Glenview, IL: Wound, Ostomy, and Continence Nurses Society (WOCN); 2003: Bergstrom N, Braden B, Laguzza A, Holman A. The Braden Scale for predicting pressure sore risk. Nurs Res 1987;36: Gawande A. The Checklist Manifesto - How To Get Things Right. New York: Metropolitan Books; Haynes A, Weiser TG, Berry WR, et al. A Surgical Safety Checklist to Reduce Morbidity and Mortality in a Global Population. N Engl J Med 2009;360: Padula WV, Mishra MK, Makic MB, Sullivan PW. Improving the Quality of Pressure Ulcer Care with Prevention: a cost-effectiveness analysis. Med Care 2011 Apr;49:

217 28. Nelson EC, Batalden PB, Godfrey MM. Quality by Design. San Francisco, CA: Jossey-Bass; Magahalles MD. Risk factors for pressure ulcers in hospitalized elderly without significant cognitive impairment. Wounds 2007;19: Sterner E, Lindholm C, Berg E, Stark A, Fossum B. Category I pressure ulcers: how reliable is clinical assessment? Orthop Nurs 2011 May-Jun;30: Cho I, Park HA, Chung E. Exploring practice variation in preventive pressure-ulcer care using data from a clinical data repository. Int J Med Inform 2011 Jan;80: Baumgarten M, Margolis D, Orwig D, Hawkes W, Rich S, Langenber P, et al. Use of pressure-redistributing support surfaces among elderly hip fracture patients across the continuum of care: adherence to pressure ulcer prevention guidelines. Gerontologist 2010 Apr;50: Glouberman S, Zimmerman B. Complicated and Complex Systems: The Reform of Medicare in Canada: discussion paper number 8. The Commission on the Future of Health Care in Canada; 2002 July. 34. Gonzales R, Handley MA, Ackerman S, O'Sullivan PS. A Framework for Training Health Professionals in Implementation and Dissemination Science. Acad Med 2012 Mar;87: Padula WV, Mishra MK, Makic MB, Sullivan PW. In reply: Improving the Quality of Pressure Ulcer Care with Prevention. Med Care 2012 Feb;50: Berwick DM. Foreword: The Clinical Microsystem. In: Nelson EC, Batalden PB, Godfrey MM, ed. Quality by Design: a clinical microsystem approach. San Francisco, CA: Jossey-Bass; Pronovost PJ, Vohr E. Safe Patients, Smart Hospitals: How One Doctor's Checklist Can Help Us Change Health Care from the Inside Out. New York, NY: Penguin; Nelson EC, Godfrey MM, Batalden PB, Berry SA, et al. Clinical microsystems, part 1: The building blocks of health systems. Jt Comm J Qual Patient Saf 2008 Jul;34: Quinn JB. Intelligent enterprise: A knowledge and service based paradigm for industry. New York: Free Press;

218 40. Godfrey MM, Melin CN, Muething SE, Batalden PB, Nelson EC. Clinical microsystems, Part 3: Transformation of two hospitals using microsystem, mesosystem, and macrosystem strategies. Jt Comm J Qual Patient Saf 2008 Oct;34: Mohr JJ, Batalden PB. Improving safety on the front lines: the role of clinical microsystems. Qual Saf Health Care 2002;11: Reiter KL, Kilpatrick K B, Greene S B, Lohr K N, Leatherman S. How to develop business case for quality. Int J Qual Health Care 2007 Feb;19: Committee on Comparative Effectiveness Research Prioritization. Initial National Priorities for Comparative Effectiveness Research. Washington, DC: Institute of Medicine; Trisolini MG. Theoretical Perspectives on Pay for Performance. In: Cromwell J, Trisolini MG, Pope GC, Mitchell JB, Greenwald LM, ed. Pay for Performance in Health Care: Methods and Approaches. Research Triangle Park, NC: RTI Press; 2011 Mar. 45. Hader R. Best-practice protocols: get your board on board. Nurs Manage 2007 Jun;38: Berwick DM. What 'patient-centered' should mean: confessions of an extremist. Health Aff (Millwood) 2009 Jul-Aug;28:w Nelson WA, Gardent PB, Shulman E, Splaine ME. Preventing ethics conflicts and improving healthcare quality through system redesign. Qual Saf Health Care 2010;19: Berwick DM, Goldmann DA. Need for Multiple Designs. In: Hiatt H, James BC, Lynn J, Sahney VK, editor. 20th Annual National Forum on Quality Improvement in Health Care; 2008; Nashville, TN: Institute for Healthcare Improvement; Wennberg JE, Fisher ES, Skinner JS, Broner KK. Extending the P4P Agenda, Part 2: How Medicare can Reduce Waste and Improve the Care of the Chronically Ill. Health Aff (Millwood) 2007;26: Deficit Reduction Act of Pub L National Pressure Ulcer Advisory Panel. Not All Pressure Ulcers are Avoidable. Washington, DC: NPUAP; Braden B, Bergstrom N. A conceptual schema for the study of etiology of pressure sores. Rehabilitation Nursing 1987;12:

