The Relationship between Nurse Staffing and Patient Satisfaction in Emergency Departments

Transcription

1 The Relationship between Nurse Staffing and Patient Satisfaction in Emergency Departments by Imtiaz Daniel A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Institute of Health Policy, Management and Evaluation University of Toronto Copyright by Imtiaz Daniel, 2012

2 ii The Relationship between Nurse Staffing and Patient Satisfaction in Emergency Departments Abstract Imtiaz Daniel Doctor of Philosophy Institute of Health Policy, Management and Evaluation University of Toronto 2012 Patient satisfaction is a key outcome measure being examined by researchers exploring the relationships between patient outcomes and hospital structure and care processes. Only a few non-generalizable studies, however, have explored the relationship of nurse staffing and patient satisfaction with nursing care in emergency departments of hospitals. This dissertation aims to address that gap. Using more than 182,000 patient satisfaction surveys collected over a five-year period from 153 emergency departments (EDs) in 107 hospitals throughout Ontario, this study explores the relationship between nurse staffing and patient perceptions of nursing care in a range of Canadian ED settings, including urban and rural, community and academic, and small and large healthcare institutions with varying sizes and case mix. Using an established conceptual framework for investigating the relationship between nurse staffing and patient outcomes, nineteen nurse staffing variables were initially investigated. Ultimately, however, only five staffing variables were used in the multi-level regression analyzes. These five variables included registered nurse (RN) proportion, RN agency proportion, percent full-time nurse worked hours, RN worked ii

3 iii hours per patient length of stay and registered practical nurse (RPN) worked hours per length of stay. Emergency department case mix index, patient age and gender, hospital peer group, size, wait times, cleanliness of the emergency department, physician courtesy, and year of measurement were controlled to account for their effect on the relationship between nursing staffing and patient satisfaction in the ED. The study revealed a subset of six patient satisfaction variables representing the overall variation in patient satisfaction with nursing care in the ED. Although RN proportion and RPN worked hours per length of stay were found to have a statistical association with patient satisfaction in the ED, the association was weak and not administratively actionable. Interpersonal and environmental factors such as physician and nurse courtesy, ED cleanliness and timeliness, however, were areas which hospital administrators should consider since they were highly associated with patient satisfaction in EDs. iii

4 iv Acknowledgments I would like to express my sincere gratitude to my supervisor, Jan Barnsley, and my committee members, Linda McGillis Hall, George Pink, and Antoni Basinski for providing support and guidance throughout the research process. Special thanks to Jan for keeping me focused and moving forward. I also wish to thank the Ontario Ministry of Health and Long-Term Care, the Canadian Institute of Health Information, and the Ontario Hospital Association for providing access to data required for this research. Special thanks to Carey Levinton and Kevin Yu for their advice on data linkage and analysis. I am grateful to Carol Brewer from the University of Buffalo and Sean Clarke from the Faculty of Nursing for serving as my examiners and for their valuable suggestions for subsequent research. I am very grateful to my late father, Sonny, my beloved mother, Shaira, and my entire family for their support, encouragement and assistance. My wife, Yen, provided inspiration, encouragement and motivation. Without her, this thesis would not have been possible. To my daughter, Kaitlyn: I adore you and appreciated having you in my office, colouring my text books, while I was working on my thesis. iv

5 v Table of Contents ACKNOWLEDGMENTS... IV LIST OF TABLES... VI LIST OF FIGURES...VII LIST OF APPENDICES... VIII CHAPTER 1 INTRODUCTION STATEMENT OF THE PROBLEM AIM OF STUDY SIGNIFICANCE OF STUDY CONCEPTUAL FRAMEWORK HYPOTHESES...8 CHAPTER 2 LITERATURE REVIEW OVERVIEW BACKGROUND - EMERGENCY DEPARTMENTS STAFFING MODELS NURSE STAFFING METHODOLOGY FACTORS INFLUENCING NURSE STAFFING NURSE STAFFING MEASURES PATIENT SATISFACTION WITH NURSING CARE INSTRUMENTS FOR MEASURING PATIENT SATISFACTION WITH NURSING CARE FACTORS ASSOCIATED WITH PATIENT SATISFACTION WITH NURSING NURSE STAFFING THEORETICAL FRAMEWORKS SUMMARY...51 CHAPTER 3 METHODS AND PROCEDURES OVERVIEW STUDY DESIGN SAMPLE POWER ANALYSIS DATA COLLECTION DATA ACCESS DATA ANALYSIS...67 CHAPTER 4 RESULTS OVERVIEW PATIENT SATISFACTION EMERGENCY DEPARTMENT CHARACTERISTICS RESEARCH QUESTIONS ANALYSIS SUMMARY CHAPTER 5 DISCUSSION AND CONCLUSION OVERVIEW STUDY VARIABLES FINDINGS IN RELATION TO THE CONCEPTUAL FRAMEWORK STUDY IMPLICATIONS LIMITATIONS OF THE STUDY FUTURE RESEARCH CONCLUSION REFERENCES v

6 vi List of Tables Table 2-1. Factors Influencing Nurse Staffing Policies Table 2-2. Nurse Staffing Measures Table 2-3. Nurse Staffing Variables from Consensus Panel Table 2-4. Summary of the Impact of Nurse Staffing on Patient Length of Stay Table 2-5. Characteristics of Good Nursing Care (Larrabee Table 2-6. Nine indicators of the ED Patients Perception of Care - NRC-Picker Survey Table 3-1. Definition of Terms Table 3-2. Emergency Department by Hospital Type Table 4-1. Patients Surveyed by Gender Table 4-2. Patients Surveyed by Age Group Table 4-3. Patient Satisfaction Variables over the study period Table 4-4. Patient Satisfaction by Gender Table 4-5. Patient Satisfaction by Peer Group Table 4-6. Patient Satisfaction by Age Groups Table 4-7. Correlation Table Patient Satisfaction Table 4-8. PCA Factor Loadings Table 4-9. Variance Explained by Each Variable Table Nursing Staffing Categories Table Emergency Department Characteristics by Hospital Type Table Control Variables by Year Table Control Variables by Peer Group Table Correlations between Control Variables and Patient Satisfaction Table List of Variables Assessed In Regression Analyses Table Variables Used in Linear Mixed Models Table Linear Mixed Model: Patient Satisfaction with Nursing Care (Aggregate Score) Table Linear Mixed Model: Overall Patient Satisfaction with Care Received in the ED EDSAT Table Linear Mixed Model: Recommending the ED EDREC Table Linear Mixed Models Results Table Linear Mixed Models Results with Standardized Coefficients vi

7 vii List of Figures Figure 1. Conceptual Framework of Nurse Staffing and Patient Outcomes (Kane et al., 2007)... 6 Figure 2. Conceptual Framework of Nurse Staffing and Patient Satisfaction... 7 Figure 3. Conceptual Framework of Nurse Staffing and Patient Satisfaction Figure 4. Predicted Satisfaction Scores for a typical ED vii

8 viii List of Appendices Appendix A. Literature Review Search Appendix B. Outcomes Model for Healthcare Research Appendix C. Quality of Care Dynamic Model Appendix D. Theoretical Model of the Relationships between Context, Structure (professional practice), and Effectiveness (outcomes) Appendix E. NRC+Picker Sampling Plan Appendix F. OHRS Staffing Accounts Appendix G. Technical Specifications Appendix H. NACRS Database Appendix I. Patient Satisfaction Descriptive Statistics Appendix J. Patient Satisfaction Principal Component Analysis Appendix K. Patient Satisfaction Correlation Table Appendix L. Nursing Staffing Categories by Hospital Type Appendix M. Staffing Variables Correlation Table Appendix N. Staffing Variables Appendix O. Nurse Staffing and Patient Satisfaction with Nursing Care Appendix P. Predicted Patient Satisfaction for a Typical ED viii

9 1 Chapter 1 Introduction 1.1 Statement of the problem Understanding the relationship between nurse staffing and patient satisfaction is important for policy makers and administrators who want to manage effectively the scarcity of nursing staff in a fiscally constrained environment. Due to the rising cost of healthcare and nursing shortages, nurse staffing in emergency departments (EDs) has become a high priority issue for policy makers and healthcare administrators. Predicting the staffing needs of an ED, however, is difficult for several reasons. First, the volume of patients can vary significantly during the day and from day to day, and ED administrators must predict how much standby time is required to meet the service demands of the department. Second, EDs must provide initial treatment for a broad spectrum of illnesses and injuries, some of which require immediate attention because they may be life-threatening. Finally, if the hospital has no available inpatient beds, admitted patients could stay in the ED for an extended time (Schull et al., 2002). Ensuring that patients not only receive adequate care, but that they also leave the ED feeling satisfied with their experience, is therefore challenging in an era of cost control and staffing issues. The relationship between nurse staffing and patient satisfaction is also a concern for researchers. One of the first studies to examine this relationship was done by Abdellah and Levine (1958) and involved 60 large general hospitals. Gathering data from 20,000 patients and staff in inpatient wards, Abdellah and Levine found that patient

10 2 satisfaction was higher when professional registered nurse (RN) hours were higher, but when the total nursing hours (which include RN, licensed practical nurses [LPNs], and nursing assistants) were higher, patient satisfaction was lower (Abdellah & Levine, 1958). Many studies have since investigated the relationship of nurse staff mix models of regulated and unregulated staff and nurse-to-patient ratios to patient outcomes (Blegen et al., 1998; Aiken et al., 2002; Needleman et al., 2002; Halm et al., 2005; Clarke, 2007; Kane et al., 2007). In particular, recent studies on nurse staffing have focused on topics such as the adequacy of hospital nursing staff (Aiken et al., 1996; Unruh & Fottler, 2006), the effects of restructuring (Brewer & Frazier, 1998; Mark et al., 2000), and nurse staffing in relation to (or as a predictor of) patient outcomes (Aiken et al., 1994; Aiken et al., 2002; Needleman et al., 2002; Mark et al., 2004). Most studies investigating the relationship between nurse staffing and patient satisfaction with nursing care, however, were performed in inpatient wards of hospitals and only a few non-generalizable studies can be found in the ED setting. As a result, a study that investigates the relationship between nurse staffing and patient satisfaction in many EDs over a period of time would allow administrators and policy makers to not only understand how to improve patient satisfaction, but to do so in an effective manner. 1.2 Aim of study This study will investigate patient satisfaction with nursing care in EDs and its relationship to nurse staffing. Since there are few non-generalizable studies exploring the relationship between nurse staffing and patient satisfaction in the ED, this research will begin to fill this void by including data from a large number of EDs with different

11 3 case mix and size. In this study, EDs from 107 hospital corporations are examined over a five-year period. The patient-level sample consists of 182,022 patients who were discharged from Ontario s EDs during the period of 2005/06 to 2009/10 and who also completed a patient satisfaction survey that contained the following questions that address patient satisfaction with nursing care: 1. When you had important questions to ask a nurse, did you get answers you could understand? (Answer) 2. If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? (Explain) 3. Did you have confidence and trust in the nurses treating you? (Trust) 4. Did nurses talk in front of you as if you weren t there? (Respect) 5. How would you rate the courtesy of your nurses? (Courtesy) 6. How would you rate the availability of your nurses? (Availability) 7. How would you rate how well the doctors and nurses worked together? (Dr- Nurse working relationship) By examining over 100 EDs in Ontario that served over 182,000 patients, this study seeks to determine to what extent specific aspects of nurse staffing relate to: 1) patient satisfaction with nursing care; 2) overall satisfaction with care received in the ED; and 3) whether the patient would recommend this ED to friends and family. The unit of analysis for this question is the hospital level. This chapter will discuss the hypotheses, significance, and conceptual framework used in the study. 1.3 Significance of study Patient satisfaction has become a well-established outcome indicator of health care used by accreditation agencies, such as the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) and the National Committee on Quality Assurance

12 4 (NCQA) (Fahad Al-Mailam, 2005). It has been described as the acid test through which the healthcare delivery system must pass when evaluating the effectiveness of nursing (Bond & Thomas, 1992). Due to the increasing focus on patient-centred care, which includes taking the patients views into account, patient satisfaction has become more important, making it a key indicator of the quality of nursing care (Johansson et al., 2002). In the last decade, there has been a shift in research away from productivity studies to exploring the relationship between the quality of patient care, nurse staffing levels, and staff mix (McGillis Hall, 2005). The resulting change in direction is a response to a call for more empirical studies to explore nurse staffing and patient outcomes. This call came from a hallmark report in the United States by the Institute of Medicine (IOM) Committee on the Adequacy of Nurse Staffing in Hospitals and Nursing Homes (McGillis Hall, 2005). Many of the nursing studies resulting from this report, however, focused on the inpatient setting and used different measures of staffing, databases, and risk adjustment methods. As a result, those study findings are not consistent (Mark, 2006). Factors such as nurse shortages, the growing demand for hospital emergency services, tight fiscal constraints, and the desire to have patients satisfied with the care experience, suggest that more studies are needed to understand the relationship between nurse staffing and patient satisfaction in EDs. ED administrators are faced with a growing volume of patients seeking care, higher acuity of patients, and rising fiscal pressures. To address these issues, administrators have implemented different staffing models to control cost in an environment of higher and more complex patient volumes. Although patient satisfaction is multidimensional and complex, patient satisfaction with nursing care has been found

13 5 to be the most important component of overall satisfaction with inpatient hospital care (Strasser & Davis, 1991). In addition, funders of EDs, such as Ontario Ministry of Health and Long-Term Care, closely monitor patient satisfaction. As policy makers explore new funding systems that take into consideration the patient s perspective, understanding the nurse staffing patient satisfaction relationship at the gateway to the acute care system has become extremely important to health administrators, managers, and staff. This study provides healthcare decision makers with vital information on this relationship as an indicator of the quality of care in EDs. 1.4 Conceptual Framework The conceptual framework for this research is adapted from the Nurse Staffing and Patient Outcomes Model developed by Kane et al. (2007) to explain the relationship between nurse staffing and outcomes of care (see Figure 1). The Kane et al. s framework focuses on two types of outcomes: nurse outcomes and patient outcomes. The researchers argue that nurse outcome variables can interact with nurse staffing variables to affect patient outcomes, and that nurse characteristics and patient factors can influence nurse staffing. Patient factors and hospital organizational factors were included in the Kane et al. s framework because these factors may influence the effect of nurse staffing on patient outcomes. As a result, Kane et al. argue that patient outcomes subsequently will affect patient length of stay (LOS) since greater complication rates will increase LOS.

14 6 Figure 1. Conceptual Framework of Nurse Staffing and Patient Outcomes (Kane et al., 2007) In the current study, Kane et al. s (2007) framework is adapted to focus on aspects of care addressed in the literature exploring the relationship between nurse staffing and patient satisfaction in EDs (see Figure 2). The present study investigates three different aspects of nurse staffing: 1) Intensity of Care examined by hours per visit by staff category (RN, registered practical nurse RPN, Nurse Practitioner NP, and agency nurse) and hours per patient length of stay by staff category; 2) Nurse staff mix examined by measuring skill mix by staff category (RN, RPN, agency nurse, NP); and 3) Staff Adequacy examined by nurse/patient staffing ratio for each staff category.

15 7 Figure 2. Conceptual Framework of Nurse Staffing and Patient Satisfaction Patient Characteristics Age Gender Hospital Organizational Characteristics Size (# of ED visits) Type of hospital (teaching, small community) ED wait (% of patients seen within recommended timeframe) ED Case mix index ED Cleanliness Nurse Staffing Intensity of Care: RN hours per visit, RPN hours per visit, Agency Nurse hours per visit, NP hours per visit, staff hours per visit, RN hours per length of stay, RPN hours per length of stay, Agency Nurse hours per length of stay, NP hours per length of stay, staff hours per length of stay. Skill Mix: RN skill mix, RPN skill mix, Agency nurse skill mix, NP skill mix. Staff Adequacy: RN/Patient staffing ratio, RPN/Patient staffing ratio, Agency Nurse/Patient staffing ratio, NP/Patient staffing ratio, Staff/Patient staffing ratio. Nurse Characteristics Age Education Level Nurse Experience (yrs) Full time/part time, employment mix Emergency Department Care Physician Characteristics Physician Courtesy Patient Satisfaction Outcomes Overall Satisfaction - Overall, how would you rate the care you received in the Emergency Department? Would you recommend this emergency department to your family and friends? Nursing Care When you had important questions to ask a nurse, did you get answers you could understand? (Answer) If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? (Explain) Did you have confidence and trust in the nurses treating you? (Trust) Did nurses talk in front of you as if you weren t there? (Respect) How would you rate the courtesy of your nurses? (Courtesy) How would you rate the availability of your nurses? (Availability) How would you rate how well the doctors and nurses worked together? (Dr-Nurse working relationship) The conceptual framework includes patient factors and hospital organizational factors that may influence the effect of nurse staffing on the selected outcomes. The patient factors include age and gender. Hospital organizational factors include the number of ED visits as an indicator of size, type of hospital, ED case mix index, ED cleanliness, and proportion of patients seen within targeted length of stay timeframe. Hospital factors and nurse characteristics can affect the relationship of nurse staffing on patient outcomes (Aiken et al., 1994; Aiken et al., 2002). In light of this, and because physician courtesy may also affect the relationship being investigated, the

16 8 hospital factors, nurse characteristics, and physician courtesy were included to moderate the effects of nurse staffing variables on patient satisfaction outcome variables. 1.5 Hypotheses This study seeks to determine to what extent specific aspects of nurse staffing relate to: 1) patient satisfaction with nursing care; 2) overall satisfaction with care received in the ED; and 3) whether the patient would recommend this ED to friends and family. The study draws on existing administrative and patient satisfaction survey data from Ontario s EDs to test the following hypotheses: Hypothesis 1: There is a positive relationship between RN proportion, nurse-to-patient ratio, nursing hours per patient visit and each patient satisfaction with nursing care variable (i.e., Answer, Explain, Trust, Respect, Courtesy, Availability, and Dr-Nurse working relationship). Hypothesis 2: There is a positive relationship between RN proportion, nurse-to-patient ratio, RN hours per patient visit and overall satisfaction with care received in the ED. Hypothesis 3: There is a positive relationship between RN proportion, nurse-to-patient ratio, RN hours per patient visit and whether the patient would recommend the ED to friends and family.

17 9 In summary, this study examines the relationship between nurse staffing and patient satisfaction in emergency departments. Using an adapted framework of nurse staffing and patient outcomes developed by Kane et al. (2007), the underlying structure of patient satisfaction with nursing care and the presence, magnitude, and direction of the relationships between nurse staffing in the ED and patient satisfaction variables are assessed. The nurse staffing variables include the intensity of care, skill mix, and staff adequacy. Patient satisfaction variables include variables with elements of nursing care related to respect, courtesy, promptness, interpersonal relations, response to patient questions, and explanation of actions taken. Overall satisfaction variables include elements of nursing care related to overall satisfaction with care and recommending this ED to family and friends.

18 10 Chapter 2 Literature Review 2 Overview This literature review surveys the state of study on nurse staffing and patient satisfaction with nursing care in emergency departments. Articles published over the last three decades were examined if they focused on research related to the relationship between nurse staffing and patient satisfaction with nursing care. Even though there are many studies exploring the relationship of nurse staffing and outcomes in the last ten years, very few examine the relationship between nurse staffing and patient satisfaction with nursing care in the ED. For that reason, the search was expanded to include all inpatient settings. The studies reviewed were primarily observational in nature and measured staffing levels and patient satisfaction. Appendix A gives more details on keywords used in the literature search for this review. This review is organized in two major areas: (a) nurse staffing and (b) patient satisfaction with nursing care. The chapter will: review the nature of services and models of care in EDs; define the concept of nurse staffing; examine the factors influencing nurse staffing; review measures of nurse staffing; review the concepts and measures of patient satisfaction, as well as the factors affecting patient satisfaction measurement;

19 11 examine the empirical evidence exploring the relationship between nurse staffing and patient satisfaction; and examine nursing theoretical frameworks. 2.1 Background Emergency Departments EDs are often the gateways to hospitals, and they have a higher census than inpatient areas (Hall & Press, 1996). Patients presenting in EDs can require a range of health services, from specialized care for life-threatening problems to primary care for non-urgent problems. As a result, nurses highly trained in caring for major traumas may also be caring for patients with non-urgent health needs. Moreover, the unique nature of the ED encounter which is brief, impersonal and emotionally intense requires a good understanding of the staffing strategies that influence patient outcomes. Each year, 20% of Ontarians visit an ED at least once. In 2009/10, there were 5.4 million visits to the ED, with 73% of the resuscitation and emergent patients waiting up to 8 hours, 75% of urgent patients waiting up to 6 hours, and 85% of the semi-urgent and non-urgent waiting up to 4 hours (OHQC, 2010). Approximately 10% of Ontario s ED patients are admitted to hospital for care, but more than 80% return to the community after receiving care in the ED (CIHI, 2011). In the last two decades, hospital restructuring has been very prominent. In the mid-1990s, overcrowding in EDs became a concern for Ontario s hospitals. The growing ED volume of patients, increases in the number of non-urgent patients, and overcrowding of EDs led health policy makers and healthcare organizations to make changes to the delivery of emergency care services by establishing hospital-managed

20 12 urgent care centres and creating fast-track services. In addition, many hospital administrators facing funding shortfalls reduced their numbers of regulated nurses (Aiken et al., 1996; CNA, 2004) and many registered nurse positions were replaced with less-skilled positions (CNA, 2004) Types of Emergency Departments Emergency departments vary in size and in the types of patients seen. In Ontario, some EDs are regional referral centres and receive severely ill patients from other hospitals, while other EDs are the only source of care that is available 24 hours a day, seven days per week. Traditionally, the ED has been for urgent medical care, but in the last decade there has been a rise in the use of EDs by patients needing nonurgent care (CIHI, 2011). A common theory for the change in utilization has been the shortage of primary care, as a large number of ED users did not have access to primary care when needed (Han et al., 2007). In the United States, the rise in non-urgent ED visits has been attributed both to cuts in access to primary care services and to individuals who lack health insurance, a regular place of medical care, or both (Tyrance et al., 1996). Interestingly, there are also a significant number of non-urgent care ED patients in jurisdictions such as Canada and Great Britain where there is universal access to primary health care (Beland et al., 1998). Non-urgent care visits in EDs are a very challenging issue facing health care organizations, policy advisors, and patient advocates. Almost half of all ED visits are non-urgent care patients (Williams & Bamezai, 2005), and this increase in non-urgent visits has resulted in ED overcrowding, longer wait times, and heavy staff workloads (Korn & Mansfield, 2008). To assist EDs, Ontario established hospital-managed urgent

21 13 care centres (UCCs) so that unscheduled patients presenting with acute or episodic conditions can be treated in a setting other than EDs. In Ontario, UCCs were established as a result of the Hospital Restructuring Commission Services directives (HSRC, March 1997). Eleven UCCs were developed during a period when ED overcrowding, ambulance diversions, and a growing perception of physician and nurse shortages was attracting significant public scrutiny. The goal of these UCCs is to assess, treat and/or plan a patient s care within 60 to 90 minutes following arrival. Comprehensive EDs are open 24 hours a day, seven days a week, and they provide care to patients who arrive by ambulance or other means, while UCCs although located in hospitals or ambulatory care centres have restricted hours and do not generally care for patients arriving by ambulance. They are, however, staffed by the same types of personnel who work in comprehensive EDs Nurse Shortages In the last decade nurse shortages in the U.S. and Canada have resulted in vacancies in hospitals, long wait times, adverse events, and untenable work environments for nursing staff. Canadian hospitals experienced registered nurse shortages in the 1990s (Aiken et al., 2001). In 2004, the Canadian Nurses Association predicted that Canada would need 60,000 full-time equivalent (FTE) registered nurses by 2022 to meet healthcare needs (CNA, 2004). In 2007, the nurse shortage in the United States was expected to reach 260,000 full-time equivalents by 2025 (Clarke, 2007). Recent studies have provided evidence suggesting however that the nurse shortage trends have reversed over the past decade (Auerbach et al., 2011; Staiger et

22 14 al., 2012). Between 2002 and 2009, there has been a large surge in the number of younger RNs entering nursing resulting in a projected growth of the nursing workforce in the U.S. in the next two decades. In fact, the growth between 2005 and 2010 is the largest expansion of the RN workforce over a 5-year period ever observed in the last four decades. The sudden rise in RN employment may be due to several factors including the economic downturn. Staiger et al. (2012) stated that many RNs who were not working or were working on a part-time basis may rejoin or change to full-time status to ensure better personal economic security. These researchers predict that many of the RNs who entered the workforce between 2005 and 2010 are likely to withdraw as job recovery takes place and unemployment rates fall (Staiger et al., 2012). With an expected wave of retirement of RNs in the next five years, another shortage is project by Staffing Models A combination of nurse shortages, an increasing number of patients, and increasing clinical responsibilities for nurses have resulted in a range of staffing models being introduced in recent years. These models include changes in nurse staffing levels, the nursing skill mix, staff allocation models with varying nurse staff levels (or nurse-to-patient ratios), shift patterns, and the use of overtime and agency staff. This review of the literature will focus on the different models of care in the emergency department, methodology used for staffing in hospitals, factors affecting nurse staffing, and measures of nurse staffing.

23 Emergency Department Models of Care Models of care are applied to an ED to assist with the management of specific patient profiles. Six contemporary models of care have been identified that are effective and appropriate for EDs. These models are: fast track, short stay unit, streaming, care coordination, rapid assessment team, and psychiatric liaison (PricewaterhouseCoopers, 2008). EDs have implemented a mix of these models, depending on the number of visits, case mix, remoteness, skill mix, and experience of the staff. Fast track is a model of care whereby patients with less urgent medical conditions are "streamed" into a dedicated space for treatment (Drummond, 2002; Yoon, 2003). These patients are treated by a dedicated clinical team with the aim of reducing patient discharge time from the ED to two hours. Studies found that the implementation of fast track zones in large and middle-sized EDs with a high volume of low-complexity patients resulted in a significant improvement in quality, safety, and efficiency outcomes (Drummond, 2002; Yoon, 2003; Rodi et al., 2006; Considine et al., 2008; Kwa & Blake, 2008). Short stay units (SSUs) are also called emergency medical units (EMUs) and clinical decision units (CDUs). These units are developed for ED patients who require observation and specialist assessment, and whose length of stay is expected to be limited (to less than 24 hours, for example) (Abenhaim et al., 2000; Daly et al., 2003; NSW, 2006; Konnyu et al., 2011). SSUs are effective in improving patient flow through an ED, limiting patient length of stay to six hours, and avoiding admissions to inpatient units.

24 16 Streaming is a model of care where patients are separated into different streams based on complexity and/or acuity and disposition (FitzGerald et al., 2010). Streaming has been shown to improve quality, safety, and efficiency outcomes in EDs, and it has been implemented in large and mid-sized EDs with visits per annum ranging from 35,000 to in excess of 60,000 (King et al., 2006). Care Co-ordination Teams (CCT) and Geriatric Consultation Teams are implemented in EDs to reduce admissions, length of stay, and re-presentations for complex patients such as the elderly, people living alone, those requiring assistance with activities of daily living, the homeless, and those with drug and alcohol problems. Implementation of these teams has been associated with a significant reduction in admissions, as well as high satisfaction among patients and staff (Sinoff et al., 1998; OHA, 2003). Rapid Assessment Teams (RATs) have been implemented in EDs to provide an early comprehensive medical assessment. This has resulted in early initiation of diagnostic tests, pain management and treatment, and the opportunity for immediate discharge where appropriate (Bullard & Villa-Rowe, 2010). The RAT comprise of a triage nurse and an independent clinician, with the triage nurse referring appropriate patients to the RAT clinician for early assessment (Leaman, 2003). This model of care is used in EDs with enough experienced medical doctors to cover both the ED and the triage area (PricewaterhouseCoopers, 2008). Psychiatric liaison roles provide psychiatric assessment and care for patients who are identified with potential mental health problems. Examples of psychiatric liaison roles include mental health triage and consultancy service nurses, mental health liaison nurses, and psychiatric nurses. These psychiatric liaison roles are beneficial in

25 17 EDs where the staff may not have the expertise in assessing and treating mental health patients, and their implementation has been associated with improvements in efficiency measures, such as waiting times and length of stay (NSW, 2006). The increased demand for ED care (as well as its very nature in treating unscheduled patients presenting with acute or episodic conditions) has prompted a change in workforce models with the introduction of nurse practitioners. The nurse practitioner s role expands the traditional nursing role and allows them to take on tasks traditionally performed by physicians, including the prescription of medication, initiation of diagnostic imaging and laboratory tests, referral of patients to specialists, and admitting and discharging patients (Tye et al., 1998). NPs were introduced to streamline care for those who are non-emergent patients, thus improving the efficiency and care of the physician in the ED (McGee & Kaplan, 2007). Generally, NPs assume responsibility for patients with minor injuries and operate independently within ED teams (Byrne et al., 2000). NPs were found to provide equal or better care than junior doctors, were better at recording medical histories, and had fewer unplanned follow-ups (Sakr et al., 1999). Overall, patients who saw an emergency NP were satisfied with their care and were both significantly more likely to have received health education and to be less worried about their health than those who saw a doctor in the ED (Byrne et al., 2000). 2.3 Nurse Staffing Methodology There are many definitions of the concept of nurse staffing in the literature, but there is a convergence around a common set of elements that include appropriateness of the amount of nursing staff, skill level of the nursing staff, mix of the nursing staff,

26 18 number of patients cared for on the assignment, cost efficiency, and effectiveness (McGillis Hall, 2005). Furthermore, nurse staffing methodology can be described as a standardized approach used to determine the appropriate number and mix of nursing personnel required to provide nursing care that meets the workload demand of the patient care unit. Staffing methodologies can be classified into four areas based on the level of logic and abstract reasoning involved in the construct (Halloran & Vermeersch, 1987). The literature on nurse staffing methodologies can be categorized into four groups: descriptive, industrial engineering, management engineering, and operations research. Descriptive methodology is based on experience and judgment, where subjective decisions are made regarding the appropriate number and mix of nursing staff. The staffing decisions are essentially made using varying degrees of knowledge, training, and analytical skill. Descriptive methodology results in ratios, formulas, or proportions being developed using a wide range of techniques that vary from guesswork to statistical analysis using empirical data. The weakness of this methodology is the lack of consistency among users (Halloran & Vermeersch, 1987). The industrial engineering methodology for nurse staffing was developed in the 1950s from techniques designed to improve productivity. These techniques included work measurement, task analysis, work distribution, and reorganization. The goal of industrial engineering method is to determine which tasks should be performed by nurses, which tasks can be transferred to other staff, and how efficiency can be improved through the mechanization of tasks. The solution to the problem was the redistribution of work from scarce resources to more abundant resources. The fallacy of

27 19 this concept, however, is the assumption that nursing is only a list of tasks and that lesser skilled staff can produce the same quality of care. Both descriptive and industrial engineering methodologies are used widely and have contributed to the subsequent methodological developments used in staffing decisions. The management engineering methodology encompasses industrial engineering concepts and techniques, including work measurement, methods improvement, and work simplification. In addition, the management engineering methodology has concepts and techniques from operations research, most notably variations in nursing work load, patient classification, and mathematical modeling. The conceptualization of nursing is meaningful to both nurses and administrators since this methodology has a clear and consistently applied protocol. Operations research methodology is the most complex and is developed from mathematical models to describe existing nurse staffing patterns for use in present and future decisions. Models are developed by abstracting information on the patients and the hospital system. Operations research methodology can manage complex situations and identify consequences of critical decisions, and it has also been used to determine nurse staffing patterns on inpatient care units. The concept of patient classification was pioneered by Connor et al. in 1961, and they were the first to propose that nursing workload on an inpatient unit varies with the degree of patient care required (Halloran & Vermeersch, 1987). These four nurse staffing methodologies show a gradual increase in both sophistication and collection of reliable and pertinent data, but one common element is that quality measurement of care which includes patient satisfaction is non-existent in all four methodologies of staffing reviewed. Both hospital administrators and the

28 20 public are aware of the challenges of nursing personnel shortages, workload, and the high cost of nursing services. There is, however, no scientifically based methodology that will assist managers and hospital administrators to allocate efficiently scarce nurse resources to promote quality patient outcomes (Yankovic & Green, 2008). 2.4 Factors Influencing Nurse Staffing Kane et al. (2007) reviewed the literature and highlighted eight policies related to nurse staffing in hospitals. These policies were related to: 1) staffing ratio, or the number of patients cared for by one nurse by job category (RN, LPN, UAP); 2) staffing hours per patient day, or total number of nursing staff hours per patient day; 3) staff mix or proportion of hours worked by each skill mix category (RN, LPN, UAP); 4) shift rotations, or scheduling nursing staff to work different work shifts (days, evening, nights) during a defined period of time; 5) shift rotation, or length of the shifts; overtime or policies permitting additional worked hours over (for example, 40 hours/week, weekend staffing or frequency of weekends worked); 6) the use of agency or temporary nurses; full-time/part-time mix, or the number and type of full-time and part-time; 7) floating to nursing units or policies regarding when nurses can work on other units; and

29 21 8) internationally educated nurses or policies regarding the hiring and use of nurses educated in a foreign country. Kane et al. (2007) found nurse staffing policies can be influenced by patient care unit factors, for example, patient flow fluctuations may determine length of shift policies. In addition, the researchers found that nurse staffing policies can be influenced in hospitals in which nurses were unionized or of the age and/or tenure of nurses. Nurse staffing strategies, however, result from staffing policies. In 1999, the American Nurses Association (ANA) developed nine principles for examining appropriate nurse staffing and three categories of factors that should be considered when making nurse staffing decisions. These categories were: a) patient care unit-related, b) staff related, and c) organization related (ANA, 1999). The factors for patient care unit-related include both staffing for the individual patient and the aggregate patient care needs of the unit. The staff-related factors, such as the education and experience level of the nurses, are determined by the patient population being served. Hospital-related factors, such as type and technology level, along with patient care unit factors and nursing organization factors, for example, management and leadership, both affect nurse staffing policies (Kane et al., 2007). Table 2.1 summarizes the factors found by the ANA and other researchers (Mark et al., 2000; Aiken et al., 2001; McGillis Hall et al., 2003; McGillis Hall, 2005; Kane et al., 2007) that influence nurse staffing policies.

30 22 Table 2-1. Factors Influencing Nurse Staffing Policies Factors Influencing Nurse Staffing Policies Patient physical and psychosocial Staff Related Organizational related Primary Diagnosis Age Type, ownership, and mission Age Experience with the specific patient Effective and efficient support services population Comorbidity Level of nurses experience (e.g., novice to expert levels) Access to timely relevant information (linked to patient outcomes) Functional status Education and preparation (e.g. certification) Orientation programs and ongoing competency assessment mechanism Communication ability Language capabilities Technological preparation or technology level Cultural and linguistics Tenure on the unit Adequate time for collaboration Severity and urgency Level of control in the practice Care coordination environment Scheduled procedures (patient Degree of involvement in quality Supervision of unregulated workers flow/census fluctuations) initiatives The ability to meet health care requisites Immersion in activities such as nursing research Process to facilitate transitions during mergers Availability of social supports Involvement in interdisciplinary and collaborative activities regarding patient Mechanisms for reporting unsafe conditions (risk management) needs Number and competencies of clinical and non-clinical support staff Logical method for determining nurse staffing levels and staff mix. Contract Nurses Many studies have examined the effects of changes in categories of nurse staffing patterns on a number of outcomes, such as rates of in-hospital mortality, rates of nosocomial infections, and rates of pressure ulcers (al-haider & Wan, 1991; Blegen et al., 1998; Dimick et al., 2001; Aiken et al., 2002; Needleman et al., 2002; Tourangeau, 2002; Aiken et al., 2003; Aiken et al., 2003; Halm et al., 2005; Kane et al., 2007; Cho et al., 2008). Fewer studies, however, have been found that examined the relationship between nurse staffing and patient satisfaction (Bolton et al., 2003; Wolf et al., 2003; Merkouris et al., 2004; Schmidt, 2004; Chan & Chau, 2005). 2.5 Nurse Staffing Measures There is no instrument that truly measures nurse staffing, but researchers have assessed nurse staffing using methods that focus on: a) staff compliment, and b) the

31 23 mix of staff employed in the organization or unit (McGillis Hall, 2005). Unfortunately, the reliability, validity, and sensitivity of these measures cannot be assessed. In addition to the numerical nurse staffing methods, few studies have used demographic characteristics of nurse staff education and experience, for example to measure nurse staffing (Blegen & Vaughn, 2001; O'Brien-Pallas et al., 2002; Aiken et al., 2003; McGillis Hall et al., 2003; McGillis Hall, 2005). Table 2.2 shows some of the common measures of nurse staffing used in studies reviewed. Table 2-2. Nurse Staffing Measures Measures of Nurse Staffing Proportion of Registered Nurses Blegen et al., 1998; Blegen & Vaughn, 1998; Mark et al., 2000; Needleman et al., 2002 Nursing Hours Per Patient Day Blegen et al., 1998; Blegen & Vaughn, 1998; Cho et al., 2008 Ratio of Registered Nurses to Patients Kovner & Gergen, 1998 ; (Spetz et al., 2000); Aiken et al., 2002 Number of Full-Time Equivalents (FTEs) Blegen et al., 1998; Blegen & Vaughn, 1998; Mark et al., 2000 Percentage of Full-Time, Part-Time and CIHI, 2008 Casual Mix of Nursing Staff McGillis Hall, 1997; McGillis Hall L et al., 2003; McGillis Hall, 2005 ; (Unruh, 2003) Level of Education and Experience Blegen & Vaughn, 2001; O'Brien-Pallas et al., 2002; Aiken et al., 2003; McGillis Hall, 2005 Controversy exists regarding the best measures of nurse staffing. An international panel of experts was surveyed recently to get their opinion about specific staffing measures (Van den Heede et al., 2007). The goal of the exercise was to develop a comprehensive set of variables for future examinations of the association between hospital nurse staffing and patient outcomes. Using a Delphi approach, consensus was reached on ten nurse staffing measures and 29 background variables. Table 2.3 shows three groups of staffing measures (number of nurse staff in relation to

32 24 patient volume, types of staff to be considered as a measure of the number of nurse staff, and skill mix indicators) identified by the panel. Consensus was not reached, however, for two variables: the number of full-time equivalents employed in an organization or unit, and the proportion of RNs to all licensed nurse staffing. More details on these variables are discussed in the following sections. Table 2-3. Nurse Staffing Variables from Consensus Panel Nurse Staffing Variables Number of nurse staff in relation to patient volume Nurse-to-patient ratio Nursing hours per patient day Types of staff to be considered in a measure of the number of nurse staff nursing staff licensed nursing staff RN staff Skill mix indicators Proportion of licensed nursing staff to total nursing staff Proportion of RNs to total nursing staff Proportion of RNs with a Bachelor s degree Proportion of RNs with a Master's degree Nurse to Patient Ratio The nurse-to-patient ratio measures the number of patients cared for by a nurse. One of the limitations of this measure is that the nurse-to-patient ratio relies on a general ratio, which may include all nurses assigned to a unit, including non-clinical time (Kane et al., 2007). The measure typically relies on less precise data about total nurse staffing to patient volume that is derived from administrative databases, averaging annual nurse-to-patient ratios at the hospital or unit level (Bolton et al., 2001). Researchers have measured RN-to-patient ratio by surveying nurses in the last shift worked (Aiken et al., 2002). This survey method has an advantage over using

33 25 administrative data to calculate the ratio in that data is obtained from the nurses who cared for the patients. There are more specific measures of the nurse-to-patient ratio which include RNto-patient ratio, LPN (or RPN)-to-patient ratio and unlicensed-assistive-staff-to-patient ratio. The ratio of patients per RN per shift ratio is frequently used as a measure of nurse staffing in studies examining the effect of staffing on outcomes (Shortell et al., 1994; Aiken et al., 2002). Nurse-to-patient staffing requirements have been mandated in the U.S. In 2004, California implemented a minimum nurse-to-patient staffing ratio requirement in acute care hospitals that set the emergency department nurse-to-patient ratio at 1-to-4 for general emergency, 1-to-1 for trauma and triage, and 1-to-2 for critically ill patients. Many other states have introduced or enacted nurse staffing legislation and/or adopted regulations addressing nurse staffing (Aiken et al., 2010). These requirements, legislation, and regulations are in response to the concern of the adequacy of nurse staffing in hospitals. The Emergency Nurse Association (ENA), however, has rejected these nurse staffing levels and subsequently developed best-practice staffing guidelines that take into consideration patient census, patient acuity, and patient length of stay (ENA, 2003). Studies have shown that increased nursing workload is significantly associated with increased mortality, nurse burnout, and job dissatisfaction (Aiken et al., 2002; Needleman et al., 2002; Kane et al., 2007; Van den Heede et al., 2009). Although the association with the increase in RN staffing in California hospitals and improved outcomes is difficult to assess, Aiken et al. (2010) examined whether nurse staffing using state-mandated minimum nurse-to-patient ratios differed from nurse staffing in

34 26 two states that did not have legislation. This study revealed that nurses in California hospitals cared for one less patient on average per shift than the two states without legislation. Furthermore, lower patient-to-nurse ratios were associated with significantly lower mortality, burnout among nurses, and job dissatisfaction. Research has shown the consequences of the shortage of nurses. Hospitals with high nurse-to-patient ratios have been found to have lower mortality rates (al- Haider & Wan, 1991; Aiken et al., 2002; McGillis Hall, 2005; Kane et al., 2007). In addition, nurse-to-patient ratios have been found to be related to process measures, such as failure to rescue rates, adverse events, medical complications, postoperative respiratory, and cardiac complications (Clarke, 2007). In a study of three adult medical surgical units within a university teaching hospital, patient satisfaction was found to increase when the number of nursing hours per patient increased (Seago et al., 2006). Although this study showed that nurse staffing can affect patients perceptions of the healthcare experience, the study cannot be generalized Nursing Hours per Patient Day Nursing hours per patient day is defined as the total number of productive hours worked by all nursing staff with direct care responsibilities per patient day (a patient day is the number of days any one patient stays in the hospital) (Kane et al., 2007). Unfortunately, different methods have been used to estimate nurse hours per patient day. Some investigators assume a 40-hour week and 52 working weeks per year (2,080 hours per year). Others use more conservative estimates, such as 37.5 hours per week for 48 weeks = 1,800 hours per year (Kane et al., 2007).

35 27 The ANA calculates the numerator, or nursing hours, as the number of productive hours worked by nursing staff assigned to the unit who have direct patient care responsibilities for more than 50% of their shift (American Nurses Association (ANA), 2007). Productive hours are defined as the actual direct hours worked, excluding vacation, sick time, orientation, education leave, or committee time. Direct patient care responsibilities include both patient-centered nursing activities in the presence of the patient and patient-related activities that occur away from the patient, such as medication administration, nursing treatments, nursing rounds, admission, transfer, discharge activities, patient teaching, patient communication, coordination of patient care, documentation, and treatment planning. According to the ANA s methodology, nursing care hours are reported each month for registered nurses (RNs), licensed practical nurses, licensed vocational nurses (LPNs/LVNs), and unlicensed assistive personnel (UAP). The denominator, or patient days, is calculated from the hospital via multiple census reports. Patient censuses are collected multiple times per day by hospitals. These patient censuses are then averaged to get the daily average census, and a sum of the daily average censuses is subsequently calculated to determine patient days for the month on the unit. Nurse hours per patient day reflect average staffing across a 24-hour period. As a result, the measure does not reflect fluctuations in patient census, scheduling patterns during different shifts (even the length of shifts varies), and periods of the year (Kane et al., 2007). These issues are amplified in the emergency department, where there can be a varying patient census in a given day. This measure also does not account for the time nurses spend in meetings, educational activities, and administrative work. Furthermore, while nurse hours per patient day gives an indication of the hours of care

36 28 available for actual patient care, it is limited in that it does not assist in identifying whether the nursing hours were adequate for the complexity of the patient care needs (McGillis Hall, 2005). The measure also does not take into consideration the mix of the nursing staff. For these reasons, the ANA (1999) questions the usefulness of the concept of nursing hours per patient day. It argues that this measure should not be used by a hospital to compare itself to other hospitals since the results are not adjusted to take into consideration factors such as the patient s age or severity of illness, either of which may require more nursing care hours. Also, the frequency of admissions and discharges, as well as the hospital layout, may also affect the nurse staffing needs. In summary, nurse-to-patient ratio and nursing hours per patient day are the two general measures of nurse staffing used in studies, and they were selected by the international panel of nursing researchers as appropriate measures of nurse staffing (Van den Heede et al., 2007; Van den Heede et al., 2009). The nursing hours per patient day addressed hours of care provided by nursing staff averaging FTEs of different nurse categories at the hospital level (Mark et al., 2004) and sometimes only included productive hours worked in direct care (Bolton et al., 2001). As discussed previously, however, the nurse-to-patient ratio relies on the less precise data of total nurse staffing to patient volume that is derived from administrative databases. The ratio of patients per RN per shift variable was more frequently used and provided greater evidence of the effect, but both the ratio of patients per RN per shift and nurse-to-patient ratio show generally the same trends (Kane et al., 2007).

37 Nursing Staff Mix Although staff mix and staff mix models are well-described in the literature, few studies are empirically based (McGillis Hall, 1997). Skill mix or staff mix has been described by (Needham, 1996) in accordance to the Royal College of Nursing as being: the balance between trained and untrained, qualified and unqualified and supervisory and operative staff within a service area the optimum skill mix is consistent with the efficient deployment of trained, qualified and supervisory personnel and the maximization of contributions from all staff. (127) This measure is defined as a proportion of productive (i.e. related to direct patient care) hours worked by each skill mix category (RN, LPN, UAP) (Kane et al., 2007). Staff mix may include combinations of RNs, registered practical nurses (RPN), or licensed practical nurses (LPNs), as well as health care aides, nurse aides, and unlicensed assistive personnel (UAP) or multi-skilled workers (McGillis Hall, 1997). The majority of studies reviewed focused on registered nurses working in acute care hospital settings. Kane et al. (2007) commented in their systematic review that the evidence on the association between RPN or LPN and UAP personnel and outcomes is limited and controversial. The proportion of registered nurses is considered to be the direct nursing care hours provided to patients by RNs. This measure has been calculated in different ways. For example, Blegen et al. (1998) used a two-step approach by first calculating RN hours as the direct patient care hours provided by a nurse, divided by total patient days on the unit. Subsequently, the proportion of RN was

38 30 calculated as RN hours per patient day divided by the total hours provided by all nursing staff per patient day on the unit. Other researchers, however, calculated RN proportion as the number of FTE RN staff divided by the number of FTE on the unit (Mark et al., 2003) Number of Full Time Equivalents A count of the number of FTEs is another measure of nurse staffing. This measure can be further broken down by category of staff, such as registered nurse FTEs, registered practical nurse FTEs, and unregulated worker FTEs. There is difference between the head count and FTE calculation: FTE represents the number of positions in the unit, but the number of staff (head count) can be higher since a position can be filled by part-time and casual staff. Thus, when this measure is used, researchers have either linked it to the total number of employees employed, or they have presented the percentage of FTE hours made up of full-time, part-time and/or casual staff (Blegen et al., 1998; Mark et al., 2000; Blegen & Vaughn, 2001) Percent of Full Time, Part Time or Casual Staff This measure has grown in stature in the last decade because of the debate about using casual staffing in hospitals. The Registered Nurses Association of Ontario (RNAO) has advocated for increasing full-time employment in hospitals to be a minimum of 70% (RNAO, 2005). No empirical literature, however, exist that associates the number or percent of full-time, part-time, or casual nursing staff to outcomes (McGillis Hall, 2005). This measure was also not selected by the international panel of researchers in their selection of nurse staffing measures (Van den Heede et al., 2009).

39 Level of Education and Amount of Experience Other staffing variables have been considered when exploring nurse staffing and patient outcomes. Researchers have selected the education level and the experience of nurses as important background variables (Van den Heede et al., 2009). Kane et al. (2007) found a significant negative correlation between the percentage of nurses with Bachelor of Science in Nursing (BSN) degrees and the incidence of deaths related to health care (r = -0.46, p = 0.02). The crude rates of complications were found to be associated with a reduction of 1.13 percent (95 percent CI ) for each additional year of nurse experience in surgical patients in the ICU (Aiken et al., 2003). Furthermore, an increase of 1 percent in the proportion of nurses with BSN degrees was associated with a reduction in the rate of failure to rescue by 0.04 percent (95 percent CI ). The authors reported a 5 percent reduction in failure to rescue corresponding to a 10 percent increase in the proportion of nurses with BSN degrees (RR 0.95, 95 percent CI ). Similarly, (Aiken et al., 2003) found hospitals reported lower rates of surgical mortality and failure to rescue if they had higher proportions of nurses with BSN degrees. Having more experienced nurses was found to be associated with lower medication errors and fall rates (Blegen & Vaughn, 2001). McGillis Hall (2005) reported similar results, with less-experienced nurses being associated with higher amounts of wound infections on a unit. In their comprehensive review of the nurse staffing literature, however, Kane et al. (2007) found studies that did not show significant changes in pressure ulcers, patient falls, or urinary tract infections in relation to nurse experience and education.

40 Other Factors Thungjaroenkul et al. (2007) performed a systematic review of the literature on the impact of nurse staffing on hospitals costs and patient length of stay. The reviewers found the relationships between nurse staffing, hospital costs, and length of stay were mixed. The studies also found a range of methods and definitions of costs and length of stay. Although it was difficult to conclude the effects of nurse staffing, the evidence suggested that significant reductions in cost and length of stay may be possible with higher ratios of nursing personnel in hospital settings (Thungjaroenkul et al., 2007). Ten out of the 13 studies showed that the ratio of RNs to patients, nursing staff mix and hours per patient day were significantly related to patient LOS. The researchers found no studies that evaluated the effect of RN staff experience and RN staff education on LOS. Table 2.4 shows the impact of nurse staffing on patient length of stay, as reported by Thungjaroenkul et al. (2007). Table 2-4. Summary of the Impact of Nurse Staffing on Patient Length of Stay Effect of Nurse Staffing on Patient Length of Stay Nurse Staffing Variables Length of Stay Measures Sources Significant Findings Ratio of RNs to patients Days of admission Amaravadi et al. (2000) Negative relationship Pronovost et al. (1998) Negative relationship Lichtig et al. (1999) NS Ratio of RNs to other nursing staff Hours per patient day Ratio of actual and expected LOS hours Days of admission Days at midnight census Ratio of actual and expected LOS Not identified Days of admission Not identified (Source: Thungjaroenkul et al. (2007)) Lassnigg et al. (2002) Pratt et al. (1993) Cho et al. (2003) Newhouse et al. (2005) Barkell et al. (2002) Lichtig et al. (1999) Needleman et al. (2006) Cho et al. (2003) Schultz et al. (2003) Behner et al. (1990) Negative relationship NS Negative relationship NS Negative relationship Negative relationship Negative relationship Negative relationship Negative relationship Negative relationship

41 Patient Satisfaction with Nursing Care This section of the review of the literature presents an overview of customer and patient satisfaction in the marketing and health services literature, the definition of patient satisfaction, the methodological issues in measuring patient satisfaction, and the factors affecting patient satisfaction. Patient satisfaction with nursing care has been defined as the patients subjective evaluation of the cognitive-emotional reaction that results from the interaction of their expectations of ideal nursing care and their perception of actual nursing care (Risser, 1975; Eriksen, 1995; Johansson et al., 2002). Unfortunately, consensus on a common conceptual definition of patient satisfaction is still lacking (Fitzpatrick, 1991; Bond & Thomas, 1992; Cleary et al., 1992; Williams, 1994). Laschinger et al. (2011) argue that few studies have demonstrated empirical support for the concept of patient satisfaction. In fact, researchers have commented that patient satisfaction, being multi-dimensional in nature, has been measured in many different ways because there is no consensus on the domains to be included (Hall & Dorman, 1990; Chang, 1997; Sitzia & Wood, 1997; Merkouris et al., 1999). Patient satisfaction is nonetheless important to hospital administrators since it is the arbitrator between patient s perception of quality of care and his/her future intentions to reuse the service or recommend the service to others (Laschinger et al., 2011). Furthermore, perception of quality can be defined as a long-term attitude developed over time, whereas patient satisfaction can be defined as a short-term response to a specific experience. So unlike healthcare marketers, who are interested in a patient s future desire to recommend the healthcare provider to others (or to return themselves),

42 34 nurses focus on utilizing patient satisfaction data to improve the patients health status. Therefore, patient satisfaction can be treated as both an outcome measure (satisfaction with health status following treatment) and a process measure (satisfaction with the way in which care was delivered) (Coulter et al., 2009). Research frameworks for patient satisfaction with nursing have been presented in the literature (Greeneich, 1993). Concepts used in these frameworks included explanations, concern, mutual goal settings, receptiveness to patients expressions of feelings, technical competence, nursing knowledge, communication, equity of treatment, and the giving of information (Bursch et al., 1993). With respect to the emergency department, five important variables that correlate with overall ED patient satisfaction are waiting time before being examined, nursing care, physicians concern, how organized the staff was, and the information provided by physicians and nurses concerning the patients illnesses (Bursch et al., 1993; Krishel & Baraff, 1993; Sandovski et al., 2001). Boudreaux et al. (2004) highlighted that the studies on patient satisfaction with ED care have inconsistent findings, thus firm conclusions are not possible. The researchers found several methodological issues that cause these discrepancies. Outcomes were not standardized in ED studies, Some studies, for example, use ratings of overall satisfaction and likelihood of recommending the ED to others. Although the outcomes are often conceptually similar and highly correlated, they are not necessarily identical. Another issue for studies of nurse staffing and patient satisfaction with nursing care is the lack of a universally accepted pool of indicators. Studies vary in the number, type and nature of the predictors used, therefore inconsistencies among findings are not

43 35 surprising (Boudreaux et al., 2004). Boudreaux et al. (2004) commented that research methodologies differ among studies with some studies using postal surveys while others used telephone surveys, proxy raters and the time elapsed since the ED visit. In addition, studies used highly correlated predictors, thus resulting in noncollinearity issues for regression analyses. Lastly, there is a tendency for researchers to use the traditional p-value cut-off strategy to interpret the results of statistical analyses, which may artificially inflate descrepancies between studies. Information sharing is an important predictor of satisfaction for both inpatient and ED setting (Nerney et al., 2001). In a study of the factors associated with older patients satisfaction with care in an ED, anxiety and concerns felt by patients were alleviated through effective communication (Nerney et al., 2001). Furthermore, staff interpersonal skills, communication skills, and the provision of information are predictors of patient satisfaction in EDs (Boudreaux et al., 2004; Boudreaux & O'Hea, 2004; Taylor & Benger, 2004). Nurses have consistently emphasized the importance of emotional care for patients (Jacox et al., 1997; Boudreaux et al., 2000; Darby, 2002; Johansson et al., 2002; Aiello et al., 2003; Boudreaux & O'Hea, 2004; Al-Mailam, 2005; Liu & Wang, 2007). This, however, is contrast to what the patients themselves feel. In a survey to determine patient s satisfaction with nursing, patients were found to expect the following nursing qualities: a friendly personality, kindness, a fast response to the needs of the patient, and adequate time to provide the needed care (Fitzpatrick, 1991). Patients also consider technical care and providing explanations regarding their condition as important (Megivern et al., 1992; Sitzia & Wood, 1997; Schmidt, 2003; Chan & Chau, 2005). Despite this perception, a study by (Donabedian, 1980) highlights that patients

44 36 have a limited knowledge of the technical aspect of care. Other researchers have supported that conclusion, showing that patients were more concerned with the interpersonal skills of staff than with their technical skills and competency (Nelson E & C., 1993; O'Connell et al., 1999; Taylor & Benger, 2004).. Patients expectations of the nurse are related to the nurse s knowledge and competence as well as personal care (Johansson et al., 2002). Moreover, patients expect nurses to act as a companion and adviser, be empathetic to their needs, have good communication skills, provide the necessary information, and direct the patient both emotionally and physically. These aspects of patients expectations are important to consider in the measurement of patient satisfaction with nursing care. 2.7 Instruments for Measuring Patient Satisfaction with Nursing Care There are several instruments used to collect patient satisfaction, but many of them are not based on theoretical models. Researchers are unable to compare results across settings because of the lack of a standardized instrument with sound psychometric properties (Dansky & Miles, 1997). In 1975, Risser created the first standardized measure of patient satisfaction. It had three dimensions of satisfaction: a) technical/professional behaviours (i.e. nursing knowledge and techniques), b) a trusting relationship (i.e. communication and interpersonal skills), and c) an educational relationship (i.e. information-sharing about patient condition and care processes) (Risser, 1975). Over the years, Risser s measures have been adapted by many researchers to create other instruments to measure patient satisfaction (Hinshaw &

45 37 Atwood, 1982; Oberst, 1984; LaMonica et al., 1986; Larson & Ferketich, 1993). In fact, newer instruments were created by modifying and extending the three dimensions of Risser's original instrument to reflect nursing behaviors expected in the acute care setting'' (LaMonica et al., 1986 p. 44). Nonetheless, all of these instruments have limitations associated with the conceptual complexity of patient satisfaction (Dozier et al., 2001). Notwithstanding the conceptual limitations, there are a number of valid and reliable instruments used to measure patient satisfaction with nursing care (Strasen, 1989; Chang, 1997; Larrabee & Bolden, 2001). In a review of 53 studies on patient satisfaction with nursing, Chang (1997) found 13 different, published instruments between 1992 to Having many different instruments measuring patient satisfaction can be problematic. Chang expressed concerns about variation in the scales used, the difference in domains covered between instruments, and the variation in data collection methods. For these reasons, Chang concluded that it was difficult to draw comparisons between studies. Larrabee and Bolden (2001) performed a literature review and found 40 instruments designed to measure patient satisfaction with nursing care had been published between 1966 to Using a convenience sample of 199 hospitalized adult patients in a public hospital in South Central United States, the authors reported patients identified five themes of good nursing care: providing for my needs, treating me pleasantly, caring about me, being competent, and providing prompt care. Table 2.5 shows the concepts for each theme.