219 53. Comfort E. Reducing Pressure Ulcer Incidence through Braden Scale Risk Assessment and Support Surface use. Adv Skin Wound Care 2008;21: Allman RM. Pressure ulcers in the elderly. N Engl J Med 1989;320: Bergstrom N, Braden B. A prospective study of pressure sore risk among institutionalized elderly. J Am Geriatr Soc 1992 Aug;40: Inman KJ, Sibbald WJ, Rutledge FS, Clark BJ. Clinical utility and costeffectiveness of an air suspension bed in the prevention of pressure ulcers. JAMA 1993;269: Farguhar M. AHRQ Quality Indicators. In: Hughes RG, ed. Patient Safety and Quality: An Evidence-based Handbook for Nurses. Rockville, MD: Agency for Healthcare Research and Quality; 2008 Apr. 58. Krapfl LA, Mackey D. Medicare changes to the hospital inpatient prospective payment systems: commentary on the implications for the hospitalbased wound care clinician. J Wound Ostomy Continence Nurs 2008 Jan- Feb;35: Ayello EA, et al. By the numbers; Braden score interventions. Adv Skin Wound Care 2004;17: Braden B, Bergstrom N. Predictive validity of Braden Scale for pressure sore risk in a nursing home population. Res Nurs Health 1994;17: Lyder CH, Yu C, Stevenson D, Mangat R, Empleo-Frazier O, Emerling J, McKay J. Braden Scale for the prediction of pressure ulcer risk in blacks and Latino/Hispanic elders: a pilot study. Ostomy Wound Manage 1998 Mar;44:42S- 9S; discussion 50S. 62. Norton D. Calculating Risk: reflections on the Norton Scale. Adv Wound Care 1996 Nov-Dec;9: Noonan C, Quigley S, Curley MA. Using the Braden Q Scale to Predict Pressure Ulcer Risk in pediatric patients. J Pediatr Nurs 2011 Dec;26: Keast D, Parslow N, Houghton PE, Norton L, Fraser C. Best Practice Recommendations for the Prevention and Treatment of Pressure Ulcers: Update Adv Skin Wound Care 2008;20: Reddy M, Gill SS, Rochon PA. Preventing Pressure Ulcers: a systematic review. JAMA 2006 Aug 23;296: Brem H, Lyder C. Protocol for the successful treatment of pressure ulcers. Am J Surg 2004 Jul;188 (1A Suppl):9S-17S. 203