46 38 Table 2-5. Characteristics of Good Nursing Care (Larrabee and Bolden, 2001) Theme Theme 1: Providing for needs Theme 2: Treating Me Pleasantly Theme 3: Caring about Me Theme 4: Being Competent Theme 5: Providing Prompt Care Concept Taking care of me Checking on me Responding to my requests Providing comfort Giving accurate information Providing a pleasant environment Treating me nicely Respecting me Having a positive attitude Treating me with patience Being there for me Showing caring or concern Using knowledgeable skills Striving for excellence More recently, there are commercially available survey instruments by the Picker Institute and Press Ganey Associates used by hospitals to measure patient satisfaction in hospitals (Hall & Press, 1996). These survey instruments were developed from data retrieved in focus groups with patients and providers. The NRC-Picker survey, used in EDs in both Ontario and British Columbia, consists of the following categories: patient preferences, coordination of care, information and education, physical comfort, emotional support, family and friends involvement, and continuity and transition. Although the commercial tools addressed some of the issues of the earlier tools, they are criticized as being too global and failing to measure specific aspects of nursing care. Over the last two decades, researchers have developed and refined dimensions of patient satisfaction to categorize variables (Ware et al., 1983; Eriksen, 1995; Chang, 1997; Larrabee & Bolden, 2001). To assist with organizing the nursing questions into structure, process and outcome, a framework established by Chang (1997) was utilized to organize the variables from the patient satisfaction survey. Table 2.6 shows the

47 39 subset of nine indicators of the emergency department patients perception of outcome of care from the NRC-Picker survey used in Ontario. There are seven indicators that relate to patients perception of nursing care outcomes. Item #8 and item #9 reflect the patients' perspective of the overall satisfaction with care in the ED. Table 2-6. Nine indicators of the ED Patients Perception of Care - NRC- Picker Survey. Item Items in Survey (Variables) Chang (1997) Indicators Outcome Indicators 1 When you had important questions to ask a nurse, did you get answers you could understand? 2 If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? 3 Did you have confidence and trust in the nurses treating you? 4 Did nurses talk in front of you as if you weren t there? 5 How would you rate the courtesy of your nurses? 6 How would you rate the availability of your nurses? 7 How would you rate how well the doctors and nurses worked together? 8 Would you recommend this emergency department to you family and friends? 9 Overall, how would you rate the care you received in the Emergency Department? Patients felt that they had better understanding of illness, received useful or helpful information, and knew how to care for themselves when they went home. Patients felt that nurses reduced their fears and concerns. Patients felt that nurses met their needs and nursing care was helpful. Patients felt that nurses made them comfortable, clean, and refreshed. Patients felt calm, better, and secure after receiving nursing care. Patients felt that nurses met their needs and nursing care was helpful. Intend to return to this hospital if needed in the future and recommend this hospital to friends and relatives. Overall satisfaction or quality. The survey questions from the NRC-Picker tool correspond to Chang s specific items with the exception of item #7. Using Chang s mapping, the NRC-Picker questions correspond to outcome variables. For items #1, #2, #8 and #9, the NRC-Picker questions measured the same concepts as Chang s indicators. For item #3, the NRC- Picker question "did you have confidence and trust in the nurses treating you" was

48 40 more specific than Chang s indicator but measured a similar concept of confidence and trust in the nurse. Item #4 " the nurses talking in front of you as if you weren t there" is more specific than Chang s indicator of respect, privacy and dignity. Item #5 "the courtesy of your nurses" is more specific than Chang s indicator, which measures if the patient felt calm, better and secure after nursing care. Item #6 "the availability of the nurse" measures if the patient felt the nurse was available when he or she needed care, emotional support, or physical assistance. So Item #6 is similar to Chang s indicator which measures if the patient felt his or her needs were met. In this review, the items from the NRC-Picker survey correspond with the outcome indicators from Chang (1997) with few differences. 2.8 Factors Associated with Patient Satisfaction with Nursing Studies have shown that patient satisfaction with nursing care in the inpatient setting is strongly associated with (and is an important predictor of) a patient s overall satisfaction with the hospital care (Johansson et al., 2002; Bolton et al., 2003; Larrabee et al., 2004; Al-Mailam, 2005; Chan & Chau, 2005; McGillis Hall, 2005; Kane et al., 2007). Some researchers have argued that published research on factors influencing patient satisfaction with nursing care has typically failed to assess the extent to which individual (patient-specific), patient-provider, and departmental (provider-specific) influences interrelate in affecting patient satisfaction (Coulter et al., 2009). These factors can be classified as patient characteristics, interpersonal and structural factors, nurse job satisfaction, and nursing work environment.

49 41 Patient characteristics such as cultural background, age, sex, and education have been found to be related to patient satisfaction ratings (Bacon & Mark, 2009). Unfortunately, the findings are mixed since some studies did not find any relationships between patient satisfaction and demographic variables (Laschinger et al., 2011). The perceived quality of the interactions between the patient and nurse has been found to be related to patient satisfaction. There are four interpersonal factors associated with patient satisfaction: 1) involving patients in decision making about their care and respecting their right to convey their thoughts or opinions about their care options; 2) providing information about patients conditions and explanations of symptoms they may experience; 3) using a compassionate care approach; and 4) creating an equitable relationship that ensures fairness (Laschinger et al., 2011). The kindness and warmth of nurses, their technical skills, the amount of information they provided to patients, the time they spent with their patients, and the respect they provided to the relatives and friends of patients have all been found to be associated with enhancing the level of patients satisfaction and their experiences of nursing care (Alhusban & Abualrub, 2009). Structural factors have been found to affect patient satisfaction with nursing care. These factors included the perception of nurses competency and method of delivery of nursing care, such as critical pathways and professional practice models. The impact of nurses job satisfaction on patient satisfaction with nursing care has been examined. Patients on a unit with nurses who reported high job satisfaction themselves were found to report higher levels of overall satisfaction with their care (Laschinger et al., 2011). Conversely, Larrabee et al. (2004) were unable to find a significant relationship between

50 42 nurses job satisfaction and patient satisfaction. Unfortunately, these findings are not consistent throughout the literature. Studies have shown staffing levels in inpatient units of hospitals have significantly influenced patient satisfaction (Sovie & Jawad, 2001; McGillis Hall et al., 2003), but more research is required to establish the link between healthcare provider working conditions and patient outcomes (Laschinger et al., 2011). In 2009, however, researchers reported that increased satisfaction on units was associated with greater support services for nursing (Bacon & Mark, 2009). Another factor that affects patients rating their satisfaction with the quality of care is the inability to dissociate their view of nurses from the hospital (Mahon, 1996). For example, patients may not differentiate between nursing care and the hotel functions provided in a hospital (Dozier et al., 2001). So while patients may feel capable of rating the hotel functions of their hospital experience, they assume that the therapeutics and care they receive are what they should be receiving (Williams, 1994; Dozier et al., 2001). Therefore, the patient s ratings of nursing care instead may be based on his/her assessment of housekeeping or dietary services. As discussed previously, many instruments used to measure patient satisfaction do not adequately capture nursing activities (Chang, 1997). This can be problematic for researchers examining the link between patient satisfaction, nursing activities, and/or changes in staff mix and staffing ratios. Furthermore, patients have difficulty in differentiating nurses from other hospital staff. This issue threatens the reliability and validity of the measurement of patient satisfaction with nursing care (Pasoe, 1983). Other factors that affect patient satisfaction with nursing care are discussed in the next sections.

51 Gender In their review of the literature, researchers have found that men are more satisfied with their care than women (Lövgren et al., 1998; Johansson et al., 2002; Alhusban & Abualrub, 2009; Arnetz & Arnetz, 1996.), while other studies reported women to be more satisfied (Lövgren et al., 1998; O'Connell et al., 1999; Ahmad & Alasad, 2004; Chan & Chau, 2005; Alhusban & Abualrub, 2009). Still some studies reported that gender was not associated with satisfaction at all (Barbara et al., 1999); (Wallin et al., 2000; Liu & Wang, 2007). There is a lack of consensus on the association of patient satisfaction and gender Age The age range of the patients attending an ED can be wide (Sandovski et al., 2001). Although the literature revealed that the associations between patient satisfaction, their education level, and length of stay were consistent, there are inconsistent results in regard to the association between patient satisfaction and age or gender. Age has been found to be significantly related to patient satisfaction (Johansson et al., 2002; Chan & Chau, 2005; Liu & Wang, 2007). Older patients were found more satisfied than younger patients (Mahon, 1996; O'Connell et al., 1999; Liu & Wang, 2007; Alhusban & Abualrub, 2009), while younger patients were found to have significantly lower satisfaction with ED care (Hansagi et al., 1992; Sun et al., 2000; Sandovski et al., 2001). Other studies, however, reported that age was not associated with satisfaction (Barbara et al., 1999; Wallin et al., 2000; Ahmad & Alasad, 2004; Alhusban & Abualrub, 2009).

52 44 Patient satisfaction with the triage nurse was examined in the emergency department (Chan & Chau, 2005). Similar to the studies of inpatient wards, Chan and Chau (2005) found a statistically significant correlation between age and patient satisfaction, with older people tending to report a higher level of patient satisfaction. Despite this, however, an earlier study on patient satisfaction with emergency nursing care found no significant correlation between age and patient satisfaction with emergency nursing care, although no explanation was given as to why there was such a difference (Raper, 1996) Nurse Staffing Education and Experience Kane et al. (2007) reviewed observational studies to examine the relationship between nurse staffing and outcomes, including patient satisfaction. The researchers commented that there is limited evidence to suggest that better nurse staffing is associated with patient satisfaction with nursing care in inpatient units. Kane et al. (2007) argued that inpatient units with a high proportion of RNs were associated with high ratings of satisfaction. In fact, they reported that surgical patients in units using a total patient care model with a larger proportion of RNs were more satisfied compared with a team nursing model which included fewer RNs (84.6 ± 13 vs ± 13 scores on the Parkside Patient Satisfaction Survey) (Kane et al., 2007). Baccalaureate-level nursing education was strongly associated with quality of care (Blegen & Vaughn, 2001). Similarly, in a study of medical units that had higher proportions of RNs with BSN degrees, patients expressed satisfaction with care 1.5 times more often (Minnick et al., 1997). In addition, an increase by one percent in the proportion of nurses with BSN degrees was associated with greater satisfaction by 13.6

53 45 ± 3.6 patient satisfaction scores (Seago et al., 2006). There is some evidence from a small number of observational studies that an increase in nurses with BSN degrees may reduce the risk of hospital-related mortality (Kane et al., 2007), but few studies were found that explore this effect on patient satisfaction in the emergency department. Chan and Chau (2005) examined the relationship between triage nurse characteristics and patient satisfaction in emergency departments. Although no statistically significant relationship was found between patient satisfaction and the educational level of triage nurses, patients reported a slightly higher level of satisfaction when triaged by a nurse who had completed an additional nursing course Skewness of Ratings Many researchers have commented about the positive skewness and lack of variability of most patient satisfaction ratings (Munro et al., 1994; O'Connell et al., 1999; Laschinger et al., 2011). In fact, these data characteristics create problems in examining statistical comparisons and relationships. O Connell et al. (1999) questioned the sensitivity of the instruments in measuring the concept of patient satisfaction, as it may throw doubt on the validity of the results. Furthermore, O Connell et al. (1999) commented that the limited variance and positively skewed distributions could be an indication of a lack of confidence among patients in their ability to judge nursing care activities, while patients who believe they are able to judge care may be reluctant to do so when hospitalized or needing emergency care. In these situations, patients are more likely to avoid criticizing or poorly rating the healthcare providers on whom they depend for survival (Williams, 1994). Patients also may not express dissatisfaction with nursing

54 46 care because this may be their first exposure to care, or they may be afraid of reprisal in the event they need to use the service in the future (Fitzpatrick, 1991) Patient Satisfaction Response Timing of the patient satisfaction survey have been found to affect both ratings and response rates (Lin, 1996). When patients are surveyed upon discharge, the return rates are higher than when surveyed several weeks following their discharge. It is suggested that less-satisfied patients are less likely to return the survey (Ley et al., 1976), but some researchers have argued that in fact it may be more-satisfied patients who are less likely to return the survey (Ware et al., 1983). More research is required to understand this issue. The format of the survey response can affect the patient satisfaction data. Researchers have found that the three- or five-point scales are preferred to an agree/disagree scale since the former will increase the variance of the score (Laschinger et al., 2011). In addition, when, where, and how patients are asked for their opinions have been found to influence both response rate and bias of responses (Bond & Thomas, 1992). A literature review of methodological issues in patient satisfaction concluded that interviews with patients were preferable to self-completion questionnaires (French, 1981), but this method can be more expensive Waiting Time People generally do not like to wait to see a clinician, and it is worse when they are anxious and uncomfortable (McMillan et al., 1986). The time spent by a patient waiting for health services can be psychologically painful because the patient has to

55 47 give up more productive and rewarding activities. Some researchers have reported that customer satisfaction is inversely proportional to waiting time (Davis & Vollmann, 1990). Davis and Vollmann (1990) conducted a study in restaurants where they observed that waiting time was correlated to self-report of satisfaction. The researchers found that the longer a customer waited, the less satisfied he or she became with the service. Furthermore, the researchers also noted that other variables may moderate the relationship, such as the customer's prior experience, their expected waiting time, the situational context, the time of day, the day of the week, and the importance of time to the customer. Some researchers argue that the perceptions regarding waiting times predict patient satisfaction but that actual waiting time does not (Thompson et al., 1996). Interestingly, patients in the ED who actually waited longer than expected were found to have significantly lower satisfaction scores than patients whose actual waiting times were the same or less than expected (Mowen et al., 1993). In other studies, long waiting times were not a significant predictor of patient satisfaction (Kurata et al., 1992; Monzon et al., 2005). Dansky and Miles (1997) examined the waiting times while in the waiting room in an ED, the waiting time in the treatment room, the waiting time to see the clinician, and the total time in an ED. Only the total time waiting to see the clinician was significant in the model of overall satisfaction with the urgent care department. This finding is similar to the finding of Sandovski et al. (2001) who reported a significant negative correlation between patient satisfaction and waiting time to examination by the ED physician. Dansky and Miles (1997) concluded that waiting time to see the clinician significantly predicted satisfaction with clinicians and therefore overall satisfaction in the

56 48 urgent care department, but it did not predict satisfaction with staff. Emergency patients with trauma or life-threatening injuries are more satisfied than urgent and non-urgent patients. Since it is more likely that non-urgent patients will not be treated more quickly than trauma patients, hospitals have improved the waiting experience by ensuring the waiting area is close to refreshments, providing magazines and privacy, and advising them of the estimated waiting times (McMillan et al., 1986) Other Factors Affecting Patient Satisfaction To isolate the relationship between nurse staffing and patient satisfaction, characteristics of the organization and environment should be included in the model as control variables. The selection of such control variables is based on evidence in the literature. These control variables represent alternative explanations of the nurse staffing and patient satisfaction relationship and are of two types internal organizational characteristics and environmental characteristics. In a study to assess the effects of four nursing strategies on the efficiency of patient care, the researchers used a number of control variables (Bloom et al., 1997). The four nursing strategies were: (1) the use of temporary nursing agencies; (2) the use of part-time nurses; (3) increased skill mix of the nursing staff (proportion of registered nurses); and (4) increased experience mix of the nursing staff. Characteristics of the organization and environment were control variables in the analyses. The six organizational control variables were: organizational size, ownership/control (e.g. church, for profit), teaching status, operating capacity (occupancy rate), and input complexity/uncertainty (length of stay). The seven environmental controls were: geographic region, urban/rural status,

57 49 regulatory intensity by state, local economic climate, hospital wage rates, hospital competition within a service area, and supply of nursing labor within the community. 2.9 Nurse Staffing Theoretical Frameworks In the late 1990s, research interests moved to examining nurse staffing in relation to patient outcomes (McGillis Hall, 2005). This shift emerged because there was a lack of evidence regarding the relationship between the quality of care and nurse staffing levels and staff mix as reported by the Institute of Medicine (IOM) Committee on the Adequacy of Nurse Staffing in Hospitals and Nursing Homes. The IOM Committee report discussed the lack of empirical evidence regarding the relationship between the quality of patient care, staffing levels, and staff mix. As more studies began to emerge, researchers developed theoretical frameworks to examine nurse staffing and outcomes that include patient satisfaction. Nurse staffing frameworks fall within outcomes research. Research in this area has been described using frameworks augmented from Donabedian s conceptual model for quality of care (Donabedian, 1966). For four decades, Donabedian's (1966) concepts have steered outcome research to evaluate and compare health care quality. Research on evaluating quality of care began with an emphasis on structures (having the right things), before shifting to processes (doing things right) and then to outcomes (having the right things happen) (Mitchell et al., 1998). Over the years, more variables have been added to Donabedian s (1966) conceptual model to examine nurse staffing and quality of care. Three established models used in the literature that adapted Donabedian s conceptual model are: Outcomes Model for Healthcare Research, Quality

58 50 of Care Dynamic Research Model, and the Nursing Role Effectiveness Model (McGillis Hall, 2005). Appendix B, C and D show these models. The Donabedian s (1966) framework implies a hierarchical analysis model. Patients are embedded in hospital units, which have both structural characteristics and processes. Furthermore, these units are within hospitals that have both structural characteristics and processes. Although studies using Donabedian s (1966) framework should use hierarchical analysis, only a few studies were published using multi-level analyses. Access to datasets that support hierarchical analysis is a major limiting factor. As discussed in previous sections of this review, nursing workforce characteristics have been studied, but few studies have included other characteristics, such as job satisfaction or turnover. Further research is recommended on work-related structure measures, which include organizational factors (such as measures of hospital commitment to quality and measures of hospital leadership). Based on these frameworks, nursing outcomes research uses data from three types of data sources. The first are large national data sets, such as hospital discharge abstracts, matched with nurse staffing data at a state or province level but not a unit level. Thus, nursing workforce measures cannot distinguish between nurses in direct patient care or those in administrative or outpatient services. Although a large number of patient outcomes are available, nursing workforce variables are limited. The second source, data from individual states/provinces, hospital surveys, or administrative data the California Nursing Outcomes Coalition Database, for example has unit-level data on both nursing workforce characteristics and patient outcomes. The third source is

59 51 data collected by researchers from convenience samples of hospitals to which they have access, but findings that are generalized from these convenience sample studies are questionable. Unfortunately, most nursing outcomes research uses cross-sectional data sets which do not allow trends or estimation of lagged effects. Understanding the trends or the lagged effects is crucial to understanding the relationship between nursing variables and patient outcomes. Using nursing theoretical frameworks can advance the knowledge of the relationship between nurse staffing attributes and patient outcomes by applying data sets that support hierarchical analyses; additional attributes of the nursing workforce; unit-level data; and large, representative, longitudinal data sets Summary Among the literature reviewed, there is consensus that patient satisfaction with nursing care is strongly associated with (and is an important predictor of) a patient s overall satisfaction with the hospital care (Johansson et al., 2002; Bolton et al., 2003; Larrabee et al., 2004; Al-Mailam, 2005; Chan & Chau, 2005; McGillis Hall, 2005; Kane et al., 2007). The quality of interpersonal relationships between nurses and patients has been shown to be a crucial aspect of nursing behaviour that influences patient satisfaction (Laschinger et al., 2011). Patients perceptions of nurses interpersonal and communication skills, friendliness, and ability to attend to the specific needs of the patient also have been found to be associated with higher satisfaction (Cleary et al., 1992; Larrabee et al., 2004). There is, however, a lack of generalizable studies

60 52 performed in the ED where the average length of stay is less than four hours that examine the relationship between patient satisfaction with nursing care and the nurse staffing models implemented.

61 53 Chapter 3 Methods And Procedures 3 Overview The study explores the relationship between nurse staffing and patient satisfaction in emergency departments using a modified Kane et al. (2007) conceptual frame. This chapter identifies the design, setting, data collection, and analysis plan of the study. 3.1 Study Design This study uses a retrospective, descriptive, correlational design to examine the relationship between ED nurse staffing and patient satisfaction. The ED study is considered to be a natural experiment in which nurse staffing varies with a resultant variation in patient satisfaction. The design to be used supports the aims of the study, the research questions and hypotheses, but the design does not permit attributing direct causal effect of nurse staffing on patient satisfaction. To do so, the requirements of causal relationships must be achieved. For causal relationships to be meaningful, there should be temporal precedence of the cause, covariation between the cause and effect, and other alternative explanations for effects should be eliminated (Cook & Campbell, 1979). Unfortunately, these conditions cannot be fully met in this study. First, decisions regarding emergency department staffing are normally done before a patient s experience is known.

62 54 Therefore, there is no precedence of cause. Second, covariation between nurse staffing in the emergency department and patient satisfaction is being assessed in this study. Studies in other areas of the hospital have shown the extent to which variables covaried may not be sufficient to meet the conditions of causal relationships (Blegen et al., 1998; Schmidt, 2003). Lastly, alternative explanations for the variation in patient satisfaction identified in the literature can be controlled in the multivariate analysis. Therefore, after assessing the conditions required to establish causal relationships, the extent to which the cause and effect varied is insufficient to establish direct causal effect of nurse staffing on patient satisfaction. For this reason, a predictive model was not pursued in this study; instead, a descriptive correlational design was used to address the research questions and hypotheses presented. The conceptual model, variables and data sources for the study are shown in Figure 3.1 and Table 3.1. Many nurse staffing variables were computed, but only six nurse staffing and characteristics variables were used in the multivariate analyses that passed multicollinearity and correlation tests as discussed in Chapter 4.

63 55 Figure 3. Conceptual Framework of Nurse Staffing and Patient Satisfaction Patient Characteristics Age Gender Hospital Organizational Characteristics Size (# of ED visits) Type of hospital (teaching, small community) ED wait (% of patients seen within recommended timeframe) ED Case mix index ED Cleanliness Nurse Staffing Intensity of Care: RN hours per visit, RPN hours per visit, Agency Nurse hours per visit, NP hours per visit, staff hours per visit, RN hours per length of stay, RPN hours per length of stay, Agency Nurse hours per length of stay, NP hours per length of stay, staff hours per length of stay. Skill Mix: RN skill mix, RPN skill mix, Agency nurse skill mix, NP skill mix. Staff Adequacy: RN/Patient staffing ratio, RPN/Patient staffing ratio, Agency Nurse/Patient staffing ratio, NP/Patient staffing ratio, Staff/Patient staffing ratio. Nurse Characteristics Age Education Level Nurse Experience (yrs) Full time/part time, employment mix Emergency Department Care Physician Characteristics Physician Courtesy Patient Satisfaction Outcomes Overall Satisfaction - Overall, how would you rate the care you received in the Emergency Department? Would you recommend this emergency department to your family and friends? Nursing Care When you had important questions to ask a nurse, did you get answers you could understand? (Answer) If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? (Explain) Did you have confidence and trust in the nurses treating you? (Trust) Did nurses talk in front of you as if you weren t there? (Respect) How would you rate the courtesy of your nurses? (Courtesy) How would you rate the availability of your nurses? (Availability) How would you rate how well the doctors and nurses worked together? (Dr-Nurse working relationship)

64 56 Table 3-1. Definition of Terms Hospital Organizational Characteristics Variable Indicator Name Data Source Size (# of ED visits) Number of ED visits SIZE OHRS Type of Hospital Teaching, Community, Small PEERGRP OCDM Type of ED 24-hr ED, Urgent Care Centre, Trauma Centre EDTYPE OHRS Severity Adjustment ED Case Mix Index CMI OCDM ED Wait Times ED Cleanliness Patient Characteristics Proportion seen with the recommended timeframe (number of CTAS I & II patients seen within 8hrs + Number of CTAS III patients seen within 6hrs + Number of CTAS IV & V patients seen within 4hrs / total number of patient visits) Was the entire Emergency Department as clean as it should have been? Three point Likert-type scale: Yes Definitely, Yes Somewhat, No EDWAIT EDCLEAN NACRS Patient Sat. Age Age (in years) PATAGEGRP Patient Sat. Gender Gender PATGENDER Patient Sat. Nurse Characteristics Nurse Staff Characteristics Nurse Staffing Nursing Intensity of Care Skill Mix Age (in years) Education Level (Diploma, BSN & higher) Nursing Experience (years in nursing = years after graduation from initial nursing program) Employment status: Percent full-time (full-time RN & RN earned hours) divided by total nursing earned hours Gender (Percent Female Nurses) NURSEAGE NURSEED NURSEEXP PERFTHRS PERFEMNURSE CIHI Nursing Database CIHI Nursing Database CIHI Nursing Database OHRS CIHI Nursing Database RN worked hours per patient visit RNWKHRS OHRS, NACRS RPN worked hours per patient visit RPNWKHRS OHRS, NACRS Agency Nurse worked hours per patient visit AGNWKHRS OHRS, NACRS Nurse Practitioner worked hours per patient visit NPWKHRS OHRS, NACRS staff worked hours per patient visit TOTSTAFFWKHRS OHRS, NACRS RN worked hours per length of stay (Annual RN worked hours divided by Annual patient length of stay) RNPLOS OHRS, NACRS RPN worked hours per length of stay (Annual RPN worked hours divided by Annual patient length of stay) RPNPLOS OHRS, NACRS Agency Nurse worked hours per length of stay (Annual Agency Nurse worked hours divided by Annual patient length of stay) Nurse Practitioner worked hours per length of stay (Annual NP worked hours divided by Annual patient length of stay) worked hours per length of stay (Annual worked hours divided by Annual patient length of stay) RN proportion (RN worked hours divided by total staff worked hours) RPN proportion (RPN worked hours divided by total staff worked hours) Agency proportion (Agency Nurse worked hours divided by total staff worked hours) Nurse Practitioner Proportion (Nurse Practitioner worked hours divided by total staff worked hours) AGNPLOS NPPLOS TOTSTAFFPLOS RNPROP RPNPROP AGNPROP NPPROP OHRS, NACRS OHRS, NACRS OHRS, NACRS OHRS OHRS OHRS OHRS

65 57 Staff Adequacy Physician Characteristics ED Doctor Courtesy RN Staff to Patient Ratio (number of RN staff / number of patients) RPN Staff to Patient Ratio (number of RPN staff / number of patients) Agency Nurse Staff to Patient Ratio (number of RPN staff / number of patients) NP Staff to Patient Ratio (number of NP staff / number of patients) Staff to Patient Ratio (total number of patient care staff / number of patients) How would you rate the courtesy of your doctors? Five point Likert-type scale: Poor, Fair, Good, Very Good and Excellent RNRATIO RPNRATIO AGNRATIO NPRATIO TOTSTAFFRATIO DRCOURTESY OHRS, NACRS OHRS, NACRS OHRS, NACRS OHRS, NACRS OHRS, NACRS Patient Sat. Patient Satisfaction with Nursing Care Variable Indicator Name When you had important questions to ask a nurse, did you get answers you could understand? If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? Did you have confidence and trust in the nurses treating you? Did nurses talk in front of you as if you weren t there? How would rate the courtesy of your nurses? How would you rate the availability of your nurses? How would you rate how well the doctors and nurses worked together? Three point Likert-type scale: Yes Always, Yes Sometimes, No, Did not have any questions Three point Likert-type scale: Yes Completely, Yes Sometimes, No, Did not have any anxieties or fears Three point Likert-type scale: Yes Always, Yes Sometimes, No Three point Likert-type scale: Yes Often, Yes Sometimes, No Five point Likert-type scale: Poor, Fair, Good, Very Good and Excellent Five point Likert-type scale: Poor, Fair, Good, Very Good and Excellent Five point Likert-type scale: Poor, Fair, Good, Very Good and Excellent ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK Patient Sat. Patient Sat. Patient Sat. Patient Sat. Patient Sat. Patient Sat. Patient Sat. Overall Patient Satisfaction Variable Indicator Name Overall, how would you rate the care you received in the ED? Would you recommend this ED to family and friends? Five point Likert-type scale: Poor, Fair, Good, Very Good and Excellent Five point Likert-type scale: Poor, Fair, Good, Very Good and Excellent EDSAT EDREC Patient Sat. Patient Sat.

66 Sample This study was conducted on EDs in Ontario s acute care hospitals. The study period was selected on the basis of the availability and quality of the data. Routinely collected administrative and patient satisfaction data for the five-year period of 2005/06 to 2009/10 were analyzed. The patient satisfaction sample consists of the 182,022 patients who were discharged from Ontario s emergency departments during the fiveyear study period and completed a patient satisfaction survey. Table 3.2 shows the type and number of hospitals with an emergency department included in the study. Some hospital corporations have multiple sites with emergency departments, and the table highlights the number of hospital corporations with an ED. As a result, the totals presented do not indicate the actual number of EDs because of the multisite issue. Furthermore, the total number of corporations varied across the five-year period of the study mainly due to hospital restructuring, mergers, and some hospitals not participating in the survey. In addition, some hospitals did not survey patients for one or two years during the study period. 107 hospital corporations reported patient satisfaction data for at least 1 year. Table 3-2. Emergency Department by Hospital Type Year 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE Peer Group Large Community Small Teaching

67 Power Analysis The number of independent variables or predictors, effect size, power, and level of significance (alpha) are important when determining the sample size and power of the study (Cohen & Cohen, 2002). A large effect size was determined to be appropriate since no known studies have examined the relationship between nurse staffing and patient satisfaction in emergency departments. According to Cohen and Cohen (2002), for multiple and multiple correlation tests, the f 2 = R 2 / (1-R 2 ) = 0.35 for a large effect size that explains 26% of the variance. There were 6 nurse staffing variables used in the regression models as discussed in Chapter 4. Since each model uses 15 independent and control variables, the sample size required for a large effect size, power of 0.80, and a 0.05 level of significance was determined using tables reported by Cohen (2002) and the following equation for sample size. n = L / f 2 + k + 1 (where n = sample size, k = number of independent variables, L = value in tables) The L value from Cohen and Cohen (2002) tables is for power of 0.8, 0.05 level of significance and 15 variables. Thus, a sample size of 70 EDs was determined to provide sufficient power to conduct this study. 3.4 Data Collection The methods and procedures to collect and manage data used for the analyses are discussed in this section. The discussion begins with the data collected at the

68 60 patient level for patient satisfaction scores, followed by the data for nurse staffing and control variables collected for each hospital ED Patient Satisfaction The patient satisfaction data in this study was obtained from the patient satisfaction survey data collected in most hospitals in Ontario using the NRC-Picker Canada Emergency Department Care Patient Satisfaction survey. The survey was developed by the Picker Institute and consists of 59 questions related to satisfaction in four categories: a) admission and discharge processes; b) doctors and medical care; c) nurses and nursing care; and d) the emergency department environment. In 2003, the NRC-Picker survey instrument was introduced in Ontario s EDs, replacing the Standardized Hospital Patient Satisfaction Survey (SHoPSS). Every year, over 100,000 individuals from approximately 100 participating hospitals in Ontario are sampled using the survey from NRC-Picker. Patients discharged between April 1, 2005 and March 31, 2010 were included in the sample. Approximately 30% of the sampled individuals returned their questionnaires every year. The sampling plan was developed collaboratively by each participating hospital corporation and NRC-Picker Canada (see Appendix E). Deciding factors influencing the agreed-upon sampling plan included budget, achieving reasonable response rates, and which sites within the corporation were of primary interest. Hospitals were then charged with the responsibility of sending patient data files to NRC-Picker every month. According to each hospital s sampling plan, a random sample was drawn from the patient data files, and surveys were mailed to the selected patients. Questionnaires were not sent to deceased patients, psychiatric patients, infants less than 10 days old,

69 61 patients with no fixed address, or patients who presented with sexual assault or other sensitive issues. Surveys that were returned without a single valid response were treated as nonresponses and dropped from the NRC-Picker. If a record had no valid response to any of the evaluative questions on the questionnaire (i.e. it only had responses to demographic-type questions), then it was considered as having insufficient data and was excluded from the subsequent analysis Emergency department variables The data for variables at the emergency department level were collected from four sources: a) the Ontario Healthcare Reporting Standards (OHRS); b) Canadian Institute for Health Information (CIHI) Nursing Database; c) the Ontario Cost Distribution Methodology (OCDM); and d) the National Ambulatory Care Reporting System (NACRS). a) Nurse Staffing Data (OHRS) All variables requiring measurement of staffing hours were calculated using data from the Ontario Healthcare Reporting Standards (OHRS). The Ontario Ministry of Health and Long-Term Care require hospitals to submit financial and statistical data annually in an electronic format using a coding structure outlined in the OHRS. To ensure the data submitted are accurate, the MOHLTC applies various edit checks and provides hospitals with verification reports so administrators can concur with the data received by the MOHLTC. Using the OHRS database, nurse staffing data for emergency departments were analyzed for the period of April 2005 to March 2010

70 62 (fiscal 2005/06 to 2009/10) for the hospitals that participated in the patient satisfaction survey. The major limitation of the OHRS is that the data can only be reported at the corporation level, so variables for hospital corporations with multiple emergency departments were computed at an aggregated level since individual site-specific values were not submitted to the MOHLTC. OHRS statistical data for the emergency department include hours worked by employment category and employment status. Staffing hours are recorded for the following categories: worked, purchased service hours, and benefit hours. These hours are recorded for the three broad categories of staff, which include management and operational support, unit-producing staff, and medical staff. The medical staffing hours were reported for only salaried ED physicians, thus the hours for fee-for-service paid physicians are not reported. In 2005, the OHRS included additional reporting requirements such as employment categories for RN, RPNs, NP, and agency nurses for staffing information. Appendix F shows the OHRS reporting framework for the staff information since The changes in the standards allowed for the calculation of the nurse staffing variables in this study. The technical specifications of the study variables drawn from the OHRS database are shown in Appendix G. In this study, worked hours for staff were calculated by combining the reported worked and purchased service hours. Worked hours do not include hours for vacation, sick time, education, orientation, and holidays. The total worked hours included only the unit producing personnel (UPP) for the emergency department. Worked hours for registered nurses (RNs), registered practical nurse (RPNs), nurse managers, clinical nurse specialists, nurse educators, and nurse practitioners who function as nurses were included in the calculation of nurse worked hours.

71 63 The nursing care hours per visit by nursing category (RN, RPN, NP, and agency nurse) were calculated by dividing the nursing worked hours for each nurse category by the number of emergency department visits in a hospital for the same time period. The worked hours were obtained from the OHRS data, while the patient activity was obtained from the National Ambulatory Care Reporting System (NACRS) database. The nursing hours per length of stay variable was calculated using nursing worked hours by the different staff category for the numerator divided by the total length of stay of patients seen for the ED in the year under consideration. The length of stay for each patient was calculated using the times reported in NACRS as the difference of the visit complete time minus the registration time or triage time depending (whichever is reported first). Each patient s length of stay was summed to compute the total length of stay for the ED in a hospital for a fiscal year. RN skill mix was calculated by dividing the total RN worked hours by the total nursing care worked hours for the same time period. Similarly, skill mix for each staff category (RPN, NP, and agency nurse) was calculated by dividing the total RPN, NP, or agency nurse worked hours by the total nursing care worked hours for the same time period. Similarly, the percentage of full-time employment was calculated by dividing the full-time worked hours for nursing care staff by the total nursing care worked hours for the ED in a hospital for a fiscal year. The nursing staff-to-patient ratio was calculated for each nurse staff category. The number of nursing staff was calculated as the total earned hours divided by The number of patients was obtained from the NACRS database.

72 64 b) Nurse Staffing Data (CIHI Regulated Nursing Professions Database) In this study, the education, experience level, and demographic information of the nursing staff were obtained from the Canadian Regulated Nursing Professions Database. This database, maintained by the Canadian Institute of Health Information (CIHI), includes data for the three regulated nursing professions in Canada: Licensed Practical Nurses, Registered Nurses, and Registered Psychiatric Nurses. In this study, the licensed practical nurses are considered registered practical nurses. The database contains demographic, education, and employment information on nursing staff. The data are collected under the terms of agreements with the Ontario College of Registered Nurses and the Ontario College of Registered Practical Nurses. These provincial regulatory authorities are responsible for data collection, which occurs during the annual registration process. The data are manually entered and a file is submitted to CIHI in a standardized format. Data are received and organized by CIHI, and calculated variables are created by combining data elements. For example, birth year is subtracted from the data year in order to create the age variable. Average age, number of years since graduation, percent of nursing staff that are females, percent of nursing staff with a Bachelor in Nursing or higher were calculated using the Regulated Nursing Professions database. Nursing staff characteristics data included both RNs and RPNs. c) Severity Adjustment from the OCDM Many nursing studies have identified the need to take the severity of the illness of patients into consideration (Blegen et al., 1998; Blegen & Vaughn, 1998). Ontario s Ministry of Health and Long-Term Care use the emergency department case mix index (CMI) developed by CIHI in their funding formulas to adjust the cost per patient visit for

73 65 patient severity. In this study, the emergency department CMIs for each hospital was used as a severity adjustment measure and the ED CMI variable was treated as a continuous level variable similar to other studies (Freeman et al., 1995; Rutledge et al., 1996). CIHI calculates the CMIs for every hospital corporation with an ED, and MOHLTC reports the CMI in every hospital s Ontario Cost Distribution Methodology (OCDM). The CMI was obtained for each ED in the study for each of the five years. The CMIs used the latest 2010 Comprehensive Ambulatory Care System (CACS) grouper and CMI weights for all years. d) NACRS Percent of ED patients seen within a recommended timeframe was calculated using CIHI s NACRS (see Appendix H for details on this database). As discussed earlier, the length of a patient's stay in the emergency department is defined as the time from registration or triage (whichever comes first) to the time the physician makes a decision to either admit or discharge (UHN, 2011). In some overcrowded emergency departments, patients can have long waits for an inpatient bed. For admitted patients, the time spent waiting in the emergency department for an inpatient bed was not included in the length of stay calculation. Ontario s Ministry of Health and Long-Term Care has established times for patients length of stay that are part of each hospital s accountability agreement. Nonadmitted patients triaged as level I and II in the Canadian Triage Acuity Scale (CTAS) must be cared for within eight hours. Similarly, non-admitted patients triaged as CTAS level III should have a length of stay less than six hours, while patients triaged as nonurgent, level IV and V should have a length of stay less than four hours (UHN, 2011).

74 66 The proportion of patients seen within the recommended timeframes for each emergency department was obtained by the Ministry of Health and Long-Term Care. The number of ED visits was used as a measure of the ED size, and the number of ED visits was measured in both the NACRS and OHRS datasets. CIHI performs data quality checks in the NACRS dataset (using a combination of data elements to identify abstracts that could be duplicates) to prevent over-coverage of patients (CIHI, 2011). In addition, the MOHLTC has mandated all EDs to submit abstracts to CIHI, so there should be minimum under-reporting. In their annual report on the data quality of NACRS, CIHI has reported no issues with under-coverage (CIHI, 2011). In 2007, the Joint Policy and Planning Committee (JPPC) reported concerns about the integrity and completeness of the OHRS ED visits. The JPPC recommended that the OHRS ED visits not be used in the 2006/07 funding allocation to hospitals because of the data integrity issues (JPPC, 2006). The number of ED visits from the NACRS dataset is being used by the MOHLTC in the hospital funding calculations and hospitals accountability agreements (MOHLTC, 2011). Due to the data quality issue in the reporting on the number of ED visits in the OHRS, however, and in order to be consistent with the methodologies used by the MOHLTC in funding allocations, the number of ED visits for this study was calculated from the NACRS database. e) Control Variables To isolate the relationship between ED satisfaction and staffing patterns, organizational characteristics are included in the model as control variables. These control variables represent competing or alternative explanations of the staffing and patient satisfaction relationship. The organizational control variables included in the

75 67 multivariate model were hospital type (teaching, large community, or small hospital), emergency department waits or proportion of patients seen within the recommended timeframe, size of ED (number of visits), ED case mix index, cleanliness of the ED, and physician courtesy. Some hospital corporations had trauma centres and urgent care centres. Another classification variable EDTYPE was also investigated. EDs were grouped into General ED, Urgent Care Center (UCC), and Trauma. Unfortunately, the data for the nurse staffing, nurse characteristics, and covariate variables were only available at the hospital corporation level. The UCCs and trauma centres were in multisite organizations and the nurse staffing data could not be separated from the general ED. Thus, EDTYPE was dropped as an explanatory variable in the study. 3.5 Data Access Data for this study was obtained through the graduate student data access programs at CIHI, Ontario MOHLTC and the Ontario Hospital Association. 3.6 Data Analysis This section discusses the research question and hypotheses, as well as the procedures and methods used to describe and analyze the data. The formation and descriptive analysis of the datasets will be discussed. The study use SAS 9.2 and SPSS 14.0 software for the various analyses.

76 Patient Satisfaction Dataset The patient satisfaction with nursing care variables include the following questions: 1. When you had important questions to ask a nurse, did you get answers you could understand? (Answer) 2. If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? (Explain) 3. Did you have confidence and trust in the nurses treating you? (Trust) 4. Did nurses talk in front of you as if you weren t there? (Respect) 5. How would you rate the courtesy of your nurses? (Courtesy) 6. How would you rate the availability of your nurses? (Availability) 7. How would you rate how well the doctors and nurses worked together? (Drnursewk) A variety of response scales are used in the NRC-Picker patient satisfaction questionnaire. Some questions employed a Likert-type scale with five response choices: Poor, Fair, Good, Very Good, and Excellent. To make interpretation of the patient satisfaction easier, the variable five-point scale was converted into a 100- point scale according to the procedure used by a number of studies (Hall & Press, 1996; Boudreaux et al., 2003; CIHI, 2008). The following scores were assigned: Poor = 0, Fair = 25, Good = 50, Very Good = 75, Excellent = 100. Other survey questions used a three point scale with responses: Yes Always, Yes Sometimes, No. These responses were assigned the following scores: Yes Always = 100, Yes Sometimes = 50, No = 0. A few questions had a viable selection; for example, the question When you had important questions to ask a nurse, did you get answers you could understand? had the response option Did not have any questions. This was an acceptable response to the question, but it was not assigned a score. The data was first checked for any missing or improbable values. For the seven patient satisfaction with nursing care variables (ANSWER, EXPLAIN, TRUST,

77 69 RESPECT, COURTESY, AVAILABILITY and DRNURSEWK), each variable was checked for missing values. The missing values for a particular nursing care variable were imputed using a linear regression, with the other nursing variables as predictors and controlling for age and gender. The patient satisfaction variables were assessed for normality, outliers, and linearity. Using Pearson s index of skewness, variables that were significantly skewed were transformed. Principal Component Analysis (PCA), which identifies variables to maximize the internal consistency, was used to ensure that the subset of seven nursing variables in the NRC-Picker survey represented the set of nursing variables in a way that sufficiently represented the overall variation in patient satisfaction with nursing care in the emergency department. The PCA analysis procedure described by Jolliffe (2002) was used. First, the inter-item correlations are analyzed; according to the PCA procedure, variables with correlation less than 0.3 and greater than 0.7 are considered for deletion (Jolliffe, 2002). The second step is the analysis of the factor loadings. In this step, variables with loading of less than 0.4 were considered for deletion. The third step is the analysis of eigenvalues and percentage of variance explained by each factor. The fourth step is the visual inspection of the scree plots. Finally, the fifth step is the analysis of impact of individual item deletion. Corrected item-total correlations of less than 0.3 were considered for item deletion. The impact on Cronbach s alpha on a deleted item was assessed to see if there was any increase in alpha for any of the variables.

78 Emergency Department Level Dataset The emergency department data analysis used data from five datasets: 1) patient satisfaction; 2) patient characteristics; 3) nurse staffing; 4) nursing staff characteristics; and 5) emergency department characteristics. 1. Patient Satisfaction dataset The patient satisfaction data consists of the patient level survey results. The dataset had a hospital corporation number that was used to link it to the other datasets. 2. Patient Characteristics The patient characteristics data were extracted from the NRC-Picker patient satisfaction data for age and gender. These two variables were part of the patient satisfaction dataset. 3. Staffing dataset The staffing dataset was created with the specific fields: nursing worked hours per visit by staff category, nursing worked hours per length of stay by staff category, skill mix by staff category, and the ratio of nursing staff to patients. Each variable was reported by fiscal year for each hospital corporation. 4. Nursing staff characteristics The data for the nursing staff characteristics was obtained from the CIHI Nursing Database. The data was sorted by postal code and linked to the postal code of each emergency department. Each ED had its unique postal code which facilitated the data linkage. The data was aggregated for multi-site hospital corporations. The fields in the dataset included nurses age and gender, educational level, emergency department

79 71 experience, and full-time/part-time status. The full-time /part-time status was calculated from the OHRS data. 5. Emergency department characteristics dataset The data for the emergency department characteristics were obtained from the OCDM, OHRS and NACRS. The variables included ED case mix, hospital type, and proportion of patients seen on time. Satisfaction with cleanliness of the ED, and physician courtesy were obtained from the NRC-Picker survey database. Finally, a merged dataset was created at the hospital corporation level in SAS using the hospital corporation OHRS facility number. This dataset was reviewed for any missing data Research Question This study seeks to determine to what extent specific aspects of nurse staffing relate to: 1) patient satisfaction with nursing care; 2) overall satisfaction with care received in the ED; and 3) whether the patient would recommend this ED to friends and family. The study draws on existing administrative and patient satisfaction survey data from Ontario s EDs to test the following hypotheses: Hypothesis 1: There is a positive relationship between RN proportion, nurse-to-patient ratio, nursing hours per patient visit and each patient satisfaction with nursing care variable (i.e., Answer, Explain, Trust, Respect, Courtesy, Availability, and Dr-Nurse working relationship).

80 72 Hypothesis 2: There is a positive relationship between RN proportion, nurse-to-patient ratio, RN hours per patient visit and overall satisfaction with care received in the ED. Hypothesis 3: There is a positive relationship between RN proportion, nurse-to-patient ratio, RN hours per patient visit and whether the patient would recommend the ED to friends and family Data Analysis The unit of analysis for the study is the emergency department and the merged dataset with staffing, patient satisfaction, patient characteristics, and nursing staff characteristics was used for analysis. A correlation matrix was constructed for all variables, and values with p-values were considered statistically significant if less than or equal to.05. The correlation matrix included variables for intensity of care, skill mix, staff adequacy, and the nine patient satisfaction variables (the seven patient satisfaction with nursing care variables and the two patient satisfaction outcomes variables). Descriptive statistics were calculated for each variable. Frequencies were obtained for the patient characteristics. Analysis of variance (ANOVA) was used to test if there are any differences, at the 0.05 level, for age between the years to be studied, and cross-tabs with Chi-Squared statistics for gender were calculated. If differences were found, these variables were considered for inclusion in the multivariate regression analysis. Similarly, nursing staff characteristics of educational level, nursing and emergency department experience, employment status, and emergency department characteristics were assessed for differences between the years. Any significant

81 73 differences were considered for inclusion as confounding variables based on differences between years. The patient satisfaction outcome variables were reported in a five-point Likert scale. Since this data is ordinal, ordered multinomial logistic models were considered for the study. The reliability of ordered multinomial regression models, however, is contingent upon the assumptions that there is sufficient data in each category of the dependent variable. As discussed in chapter 4, the patient satisfaction data was negatively skewed in this study and not enough data fell into some of the categories of the dependent variable. In addition, interpreting the cut points can be challenging using logistic models. The ordinal data in this study was therefore treated as interval-scaled data, a common practice among social science researchers (Hoelzle, 2011). For this reason the patient satisfaction variable five-point scale was converted into a 100-point scale according to the procedure used by a number of studies (Hall & Press, 1996; Boudreaux et al., 2003; CIHI, 2008). The data for the multivariate analyses were structured into a linear mixed model in which repeated measurements of patient satisfaction rating over time were nested in emergency departments. To take into account that observations in a hospital may tend to be correlated, multilevel models were used since the correlation of observations in the same cluster violates the assumptions of traditional linear regression (Cohen & Cohen, 1983; Tabachnick & Fidell, 1996; Cohen & Cohen, 2002). Multilevel models incorporate random components of cluster effects in the statistical model. In addition, multilevel models address the under specification of the traditional linear regression, which can lead to underestimation of standard errors and an increased likelihood of reporting statistics as statistically significant. Unequal sample sizes within clusters and

82 74 missing data also can be analyzed using multilevel models with repeated measures data (Tabachnick & Fidell, 1996). In the original plan, nine multivariate linear mixed regression models were expected to be developed to assess the relationship between nurse staffing and a) patient satisfaction with nursing care (i.e., Answer, Explain, Trust, Respect, Courtesy, Availability and Dr-Nurse working relationship), b) overall satisfaction in the emergency department, and c) recommending this emergency department to friends and family. After the PCA, however, the patient satisfaction with nursing care (Aggregate Score) was used as the main nursing care satisfaction variable, which is the average of the six nursing care variables (Answer, Explain, Trust, Courtesy, Availability, and Dr-Nurse Working Relationship) since the factor loading were above 0.8. Thus, three hierarchical regressions were developed in the study to answer the research question. The seven regressions with the dependent variables being the patient satisfaction with nursing care variables (i.e. Answer, Explain, Trust, Respect, Courtesy, Availability and Dr-Nurse working relationship) were also developed to further understand the relationship between patient satisfaction with nursing care and nurse staffing. The results of these seven regressions are shown in the Appendix O, but this study focused only on the three regression models with the dependent variable being patient satisfaction with nursing care (Aggregate Score), overall satisfaction with care, and recommending the ED to friends and family. Thus, original hypothesis 1 was not tested. The three patient satisfaction outcomes or dependent variables can be expressed using a pair of linked models: one at the patient level (level-1) and another at the ED-level (level-2). For the model analysis, X = independent variable (Intensity of Care, Skill Mix and Staffing Adequacy) and Y = dependent variable.

83 75 Assumptions of multivariate linear mixed regression procedures were tested using SAS. Variables were assessed first to ensure they did not violate the assumptions of normality, outliers, heteroscedacity, autocorrelation, and multicollinearity. Regression procedures assume that variables have normal distributions, since non-normally distributed variables can distort relationships and significance tests. Normality was assessed visually by constructing and reviewing histogram distributions as well as a Kolmogorov-Smirnov test of normality was performed on the patient satisfaction variables. For the variables calculated at the EDlevel, normality was observed using the Shapiro-Wilk test. If skewness values were greater than 1 for any nursing and control variables, the variable was considered for deletion. Kurtosis of the distribution, which is a measure for peakedness, was assessed to identify distributions with long or short tails. If a distribution had kurtosis values significantly different than zero, the tails are longer or shorter than a normal distribution. Extreme values for skewness and kurtosis are values greater than +3 or less than -3. Outliers were identified by visual inspection of the data using histograms, frequency distributions, and box-plots for each year, in addition to identifying any data above or below the mean by 1.5 times the interquartile range (Larson, 2006). Outliers were recoded to a value calculated as the mean plus or minus 1.5 times the interquartile range for that variable. Linearity was assessed using bivariate scatterplots of the dependent variables against the independent variables. Multicollinearity, a high correlation between two or more independent variables, can lead to failure of significance of the regression coefficients and failure of the model to converge (Cohen & Cohen, 1983; Tabachnick & Fidell, 1996). The Variance Inflation Factor (VIF) was used to quantify the severity of multicollinearity. Multicollinearity

84 76 between independent variables was assessed by regressing each independent variable on each other. If VIF values exceed 4, multicollinearity is high. The VIF of 10 was used as a cutoff with any VIF value greater than 10 considered to multicollinear. If multicollinearity between two variables was identified, one of the variables was removed. The Lasso, or least absolute shrinkage and selection operator technique, was used to reduce the number of nurse staffing and control variables. LASSO is a shrinkage and selection method for linear regression that minimizes the usual sum of squared errors with a bound on the sum of the absolute values of the coefficients (Tibshirani, 1996). In a regression, the ordinary least squares estimates are obtained by minimizing the residual squared error but these estimates often have low bias and large variance. Prediction accuracy can be improved by shrinking or setting to 0 some coefficients. At the expense of a little bias, the variance of the predicted values is reduced and hence the overall prediction accuracy may improve. In addition, the interpretation is better since with a large number of predictors, having a smaller subset that exhibits the strongest effects is desirable. Tibshirani (1996) stated that the two standard techniques for improving the regression estimates are subset selection and ridge regression, both of which have drawbacks. Subset selection provides interpretable models but can be extremely variable so small changes in the data can result in very different models being selected. Ridge regression shrinks coefficients and hence is more stable however it does not set any coefficients to 0 and hence does not give an easily interpretable model. Lasso shrinks some coefficients and sets others to 0 and hence tries to retain the good features of both subset selection and ridge regression. For these reasons, the Lasso technique was used.

85 77 In the original plan, the hypotheses of the study included three nursing staffing variables: RN proportion, nurse-to-patient ratio, and nursing hours per patient visit. This plan to use these three nurse staffing variables in the regressions was altered after the LASSO, and correlation and multicollinearity checks of the nursing variables were done. The nursing variables obtained after the LASSO procedure that were not multicollinearity were used as the independent variables in the three regression models. For the confirmed set of variables, three hierarchical regression models were constructed. For each regression model, the following were produced for evaluation: regression coefficients, standard error of the estimate, analysis of variance table, predicted values, and residuals (Cohen & Cohen, 2002). When regression models were constructed, normality within the regression analysis was examined by looking at scatterplots of the standardized residuals against the predicted values of the independent variables. If the data (and the residuals) are normally distributed, the residuals scatterplot will show the majority of residuals at the center of the plot for each value of the predicted score, with some residuals trailing off symmetrically from the center. Heteroscedacity was assessed by plotting the standardized and studentized residuals against the predicted values of the independent variables. The Durbin- Watson test was used to assess autocorrelation of the residuals. A Durbin-Watson statistic substantially less than 2 (i.e. 0 to 1) or greater than 2 was used to identify autocorrelation. For the three hierarchical regression models, the magnitude and the level of significance of the estimates were observed. The effect of the nurse staffing on patient satisfaction was investigated by calculating the predicted satisfaction for an ED. Using the median values for the regression covariates, the predicted satisfaction scores were

86 78 calculated using the estimates from the regression models. In this analysis, a typical ED was simulated using the median values, and the ED patient satisfaction score was predicted using the actual nurse staffing variables data. The variation of the predicted patient satisfaction scores for the EDs in the study was calculated to investigate the effect of the nurse staffing on patient satisfaction in a typical ED. In addition, a similar analysis was performed using the median values for the nursing variables in the study and the actual covariate data for an ED to predict the patient satisfaction score for that ED with typical nurse staffing. The variation of the predicted satisfaction for the EDs in the second analysis was used to investigate the effect of the covariates, such as EDWAIT, CMI and DRCOURTESY, in an ED with a typical nurse staffing.