220 67. Bates-Jensen BM. The Pressure Sore Status Tool a few thousand assessments later. Adv Wound Care 1997;10: Stotts NA, Rodeheaver GT, Thomas DR, et al. An instrument to measure healing in pressure ulcers: development and validation of the Pressure Ulcer Scale for Healing (PUSH). J Gerontol 2001;56:M795-M Cannon B, Cannon, JP. Management of pressure ulcers. Am J Health- Syst Pharm 2004;61: Flack S, Apelqvist J, Keith M, Trueman P, Williams D. An economic evaluation of VAC therapy compared with wound dressings in the treatment of diabetic foot ulcers. Journal of Wound Care 2008;17: Singer AJ, Clark RAF. Cutaneous Wound Healing. N Engl J Med 1999 Sep;341: Breslow RA, Hallfrisch J, Guy DG, et al. The importance of dietary protein in healing pressure ulcers. J Am Geriatr Soc 1993; Bergstrom N, Braden B. Predicting Pressure Sores in Those at Risk. Am J Nursing 1989 Jan;89: Makic MB. Evidence-based practice for the prevention of pressure ulcers. Denver, CO: University of Colorado; Lindenauer PK, Lagu T, Shieh MS, Pekow PS, Rothberg MB. Association of Diagnostic Coding with Trends in Hospitalizations and Mortality of Patients with Pneumonia, JAMA 2012;307: Rothwell PM. External validity of randomised controlled trials: "To whom to the results of this trial apply?". Lancet 2005;365: Marko NF, Weil RJ. The role of observational investigations in comparative effectiveness research. Value Health 2010 Dec;13: Desneves K, Todorovic BE. Pressure Ulcers in America: prevalence, incidence and implications for the future. An executive summary of the National Pressure Ulcer Advisory Panel. Adv Skin Wound Care 2001;14: Padula WV, Mishra MK, Weaver CD, Yilmaz T, Splaine ME. Building Information for Systematic Improvement of the Prevention of Hospital-acquired Pressure Ulcers with Statistical Process Control Charts and Regression. BMJ Quality & Safety 2012 Jun;21: LeMaster K. Reducing Incidence and Prevalence of Hospital-Acquired Pressure Ulcers at Genesis Medical. The Joint Commission Journal on Quality and Patient Safety 2007;

221 81. Morton A, Mengersen K, Waterhouse M, Steiner S. Analysis of aggregated hospital infection data for accountability. J Hosp Infection 2010;76: Kottner J, Halfens R. Using Statistical Process Control for Monitoring the Prevalence of Hospital-acquired Pressure Ulcers. Ostomy Wound Manage 2010 May;56: Carey R. Improving Healthcare with Control Charts. Milwaukee, WI: ASQ Quality Press; Beckrich K, Aronovitch SA. Hospital-Acquired Pressure Ulcers: A Comparison of Costs in Medical vs. Surgical Patients. Nurs Econ 1999 Sep- Oct;17: Xakellis G, Frantz RA. The cost of healing pressure ulcers across multiple care settings. Adv Skin Wound Care 1996 Nov-Dec;9: Pappas S. The Cost of Nurse-Sensitive Adverse Events. JONA 2008;38: Mackey D. Support Surfaces: Beds, Mattresses, Overlays--Oh My! Nurs Clin N Am 2005;40: Iglesias C, Nixon J, Cranny G, et al. Pressure relieving support surfaces (PRESSURE) trial: cost effectiveness analysis. BMJ 2006;332: Xakellis G, Frantz RA, Lewis A, Harvey P. Cost-effectiveness of an intensive pressure ulcer prevention protocol in long-term care. Adv Wound Care 1998;11: Pham B, Stern A, Chen W, Sander B, et al. Preventing pressure ulcers in long-term care: a cost-effectiveness analysis. Arch Intern Med 2011 Nov;171: Griffin FA. Reducing Surgical Complications. Jt Comm J Qual Patient Saf 2007 Nov;33: Leape LL, Brennan TA, Laird N, Lawthers AG, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. N Engl J Med 1991 Feb;324: Gawande AA, Thomas EJ, Zinner MJ, Brennan TA. The incidence and nature of surgical adverse events in Colorado and Utah in Surgery 1999 Jul;126: Bosk CL. Forgive and Remember: Managing Medical Failure. Chicago: University of Chicago Press;

222 95. Zimmerman B, Lindberg C, Plsek P. A complexity science primer. In: Zimmerman B, et al, ed. Edgeware, insights from complexity science for health care leaders. Irving, TX: VHA Inc; 2001: Weinstein MC, Skinner JA. Comparative effectiveness and health care spending - implications for reform. N Engl J Med 2010;362: Llewellyn-Thomas HA. Investigating patients' preferences for different treatment options. Can J Nurs Res 1997;23: Kohn LT, Corrigan JM, Donaldson MS, ed. To Err is Human: Building a Safer Health System. Washington, D.C.: National Academy Press; Leape LL, Lawthers AG, Brennan TA, et al. Preventing Medical Injury. Qual Rev Bull 1993;19: Berwick DM. A user's manual for the IOM's 'Quality Chasm' report. Health Aff (Millwood) 2002 May-Jun;21: McCannon CJ, Hackbarth AD, Griffin FA. Miles to Go: An Introduction to the 5 Million Lives Campaign. Jt Comm J Qual Patient Saf 2007 Aug;33: Berwick DM, Calkins DR, McCannon CJ, Hackbarth AD. The 100,000 lives campaign: setting a goal and a deadline for improving care quality. JAMA 2006 Jan;295: Lewis C. 5 Million Lives Campaign. J Vasc Nurs 2007;25: Leasure AR, Stirlen J, Thompson C. Barriers and Facilitators to the Use of Evidence-Based Best Practices. Dimens Crit Care Nurs 2008;27: Makic MB, VonReuden KT, Rauen CA, Chadwick J. Evidence-based practice habits: putting more sacred cows out to pasture. Crit Care Nurs 2011 Apr;31: IHI Shares results of 5 Million Lives Campaign. Hospital Peer Review 2008 Dec;33: Berwick DM, Hackbarth AD, McCannon CJ. IHI Replies to "The 100,000 Lives Campaign: A Scientific and Policy Review. Jt Comm J Qual Patient Saf 2006 Nov;32: Golladay KK, Collins AB, Ashcraft A, Dorrill RA, et al. Adverse Events in Hospitals: Methods for Identifying Events. In: Department of Health & Human Services, ed. Office of Inspector General. Washington, DC;

223 109. Williams SC, Watt A, Schmaltz SP, Koss RD, Loeb JM. Assessing the reliability of standardized performance indicators. Int J Qual Health Care 2006 Jun;18: Centers for Medicare and Medicaid Services (CMS) HHS. Medicare Program: Changes to the hospital inpatient prospective payment systems and fiscal year 2009 rules. Fed Regis 2008;73: Gray M. Context for WOC practice: must-read information for WOC nurses: pay for performance in the acute care setting. J Wound Ostomy Continence Nurs 2008 Jan-Feb;35: Stone PW, Glied SA, McNair PD, Matthes N, et al. CMS Changes in Reimbursement for HAIs. Med Care 2010 May;48: Division of Quality Evaluation and Health Outcomes. Quality Measures Compendium. In: CMS, ed. Washington, DC; 2007 Dec Schaffer A. Fighting Bedsores with a Team Approach. The New York Times Trisolini MG. Introduction to Pay for Performance. In: Cromwell J, Trisolini MG, Pope GC, Mitchell JB, Greenwald LM, ed. Pay for Performance in Health Care: Methods and Approaches. Research Triangle Park, NC: RTI Press; 2011 Mar Golden B, Sloan F. Physician pay for performance: Alternative perspectives. In: Sloan F, Kasper H, ed. Incentives and choice in health care. Cambridge, MA: MIT Press; Shortell SM, Kaluzny A. Physician group cultural dimensions and quality performance indicators: Not all is equal. Clifton Park, NY: Thomson Delmar Learning; Kimberly J, Minvielle E. Quality as an organizational problem. In: Mick S, Wyttenbach M, ed. Advances in health care organization theory. San Francisco: Jossey-Bass; Town R, Wholey DR, Kralewski J, Dowd B. Assessing the influence of incentives on physicians and medical groups. Med Care Res Rev 2004;61:80S- 118S Landon BE, Wilson IB, Cleary PD. A conceptual model of the effects of health care organizations on the quality of medical care. JAMA 1998;279: Smalarz A. Physician group cultural dimensions of quality performance indicators: Not all is equal. Health Care Manag Rev 2006;31:

224 122. Fisher ES. The paradox of plenty: implications for performance measurement and pay for performance. Managed Care 2006;15: Nelson EC, Batalden PB, Godfrey MM, Lazar JS. Value by Design. San Francisco: Jossey-Bass; Deming WE. Out of Crisis. Cambridge, MA: MIT Center for Advanced Engineering Study; Nelson EC, Batalden PB. Patient-based quality measurement systems. Quality Management in Health Care 1993;2: Lyder CH, Grady J, Mathur D, Petrillo MK, Meehan TP. Preventing pressure ulcers in Connecticut hospitals by using the plan-do-study-act model of quality improvement. Jt Comm J Qual Safe 2004;30: Wheeler JR, White B, Rauscher S, Nahra TA, Reiter KL, Curtin KM, Damberg CL. Pay-for-performance as a method to establish the business case for quality. J Health Care Finance : Nelson EC, Splaine ME, Plume SK, Batalden PB. Good measurement for good improvement work. Qual Manag Health Care 2004 Jan-Mar;13: Lindenauer PK. Effects of quality improvement collaboratives are difficult to measure using traditional biomedical research methods. BMJ 2008;336: Damschroder LJ, Aron DC, Keith RE, Kirsh SR, Alexander JA, Lowery JC. Fostering implementation of health services research findings into practice: a consolidation framework for advancing implementation science. Implement Sci 2009 Aug 7;4: Fink R, Gilmartin H, Richard A, Capezuti E, Boltz M, Wald H. Indwelling urinary catheter management and catheter-associated urinary tract infection prevention practices in Nurses Improving Care for Healthsystem Elders hospitals. Am J Infect Control 2012 Oct;40: Wald HL, Richard A, Dickson VV, Capezuti E. Chief nursing officers' perspectives on Medicare's hospital-acquired conditions non-payment policy: implications for policy design and implementation. Implement Sci 2012;7: Batalden P, Bate P, Webb D, McLoughlin V. Planning and leading a multidisciplinary colloquium to explore the epistemology of improvement. BMJ Qual Saf 2011 Apr;20 Suppl 1:i Mohr JJ. Forming, Operating, and Improving Micro-systems of Health Care. Hanover, NH: Dartmouth College; 2000 May. 208

225 135. Castellanos-Ortega A, Suberviola B, Garcia-Astudillo LA, Holanda MS, Ortiz F, Llorca J, Delgado-Rodriguez M. Impact of the Surviving Sepsis Campaign protocols on hospital length of stay and mortality in septic shock patients: Results of a three-year follow-up quasi-experimental study. Crit Care Med 2010 Apr;38: Spetz J, Keane D. Information technology implementation in a rural hospital: a cautionary tale. J Healthc Manag 2009 Sep-Oct;54: van Loon E, Zuiderent-Jerak T. Framing Reflexivity in Quality Improvement Devices in the Care for Older People. Health Care Anal 2012 Jun;20: Chandra A, Jena AB, Skinner JS. The Pragmatist's Guide to Comparative Effectiveness Research. J Econ Perspectives 2011 Spring;25: Iglehart JK. Prioritizing Comparative-Effectiveness Research - IOM Recommendations. N Engl J Med 2009 Jul;361: Chou R, Dana T, Bougatsos C, Blazina I, Starmer AJ, Reitel K, Buckley DI. Pressure Ulcer Risk Assessment and Prevention: A Systematic Comparative Effectiveness Review. Ann Intern Med 2013 Jul;159: Mohlenbrock WC, Kish TM. Stretching dollars without compromising care: two industry veterans lay out a pathway to physician-directed best-practice improvements through comparative effectiveness research. Health Management Technology 2011 Jan;32: Boesch RP, Myers C, Garrett T, Nie AM, et al. Prevention of Tracheostomy-related Pressure Ulcers in Children. Pediatrics 2012 Feb;129:e Ferrer R, Artigas A, Levy MM, Gonzalez-Diaz G, et al. Improvement in process of care and outcome after a mutlicenter severe sepsis education program in Spain. JAMA 2008 May;299: Campbell DT, Stanley JC. Experimental and Quasi-Experimental Designs for Research. Boston: Houghton Mifflin Company; Wooldridge J. Introductory Econometrics: A Modern Approach. 4th ed. Australia: South-Western Cengage Learning; Simon SD. Is the randomized clinical trial the gold standard of research? J Androl 2001 Nov-Dec;22: Levy H, Meltzer D. The impact of health insurance on health. Annu Rev Public Health 2008;29:

226 148. Manning WG, Newhouse JP, Duan N, Keeler EB, Benjamin B, Leibowitz A, et al. Health insurance and the demand for medical care: Evidence from a randomized experiment. Santa Monica, CA: RAND Corporation; Dreyer NA. Making observational studies count: shaping the future of comparative effectiveness research. Epidemiology 2011 May;22: Fisher E. The paradox of plenty: implications for performance. Managed Care 2006 Oct;15: Catalano R, Serxner S. Time series designs of potential interest to epidemiologists. Am J Epidemiol 1987;126: Aday LA, Cornelius LJ. Designing and Conducting Health Survey. 3rd ed. San Francisco, CA: Jossey-Bass; Dillman DA. Why Choice of Survey Mode Makes a Difference. Public Health Reports 2006 Jan-Feb;121: Tabachnick BG, Fidell LS. Discriminant Analysis. In: Using Multivariate Statistics. 5th ed. Boston: Pearson; 2007: Tabachnick BG, Fidell LS. Analysis of Covariance. In: Using Multivariate Statistics. 5th ed. Boston: Pearson; 2007: Pieper B, Sugrue M, Weiland M, Sprague K, Heimann C. Occurence of skin lesions/conditions in ill persons. Dermatol Nurs 1997 Aug;9: Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. Philadelphia: Lippincott, Williams and Wilkins; Berlowitz DR, Ash AS, Brandeis GH, BrandHK, et al. Rating long-term care facilities on pressure ulcer development: importance of case-mix adjustment. Ann Intern Med 1996 Mar;124: Shadish WR, Cook TD, Campbell DT. Experimental and Quasi- Experimental Designs for Generalized Causal Inference. Belmont, CA: Wadsworth; Tesher AN, Branda ME, O'Byrne TJ, Naessens JM. All At-Risk Patients Are Not Created Equal: Analysis of Braden Pressure Ulcer Risk Scores to Identify Specific Risks. J Wound Ostomy Continence Nurs 2012;39:

227 APPENDIX A KEY TERMS AND DEFINITIONS The following list contains the conceptual and operational definitions of the key terms related to the research question: 1. Adoption Conceptual: voluntary selection of an intervention or process to assist or enhance in clinical care Operational: open acceptance of a QI intervention or strategy to enhance HAPU prevention, while understanding potential cons 2. Best-practices Framework Conceptual: a classification system for QI interventions separated by impact on clinical practice and culture Operational: four-category QI intervention system developed by 3. Domain Nelson et al. (2007) into Leading Organizations, Staff, Information Technology, and Performance & Improvement Conceptual: a component of a framework. Operational: one of the four-categories of the best-practices 4. Dynamic framework Conceptual: a practice or protocol that is inclusive of recommended guidelines and diverse mechanisms of prevention 211

228 Operational: a QI strategy that is comprehensive in terms of both 5. Element scope and scale Conceptual: the building block of a domain Operational: the interventions of a best-practices domain in the 6. Implementation best-practice framework Conceptual: forced use of a policy or device for clinical care Operational: a device in HAPU prevention that is forced upon clinicians without consent, such as CMS reimbursement policy or HAPU prevention protocol 7. Information & Information Technology Conceptual: a QI intervention that improves clinician accessibility to patient records, practice guidelines and notifications through computer technology Operational: data tracking of pressure ulcer outcomes in an electronic health record (EHR); a computerized Braden Scale; a computerized alarm reminder for a clinician to initiate a prevention protocol 8. Intervention Conceptual: a device or concept adopted to change the process of clinical practice and improve care 212

229 Operational: a device or concept that changes a clinician s approach to caring for patients, with respect to preventing HAPUs 9. Leading Organizations Conceptual: a QI intervention that improves interdisciplinary clinical organization by supporting assigned hierarchies, and providing inexperienced clinicians with access to knowledge of those who are more experienced Operational: a leadership retreat to promote HAPU prevention; an interdisciplinary approach to HAPU prevention of assigned responsibility to physicians and nurses; a promotional system for public awareness of HAPU prevention; regular discussions between unit managers and their clinicians to review HAPU prevention techniques 10. Length of Stay Conceptual: The number of days a patient spends admitted as a hospital inpatient Operational: Total days spent as a hospital inpatient for primary diagnosis, as well as with a HAPU 11. Performance and Improvement Conceptual: A QI intervention that improves HAPU prevention through improved contact with the patient and consistent initiation of the prevention protocol 213

230 Operational: Increased use of a risk assessment tool (e.g. Braden Scale); a checklist nurses can use to track initiation of components of a prevention protocol or supported QI interventions; adoption of a new bed, support surface, and/or underpad; development of an improved nutrition or repositioning regimen for patients 12. Pressure Ulcer Conceptual: a staged skin condition cause by poor blood circulation and bacterial infection that leads to tissue deterioration Operational: a stage III or IV pressure ulcer acquired by a patient during hospital stay which leads to additional length of stay 13. Prevention Protocol Conceptual: an evidence-based set of clinical guidelines applied to high-risk inpatients for effective prevention of pressure ulcers Operational: AHRQ standard 5-step procedure that must be initiated consistently to prevent HAPUs, including daily risk assessment, patient repositioning every 3-6 hours, management of moisture and incontinence, a nutrition regimen, and improved support surfaces 14. Quality Improvement Conceptual: study of intervention and/or system redesign for better patient outcomes, better systematic performance, and better professional services 214