87 79 Chapter 4 Results 4 Overview This chapter presents a description of the patient satisfaction data and emergency department level data, which include both nursing staffing and organizational data, followed by the results of the multivariate analyses. 4.1 Patient Satisfaction In this study, the patient satisfaction data was the only dataset with patient-level records. Data related to patient gender and age group are presented in Table 4.1 and Table 4.2. No missing values were reported for these two variables. Over the five-year period 2005/2006 to 2009/2010, 55.2% (100,532) of the 182,022 ED patients surveyed were female. The patients were categorized into six age groups, and 29.2% were elderly patients aged 65 years and older. Elderly patients were the most predominant age group of ED patients surveyed. Table 4-1. Patients Surveyed by Gender Patient Gender Female Male Count % within Year Count % within Year Count % within Year Year 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 18,988 19,256 19,207 19,232 23, , % 54.8% 54.9% 56.2% 55.7% 55.2% 15,820 15,895 15,771 15,011 18,993 81, % 45.2% 45.1% 43.8% 44.3% 44.8% 34,808 35,151 34,978 34,243 42, , % 100.0% 100.0% 100.0% 100.0% 100.0%

88 80 Table 4-2. Patients Surveyed by Age Group Patient Age Group Under and over Count % within Year Count % within Year Count % within Year Count % within Year Count % within Year Count % within Year Count % within Year Year 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 6,353 6,194 5,631 5,593 6,653 30, % 17.6% 16.1% 16.3% 15.5% 16.7% 5,068 5,175 4,654 4,525 5,429 24, % 14.7% 13.3% 13.2% 12.7% 13.7% 3,945 3,961 3,606 3,293 3,820 18, % 11.3% 10.3% 9.6% 8.9% 10.2% 4,966 5,054 5,025 4,932 6,245 26, % 14.4% 14.4% 14.4% 14.6% 14.4% 5,154 5,285 5,608 5,500 7,159 28, % 15.0% 16.0% 16.1% 16.7% 15.8% 9,322 9,482 10,454 10,400 13,536 53, % 27.0% 29.9% 30.4% 31.6% 29.2% 34,808 35,151 34,978 34,243 42, , % 100.0% 100.0% 100.0% 100.0% 100.0% The descriptive statistics for the patient satisfaction variables are shown in Table 4.3. Many patients did not respond to the following questions: i. When you had important questions to ask a nurse, did you get answers you could understand? (Answer); and ii. If you had any anxieties or fears about your condition or treatment, did a nurse discuss them with you? (Explain). The percentages of patients who did not respond (missing cases) to these questions were 27.39% for ANSWER and 46.97% for EXPLAIN over the five-year period of the study. Missing values for these variables were imputed using information from the other patient satisfaction with nursing care variables (i.e. TRUST, RESPECT, COURTESY, AVAILABILITY and DRNURSEWK) and a generalized linear model with the patient age and sex as control variables.

89 81 Table 4-3. Patient Satisfaction Variables over the study period Year ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC 2005/2006YE N 33,303 33,074 33,515 33,619 33,669 33,658 33,389 33,987 33,926 % Response 95.68% 95.02% 96.29% 96.58% 96.73% 96.70% 95.92% 97.64% 97.47% Mean Std. Deviation /2007YE N 33,599 33,343 33,861 33,927 34,059 33,963 33,729 34,254 34,191 % Response 95.58% 94.86% 96.33% 96.52% 96.89% 96.62% 95.95% 97.45% 97.27% Mean Std. Deviation /2008YE N 33,307 32,957 33,542 33,673 33,757 33,709 33,465 34,043 33,968 % Response 95.22% 94.22% 95.89% 96.27% 96.51% 96.37% 95.67% 97.33% 97.11% Mean Std. Deviation /2009YE N 32,642 32,345 32,844 32,990 33,082 32,987 32,798 33,403 33,270 % Response 95.32% 94.46% 95.91% 96.34% 96.61% 96.33% 95.78% 97.55% 97.16% Mean Std. Deviation /2010YE N 40,823 40,416 41,067 41,183 41,276 41,261 41,019 41,748 41,620 % Response 95.29% 94.34% 95.86% 96.13% 96.34% 96.31% 95.74% 97.45% 97.15% Mean Std. Deviation N 173, , , , , , , , ,975 % Response 95.41% 94.57% 96.05% 96.36% 96.61% 96.46% 95.81% 97.48% 97.23% Mean Std. Deviation The scores for each patient satisfaction variable ranged from 0 to 100. Over the five year period, the mean score for RESPECT was highest at 90.62, followed by TRUST (83.02), ANSWER (80.15), EDREC (72.99), COURTESY (71.43), DRNURSEWK (69.73), EDSAT (67.08), EXPLAIN (64.25), and AVAILABILITY (61.35). The distribution of scores for each variable was negatively skewed. The skewness values of the distributions range from to A Kolmogorov-Smirnov test of normality was performed on the patient satisfaction variables, and all variables were normally distributed at the hospital level of analysis. More details are shown in Appendix I. Table 4-4. Patient Satisfaction by Gender GENDER Female Male N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC 95,468 94,754 96,209 96,664 96,854 96,680 95,815 97,799 97, ,206 77,381 78,620 78,728 78,989 78,898 78,585 79,636 79, , , , , , , , , ,

90 82 Table 4.4 shows the descriptive statistics on the patient satisfaction variables by gender. The means for males were statistically higher on each patient satisfaction variable compared to females (p<0.01) with one exception: no statistical difference (p=0.641) was found with patient gender for the variable RESPECT. RESPECT had a mean patient satisfaction score of for females and for males. See Appendix I for more details. Patient satisfaction scores between the three hospital peer groups teaching, large community, and small hospitals were significantly different (p<0.01), with small hospitals having highest patient scores compared to large community and teaching hospitals for each variable. Table 4.5 shows the patient satisfaction variables by peer group and descriptive statistics. See Appendix I for the F-statistics and p-values. Table 4-5. Patient Satisfaction by Peer Group Hospital Peer Group Large Community Small Teaching N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC 117, , , , , , , , , ,658 25,380 25,838 25,934 25,987 25,931 25,873 26,146 26, ,300 30,044 30,499 30,586 30,631 30,601 30,426 31,049 30, , , , , , , , , , Table 4.6 shows the differences in patient satisfaction scores across the various patient age groups. In the ED, patients over 65 years had the highest mean satisfaction scores across all patient satisfaction variables, except RESPECT and EXPLAIN (slightly higher in group).

91 83 Table 4-6. Patient Satisfaction by Age Groups Patient Age Group Under and over N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC 29,422 29,127 29,536 29,604 29,596 29,592 29,448 29,792 29, ,192 24,027 24,268 24,320 24,289 24,293 24,173 24,378 24, ,031 17,906 18,097 18,139 18,139 18,129 18,060 18,243 18, ,212 25,039 25,372 25,375 25,415 25,418 25,282 25,652 25, ,283 27,136 27,525 27,655 27,680 27,669 27,479 28,063 28, ,534 48,900 50,031 50,299 50,724 50,477 49,958 51,307 51, , , , , , , , , , Analysis of variance between groups was used to identify differences in each satisfaction variable between groups defined by gender, age groups, and hospital peer groups. Statistically significant differences were found within patient age group and patient gender (p<0.001). There were differences found at the 0.05 significance level between years using cross-tabs with the Chi-Squared statistic for patient gender and patient age groups. Therefore, patient age group, gender and year of measurement were included in the multivariate analyses. Table 4.7 shows the descriptive statistics and correlation coefficients for all study patient satisfaction variables. There were statistically significant correlations (p<0.01) found between overall satisfaction with care in the ED and each nursing care satisfaction variable. Correlations with overall satisfaction with care in the ED were as follows: DRNURSEWK (.801), COURTESY (.708), AVAILABILITY (0.70), EXPLAIN (0.58), ANSWER (0.57), TRUST (0.57), and RESPECT (0.28). Overall satisfaction with care in the ED was significantly associated with patients recommending the ED they attended (0.72, p<0.01).

92 84 Table 4-7. Correlation Table Patient Satisfaction ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC Pearson Correlation N Pearson Correlation N Pearson Correlation N Pearson Correlation N Pearson Correlation N Pearson Correlation N Pearson Correlation N Pearson Correlation N Pearson Correlation N **. Correlation is significant at the 0.01 level (2-tailed). ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC 1.654**.709**.307**.654**.593**.557**.572**.518** ** 1.667**.242**.664**.632**.573**.583**.502** **.667** 1.308**.659**.573**.549**.567**.529** **.242**.308** 1.361**.273**.261**.277**.236** **.664**.659**.361** 1.770**.698**.708**.548** **.632**.573**.273**.770** 1.680**.703**.576** **.573**.549**.261**.698**.680** 1.801**.644** **.583**.567**.277**.708**.703**.801** 1.719** **.502**.529**.236**.548**.576**.644**.719** Note: Patient level unit of analysis The seven nursing care variables (ANSWER, EXPLAIN, TRUST, RESPECT, COUTESY, AVAILABILITY and DRNURSEWK) were investigated to see if this set of variables represented patient satisfaction with nursing care in the ED. Principal Component Analysis (PCA) was performed. First, using the correlation statistics in Table 4.7, variables with correlation less than 0.3 were considered insignificant, while those with correlation greater than 0.7 were considered as redundant and could be deleted according to the PCA procedure (Jolliffe, 2002). A few variables had values outside these limits. At the lower bound, RESPECT was correlated between 0.24 (EXPLAIN) and 0.36 (COURTESY) with the other six patient satisfaction with nursing care variables. At the upper bound, COURTESY is correlated at 0.77 with AVAILABILITY. The factor loadings are presented in Table 4.8. Variables with loading of less than 0.4 were considered for deletion (Jolliffe, 2002), but the loading (except for RESPECT) ranged from for DRNURSEWK to for COURTESY. In addition, the factor loadings without RESPECT are shown.

93 85 Table 4-8. PCA Factor Loadings Variable Factor Loading (Including RESPECT) Factor Loadings (Excluding RESPECT) ANSWER EXPLAIN TRUST RESPECT N/A COURTESY AVAILABILITY DRNURSEWK Note: Patient level unit of analysis Analyzing the eigenvalues, only one principal component emerged that represented 62.34% of the variance explained by the set of seven patient satisfaction with nursing care variables. Appendix J presents the eigenvalues, the Scree Plot, and more details of the principal component analysis. The impact of item deletion was assessed using item-total statistics. Item-total correlations of less than 0.30 were considered for item deletion, and the impact on alpha if the item is deleted was assessed. As shown in Appendix J, all variables had corrected item-total correlations of greater than 0.3, with RESPECT having the lowest value of From the PCA analysis, the patient satisfaction with nursing care in ED is best represented by six variables (ANSWER, EXPLAIN, TRUST, COURTESY, AVAILABILITY and DRNURSEWK), removing RESPECT. The principal component analysis was repeated with all the patient satisfaction with nursing care variables, except RESPECT. Analyzing the eigenvalues, only one principal component emerged that represented 70.19% of the variance explained by the set of six variables, compared to 62.34% with the completed set of variables including RESPECT. The impact of item deletion was re-assessed using the item-total statistics. All the variables had corrected item-total correlations of greater than 0.3, with the lowest

94 86 values for DRNURSEWK (0.72) and TRUST (0.75). The overall alpha (0.91) was not improved with the deletion of any single patient satisfaction with nursing care item. Therefore, the subset of patient satisfaction variables that represents the overall variation in patient satisfaction with nursing care is comprised of six patient satisfaction with nursing care variables. Appendix K shows the patient satisfaction with nursing care (aggregate score) correlations with the other patient satisfaction variables. The patient satisfaction with nursing care component was used as a dependent variable in the multivariate analysis discussed later. Table 4.9 shows the impact of each nursing variable on overall satisfaction with care in the ED (EDSAT) and recommending the ED (EDREC) variables in terms of the variance explained by each nursing care variable. Patients deem that a good doctor and nurse working relationship accounts for 64% of the variance in the overall satisfaction with ED care. The COURTESY and AVAILABILITY nursing satisfaction variables accounted for the next highest levels of variance. These findings give an insight into what patients value as important, relative to their overall satisfaction with ED care. Table 4-9. Variance Explained by Each Variable Variable Variance (R 2 ) -EDSAT Variance (R 2 ) - EDREC ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK NURSING CARE (Aggregate Score) Note: Patient level unit of analysis

95 Emergency Department Characteristics The ED staffing, nurse characteristics, and covariate variables data were drawn from several administrative datasets for hospitals (NACRS, OHRS, OCDM and CIHI Nursing Database) reporting at the ED level. There were no missing data for the staffing variables, but there were missing data for nurse characteristics for 5 EDs, which accounts for 10 data points in the five year dataset of 499 hospital level observations. Outliers were identified by visual inspection of the data, using box-plots for each year and identifying any data above or below the mean by 1.5 times the interquartile range. Outliers were re-coded to a value calculated as the mean, plus or minus 1.5 times the interquartile range for that variable. Normality was observed using the Shapiro-Wilk test, normal probability plots, and histograms for each year. Nurse age, education, and employment status were normally distributed, but some of the staffing variables were not normally distributed. Many small hospitals did not employ RPNs, agency nurses, or nurse practitioners in the ED. Only one small hospital in the study dataset reported using agency nurses. The percentage of teaching hospitals using nurse practitioners has been increasing from 40% to 63%, and more large community and small hospitals have been employing RPNs over the five year period. Appendix L shows more details. Table 4.10 shows the worked hours per visit by nursing staffing categories. Although these variables have not been adjusted for the severity of the patients, the table highlights that small hospitals reported fewer agency nurses and nurse practitioners worked hours per visit than community and teaching hospitals.

96 88 Table Nursing Staffing Categories Hospital Peer Group Large Community Small Teaching Year 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE RNWKHRS RPNWKHRS AGNWKHRS NPWKHRS Mean Mean Mean Mean Note: ED level unit level of analysis Descriptive statistics including mean, standard deviation, and range for each nurse staffing variable are shown in Table Nurse staff characteristics are shown first, followed by intensity of care, skill mix, and staff adequacy variables. In addition, emergency departments were grouped into three types of hospitals: small hospitals, teaching or academic hospitals, and large community hospitals. Descriptive statistics for each variable by hospital type are also presented. Using ANOVA between groups, the patient satisfaction variable scores were statistically different among the hospital peer groups. Thus, hospital peer group was included in the multivariate analyses.

97 89 Table Emergency Department Characteristics by Hospital Type Nurse Staff Characteristics Intensity of Care Skill Mix Staff Adequacy Patient Satisfaction Note: ED level unit level of analysis Peer Group Large Community Small Teaching Mean Std. Deviation Mean Std. Deviation Mean Std. Deviation Mean Std. Deviation NURSEAGE NURSEED NURSEEXP PERFEMNURSE PERFTHRS RNWKHRS RPNWKHRS AGNWKHRS NPWKHRS TOTSTAFFWKHRS RNHPLOS RPNHPLOS NPHPLOS AGNHPLOS TOTSTAFFHPLOS RNPROP RPNPROP AGNPROP NPPROP RNRATIO RPNRATIO AGNRATIO NPRATIO TOTSTAFFRATIO ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC Nurse Characteristics Over the period from 2005/2006 to 2009/2010, the average age of the ED nursing staff (which includes both registered nurses and registered practical nurses) was 43.8 years. Teaching hospitals had the lowest average nursing age (40.1 years), while small hospitals had the highest average age (47.1 years). On average, 23.7% of the nursing staff had a baccalaureate or higher degree. Over the study period, this percentage has been increasing. Teaching hospitals had the highest percentage of nurses with a baccalaureate or higher degree (38.3%), compared to 18.0% in small hospitals and 21.6% in large community hospitals.

98 90 The average nursing experience of the ED nursing staff ranged from 4.0 to 33.6 years over the five years, with an overall average of 19.8 years (see Appendix N). The average nursing experience across the different hospital types ranged from 15.5 years for teaching, 19.2 years for large community, and 23.6 years for small hospitals. The mean percentage of female nurses ranged from 93.9 to 95.7 percent across the different hospital peer group. The mean percentage of full-time nursing worked hours ranged from 59.9% to 68.2%. Small hospitals had the lowest full-time nursing worked hours (59.91%) while teaching and large community hospitals had higher percentages (68.2% and 65.9%, respectively). In summary, ANOVA showed that all nursing characteristics variables in the study were significantly different (p<0.01) by hospital type Nursing Intensity of Care The average annual RN worked hours per visit ranged from 1.31 to 1.41 hours per visit, with a steady increase over the five-year period. The average annual RPN worked hours per visit ranged from 0.04 to 0.08 hours per visit. Teaching and large community hospitals had the highest agency nurse worked hours per visits compared to small hospitals. There were significant differences (p<0.001) between the years for RPN worked hours per visit, with means ranging from 0.04 to 0.08 hours per visit. RN worked hours per visit, NP worked hours per visit, and agency nurse worked hours per visit did not vary significantly (p>0.01) between the years. See Appendix N Intensity of Care for more details on the differences. Registered nurse hours per patient length of stay (RNHPLOS) was calculated using the annual registered nurse worked hours and divided by the annual total length

99 91 of stay of the patient reported at the emergency department level. The average annual RNHPLOS decreased from 0.54 to 0.38 from 2005/2006 to 2007/2008, with a gradual increase to 0.45 in 2009/2010 (see Appendix N Intensity of Care). Over the five-year study period, small hospitals had the highest average RNHPLOS of 0.63, compared to large community and teaching hospitals, which had an average of 0.4. The annual average registered practical nurse patient length of stay (RPNPLOS) ranged from to Small hospitals had the highest average RPNPLOS over the five-year study period. The annual average nurse practitioner (NPHPLOS) ranged from to 0.002, with large community hospitals and teaching hospitals having Similar to NPHPLOS, annual average agency nurse hours per patient length of stay (AGNPLOS) was low, ranging from to Small hospitals had the lowest annual mean AGNPLOS of 0.0. The annual mean total staff hours per patient length of stay ranged from to 0.591, and small hospitals had the highest mean (0.658) Skill Mix For the period from 2005/2006 to 2009/2010, the average proportion of RN worked hours to the total staff worked hours (RNPROP) in EDs was approximately The skill mix proportions for the various employment categories (RNPROP, RPNPROP, AGNPROP and NPPROP) have been relatively consistent across the five years of the study, with no statistical difference (p>0.01) between the years. The average RN proportion was highest in small hospitals (0.96), compared to teaching (0.87) and large community (0.89) hospitals. See Appendix N Skill Mix section for more details.

100 Staff Adequacy The average RNRATIO variable (the ratio of the number RN staff to patients) increased steadily but not significantly (p>0.01) over the years in the study, rising from 7.9 RNs per 10,000 patients to 8.6 RNs per 10,000 patients. Similarly, the ratio of the average total number of staff to patient increased from 8.7 full-time equivalents to 9.7 full-time equivalents per 10,000 patients. Over the study period, teaching hospitals had an average of 12.9 RNs per 10,000 patients, compared to small and community hospitals, which had 5.5 RNs and 8.5 RNs per 10,000 patients, respectively. RNRATIO, AGNRATIO, NPRATIO and TOTSTAFFRATIO did not vary significantly (p>0.01) over the five-year period. All staff adequacy ratios in the study were significantly different between the hospital peer groups. There were high, statistically significant, correlations between the nurse staffing variables. Appendix M shows the correlation coefficients for the nursing variables. Highly correlated variables are considered redundant and were considered for deletion. Using a data reduction process discussed later, one of the highly correlated variables was dropped in the multivariate analyses Covariates Five ED confounding variables are presented in Table 4.12 and Table The number of hospital corporations with patient satisfaction survey results in the study varied from 96 in 2005/2006 to 103 in 2007/2008. In this study the number of ED visits for this study was calculated from the NACRS database. The number of ED visits was measured in both the NACRS and OHRS datasets. The numbers of visits reported in

101 93 these two datasets, however, were found to be different. If the difference was greater than 5% or more than 500 visits, the number of ED visits from each dataset for that year of the study was investigated by reviewing the volume trends over the five-year period. There were two cases where the NACRS dataset underreported as much as 89% and 682% below the number reported in the OHRS database. In these two cases, the number of visits from the OHRS dataset was used after reviewing the hospital trend data. The average number of visits has been increasing every year, from 48,810 visits in 2005/2006 to 51,079 visits in 2009/2010. This increase has not been significant (p>0.05). The EDs in this study vary in size with significant differences between the hospital peer groups, with small hospitals group reporting an average number of visits of 16,078. The annual number of visits per hospital ranged from 3,778 to 200,130 visits for a multi-site corporation. The average ED case mix index has also been increasing during the five years of the study, but not significantly (p>0.05). Teaching hospitals had the highest average CMI: EDWAIT variable measures the proportion of ED visits seen within the established times based on the Canadian Triage Acuity Scale (CTAS). The average EDWAIT was approximately.8554 for EDs in the study over the five years. The average EDWAIT was significantly different (p<0.05) between the years, with an increase in annual mean EDWAIT in 2008/2009 and 2009/2010. Significant differences were reported among the hospital peer groups, with small hospitals group having the highest average EDWAIT (0.9533) compared to community and teaching hospitals ( and , respectively).

102 94 The EDCLEAN variable measures the average ED score of patients who felt that the ED they visited was clean. Over the study period, the EDCLEAN variable for the EDs included in the study had a mean of 83.12, and there were significant differences among the hospital peer groups. The average ED score of patients who felt the attending physician was courteous was 72.5 over the five-year period. No significant difference over the five years for the DRCOURTESY variable, but there were significant differences among the peer groups, with small hospitals in the study having the highest average score with physician courtesy. Table Control Variables by Year # of Visits EDCMI EDWAIT EDCLEAN DRCOURTESY N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation Year 2005/ / / / / ,810 48,562 48,793 50,490 51,079 49,549 36,162 36,060 36,169 36,213 38,017 36, Note: ED level unit level of analysis

103 95 Table Control Variables by Peer Group # of Visits EDCMI EDWAIT EDCLEAN DRCOURTESY N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation N Mean Std. Deviation Peer Group Large Community Small Teaching ,751 16,078 70,997 49,549 35,650 6,385 32,994 36, Note: ED level unit level of analysis In summary, using ANOVA between groups, the covariate variable scores were statistically different among the hospital peer groups. Only EDWAIT scores were statistically different across the five years. The correlation coefficients for the control variables and patient satisfaction variables are shown in Table The correlation coefficients for the size of the ED (number of visits) and each patient satisfaction variable were negative and statistically significant. Similarly, EDCMI and each patient satisfaction variable had negative correlation coefficients that were negative and statistically significant. The correlation coefficients for EDWAIT, EDCLEAN, DRCOURTESY, and each of the patient satisfaction variables were positive and statistically significant, ranging from 0.61 for EDREC to 0.86 for EDSAT.

104 96 Table Correlations between Control Variables and Patient Satisfaction # Of Visits EDCMI EDWAIT EDCLEAN DRCOURTESY ANSWER ** ** 0.680** 0.742** 0.755** EXPLAIN ** -0.68** 0.668** 0.749** 0.779** TRUST ** ** 0.72** 0.752** 0.761** RESPECT ** ** 0.633** 0.659** 0.715** COURTESY ** ** 0.666** 0.759** 0.807** AVAILABILITY ** ** 0.716** 0.79** 0.753** EDSAT ** ** 0.699** 0.782** 0.856** EDREC ** ** 0.614** 0.764** 0.811** NURSING (FACTOR SCORE) ** ** 0.698** 0.782** 0.814** ** Correlation is significant at the 0.01 level (2-tailed). Note: ED level unit level of analysis 4.3 Research Questions Analysis A series of linear mixed models were constructed to assess the effect of nurse staffing on patient satisfaction outcomes in the emergency department. The initial list of nursing characteristics and staffing variables, shown in Table 4.15, were checked for correlation and multicollinearity. Table 4.15 shows the final list of independent variables derived from the LASSO procedure: NURSEEXP, PERFTHRS, RNHPLOS, RPNHPLOS, RNPROP, and AGNPROP. This final set of variables was re-checked for correlation and multicollinearity. The hypotheses in the initial plan included RN proportion, nurse-to-patient ratio, and nursing worked hours per patient visit. As shown in Table 4.15, other nursing intensity and skill mix measures were modeled to examine the relationship between nursing staffing and patient satisfaction.

105 97 Table List of Variables Assessed In Regression Analyses Nurse Staff Characteristics Nurse Staff Characteristics Nurse Staffing Intensity of Care Age (in years) Education Level (Diploma, BSN, & higher) Nursing Experience (years in nursing = years after graduation from initial nursing program) Employment status Percent full-time (full-time RN & RN earned hours) divided by total nursing earned hours Gender (Percent Female Nurses) RN worked hours per patient visit RPN worked hours per patient visit Agency Nurse worked hours per patient visit Nurse Practitioner worked hours per patient visit staff worked hours per patient visit RN worked hours per patient length of stay (Annual RN worked hours divided by Annual Length of Stay) RPN worked hours per patient length of stay (Annual RPN worked hours divided by Annual Length of Stay) NP worked hours per patient length of stay (Annual NP worked hours divided by Annual Length of Stay) Agency Nurse worked hours per patient length of stay (Annual Agency Nurse worked hours divided by Annual Length of Stay) Variable Label NURSEAGE NURSEED NURSEEXP PERFTHRS PERFEMNURSE RNWKHRS RPNWKHRS AGNWKHRS NPWKHRS TOTSTAFFWKHRS RNHPLOS RPNHPLOS NPHPLOS AGNHPLOS Final Model (from LASSO) NURSEEXP PERFTHRS RNHPLOS RPNHPLOS staff worked hours per patient length of stay (Annual staff worked hours divided by Annual Length of Stay) TOTSTAFFPLOS Skill Mix Staff Adequacy RN proportion (RN worked hours divided by total staff worked hours) RPN proportion (RPN worked hours divided by total staff worked hours) Agency proportion (Agency Nurse worked hours divided by total staff worked hours) Nurse Practitioner Proportion (Nurse Practitioner worked hours divided by total staff worked hours) RN Staff to Patient Ratio (number of RN staff / number of patients) RPN Staff to Patient Ratio (number of RPN staff / number of patients) NP Staff to Patient Ratio (number of NP staff / number of patients) Staff to Patient Ratio (total number of patient care staff / number of patients) RNPROP RPNPROP AGNPROP NPPROP RNRATIO RPNRATIO NPRATIO TOTSTAFFRATIO RNPROP AGNPROP

106 98 For the model analysis, the independent variables control variables and dependent variables are shown in Table 4.16 below. The three regressions each had a different dependent variable: patient satisfaction with nursing (Aggregate Score), overall satisfaction with care in the emergency department, and recommending the emergency department to friends and family. Table Variables Used in Linear Mixed Models Dependent Variables Independent Variables Control Variables Patient Satisfaction with Nursing RNHPLOS (Intensity of Care) Size of ED Care (Aggregate Score) Overall Patient Satisfaction with Care RPNHPLOS (Intensity of Care) ED Wait Times in the Emergency Department. Recommending the Emergency Department to Friends and Family AGNPROP (Skill Mix) Severity or Case Mix Index of the ED RNPROP (Skill Mix) Age group of Patient PERFTEHRS (Nurse Characteristic) Gender of Patient RNEXP (Nurse Characteristic) ED Cleanliness Doctor Courtesy Hospital Peer Group Year of Measurement Patient satisfaction with nursing care variables (i.e., Answer, Explain, Trust, Respect, Courtesy, Availability and Dr-Nurse working relationship) were expected to be more positive in EDs with higher RN proportion, nurse-to-patient ratio, and nursing hours per patient visit. Due to the changes in the study, this hypothesis could not be tested. Nurse-to-patient ratio and nurse hours per patient visit were not used in the regression models because of multicollinearity and correlational issues. Despite this, RPN hours per patient length of stay, RN hours per patient length of stay, and RN agency nurse proportion were used as independent staffing variables. Also, a patient satisfaction with nursing care (Aggregate Score), which included six nursing variables except RESPECT, was used as the dependent variable in the multivariate analysis. So

107 99 conceptually the hypothesis was supported: there is a positive relationship between intensity of care and RN skill mix and patient satisfaction with nursing using the aggregate score. Three final models were established, with (1) patient satisfaction with nursing care (Aggregate Score), (2) overall satisfaction with received in the ED (EDSAT), and (3) recommending this ED to family and friends (EDREC) as dependent variables. Hierarchical linear modeling was used for all multivariate models using the SAS MIXED procedure. A two-level model was hypothesized to explore the relationship between nurse staffing and each patient satisfaction with nursing variable, overall satisfaction with care, and recommending the ED variables. The first level was the patient in ED and the second level was the ED with repeated measurements over time. The intraclass correlation coefficient (ICC) was calculated to assess if a two-level model was required for each of the three regressions. For each dependent variable, the ICC was calculated by using an unconditional intercept model. The ICC reveals what portion of the total variance is attributable to ED-level characteristics, or what portion of the total variance is between EDs. The ICC calculated for the following dependent variables were: Patient Satisfaction with Nursing Care Aggregate Score (0.06), EDSAT (0.07), AND EDREC (0.06). The ICC indicated that there is clustering of each dependent variable score within EDs, and therefore multi-level modeling should be used. Two levels of analyses were conducted at the patient and the ED levels. The first level measured the associations between patient satisfaction and nurse staffing at the level of the individual patient. The intercept at this level accounts for potentially correlated errors attributable to similarities among patients treated at the same hospital.

108 100 The second level of analyses measured the associations between mean ED satisfaction and nurse staffing measures at the ED level, with each model using a random intercept. Three multivariate linear mixed regression models were developed to assess the relationship between nurse staffing and patient satisfaction in the ED. The patient satisfaction outcome EDSAT (Y ijk ) can be expressed using a pair of linked models: one at the patient level (level-1) and another at the ED level (level-2). At level 1, the patient s outcome can be expressed as the sum of an intercept for the patient s ED (b 0jk ) and a random error (r ijk ) associated with the ith patient in the jth ED with s 2 representing the variance among patients within EDs and in the kth year. Level-1 predictors are Age group and Gender. Y ijk = b 0jk + b 1 AGEGRP ijk + b 2 PATGENDER ijk + b 3 Year k + b k + r ijk where r ijk =N(0,s 2 ) and AGEGRP, PATGENDER and YEAR are dummy variables. At level 2 (the ED level), the ED level intercepts were expressed as the sum of an overall mean (γ 00 ) and a series of random variations from that mean. b 0jk =γ 0 + γ 01 RNEXP jk + γ 02 AGNPROP jk + γ 03 RNHPLOS jk + γ 04 PERFTHRS jk +γ 05 RPNHPLOS jk + γ 06 RNPROP jk + γ 07 RNRATIO jk + r jk Each regression model was adjusted for heterogeneity of ED patient by entering terms into the model pertaining to demographic characteristics of patients. Each association was controlled for patient age, patient gender, ED cleanliness, ED physician courtesy, ED wait times, and ED case mix index. The control variables were entered as independent variables in the model. All dependent variables were checked for normality of distribution. The distributions of variables were checked for normality using the Shapiro-Wilk test for level-2 variables and the Kolmogorov-Smirnov test for the patient satisfaction variables (level-1).

109 101 Model assumptions of linearity and normality were checked by analyzing the residuals for each regression model using scatterplots of the residuals and the predicted dependent variable. Influential hospitals were analyzed, and one hospital corporation was found to be influential across all three models. This organization had five general EDs of different sizes and one urgent care center. This organization was removed from the regression analyzes. The patient satisfaction with nursing care using the aggregate score of six variables as the dependent variable is discussed in the next section. Models, however, of the seven patient satisfactions with nursing care dependent variables ANSWER, EXPLAIN, TRUST, RESPECT, COURTESY, AVAILABILITY AND DRNURSEWK variables were developed for additional information. See Appendix O for the results of these models Nurse Staffing and Patient Satisfaction with Nursing Care Table 4.17 shows the results of the model using the patient satisfaction with nursing care (aggregate score). Patient gender, patient age, cleanliness of the ED, attending physician courtesy, and proportion of patients seen within recommended timeframes were significantly associated with patient satisfaction with nursing care. Statistically significant associations were found between ED size and case mix index and the patient satisfaction with nursing care (aggregate score). Compared to teaching hospitals, large community hospitals had significantly different levels of patient satisfaction with nursing care (p<0.05).

110 102 On average, RN skill mix (RNPROP) and RPN worked hours per length of stay (RPNHPLOS) were positively associated with patient satisfaction with nursing care, with estimates of (p<0.0001) and (p<0.0001) respectively. Thus, for each percent increase in RPN worked hours per length of stay, there was an associated increase in patient satisfaction with nursing care of about.143 on a scale of 0 to 100. For each one percent increment in RN staff skill mix, however, there was an associated increase in patient satisfaction with nursing care of on a scale of 0 to 100. The percent of full-time nursing worked hours was negatively associated with patient satisfaction with nursing care, with an estimate of (p<0.05). Each one percent increase in full-time nursing staff was associated with a decrease in patient satisfaction with nursing care of approximately on a scale of 0 to 100. No statistically significant associations were found between agency proportion, nurse experience, RN worked hours per length of stay, and patient satisfaction with nursing care (p>0.05).

111 103 Table Linear Mixed Model: Patient Satisfaction with Nursing Care (Aggregate Score) Effect Description Units Estimate Standard Error Pr > t Intercept <.0001 Year Year Year Year Year 5 0 a.. Hospcmi Gender Female <.0001 Gender Male 0 a.. Patagegrp Under <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp Patagegrp Over 65 0 a.. Peer Group Large Community Peer Group Small Peer Group Teaching 0 a.. EDClean <.0001 # of Visits -7.59E DRcourtesy <.0001 EDWAIT <.0001 NURSEEXP RNPROP <.0001 AGNPROP PERFTHRS RNHPLOS RPNHPLOS <.0001 a Variable is set to zero and considered a reference for the dummy variable group Note: Unit level of analysis: Level 1 Patient; Level II - ED Nurse Staffing and Overall Satisfaction with Care Received in the ED EDSAT Table 4.18 shows the results of the model. Patient gender, patient age group, cleanliness of the ED, proportion of patient seen within recommended timeframes, and attending physician courtesy all had a positive and significant association with overall satisfaction with ED care. No statistically significant associations were found between

112 104 the size of the ED measured as the number of ED visits and ED case mix index and the overall satisfaction with ED care (p>0.05). Compared to teaching hospitals, large community hospitals had significantly different levels of patient satisfaction, with an estimate of (p<0.001). Patients under 54 years old had significantly different patient satisfaction scores to elderly patients over 65 years old. On average, RN skill mix (RNPROP) and RPN worked hours per length of stay (RPNHPLOS) were positively associated with overall patient satisfaction with care, with estimates of 4.98 (p<0.01) and (p<0.01) respectively. Thus, for each percent increase in RPN worked hours per length of stay, there was an increase in overall patient satisfaction with care of about.112 on a scale of 0 to 100. For each one percent increment in RN staff skill mix, there was an associated increase in overall patient satisfaction with care received in the ED of.05 on a scale of 0 to 100. The percent of full-time nursing worked hours was negatively associated with overall patient satisfaction with care with an estimate of (p<0.05). For each one percent increase in full-time nursing staff, there was an associated decrease in overall patient satisfaction with care received in the ED of approximately 0.02 on a scale of 0 to 100. No statistically significant associations were found between agency proportion, nurse experience, RN worked hours per length of stay, and overall patient satisfaction with care received in the ED.

113 105 Table Linear Mixed Model: Overall Patient Satisfaction with Care Received in the ED EDSAT Effect Description Units Estimate Standard Error Pr > t Intercept Year <.0001 Year <.0001 Year <.0001 Year Year 5 0 a.. Hospcmi Gender Female <.0001 Gender Male 0 a.. Patagegrp Under <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp Patagegrp Over 65 0 a.. Peer Group Large Community Peer Group Small Peer Group Teaching 0 a.. EDClean <.0001 # of Visits E DRcourtesy <.0001 EDWAIT <.0001 NURSEEXP RNPROP AGNPROP PERFTHRS RNHPLOS RPNHPLOS a Variable is set to zero and considered a reference for the dummy variable group Note: Unit level of analysis: Level 1 Patient; Level II - ED Nurse Staffing and Recommending the ED EDREC Table 4.19 shows the results of the linear mixed model. Patient gender, age, cleanliness of the ED, proportion of patient seen within recommended timeframes, and attending physician courtesy had a positive and statistically significant association with the patient recommending the ED they visited. No statistically significant association

114 106 was found between the size of the ED measured as the number of ED visits and ED case mix index and recommending the ED. Compared to teaching hospitals, both large community hospitals and small hospitals had significantly lower levels of patient recommendation scores with estimates of (p<0.001) and (p<0.05). Elderly patients aged 65 years and older had higher significantly higher patient recommendation score than other patients. On average, RN skill mix (RNPROP) was positively associated with recommending the ED patient satisfaction scores with estimates of (p<0.002). For each one percent increment in RN staff skill mix, there is an associated increase in patient recommendation of.07 on a scale of 0 to 100. The percent of full-time nursing worked hours was negatively associated with recommending the ED patient satisfaction scores, with an estimate of (p<0.01). For each one percent increase in full-time nursing staff, there was an associated decrease in patient recommendation of approximately 0.04 on a scale of 0 to 100. No statistically significant associations were found between agency proportion, nurse experience, RN, and RPN worked hours per length of stay with recommending the ED.

115 107 Table Linear Mixed Model: Recommending the ED EDREC Effect Description Units Estimate Standard Error Pr > t Intercept Year Year <.0001 Year <.0001 Year Year 5 0 a.. Hospcmi Gender Female Gender Male 0 a.. Patagegrp Under <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp <.0001 Patagegrp Over 65 0 a.. Peer Group Large <.0001 Community Peer Group Small Peer Group Teaching 0 a.. EDClean <.0001 # of Visits 7.64E DRcourtesy <.0001 EDWAIT <.0001 NURSEEXP RNPROP AGNPROP PERFTHRS RNHPLOS RPNHPLOS a Variable is set to zero and considered a reference for the dummy variable group Note: Unit level of analysis: Level 1 Patient; Level II - ED 4.4 Summary Descriptive statistics for over 182,000 patient surveys and 107 EDs used in the study gave valuable insight for the analyses performed to address the research questions. One principal component emerged representing 70.19% of the variance explained by the set of six patient satisfaction with nursing care variables (excluding RESPECT). Three linear mixed models were used to examine the relationship of nurse staffing in the emergency department and patient satisfaction with nursing care, overall

116 108 patient satisfaction with care received in the ED, and recommending the ED to family and friends. Table 4.20 and Table 4.21 below highlights the significant relationships found in this study. Table 4.21 shows the standardized coefficients to highlight the relative importance of the explanatory variables. Variables were standardized so that the magnitude of each association was calculated as a standardized regression coefficient that represents the number of standard deviations of change in the patient satisfaction outcome of interest per standard deviation of change in the explanatory variable. Therefore, the nurse staffing and nurse characteristics variables were centered and standardized. The standardized regression coefficient allows comparisons of magnitude across differing variables and represents an approximation of an r value.

117 109 Table Linear Mixed Models Results Patient Satisfaction with Nursing (Aggregate Score) EDSAT EDREC Effect Description Estimate Pr > t Estimate Pr > t Estimate Pr > t Intercept < Year < Year < <.0001 Year < <.0001 Year Year 5 0 a. 0 a. 0 a. Hospcmi Gender Female < < Gender Male 0 a. 0 a. 0 a. Patagegrp Under < < <.0001 Patagegrp < < <.0001 Patagegrp < < <.0001 Patagegrp < < <.0001 Patagegrp <.0001 Patagegrp Over 65 0 a. 0 a. 0 a. Peer Group Large <.0001 Community Peer Group Small Peer Group Teaching 0 a. 0 a. 0 a. EDClean < < <.0001 # of Visits -7.59E-06 < E DRcourtesy < < <.0001 EDWAIT < < <.0001 NURSEEXP RNPROP < AGNPROP PERFTHRS RNHPLOS RPNHPLOS < a Variable is set to zero; considered a reference for the dummy variable group Note: Unit level of analysis: Level 1 Patient; Level II - ED

118 110 Table Linear Mixed Models Results with Standardized Coefficients Patient Satisfaction with Nursing (Aggregate Score) EDSAT EDREC Effect Description Estimate Pr > t Estimate Pr > t Estimate Pr > t Intercept < < <.0001 Year < Year < <.0001 Year < <.0001 Year Year 5 0 a. 0 a. 0 a. Hospcmi Gender Female < < Gender Male 0 a. 0 a. 0 a. Patagegrp Under < < <.0001 Patagegrp < < <.0001 Patagegrp < < <.0001 Patagegrp < < <.0001 Patagegrp <.0001 Patagegrp Over 65 0 a. 0 a. 0 a. Peer Group Large <.0001 Community Peer Group Small Peer Group Teaching 0 a. 0 a. 0 a. EDClean < < <.0001 # of Visits -2.72E E DRcourtesy < < <.0001 EDWAIT < < <.0001 NURSEEXP RNPROP < AGNPROP PERFTHRS RNHPLOS RPNHPLOS < a Variable is set to zero; considered a reference for the dummy variable group Note: Unit level of analysis: Level 1 Patient; Level II - ED In all the models, patient age, gender, the cleanliness of the ED, and courtesy of the ED doctor variables were significantly associated with patient satisfaction. The estimates for the EDWAIT, or the proportion of patient seen within the targeted length of stay, were and for overall patient satisfaction with care and recommending the ED respectively. The size of the ED was significant in the patient satisfaction with nursing care model, but it was not significant for the EDSAT and EDREC models.

119 111 An increase in RNPROP was associated with an increase in patient satisfaction with nursing care, overall patient satisfaction with care received in the ED, and patients recommending the ED to family and friends but the magnitudes of the effect were small. An increase in RPN worked hours per patient length of stay was statistically associated with an increase in patient satisfaction with nursing care and overall patient satisfaction with care received in the ED, but once again the magnitudes of the effect were small. An unexpected finding was that a decrease in the percentage of full-time hours for nursing staff was associated with a statistically significant increase in patient satisfaction with nursing care, overall patient satisfaction with care, and patients recommending the ED. So the magnitudes of the associations between nurse staffing and patient satisfaction in EDs were significant but were small. Table 4.21 shows that DRCOURTESY, EDCLEAN, and EDWAIT are the most important control variables across all three models. The relative importance of the nurse staffing variables was found to be lower compared to these covariates. The following graphs in Figure 4 were developed to further investigate the magnitude of the nurse staffing effect on patient satisfaction in EDs. Using the latest year of data, 2009/2010, the actual and predicted average patient satisfaction scores for each ED were plotted against the size or number of visits of the ED. The first graph shows there is a large variation in the actual average patient satisfaction scores for overall care in 2009/2010. A typical ED was simulated using the median values for the covariate variables (CMI, EDCLEAN, DRCOURTESY, and EDWAIT). The predicted overall patient satisfaction with care received score for an ED was computed using the estimates from the regression model, the median values for the covariates and the actual nurse staffing values for that ED. The predicted overall

120 112 satisfaction with care received was plotted for all EDs in 2009/2010 in the second graph. This graph shows that variation in predicted overall satisfaction with care received in the ED care was less than 2%. Finally, a typical ED was simulated using the median values of the nursing variables. Similar to the second graph, the estimates from the regression model, the median values of the nursing variables and the actual covariate values were used to predict the overall patient satisfaction with care received in the ED for each ED. The third graph showed the variation of the predicted patient satisfaction scores ranged from 18% to 40%. This analysis confirms that the magnitude of the effect of nurse staffing on overall patient satisfaction with care received in the ED was small since the variation of the predicted patient satisfaction was small when using the median values of the covariates and the actual ED nurse staffing variables. Similar graphs were created for two other patient satisfaction outcome variables: patient satisfaction with nursing care and recommending the ED to family and friends. The graphs are shown in Appendix P. The results were also similar, the magnitudes of the effect of nurse staffing on patient satisfaction with nursing care and recommending the ED to family and friends were small since the variations of the predicted ED scores were low.

121 113 Figure 4. Predicted Satisfaction Scores for a typical ED Overall Patient Satisfaction in 2009/ Overall Satisfaction % ,000 40,000 60,000 80, , , , , ,000 Visits Predicted Overall Satisfaction in 2009/10 (Using median covariates values and nurse staffing variables) Predicted Overall Sat ,000 40,000 60,000 80, , , , , ,000 visits

122 114 Predicted Overall Patient Satisfaction in 2009/10 (using medians for nurse staffing and using covariates data) Predicted Overall Sat ,000 40,000 60,000 80, , , , , ,000 visits

123 115 Chapter 5 Discussion and Conclusion 5 Overview Having shown in the previous chapter the results of the multivariate models, this chapter concludes this study s investigation of patient satisfaction and its relationship to nurse staffing in the ED by discussing the findings in detail. In particular, the chapter will describe the following: 1. the relationships among variables; 2. major findings; 3. implications of the study; 4. limitations of the study 5. areas for future research; and 6. conclusions. 5.1 Study Variables In this study, the mean patient satisfaction scores were found to vary among the three hospital peer groups, with emergency departments in small hospitals having higher patient satisfaction scores than teaching and large community hospitals. Patient satisfaction scores were also different between the age groups, with patients over 65 years old having higher mean satisfaction scores than the non-elderly. These finding are consistent with previous studies that reported older patients were more satisfied than younger patients (Mahon, 1996; O'Connell et al., 1999; Liu & Wang, 2007; Alhusban & Abualrub, 2009). In this study, mean patient satisfaction scores for males were significantly higher on each patient satisfaction variable than they were for females (except for one variable:

124 116 RESPECT). Despite this, the mean patient satisfaction score for males were less than 2% higher than the mean scores for females, so the mean differences between males and females were significant, but the magnitude of the difference was quite small. Patient characteristics, such as cultural background, age, sex, and education have been found to be related to patient satisfaction (Bacon & Mark, 2009). The literature is mixed, since some studies did not find any relationships between patient satisfaction and demographic variables (Laschinger et al., 2011). Although there is a lack of consensus in the literature about the association of patient satisfaction and gender, some studies have found men to be more satisfied with their care than women (Lövgren et al., 1998; Arnetz & Arnetz, 1996.). Seven patient satisfaction with nursing care variables were explored in this study: ANSWER, EXPLAIN, TRUST, RESPECT, COURTESY, AVAILABILITY and DRNURSEWK. The correlations of each patient satisfaction with nursing care variables and the overall patient satisfaction with care in the ED variable highlighted what ED patients determine to be important relative to their overall satisfaction with care. Having nurses who are available, courteous, and able to work well with doctors highly correlated with patients overall satisfaction with care received in the ED. These findings support the importance of the interpersonal aspect of nursing practice similar to the findings in other studies (Jacox et al., 1997; O'Connell et al., 1999; Taylor & Benger, 2004). The current study findings are also consistent with previous studies that found patients expect the following nursing qualities: friendly and kind, quick to respond to patients needs, and having adequate time to provide care (Fitzpatrick, 1991). In the current study, there was considerable variance in the overall satisfaction with care received in the ED. This variance was explained by each patient satisfaction

125 117 with nursing care variable, with the exception of RESPECT. The patient satisfaction with nursing care (Aggregate Score) accounted for 61% of the variation of the overall satisfaction with care received in the ED. These findings are consistent with previous studies conducted in inpatient units of hospitals. Jacox et al. (1997) studied nine medical-surgical units in a teaching hospital and found patient satisfaction with nursing care was correlated with overall satisfaction with r=0.61. The current study overall correlation of 0.78 between patient satisfaction with nursing care (Aggregate Score) and overall satisfaction with care in the ED is similar to the results of Cleary et al. (1989) of 0.76 for surgical patients. When compared to studies identified by Kane et al. (2007) in their review of the literature on nurse staffing, the mean experience level for the nurses in the EDs in this study was high at 19.8 years. In addition, the mean RN proportion in this study of 90% was higher than studies reviewed on inpatient units; for example, Blegan et al. (1998) had an RN proportion for 42 inpatient units that ranged between 46% to 96%, with an average of 72% across all units(blegen et al., 1998). In a study of nurse staffing in California hospitals, researchers found RN proportion was 70.34% and 74.26% for medical/surgical and step-down units (Bolton et al., 2007). Thus, on average, the experience level and mix of RNs in the EDs in this study were higher than those of inpatient units studied previously. Patients who stayed longer in hospitals tend to be less satisfied than those who stayed a short time (Boudreaux et al., 2000). In addition, the length of time to see a physician is associated with patient satisfaction with emergency patients and the number of patient who left without being seen (Mowen et al., 1993; Kyriacou et al., 1999). In this current study, the proportion of patient seen within the recommended

126 118 timeframe in EDs had a positive significant association with patient satisfaction in the ED. In fact, the magnitude of the effect of wait times on patient satisfaction in the ED for nursing care, overall care and likelihood of recommending the ED to friends and family was higher than any of the nurse staffing variables explored. The EDWAIT variable is calculated at the ED level, so there may be other factors that can affect this variable. Significant differences in patient satisfaction were found between the peer groups used in this study, with small hospitals having higher patient satisfaction scores on average compared to community and teaching hospitals. EDs in small hospitals did not utilize RN agency nurses, but they still had the lowest percentage of full-time nursing staff. Small hospitals, however, had the highest RN skill mix compared to community and teaching hospitals. In Ontario, the small hospitals are located in rural areas where the ED can be used for emergency medical care as well as routine care because of access to primary care (JPPC, 2007). The relationships between small hospital staff and community could have an impact on the higher patient satisfaction scores since patients may fear any negative patient ratings could lead to closure or reduced funding to their hospital. Three covariates (DRCOURTESY, EDCLEAN and EDWAIT) were found to have a significant association and major effect on all three patient satisfaction outcomes investigated (patient satisfaction with nursing care-aggregate Score, EDSAT and EDREC). The average scores for these covariates were significantly higher for EDs in small hospitals compared to EDs in community and teaching hospitals. The combination of these covariates and other factors may explain the higher patient satisfaction in EDs in small hospitals.

127 Findings in Relation to the Conceptual Framework The current study adapted the conceptual framework used by Kane et al. (2007) to investigate the association between nurse staffing and patient satisfaction with one major change. Kane et al. (2007) included nurse outcomes such as nurse job satisfaction, nurse retention rate and nurse burnout rate in their conceptual framework. Some of the relationships hypothesized in this current study were not modeled because nurse-to-patient ratio and RN worked hours per patient visit were not included as predictors in the regression analyses. These two variables were dropped as a result of the LASSO procedure because of multicollinearity issues. There was weak support, however, for the relationships hypothesized for RN proportion and patient satisfaction, and for that reason, the conceptual model adapted from Kane et al. (2007) was only partially confirmed in this study. To test the conceptual model, the effect size, as discussed in Chapter Three, must be large enough to explain 26% of the variance of the dependent variable patient satisfaction scores (Cohen & Cohen, 2002). For the three regression models presented in this study, the reduction in variance of the dependent variables were 0.45 for the patient satisfaction with nursing care (Aggregate Score), 0.53 for the overall satisfaction with care, and 0.38 for recommending the ED to family and friends, making the effect size large enough to test the model Research Question 1 To what extent specific aspects of nurse staffing relate to patient satisfaction with nursing care?

128 120 During the analysis of the data to answer this research question, two interesting findings were observed regarding the patient satisfaction with nursing variables. First, using Principal Component Analysis, the current study identified a subset of patient satisfaction with nursing care variables that sufficiently represents the overall variation in patient satisfaction with nursing. One component was found to represent the overall variation (70.19%) in patient satisfaction with nursing care comprising six of the seven patient satisfaction variables. RESPECT was excluded because the response to the question did nurses talk in front of you as if you weren t there was not highly associated with patient satisfaction with nursing care in the ED. Also, there were no differences found between men and women for RESPECT. These findings do not deemphasize the importance of having respect for patients, but rather, RESPECT seems separate from patient satisfaction with nursing care in this study. Second, the patient satisfaction with nursing care (aggregate score) was found to be significantly associated with patient satisfaction with overall care received in the ED. This finding is consistent with other studies that revealed patient satisfaction with nursing care in inpatient units is associated strongly with (and is an important predictor of) overall satisfaction with hospital care (Johansson et al., 2002; Bolton et al., 2003; Larrabee et al., 2004; Al-Mailam, 2005; Chan & Chau, 2005; McGillis Hall, 2005; Kane et al., 2007). To investigate this research question, the strategy for the multivariate analysis changed from the original plan based on the results of the principal components analysis. Instead, the final model presented two nurse staffing variables, RNPROP and RPNHPLOS, both of which were positively significantly associated with patient satisfaction with nursing care. More importantly, however, the magnitude of the

129 121 estimates for the independent variables in this model was quite small. When interpreting patient satisfaction results, the effect size estimates should be emphasized rather than the p-values alone (Boudreaux et al., 2004). With respect to the first research question "To what extent do specific aspects of nurse staffing relate to patient satisfaction with nursing care?" the study findings were unexpected regarding the effect size estimates. There are statistical associations between RNPROP and patient satisfaction with nursing, but the magnitude of the effect was very small. On average, the mean RN skill mix is 90% for the EDs in this study. At higher levels of RN skill mix, any increase will result in only a very small increase in patient satisfaction with nursing. The magnitude of the effect of the association between RPN hours per length of stay and patient satisfaction with nursing was significant, and although the magnitude of the effect was larger than RNPROP, the magnitude of the effect is still small. Only a few EDs in this study reported RPNs worked hours. Nonetheless, for each percent increase in RPN worked hours per length of stay, there was only a small increase in overall patient satisfaction with nursing care. Therefore, there is a statistical association between two nurse staffing measures and patient satisfaction with nursing care, but it is a minute effect that is not actionable Research Questions 2 and 3 To what extent specific aspects of nurse staffing relate to overall satisfaction with care in the ED and b) whether the patient would recommend this ED to friends and family? The regression models presented two nurse staffing variables, RNPROP and RPNHPLOS, both of which were positively significantly associated with overall satisfaction with care in the ED. Thus, there is a positive relationship between intensity

130 122 of care, RN skill mix and overall satisfaction with care, and recommending the ED to friends and family. Although the study showed that RNPROP and RPNHPLOS were positively significantly associated with overall satisfaction with care in the ED, the magnitude of the effect was again small. Similar to the discussion for the previous research question, increases in RN skill mix or RPN worked hours per length of stay are associated with very small increases in patient satisfaction scores. The magnitude of the effect of nurse staffing was indeed quite small, so conceptually there was very weak support for the hypothesis. Although other studies found that a higher proportion of RNs was associated with a significantly lower rates of patient complaints in inpatient units (Blegen et al., 1998; Blegen & Vaughn, 1998), this current study found a statistical association between RNPROP and patient satisfaction, but not one that is administratively actionable by nurse managers. The RPNHPLOS variable estimates were the highest of the nurse staffing variables used in the three regression analyses. Higher RPN worked hours per length of stay was associated statistically with an increase in patient satisfaction with nursing care, overall satisfaction with care received in the ED, and recommending the ED scores. One possible explanation is that the RN job satisfaction may be higher if RNs concentrate on nursing tasks and do not perform non-nursing functions (McGillis Hall et al., 2003). The current study, however, reveals that an increase in RN skill mix and higher RPN worked hours per length of stay are associated statistically with better patient satisfaction outcomes, although the magnitudes of the effects are too small to draw any conclusions. The variation in the mean score for patient satisfaction and nurse staffing measures were also very small at the ED level. In addition, the average

131 123 RNPROP for EDs in the current study was quite high at 90%. Blegan et al. (1998) found in their study of relationships between major adverse events (MAE) and nurse staffing that as the proportion of RNs on a unit increased from 50% to 85% "the rate of MAEs declined, but as the RN proportion increased from 85% to 100% the rate of MAEs increased" (Blegen & Vaughn, 1998). In economics, diminishing marginal returns is referred to as the decrease in the marginal output of a production process as the amount of a single factor of production is increased, while the amounts of all other factors of production stay constant (Samuelson & Nordhaus, 2001). With the EDs in the study having high RN proportions, there may be no more room for improvement in patient satisfaction scores with additional increase in nurse staffing. The percent of full-time nursing worked hours variable was negatively significantly associated with each of the three patient satisfaction outcome variables: patient satisfaction with nursing (Aggregate Score), patient satisfaction with overall care received in the ED, and patient recommendation of the ED. Although significant, the magnitude of the effect for percent full-time nursing worked hours was very tiny and not meaningful with any of the three outcomes. Small hospitals had a significantly lower mean percentage of full-time nurses compared to large community and teaching hospitals. Furthermore, small hospitals had higher mean patient satisfaction scores than large community and teaching hospitals. This may have led to percent of full-time nursing variable being significant without a large effect in the multivariate models. In the three regression models, EDWAIT (or the percent of visits seen within the target lengths of stay) was positively significantly associated with patient satisfaction in the ED. Put simply, patients generally do not like to wait to see a clinician (McMillan et al., 1986) and the current study emphasizes the association of length of stay and patient

132 124 satisfaction. The waiting time to see the clinician was included in the total length of stay time in the ED. Researchers have found that only the waiting time to examination by the ED physician is significantly negatively correlated with patient satisfaction (Dansky & Miles, 1997; Sandovski et al., 2001). ED physician courtesy and cleanliness of the ED were positively significantly associated with all three patient satisfaction outcomes investigated. Research has shown patient satisfaction is correlated strongly with rating for physician courtesy (Comstock et al., 1982). Similarly, cleanliness of the hospital has been found to be associated with patient satisfaction (Sitzia & Wood, 1997). In the current study, both physician courtesy and cleanliness of the ED had a higher magnitude of effect than any other explanatory variable including the nurse staffing variables. These findings highlight the importance of interpersonal factors and environmental factors with respect to patient satisfaction but do not suggest a casual relationship. As discussed in Chapter Three, the associations explored in this study do not reflect a causal relationship. In their review of the nurse staffing literature, Kane et al. (2007) found increased staffing in hospitals was associated with better care outcomes, but the association did not reflect a causal relationship. Kane et al. (2007) commented that hospitals that invest in more nurses may also invest in other actions that improve quality. In fact, overall hospital commitment to provide high quality of care in combination with effective nurse retention strategies has shown to lead to better patient outcomes, greater patient satisfaction with overall care received and nursing care, and nurse job satisfaction (Laschinger et al., 2003; Lake & Friese, 2006; Kane et al., 2007).