231 Operational: A strategic system organized by interventions to 15. Repositioning improve consistency of preventive care for HAPUs Conceptual: Moving a patient from one side to another for the purposes of improving circulation while they are bed-ridden Operational: Moving the patient while in bed every 3-6 hours to improvement circulation and prevent the development of a pressure ulcer 16. Risk-assessment Conceptual: an instrument used frequently to assess patient risk for developing a HAPU upon admission and at further increments for extended length of stay Operational: application of the Braden, Braden Q, or Norton scales upon admission and once every 1-2 days thereafter to assess risk for developing a HAPU as part of a standard prevention protocol 17. Scale Conceptual: span of QI intervention adoption within the segment of a framework for a more comprehensive strategy Operational: increased application of QI interventions within each of the four categories of the best-practice framework 18. Scope Conceptual: span of QI intervention adoption across an entire framework for a more comprehensive strategy 215

232 Operational: a comprehensive QI strategy consisting of multiple domains of the best-practice framework, including components of Leading Organizations, Staff, Information & Information Technology, and Performance & Improvement domains 19. Staff Conceptual: a QI intervention that improves communication and camaraderie between clinicians to enhance teamwork Operational: Increased teamwork in context of pressure ulcer prevention by initiating daily team huddles, regular all-staff meetings, continuing education, and training for inexperienced staff to improve knowledge of HAPU prevention protocols and QI strategy 20. Strategy Conceptual: a set of QI interventions practiced in combination consistently to prevent HAPUs. Operational: a unique set of QI interventions to a hospital that in 21. Support Surface combination improve HAPU prevention. Conceptual: a bed or chair that a patient can lay on for extended periods Operational: a modern device that relieves stress on a patient s pressure points and reduces the likelihood of a pressure ulcer developing. 216

233 22. Underpad Conceptual: a biomedical device that soaks up excess fluid Operational: a device similar to a diaper that lays flat on a bed underneath a patient to manage levels of moisture and incontinence that may be detrimental to the development of a pressure ulcer 217

234 APPENDIX B SURVEY INSTRUMENT Quality Improvement Interventions for Pressure Ulcer Prevention in the Hospital A pressure ulcer (AKA decubitus ulcer or bed sore) is a complication of being bedbound that is mostly preventable. As of October 2008, Medicare policy no longer reimburses for the cost of hospital-acquired pressure ulcers, so many care providers are focusing on the evidence-based guidelines to prevent them. We would like to know how your hospital prepares care providers in the prevention of pressure ulcers. Please consider answers to the following questions. This survey may require participation from your colleagues in case you are unfamiliar with some components of pressure ulcer prevention. Your answers will remain anonymous. By participating in this survey you will receive feedback about your hospital s performance at pressure ulcer prevention relative to other academic medical centers. Questions 1-3: where do you work, and what is your organizational role? 1. Your name and credentials (First, Last; Credentials, e.g. MD, PhD, RN): 2. Your job title: 3. The hospital you work for (Hospital, City, State): Questions 4-6: we would like to learn your hospital s general approach to preventing pressure ulcers. 4. Does your hospital have a prevention protocol for individuals identified as being at risk for pressure ulcers? a. Yes b. No If yes, who initiates this protocol? i. Certified wound, ostomy, and continence nurses (CWOCN) ii. Staff Nurse iii. Physician iv. Physical Therapist (PT); Certified Wound Specialist PT v. Other: Would you be willing to share a copy of your protocol with us? i. Yes. If so, please upload a copy with the electronic survey. ii. No 218

235 5. Does your hospital have a certified wound, ostomy, and continence nurse (CWOCN) or certified wound care nurse (CWCN) on staff? a. Yes b. No If yes, do you know how many full-time equivalent CWOCNs/CWCNs there are? Answer: Number of Full-time Equivalent Employees 6. In general, how would you describe the knowledge of your nursing staff to prevent pressure ulcers? Select one of the following options related to their level of skill. a. Poor b. Fair c. Moderate d. Good e. Excellent 219

236 220

237 221