133 Study Implications The findings in the current study have implications for hospital administrators, policy and future research. The following section will describe those implications for each group or application Administration Hospital administrators are faced with higher acuity of patients and fiscal pressures, and this result in new staffing models being implemented in order to control cost, even as volume of patients seeking care in EDs continues to grow. Despite these pressures, hospital administrators need to reflect on patient satisfaction in the planning, delivery, and evaluation of care. The impact of any changes in staff mix or levels should be assessed, monitored, and evaluated with respect to patient outcomes, including patient satisfaction. The results of this study, therefore, need to be considered when implementing quality improvement initiatives, replacing RNs with other nursing staff, and scheduling staff in the department. This study shows that with the present nurse staffing, there is an extremely small or imperceptible improvement in patient satisfaction that can be achieved from staffing measures such as RN skill mix. Other changes, however, such as alterations in the ED physician courtesy, ED cleanliness, and reduction in patient wait times may have greater improvement on (and are more strongly associated with) patient satisfaction outcomes than changes in nurse staffing. Caution must be taken when using results which demonstrate association and not cause. Administrators need to consider initiatives to improve patient satisfaction that have high return on investment and this

134 126 study reveals that relatively small improvement in patient satisfaction can be achieved through changes to nurse staffing. More research on causation of higher patient satisfaction ratings in the ED is required so that administrators can act with greater certainty of changes to satisfaction scores. This study also emphasizes the association of nursing availability to overall satisfaction with care in the ED and highlights the importance of the interpersonal aspect of nursing practice. For that reason, there is a real need to explore effective ways for improving interpersonal aspects of patient care in the potentially high-stress environment of a hospital. Hospital administrators, nurse managers, researchers, and nurses all must investigate the factors affecting the interpersonal aspect of nursing in EDs. The current study shows the correlation between nurses availability, courtesy, ability to answer questions from patients, nurse competency, and patient satisfaction with overall care received in the ED. These results have implications for administrators as they monitor and evaluate their ED patient satisfaction scores. The workload in EDs are unpredictable, however, managers must recognize the importance of having courteous, competent staff in EDs with enough stand-by capacity so staff are able to address questions and concerns of patients at all times. To investigate patient satisfaction scores, administrators need to explore initiatives to improve the interpersonal aspects of care, environmental factors such as the cleanliness of the ED, and the overall wait times in the ED. Finally, this study highlights the limitations with aggregate level of data in reporting and evaluating patient satisfaction with nursing care. Hospital administrators and nurses need to support database systems that are not only effective, but that can

135 127 be used to improve the quality of care and patient experience through research. The database should include the amount and type of nursing resources used to care for each ED patient Policy In Ontario, the Excellent Care for All Act, 2010 puts patients first by improving the quality and value of the patient experience. This legislation magnifies the importance of the patient experiences and requires hospitals to survey patients in order to assess their satisfaction with services provided. In addition, the Centers for Medicare & Medicaid Services (CMS) have begun to publicly report patient satisfaction in the U.S. by using a standardized survey the Hospital Consumer Assessment of Healthcare Providers & Systems (HCAHPS) that was developed by the Agency for Healthcare Research and Quality. As more states consider regulating nurse-to-patient ratios and more public reporting of patient satisfaction, there is an urgent need to improve the perception patients have of their care experience. In the U.S. and U.K., hospital payments will eventually depend on patient satisfaction scores, as there is a shift to a patient-centered health system. Beginning in the fiscal year 2013, CMS is in the process of implementing a Hospital Value-Based Purchasing (VBP) program that will pay incentives to acute care hospitals based on achievement or improvement related to predefined quality measures. Patient satisfaction is a core measure of the VBP program and will include better patientclinician communication, pain management, and preparation for discharge. Similarly, 10% of the payments to NHS trusts in the U.K. will depend on patient satisfaction.

136 128 Thus, the findings of this current study are important to both policy makers and hospital administrators as the effort continues to improve quality of care and decrease cost. In the U.S., some federal or state legislatures have mandated hospitals to implement staffing plans with specific nurse-to-patient ratios. This study does not bring into dispute the importance of nurse staffing, but the current study highlights how the association between nurse staffing and patient satisfaction can be affected by different factors, such as patient and nurse characteristics, and environmental characteristics of the emergency department (such as cleanliness). If policy makers expect large changes in patient satisfaction with mandated staffing ratios, this study has shown the effect of nurse staffing in the ED on patient satisfaction may be, at the most, very small. 5.4 Limitations of the Study The major limitation of the study is its use of aggregate hospital-level data instead of patient-level data. This current study can be considered as a partial ecological study where the variables can be grouped as individual-level (properties of each patient) and ecological variables (properties of organizations). Ecological studies can be used for hypothesis-generating exercises, but they require confirmation using individual-level variables. They are also potentially prone to biases that may occur when the association that exists between variables at an aggregate level do not represent the true association that exists at an individual level (Piantadosi et al., 1988). Another limitation of the current study relates to missing data. Missing data were found at two levels: the patient level and the hospital level. Some patients did not respond to the questions in the patient satisfaction survey, therefore missing data were

137 129 imputed for two patient satisfaction with nursing variables, ANSWER and EXPLAIN. At the hospital ED level, missing data were found for the nursing characteristics for specific years and were subsequently removed from the analysis. The nurse staffing data for the 153 emergency departments included in the study were reported for 107 hospital corporations. For hospital corporations with a general ED and an UCC, for example, the nurse staffing variables were reported collectively since it was not possible to separate the data by site. For hospital corporations with different types of EDs, the effects of organizational characteristics, which could affect the relationship between nurse staffing and patient satisfaction with care, were unable to be controlled in the current study. In this study, there were no medical staffing measures because of the lack of reporting of the fee-for-service physicians. Two general measures of nurse staffing, however, were studied. One addressed hours of care provided by nursing staff of different nurse categories at the hospital level for the emergency department. The hours of care or worked hours comprised both direct times (patient-related) and nondirect times, such as lunch breaks and training times. The other nurse staffing measure was based on less precise data of total nurse staffing, averaging FTE to patient volume to create a ratio of nurse staffing to patient. In studies of nursing staffing and patient outcomes, Kane et al. (2007) found that RN per shift ratio was more frequently used and provided greater evidence of the effect, although generally they showed the same trends as nurse-to-patient ratios. RN per shift ratio and measures of only direct nursing patient care times could not be obtained from the administrative databases. Another limitation of the current study is that only two control variables for differences in served patients were included: age and gender. Patient factors has been

138 130 found to account for as much as 47.5% of the variance in patient satisfaction with nursing care (Angelo et al., 2003). In the current study, patient severity, length of stay, and the amount of nursing hours were reported at the hospital corporation level and not at the patient level. Other patient characteristics, such as education level, were not included. In addition, this study did not consider nurse ethnicity, race and language considerations. It is unknown, therefore, to what extent RN characteristics make any difference in patient satisfaction outcomes. As discussed previously, the framework devised by Kane et al. (2007) included nursing outcome variables such as nursing staff satisfaction, burnout, and turnover. In hospitals with high patient-to-nurse ratios, nurses are more likely to experience burnout and job dissatisfaction, outcomes that can, in turn, affect patient outcomes (Aiken et al., 2002). For example, researchers have found nurse outcomes can affect patient outcomes which includes patient satisfaction. For example, nurses may become dissatisfied with their work when unable to give good care, leading to reduced patient satisfaction (McNeese-Smith, 1999; Vahey DC, 2004). Furthermore, RN turnover has been found to be inversely related to patient satisfaction (Henry, 1992). Unfortunately, nursing outcomes data were not available for this study. Finally, another limitation of the current study is the omission of data about physicians. Only physician courtesy was included in this study. It is possible that EDs with higher level of nursing staff could also have greater numbers of better-qualified doctors working in the ED. In patient outcomes studies, researchers found that doctors were the most important professional group associated with reductions in mortality (Aiken et al., 2002; Aiken et al., 2003).

139 Future Research Patient satisfaction with nursing has grown in importance for researchers for a number of reasons. Assessing patient satisfaction provides a means of monitoring the quality of nursing care and evaluating effectiveness of nursing interventions. Satisfaction with nursing care is also the most important predictor of overall satisfaction with hospital care (Greeneich, 1993). The current study makes important contributions to better understand the relationship between nurse staffing and patient satisfaction in emergency departments. The results reinforce the findings by previous studies that indicate that there are differences in patient satisfaction between men and women and between elderly and non-elderly patients. These differences should be addressed in the design and data analysis of health services research on patient satisfaction, and these variables should also be controlled to address any potential confounding in future research. The present research in this study used hierarchical linear mixed models. Although the patient satisfaction outcomes were patient specific, the analysis could not be performed at the patient level because of data limitations. If staffing and acuity data for each patient were obtained by shift for the patients surveyed, the sensitivity of the study to detect the association between nurse staffing and patient satisfaction would have been increased. In addition, the current study has also shown that the proportion of patients seen within the recommended timeframe is associated with patient satisfaction in EDs. Future research examining the relationship between nurse staffing and patient satisfaction in emergency departments should include patient-level data for acuity, staffing, length of stay, as well as the time to see a physician.

140 132 Future research on the relationship between nurse staffing and patient satisfaction in emergency departments should include additional patient-level characteristics (patient education level, for example). Since patient factors have been found to account for much of the variance in patient satisfaction with nursing care (Angelo et al., 2003), including other patient characteristics will improve the model. Given the established association between nurse outcomes and patient outcomes including patient satisfaction (Aiken et al., 2002; Vahey DC, 2004) gaining further understanding about these relationships in the ED is important. Future research should examine the effect of nursing outcomes such as nurse satisfaction, burnout, and turnover on patient satisfaction in EDs. Additionally, this study included the RN characteristics of education, experience, and age, however future research should include RPN characteristics. In the current study, there were different types of EDs, such as general EDs, UCCs, or trauma centres. The type of ED was not used to control for organizational characteristics. Future research should include the type of ED, which may in turn better isolate the effects of the hospital characteristics, including case mix index and wait times, on the relationship between nurse staffing and patient satisfaction. 5.6 Conclusion Patient satisfaction is a key outcome measure being examined by researchers exploring relationships among patient outcomes, hospital structure, and care processes. No studies have examined the relationship between nurse staffing and patient

141 133 perceptions of nursing care in multiple EDs using shared definitions of nurse staffing and a common patient satisfaction survey tool. This five-year study makes an important contribution to the literature by presenting findings from over 100 EDs across urban and rural, community and academic, small and large, healthcare institutions with varying sizes and case mix, and for using common measures for both nurse staffing and patient perception of care from 182,000 patients. The current conceptual model linking nurse staffing and patient satisfaction outcomes in EDs was not fully supported. Although the hypotheses for the study could not be confirmed, RN proportion and RPN worked hours per length of stay were found to have a weak statistical association with patient satisfaction with nursing care, patient satisfaction with overall care in the ED, and the likelihood to recommend the ED to friends and family. Notwithstanding the limitations of ecological variables used in the analyses, interpersonal and environmental factors such as courtesy, cleanliness, and timeliness rather than nurse staffing should be the focus of both administrative efforts for consideration to investigate to improve patient satisfaction in EDs and further research on the subject.

142 134 References Abdellah, F. G. & Levine, E. (1958). Effect of Nurse Staffing on Satisfactions with Nursing Care, American Hospital Association. Abenhaim, H. A., Kahn, S. R., Raffoul, J. & Becker, M. R. (2000). "Program Description: a hospitalist-run, medical short stay unit in a teaching hospital." CMAJ Canadian Medical Association Journal 163(11): Ahmad, M. M. & Alasad, J. A. (2004). "Predictors of patients' experiences of nursing care in medical-surgical wards." International Journal of Nursing Practice 10(5): Aiello, A., Garman, A. & Morris, S. B. (2003). "Patient satisfaction with nursing care: a multilevel analysis." Quality Management in Health Care 12(3): Aiken, L. H., Clarke, S. P. & Cheung. R. B (2003). "Educational levels of hospital nurses and surgical patient mortality.." JAMA 290(12): Aiken, L. H., Clarke, S. P., Silber, J. H. & Sloane, D. (2003). "Hospital nurse staffing, education, and patient mortality." LDI issue brief 9(2): 1-4. Aiken, L. H., Clarke, S. P. & Sloane, D. M. (2002). "Hospital staffing, organization, and quality of care: Cross-national findings." Nursing outlook 50(5): Aiken, L. H., Clarke, S. P., Sloane, D. M., J.A., S., Busse, R. & Clarke, H. (2001). "Nurses' reports on hospital care in five countries." Health Affairs 20(3): Aiken, L. H., Clarke, S. P., Sloane, D. M., Sochalski, J. & Silber, J. H. (2002). "Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction." JAMA 288(16): Aiken, L. H., Sloane, D. M., Cimiotti, J. P., Clarke, S. P., Flynn, L., Seago, J. A., Spetz, J. & Smith, H. L. (2010). "Implications of the California Nurse Staffing Mandate for Other States." Aiken, L. H., Smith, H. L. & Lake, E. T. (1994). "Lower Medicare mortality among a set of hospitals known for good nursing care." Medical care 32(8): Aiken, L. H., Sochalski, J. & Anderson, G. F. (1996). "Downsizing the hospital nursing workforce." Health Affairs 15: al-haider, A. S. & Wan, T. T. (1991). "Modeling organizational determinants of hospital mortality." Health services research 26(3): Alhusban, M. A. & Abualrub, R. F. (2009). "Patient satisfaction with nursing care in Jordon." Journal of Nursing Management 17: Al-Mailam, F. F. (2005). "The effect of nursing care on overall patient satisfaction and its predictive value on return-to-provider behavior: a survey study." Quality Management in Health Care 14(2): American Nurses Association (ANA) (2007). "NDNQI Guidelines." Angelo, A., Andrew, G. & Scott, B. (2003). "Patient satisfaction with nursing care: a multilevel analysis." Qual Manag Health Care. 12: Arnetz, J. E. & Arnetz, B. B. ( 1996.). "The development and application of a patient satisfaction measurement system for hospital-wide quality improvement." Int J Qual Health Care. 8(6): Auerbach, D. L., Buerhaus, P. I. & Staiger, D. O. (2011). "Registered Nurse Supply Grows Faster Then Projected Amid Surge In New Entrants Ages " Health Affairs 30;12:

143 Bacon, C. T. & Mark, B. (2009). "Organizational effects on patient satisfaction in hospital medical-surgical units." Journal of Nursing Administration 39: Barbara, A., Davis, D. & Elaine, D. (1999). "Patient satisfaction wth nursing care in rural and an urban emergency department." Australian Journal of Rural Health 7: Beland, F., Lemay, A. & Boucher, M. (1998). "Patterns of Visits to Hospital-Based Emergency Rooms." Soc. Sci. Med. 47: p Blegen, M. A., Goode, C. J. & Reed, L. (1998). "Nurse staffing and patient outcomes." Nursing research 47(1): Blegen, M. A. & Vaughn, T. (1998). "A multisite study of nurse staffing and patient occurrences." Nursing economic$ 16(4): Blegen, M. A. & Vaughn, T. (2001). "Nurse experience and education." Journal of Nursing Administration 31(1): Bloom, J. R., Alexander, J. A. & Nuchols, B. A. (1997). "Nurse staffing patterns and hospital efficiency in the United States." Social science & medicine (1982) 44(2): Bolton, L. B., Aydin, C. E., Donaldson, N., Brown, D. S., Nelson, M. S. & Harms, D. (2003). "Nurse staffing and patient perceptions of nursing care." Journal of Nursing Administration 33(11): Bolton, L. B., Aydin, C. E. & Nancy Donaldson, D. S. B., Meenu Sandhu, Moshe Fridman and (2007). "Mandated Nurse Staffing Ratios in California: A Comparison of Staffing and Nursing-Sensitive Outcomes Pre- and Postregulation." Policy, Politics, & Nursing Practice 8; 238. Bolton, L. B., Jones, D. & Aydin, C. E. (2001). "A response to California's mandated nursing ratios." J Nurs Scholarsh 33(2): Bond, S. & Thomas, L. H. (1992). "Measuring patients' satisfaction with nursing care." Joumal of Advanced Nursing 17: Boudreaux, E. D., Ary, R. D., Mandry, C. V. & McCabe, B. (2000). "Determinants of patient satisfaction in a large, municipal ED: the role of demographic variables, visit characteristics, and patient perceptions." American Journal of Emergency Medicine 18(4): Boudreaux, E. D., D'Autremont, S., Wood, K. & Jones, G. N. (2004). "Predictors of emergency department patient satisfaction: stability over 17 months." Academic Emergency Medicine 11(1): Boudreaux, E. D., Mandry, C. V. & Wood, K. (2003). "Patient satisfaction data as a quality indicator: a tale of two emergency departments." Academic Emergency Medicine 10(3): Boudreaux, E. D. & O'Hea, E. L. (2004). "Patient satisfaction in the Emergency Department: a review of the literature and implications for practice." Journal of Emergency Medicine 26(1): Brewer, C. S. & Frazier, P. (1998). "The influence of structure, staff type, and managed-care indicators on registered nurse staffing." Journal of Nursing Administration 28(9): Bullard, M. J. & Villa-Rowe, C. (2010). "The role of a rapid assessment zone/pod on reducing overcrowding in emergency departments: a systematic review." Emerg Med J Bursch, B., Beezy, J. & Shaw, R. (1993). "Emergency department satisfaction: what matters most?" Annals of Emergency Medicine 22(3): Byrne, G., Richardson, M., Brunsdon, A. & Patel, A. (2000). "Patient satisfaction with emergency nurse practitioners in A&E." Journal of Clinical Nursing 9:

144 Chan, J. N. & Chau, J. (2005). "Patient satisfaction with triage nursing care in Hong Kong." Journal of Advanced Nursing 50(5): Chang, B. L. (1997). "Dimensions and indicators of patients' perceived nursing care quality in the hospital setting." Journal of Nursing Care Quality 11(6): Cho, S. H., Hwang, J. H. & Kim, J. (2008). "Nurse staffing and patient mortality in intensive care units." Nursing research 57(5): CIHI (2008). Hospital e-scorecard Report 2008: Emergency Department Care Patient Satisfaction Technical Summary, Canadian Institute for Health Information. CIHI (2011). Canadian Institute for Health Information, Highlights of Inpatient Hospitalizations and Emergency Department Visits. CIHI (2011). Data Quality Documentation for External Users: National Ambulatory Care Reporting System, Clarke, S. P. (2007). "Nurse staffing in acute care settings: research perspectives and practice implications." Joint Commission journal on quality and patient safety / Joint Commission Resources 33(11 Suppl): Cleary, P. D., S. Edgman-Levitan, S., McMullen, W. & Delbanco, T. L. (1992). "The relationship between reported problems and patient summary evaluations of hospital care." Qual Rev Bull. 18: CNA, C. N. A. (2004). Nursing Staff Mix: A Literature Review. CNA, C. N. A. (2004). Tested Solutions for Eliminating Canada's Registered Nurse Shortage. Cohen, J. & Cohen, P. (1983). Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, Lawerence Erlbaum Associates, Publishers, Hillsdale, New Jersey. Cohen, J. & Cohen, P. (2002). Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, Lawerence Erlbaum Associates, Publishers, Hillsdale, New Jersey. Comstock, L., Hooper, E., Goodwin, J. & Goodwin, J. (1982). "Physician behaviors that correlate with patient satisfaction." Journal of Medical Education 57(2): Considine, J., Kropman, M., Kelly, E. & Winter, C. (2008). "Effect of emergency department fast track on emergency department length of stay: a case control study." Emergency Medicine Journal 25( ). Cook, T. D. & Campbell, D. T. (1979). Quasi-Experimentation: Design and Analysis for Field Settings, R and McNally, Chicago, Illinois. Coulter, A., Fitzpatrick, R. & Cornwell, J. (2009). "The Point of Care Measures of patients' experience in hospital: purpose, methods and uses." The King's Fund ISBN: Daly, S., Cambell, D. A. & Cameron, P. A. (2003). "Short Stay Units and Observational Medicine: A systematic Review -." MJA 178: Dansky, K. & Miles, J. (1997). "Patient satisfaction with ambulatory healthcare services: waiting time and filling time." Hosp Health Serv Adm. 42(2): Darby, C. (2002). "Patient/parent assessment of the quality of care." Ambulatory Pediatrics 2(4 Suppl): Davis, M. M. & Vollmann, T. A. (1990). "A framework for relating wait time and customer satisfaction in a service operation." Journal of Service Marketing 4(1): Dimick, J. B., Swoboda, S. M., Pronovost, P. J. & Lipsett, P. A. (2001). "Effect of nurse-topatient ratio in the intensive care unit on pulmonary complications and resource use after hepatectomy." American Journal of Critical Care 10(6): Donabedian, A. (1966). "Evaluating the quality of medical care." Millbank Memorial Fund Quarterly 44(3):

145 137 Donabedian, A. (1980). Explorations in quality assessment and monitoring, Ann Arbor: Health Administration Press. Dozier, A. M., Kitzman, H. J., Ingersoll, G. L., Holmberg, S. & Schultz, A. W. (2001). "Development of an Instrument to Measure Patient Perception of the Quality of Nursing Care." Research in Nursing & Health, 2001, 24, 24: 506±517. Drummond, A. J. (2002). "No room at the inn: overcrowding in Ontario's emergency departments." CMAJ Canadian Medical Association Journal 4(2): ENA (2003). Emergency Nurses Association White Paper: Staffing and Productivity in the Emergency Care Setting Eriksen, L. R. (1995). "Patient satisfaction with nursing care: concept clarification." Journal of Nursing Measurement 3(1): Fahad Al-Mailam, F. (2005). "The effect of nursing care on overall patient satisfaction and its predictive value on return-to-provider behaviour: A survey study." Quality Management in Health Care 14(2): FitzGerald, G., Jelinek, G. A., Scott, D. & Gerdtz, M. F. (2010). "Emergency department triage." Emer Med J 27: Fitzpatrick, R. (1991). "Surveys of patient satisfaction: II--Designing a questionnaire and conducting a survey." BMJ 302(6785): Freeman, J. L., Fetter, R. B., Park, H., Schneider, K. C. & al., e. (1995). "Diagnosis-Related Group Refinement with Diagnsis- and Procedure-Specific Comorbodities and Complications." Medical Care 33(8): French, K. (1981). "Methodological consideration in hospital patient opinion surveys." International Journal of Nursing Studies. 18: Greeneich, D. (1993). "The link between new and return business and quality of care: Patient satisfaction." Advances in Nursing Science 16:: Hall, J. A. & Dorman, M. C. (1990). "Patient sociodemographic characteristics as predictors of satisfaction with medical care: A meta-analysis." Social science & medicine 30: Hall, M. F. & Press, I. (1996). "Keys to patient satisfaction in the emergency department: results of a multiple facility study." Hospital & Health Services Administration 41(4): Halloran, E. & Vermeersch, P. (1987). "Variability in Nurse Staffing Research." 17(2): Halm, M., Peterson, M., Kandels, M., Sabo, J., Blalock, M., Braden, R., Gryczman, A., Krisko- Hagel, K., Larson, D., Lemay, D., Sisler, B., Strom, L. & Topham, D. (2005). "Hospital nurse staffing and patient mortality, emotional exhaustion, and job dissatisfaction." Clinical Nurse Specialist 19(5): Han, a., Ospina, M., Blitz, S., Strome, T. & Rowe, B. H. (2007). "Patients presenting to the emergency department: the use of other health care services and reasons for presentation." CJEM Canadian Journal of Emergency Medical Care 9(6): Hansagi, H., Cadsson, B. & Brismar (1992). "The urgency of care need and patient satisfaction at a hospital emergency department." Health Care Management Review 17.: Henry, B. (1992). "Leadership, nurses, and patient satisfaction: a pilot study." Nurs Adm Q. 16:(3):: Hinshaw, A. & Atwood, J. (1982). "A patient satisfaction instrument. Precision by replication." Nursing research 31: Hoelzle, B. (2011). "Regression Analysis with the Ordered Multinomial Logistic Model." Paper presented at the annual meeting of the Southwest Educational Research Association, San Antonio, TX, Feb 2-4, 2011.

146 HSRC (March 1997). Health Services Restructuring Commission, Metropolitan Toronto Health Services Restructuring Report. Jacox, A. K., Bausell, B. R. & Mahrenholz, D. M. (1997). "Patient satisfaction with nursing care in hospitals." Outcomes Management for Nursing Practice 1(1): Johansson, P., Oleni, M. & Fridlund, B. (2002). "Patient satisfaction with nursing care in the context of health care: a literature study." Scandinavian Journal of Caring Sciences 16(4): Jolliffe, I. (2002). Principal Component Analysis. New York, Springer. JPPC, O. J. P. a. P. C. (2006). Summary Report of the Hospital Funding Committee on the Use of 2004/05 Cost and Activity Data. JPPC, O. J. P. a. P. C. (2007). Recommendation of CACS grouper and Modified RIWs for Emergency Department and Day Surgery. Toronto, Ontario. Kane, R. L., Shamliyan, T., Mueller, C., Duval, S. & Wilt, T. J. (2007). " Nursing Staffing and Quality of Patient Care." Evidence Report/Technology Assessment No. 151 (Prepared by the Minnesota Evidence-based Practice Center under Contract No ) AHRQ Publication No. 07-E005. Rockville, MD: Agency for Healthcare Research and Quality.: King, D. L., Ben-Tovim, D. I. & Bassham, J. (2006). "Redesigning emergency department patient flows: Application of Lean Thinking to healthcare." Emergency Medicine Australasia 18: Konnyu, K., Kwok, E., Grimshaw, J. & Moher, D. (2011). What is the evidence of the effectiveness and safety of emergency department short stay units., Ottawa Hospital Research Institute. Korn, R. & Mansfield, M. (2008). "ED overcrowding: an assessment tool to monitor ED registered nurse workload that accounts for admitted patients residing in the emergency department." Journal of Emergency Nursing 34(5): Krishel, S. & Baraff, L. J. (1993). "Effect of emergency department information on patient satisfaction." Annals of Emergency Medicine 22(3): Kurata, J., Nogawa, A., Phillips, D., Hoffman, S. & Werblum, M. (1992). "Patient and provider satisfaction with medical care." J of Fam Prac 35: Kwa, P. & Blake, D. (2008). "Fast Track: Has it changed patient care in the emergency department?" Emergency Medicine, Australasia 20: Kyriacou, D. N., Ricketts, V., Dyne, P. L., McCollough, M. D. & Talan, D. A. (1999). "A 5-year time study analysis of emergency department patient care efficiency.[see comment]." Annals of Emergency Medicine 34(3): Lake, E. & Friese, C. (2006). "Variations in nursing practice environments: relation to staffing and hospital characteristics." Nurs Res 55(1):1-9. LaMonica, E., Oberst, M., Madea, A. & Wolf, R. (1986). "Development of a patient satisfaction scale." Research in Nursing & Health 9: Larrabee, J. H. & Bolden, L. V. (2001). "Defining patient-perceived quality of nursing care." Journal of Nursing Care Quality 16(1): Larrabee, J. H., Ostrow, C. L., Withrow, M. L., Janney, M. A., Hobbs, G. R. & Burant, C. (2004). "Predictors of patient satisfaction with inpatient hospital nursing care." Research in Nursing & Health 27: Larson, M. G. (2006). "Descriptive Statistics and Graphical Displays." Journal of the American Heart Association 114:

147 139 Larson, P. J. & Ferketich, S. L. (1993). "Patients' satisfaction with nurses' caring during hospitalization." Western Journal of Nursing Research 15(6): Laschinger, H., Gilbert, S. & Smith, L. (2011). "Nursing Outcomes; The State of the Science." Jones & Barlett, Learning ISBN Laschinger, H., J;, A. & D., T.-H. (2003). "Workplace empowerment and magnet hospital characteristics: making the link." J Nurs Adm 33(7-8): Leaman, A. M. (2003). "See and treat: A management driven method of achieving targets or a tool for better patient care? One size does not fit all.." Emergency Medicine Journal 20: 118. Ley, P. J., Kinsey, S. T. & Atherton, B. (1976). "Increasing patients' satisfaction with communication." British Journal of Clinical Psychology 15: Lin, C. C. (1996). "Patient satisfaction with nursing care as an outcome variable: dilemmas for nursing evaluation researchers." Journal of Professional Nursing 12(4): Liu, Y. & Wang, G. (2007). "Inpatient Satisfaction With Nursing Care and Factors Influencing Satisfaction in a Teaching Hospital in China." J Nurs Care Qual 22(3): Lövgren, G., Sandman, P. O., Engström, B., Norberg, A. & Eriksson, S. (1998). "The view of caring among patients and personnel." Scand J Caring Sci.. 12(1): Mahon, P. Y. (1996). "An analysis of the concept 'patient satisfaction' as it relates to contemporary nursing care." Journal of Advanced Nursing 24(6): Mark, B. A. (2006). "Methodological Issues in Nurse Staffing Research." Western Journal of Nursing Research 28(6): Mark, B. A., Harless, D. W. & McCue, M.,. (2004). "A longitudinal examination of hospital registered nurse staffing and quality of care." Health Serv Res 39(2): Mark, B. A., Salyer, J. & Wan, T. T. (2000). "Market, hospital, and nursing unit characteristics as predictors of nursing unit skill mix: a contextual analysis." Journal of Nursing Administration 30(11): Mark, B. M., Salyer, J. & Wan, T. T. (2003). "Professional Nursing Practice; Impact on Organizational and Patient Outcomes." Journal of Nursing Administration 33 (4): McGee, L. A. & Kaplan, L. (2007). "Factors influencing the decision to use nurse practitioners in the emergency department." Journal of Emergency Nursing 33(5): McGillis Hall, L. (1997). "Staff Mix Models: Complementary or substitute roles for nurses." Nursing Administration Quarterly 21(2): McGillis Hall, L. (2005). "Quality Work Environments for Nurse and Patient Safety." McGillis Hall, L., Doran, D., Baker, G., Pink, G., Sidani, S., O'Brien-Pallas, L. & GJ., D. (2003). "Nurse staffing models as predictors of patient outcomes." Medical Care 41(9):: McMillan, R., Younger, M. & DeWme, L. (1986). "Sahsfachon with hospital emergency department as a funchon of patient tnage." Health Care Management Revtew 11(3): McNeese-Smith, D. (1999). "A content analysis of staff nurse descriptions of job satisfaction and dissatisfaction." J Adv Nurs. 29(6): Megivern, K., Halm, M. A. & Jones, G. (1992). "Measuring patient satisfaction as an outcome of nursing care." Journal of Nursing Care Quality 6(4): Merkouris, A., Ifantopoulosh, J., Lanara, V. & Lemonidou, C. (1999). "Patient satisfaction: a key concept for evaluating and improving nursing services." Journal of Nursing Management 7:

148 140 Merkouris, A., Papathanassoglou, E. D. & Lemonidou, C. (2004). "Evaluation of patient satisfaction with nursing care: quantitative or qualitative approach?" International Journal of Nursing Studies 41(4): Minnick, A. F., Roberts, M. J. & Young, W. B. (1997). "What influences patients' reports of three aspects of hospital services?." Med Care 35(4): Mitchell, P., Ferketich, S. & Jennings, B. (1998). "Quality health outcomes model." Journal of Nursing Scholarship 30: MOHLTC (2011). Health Based Allocation Model (HBAM) Manual. Monzon, J., Friedman, S. M., Clarke, C. & Arenovich, T. (2005). "Patients Who Leave the Emergency Department Without Being Seen by a Physician: A Control-Matched Study." Canadian Journal of Emergency Medicine 7(2) Mowen, J. C., Licata, J. W. & McPhail, J. (1993). "Waiting in the emergency room: how to improve patient satisfaction." Journal of Health Care Marketing 13(2): Munro, B., Jacobsen, B. S. & Brooten, D. A. (1994). "Re-examination of the pschometric characteristics of the La Monica-Oberst Patient Satisfaction Scale." Research in Nursing & Health 17: Needham, J. (1996). "Balancing skill mix--future paediatric health care provision." J Nurs Manag. 4(3):: Needleman, J., Buerhaus, P., Mattke, S., Stewart, M. & Zelevinsky, K. (2002). "Nurse-staffing levels and the quality of care in hospitals." The New England journal of medicine 346(22): Nelson E & C., L. (1993). "Patients' good and bad surprises: How do they relate to overall patient satisfaction?" Quality Review Bulletin 19: Nerney, M. P., Chin, M. H., Jin, L., Karrison, T. G., Walter, J., Mulliken, R., Miller, A., Hayley, D. C. & Friedmann, P. D. (2001). "Factors associated with older patients' satisfaction with care in an inner-city emergency department." Annals of Emergency Medicine 38(2): NSW, N. S. W. (2006). Clinical Services Redesign Program: Models of emergency care. North Sydney: NSW Health. Oberst, M. (1984). "Patient's perception of care." Cancer 53: O'Brien-Pallas, L., Doran, I., Murray, M., Cockerill, R., Sidan, i. S. & Laurie-Shaw, B. (2002). "Evaluation of a client care delivery model, Part 2: Variability in client outcomes in community home nursing." Nursing economic$ 20: O'Connell, B., Young, J. & Twigg, D. (1999). "Patient satisfaction with nursing care: a measurement conundrum." International Journal of Nursing Practice 5(2): OHA (2003). Geriatric Emergency Management, Ontario Hospital Association. OHQC (2010). Ontario Health Quality Council, Quality Monitor Pasoe, G. C. (1983). "Patient satisaction in primary health care: A literature review and analysis." Evaluation and Program Planning 6: Piantadosi, S., Byar, D. P. & Green, S. B. (1988). "The Ecological Fallacy." American Journal of Epidemoology 127, 5. PricewaterhouseCoopers (2008). NSW Health Emergency Department Workforce Research Project Final Report. Raper, J. L. (1996). "A cognitive approach to patient satisfaction with emergency department nursing care." Journal of Nursing Care Quality 10(4): Risser, N. L. (1975). "Development of an instrument to measure patient satisfaction with nurses and nursing care in primary care settings." Nursing research 24(1):

149 RNAO (2005). "The 70 percent solution: A progress report on increasing full-time employment for Ontario RNs." Registered Nurses' Association of Ontario. Rodi, S. W., Grau, M. V. & Orsini, C. M. (2006). "Evaluation of a fast track unit: alignment of resources and demand results in improved satisfaction and decreased length of stay for emergency department patients." Quality Management in Health Care 15(3): Rutledge, R., Osler, T. & Kromhout-Schiro, S. (1996). "Illness severity adjustments for outcomes analysis: validation of the ICISS methodology in all 821,455 patients hospitalized in North Carolina in 1996." Surgery 124(2): Sakr, M., Angus, J., Perrin, J., Nixon, C., Nicholl, J. & Wardrope, J. (1999). "Care of minor injuries by emergency nurse practitioners or junior doctors: a randomised controlled trial." The Lancet 354: Samuelson, P. A. & Nordhaus, W. D. (2001). Microeconomics, McGraw Hill. Sandovski, U., Salman, H., Bergman, M., Neiman, V., Bessler, H. & Djaldetti, M. (2001). "Patients' satisfaction with the staff function in an emergency department." European Journal of Emergency Medicine 8(2): Schmidt, L. A. (2003). "Patients' perceptions of nursing care in the hospital setting." Journal of Advanced Nursing 44(4): Schmidt, L. A. (2004). "Patients' perceptions of nurse staffing, nursing care, adverse events, and overall satisfaction with the hospital experience." Nursing Economics 22(6): Schull, M. J., Slaughter, P. M. & Redelmeier, D. A. (2002). "Urban emergency department overcrowding: defining the problem and eliminating misconceptions." Canadian Journal of Emergency Medicine 4(2): Seago, J. A., Williamson, A. & Atwood, C. (2006). "Longitudinal analyses of nurse staffing and patient outcomes: more about failure to rescue." J Nurs Adm 36(1): Shortell, S. M., Zimmerman, J. E., Rousseau, D. M., Gillies, R. R., Wagner, D. P., Draper, E. A., Knaus, W. A. & Duffy, J. (1994). "The performance of intensive care units: does good management make a difference?" Medical care 32(5): Sinoff, G., Clarfield, A., Bergman, H. & Beaudet, M. (1998). "A two-year follow-up of Geriatric Consults In Emergency Departments." Journal of the American Geriatrics Society 46(6): Sitzia, J. & Wood, N. (1997). "Patient satisfaction: A review of issues and concepts." Social science & medicine 45: Sovie, M. D. & Jawad, A. F. (2001). "Hospital restructuring and its impact on outcomes: Nursing staff regulations are premature." Journal of Nursing Administration 31: Spetz, J., Seago JA, Coffman, J., Rosenoff, E. & O'Neil, E. (2000). "Minimum Nurse Staffing Ratios in California Acute Care Hospitals." Oakland,CA: Californai Health Care Foundation. Staiger, D. O., Auerbach, D. I. & Buerhaus, P. I. (2012). "Registered Nurse Labour Supply and the Recession - Are we in a bubble?" The New England Journal of Medicine 10, Strasen, L. (1989). "Incorporating Patient Satisfaction Standards Into Quality of Care Measures." Journal of Nursing Administration: 18(11): 5-6. Strasser, S. & Davis, R. M. (1991). Measuring patient satisfaction for improved patient services, Ann Arbor: Health Administration Press. Sun, B. C., Adams, J., Orav, E. J., Rucker, D. W., Brennan, T. A. & Burstin, H. R. (2000). "Determinants of patient satisfaction and willingness to return with emergency care.[see comment]." Annals of Emergency Medicine 35(5):

150 Tabachnick, B. G. & Fidell, L. S. (1996). Using Multivariate Statistics, Harper Collins College Publishers (New York, NY). Taylor, C. & Benger, J. R. (2004). "Patient satisfaction in emergency medicine." Emergency Medicine Journal 21(5): Thompson, D. A., Yarnold, P. R., Williams, D. R. & Adams, S. L. (1996). "Effects of actual waiting time, perceived waiting time, information delivery, and expressive quality on patient satisfaction in the emergency department." Annals of Emergency Medicine 28(6): Thungjaroenkul, P., Cummings, G. G. & Embleton, A. (2007). "The Impact of Nurse Staffing on Hospital Costs and Patient Length of Stay: A Systematic Review." Nursing Economics 25(5): Tibshirani, R. (1996). "Regression Shrinkage and Selection via the Lasso." Journal of the Royal Statistical Society, Series B (Methodological), Volume 58(1). Tourangeau, A. (2002). "Nursing skill mix and experience reduce patient mortality." Hospital quarterly 5(3): Tye, C. C., Ross, F. & Kerry, S. M. (1998). "Emergency nurse practitioner services in major accident and emergency departments: a United Kingdom postal survey." Journal of Accident & Emergency Medicine 15: Tyrance, P. H., Jr.,, Himmelstein, D. U. & Woolhandler, S. (1996). "US Emergency Department Costs: No Emergency." American Journal of Public Health 86(11): p UHN (2011). H-SAA Amending Agreement. Agreement between Toronto Central Local Health Network and University Health Network. Unruh, L. (2003). "Licensed nurse staffing and adverse events in hospitals." Medical care 41(1): Unruh, L. Y. & Fottler, M. D. (2006). "Patient Turnover and Nursing Staff Adequacy." Health Services Research 41(2): Vahey DC, e. a. (2004). "Nurse burnout and patient satisfaction." Med Care 42(2 Suppl):II Van den Heede, K., Clarke, S. P., Sermeus, W., Vleugels, A. & Aiken, L. H. (2007). "International experts' perspectives on the state of the nurse staffing and patient outcomes literature." J Nurs Scholarsh 39(4)(J Nurs Scholarsh.): Van den Heede, K., Sermeus, W., Diya, L., Clarke, S. P., Lesaffre, E., Vleugels, A. & Aiken, L. H. (2009). "Nurse staffing and patient outcomes in Belgian acute hospitals: Crosssectional analysis of administrative data." International Journal of Nursing Studies 46: Wallin, E., Lundergren, P., Ulander, K. & Holistein, C. (2000). "Does age, gender, or educational background affect patients' satisfaction with short stay surgery?" Ambulatory Surgery 8: Ware, J. E., Synder, M. K., Wright, W. R. & Allison, R. D. (1983). "Defining and measuring patient satisfaction with medical care." Evaluation and Program Planning 6: Williams, B. (1994). "Patient satisfaction:a valid concept?" Social science & medicine 38: Williams, R. M. & Bamezai, A. (2005). "The costs of visits to emergency departments - Revisited." Annals of Emergency Medicine 46(5): p Wolf, Z. R., Miller, P. A. & Devine, M. (2003). "Relationship between nurse caring and patient satisfaction in patients undergoing invasive cardiac procedures." MEDSURG Nursing 12(6):

151 Yankovic, N. & Green, L. V. (2008). ""A Queueing Model for Nurse Staffing" Working Paper." Columbia University. Yoon, P. (2003). Emergency Department Fast Track System, Alberta Heritage Foundation for Medical Research. 143

152 144 Appendix A. Literature Review Search Identification of papers for review The search strategy used to identify pertinent literature was based on the conceptual model. An iterative process was developed using the results of the broader search to identify sub-categories of staff mix and patient outcomes. Table 1 shows the final textword query and describes the strategies for the research questions. The same eligibility criteria, selection of studies, and analysis were used to examine the association between nurse staffing and patient outcomes. Studies were sought from a wide variety of sources, including MEDLINE, PubMed, CINAHL, EBSCO research database, Canadian Institute for Health Information (CIHI) reports, federal reports, and Digital Dissertations. In addition to traditional medical databases, alternative databases of associated literature were also searched. Structured Internet searches were conducted for organizations that perform relevant research, such as the National Database of Nursing Quality Indicators, National Center for Health Workforce Analysis, American Nurses Association, Canadian Nurse Association, Emergency Nurse Association, Registered Nursing Association of Ontario, Canadian Practical Nurse Association, Practical Nurse Association of Ontario, Registered American Academy of Nurse Practitioners, and Provincial Ministry of Health and Long-Term Care.

153 145 Table A1. Search Strategy and Data Sources Search Strategy Data Sources Text-word queries for staff mix models or models of nursing or nurse staffing strategies or patient satisfaction were combined with subject queries for urgent care centres or ambulatory care facilities or walk-in clinics or hospitals or emergency medical services or Accident and Emergency Departments or their equivalents Medline ( ) HealthStar ( ) Social Science Abstracts ( ) Pubmed ( ) EMBASE ( ) (PAIS) ( ) Consolidated International Nursing and Allied Health Sciences Library (CINAHL) ( ) electronic databases Web-Science ( ) Searches of complete volumes published between Healthcare Management Review The Journal of Ambulatory Care Management The Canadian Journal of Emergency Medicine American Journal of Emergency Medicine Journal of Accident and Emergency Medicine Journal of Emergency Nursing Academic Emergency Medicine Annals of Emergency Medicine Searches of the on-line documents published between Searches of the on-line government documents Searches of web-sites for documents on Emergency Department staffing in Canada, United States, United Kingdom, Australia and New Zealand Searches of the Proquest Digital Dissertation Abstracts (1975 to 2011) for dissertations on either Emergency Departments, Nurse Staffing, Searches of on-line catalogues at the University of Toronto and The Institute for Clinical Evaluative Sciences in Ontario Gleaning of the reference lists of articles eventually selected for this review Personal communication with hospital managers of Ontario emergency departments, Ontario Ministry of Health and Long-Term Care, and the Ontario Nursing Secretariat. Eligibility Abstracts were reviewed to exclude studies with ineligible target populations. Reviews, letters, comments, legal cases, and editorials were also excluded. The full texts of the original epidemiologic studies were examined to define eligible independent variables (nurse staffing and strategies) and patient satisfaction. Studies outside of the emergency department were included, but studies were excluded if they did not test the associative hypotheses and did not provide adequate

154 146 information on tested hypotheses. Inclusion criteria were applied to select articles for full review. Studies needed to meet one of the following criteria: Retrospective observational cohort studies and retrospective cross sectional comparisons; Administrative cross-sectional survey and analyses; Evaluates the associations between nurse staffing and patient; Satisfaction/nurse quality measures among eligible target populations (either patients hospitalized in acute care hospitals or emergency department patients); or Ecologic studies on correlations between nurse staffing and patient satisfaction. During the period of 1980 to 2012, there were potentially relevant documents that explored nursing staffing and patient outcomes, but few studies were found that focused on the emergency department setting. Documents written in the English language and restricted to publications on nursing staffing and outcomes in Canada, the United States of America, the United Kingdom, Australia and New Zealand were considered. This broad search, which included inpatient units as well as the emergency department, identified key elements of heterogeneity. The articles were classified by the following themes: Emergency department nurse staffing; Emergency department patient satisfaction; and Nursing Staffing and patient satisfaction (inpatient units). Abstraction forms were used to collect the data and the bibliographic information was manipulated using Thomas ISI ResearchSoft EndNote (Version 6) reference manager

155 147 program. A log of ineligible studies was created using custom fields in the EndNote library. The log includes the reasons why studies are deemed ineligible (Petitti, 1994). In total, 251 papers were selected for this review and are categorized as shown in Table 2. Table A2. Paper selected by categories Topic of Paper # of Articles # of Articles Selected Nurse staffing Patient satisfaction

156 148 Appendix B. Outcomes Model for Healthcare Research Figure B-1 Outcomes Model for Healthcare Research The Effect of Nursing Care on Outcomes Health Facility Environment Mission Size Teaching Status Location Funding Status Organizational Structure Leadership Unit Qualities Culture RN-MD relationships Control over practice Autonomy Leadership RN-Patient interaction RN staffing Staff mix Nursing Qualities Intrinsic Critical thinking skills Communication skills Compassion/caring Professionalism Stress/Fatigue Competency Extrinsic Experience Education Age Physical ability Certification Patient Qualities Acuity level Disease Patient-Family dynamics Nursing Care Caregiver Dependency Comfort Education Therapeutics Monitoring/ Surveillance Integrator Nurse/Patient Nurse-to-patient Nurse/Physician Nurse/ Other Caregivers Outcomes Facility Unit Nurse Patient

157 149 Appendix C. Quality of Care Dynamic Model Table C-1 Components and Variables of the Model Components of the Model System Client Interventions Outcomes Variables Includes the traditional structure variables: size of the organization, ownership, nurse skill mix, client demographics, and technology. Client health, demographics, and disease risk factors. Clinical processes and activities that contribute to the outcomes. Results of care structures and processes: achievement of appropriate self-care, demonstration of health promoting behaviours, health-related quality of life, perception of being well cared for, and symptom management.

158 150 Appendix D. Theoretical Model of the Relationships between Context, Structure (professional practice), and Effectiveness (outcomes) CONTEXT STRUCTURE EFFECTIVENESS Hospital Characteristics Technological Complexity Case Mix Index Teaching Status Admission Volatility Size Nursing Unit Characteristics Experience Education Skill Mix Unit Size Support Services Patient Technology PROFESSIONAL PRATICE Organizational Outcomes Nursing Work Satisfaction Nursing Turnover Patient Length of Stay Patient Outcomes Patient Satisfaction Medication Errors Patient Falls Components and Variables of the Model Components of the Model Variables Structure Nurse Experience level, knowledge, and skill level. Organizational Measures of availability of nursing staff (e.g., staff mix, daily staffing levels, nurse/patient ratios) and nurse assignment patterns (e.g., functional nursing, team nursing, primary or modular nursing). Patient Age, physical function at admission, severity of presenting problem, and co-morbidity. Process Nurses Independence Role Roles for which nurses are held accountable. These include the activities of assessment, decision making, intervention, and followup. Nurse-initiated treatments are also included (e.g., physiological comfort promotion, coping assistance, self-care facilitation, activity and exercise enhancement, immobility management, and nutritional support). Outcomes affected include symptom control, functional health status, and knowledge of self-care strategies. Outcomes Nurses Dependence Role Nurses Interdependence Role Nurse-sensitive patient outcomes Functions and responsibilities associated with implementing medical orders and medical treatments initiated by the physician. Outcomes affected include adverse events such as medication errors. Activities and functions that are partially or totally dependent on the functions of other health providers. These include monitoring and reporting of changes in the patient s health conditions and coordinating health services. Outcomes affected are the quality of intra-team and interprofessional communication and coordinating care. General patient state, behaviour, or perception resulting from nursing interventions. Outcomes include prevention of complications (such as nosocomial infections), symptom control, functional health outcomes, and satisfaction with care and cost.

159 151 Appendix E. NRC+Picker Sampling Plan The Survey Process Sampling Plan Participating hospital corporations and NRC+Picker collaboratively established a sampling plan. Deciding factors influencing the agreed-upon sampling plan included budget, achieving reasonable response rates, and which sites within the corporation were of primary interest. Hospitals were then charged with the responsibility of sending patient data files to NRC+Picker every month. Patient satisfaction data were collected for all 12 months of each fiscal year. Then, according to each hospital s sampling plan, a random sample was drawn from the patient data files, and surveys were mailed. Questionnaires were not sent to deceased patients, psychiatric patients, infants less than 10 days old, patients with no fixed address, or patients who presented with sexual assault or other sensitive issues. Mailing of Questionnaires Included in each patient mailing were an explanatory cover letter, a return envelope (postage-paid), and the questionnaire itself. The first mailing went out within a couple of weeks of NRC s reception of a hospital s monthly patient data file. To increase response rates, there was a second wave of mailings to patients whose first questionnaires were not returned within three weeks of the original mailing date. Inclusion/Exclusion Criteria Surveys that were returned without a single valid response were treated as nonresponses and dropped from the analysis. If a record had no valid responses to any of the evaluative questions on the questionnaire (i.e. it only had responses to demographic-type questions), then it was seen as having insufficient data and was excluded from the subsequent analysis.

160 Appendix F. OHRS Staffing Accounts 152

161 153 Appendix G. Technical Specifications Intensity of Care RN worked hours per patient visit RPN worked hours per patient visit Agency Nurse worked hours per patient visit NP worked hours per patient visit worked hours per patient visit Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 63511*1; 63511*2; 63513*1; 63513*2, 63514*1; 63514*2, 63515*1, 63515*2, 63516*1, 63516*2, 63811*1; 63811*2, 63813*1; 63813*2, 63814*1; 63814*2, 63815*1, 63815*2 (for 2004/05 Secondary Account ) Denominator NACRS: number of patient visits Numerator OHRS; Primary Account Sector Code 1* Type S Secondary Account 63512*1; 63512*2 (for 2004/05 Secondary Account ) Denominator NACRS: number of patient visits Numerator OHRS; Primary Account Sector Code 1* Type S Secondary Account *; *; *; *, *, *; *; *; *, (for 2004/05 Secondary Account 35090; 38090) Denominator NACRS: number of patient visits Numerator OHRS; Primary Account Sector Code 1* Type S Secondary Account 638***1; 638***2; 63516*1; 63516*2 (for 2004/05 Secondary Account 38010; 38090) Denominator NACRS: number of patient visits Numerator OHRS; Primary Account Sector Code 1* Type S Secondary Account 635***1; 635***2; 638***1; 638***2 (for 2004/05 Secondary Account 7295*4; 38010; 38090) Denominator NACRS: number of patient visits

162 154 Skill Mix RN proportion (RN worked hours divided by total staff worked hours) RPN proportion (RPN worked hours divided by total staff worked hours) Agency Nurse proportion (Agency nurse worked hours divided by total staff worked hours) Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 63511*1; 63511*2; 63513*1; 63513*2, 63514*1; 63514*2, 63515*1, 63515*2, 63516*1, 63516*2, 63811*1; 63811*2, 63813*1; 63813*2, 63814*1; 63814*2, 63815*1, 63815*2 (for 2004/05 Secondary Account ) Denominator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 635***1; 635***2; 638***1; 638***2 (for 2004/05 Secondary Account 7295*4, ) Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 63512*1; 63512*2; (for 2004/05 Secondary Account ) Denominator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 635***1; 635***2; 638***1; 638***2 (for 2004/05 Secondary Account 7295*4, ) Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account *; *; *; *, *, *; *1; *; * Denominator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 635***1; 635***2; 638***1; 638***2 (for 2004/05 Secondary Account 7295*4, )

163 155 Skill Mix Staff Adequacy NP proportion (NP worked hours divided by total staff worked hours) RN Staff to Patient Ratio (number of RN staff / number of patients) RPN Staff to Patient Ratio (number of RPN staff / number of patients) Agency Nurse Staff to Patient Ratio (number of Agency Nurse staff / number of patients) (NB: agency nurses report only worked hours) Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 638***1; 638***2; 63516*1; 63516*2 (for 2004/05 Secondary Account 38010; 38090) Denominator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 635***1; 635***2; 638***1; 638***2 (for 2004/05 Secondary Account 7295*4; ) Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 63511*; 63513*; 63514*; 63515*, 63516*, 63811*; 63813*; 63814*; 63815* (for 2004/05 Secondary Account 72951*) Divided by 1950 Denominator NACRS: number of patient visits Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 63512* (for 2004/05 Secondary Account 72952*) Divided by 1950 Denominator NACRS; Number of patient visits Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account *; *; *; *, *, *; *; *; *, * (for 2004/05 Secondary Account 35090; 38090) Divided by 1950 Denominator NACRS; Number of patient visits

164 156 Staff Adequacy NP Staff to Patient Ratio (number of NP staff / number of patients) Staff to Patient Ratio (total number of patient care staff / number of patients) Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 638*; 63516* (for 2004/05 Secondary Account 380*) Divided by 1950 Denominator NACRS; Number of patient visits Numerator OHRS Data Definition Sector Code 1* Type S Primary Account Secondary Account 635*; 638* (for 2004/05 Secondary Account 7295*; 380&) Divided by 1950 Denominator NACRS; Number of patient visits

165 157 Appendix H. NACRS Database The National Ambulatory Care Reporting System (NACRS) was developed by the Canadian Institute for Health Information (CIHI) in consultation with clinicians and managers. The NACRS is a data collection tool designed to capture information on client visits to facility- and community-based ambulatory care. The data set provides information on the type of patients seen in EDs, the type of care provided, and the outcomes of care. The database includes: demographic data, clinical data, administrative data, financial data, and service-specific data elements. The Ontario Ministry of Health and Long-Term Care mandated all Ontario EDs to begin reporting clinical activity using NACRS effective July, Every time a patient is registered at an Ontario ED, a NACRS record or abstract is generated for that visit and submitted to CIHI. The NACRS Abstract is the record of ambulatory care visit activity that is submitted to CIHI s NACRS database from each facility. Each abstract is associated with a client visit and contains a list of the relevant data elements to be submitted to the NACRS for that client visit. All abstracts sent to the NACRS contain an MIS (Management Information System) functional centre to determine the functional centre where the activity occurred. Prior to , a multiple contact record (MCR) was created when an Allied Health Professional (AHP) outside of the mandated MIS functional centre in which the visit occurred provides care and/or treatment. MCRs were discontinued in the reporting year. Hospitals were instructed to collect AHP care on the main visit abstract. In FY , NACRS was re-engineered to collect diagnosis- and intervention-related information solely in the ICD-10-CA/CCI coding system. The

166 158 enhanced ICD-10-CA replaces the earlier 9 th Revision of the International Statistical Classification of Diseases (ICD-9). The Canadian Classification of Health Interventions (CCI) contains a comprehensive list of diagnostic, therapeutic, and support interventions, and replaces the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (CCP) and the ICD-9-Clinical Modification (ICD-9-CM) intervention codes. In fiscal year , NACRS diagnosis and intervention coding were classified using the ICD-9-CM/CCP classification system. Since then, all clinical data submitted to the NACRS has been coded in ICD-10-CA/CCI. The postal code is a common variable in almost all CIHI databases. Along with the PCCF (Postal Code Conversion File), any standard geographical classification can be obtained, making it possible to compare with other databases. The forward sortation area (first three digits of a postal code) is typically the lowest level of aggregation normally available to external users under CIHI s Privacy and Confidentiality Policy.

167 159 Appendix I. Patient Satisfaction Descriptive Statistics ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK EDSAT EDREC Year 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE 2005/2006YE 2006/2007YE 2007/2008YE 2008/2009YE 2009/2010YE Std. Error of Std. Error of N Minimum Maximum Mean Std. Deviation Skewness Skewness Kurtosis Kurtosis 25, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

168 160 Analysis of Variance Between Groups: Year ANOVA Table ANSWER * Year EXPLAIN * Year TRUST * Year RESPECT * Year COURTESY * Year AVAILABILITY * Year DRNURSEWK * Year EDSAT * Year EDREC * Year Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) Sum of Squares df Mean Square F Sig E E E E E E E E E E E E E E E E E E Note: Unit Patient level unit of analysis Analysis of Variance Between Groups: Patient Age Group ANOVA Table ANSWER * Patagegrp EXPLAIN * Patagegrp TRUST * Patagegrp RESPECT * Patagegrp COURTESY * Patagegrp AVAILABILITY * Patagegrp DRNURSEWK * Patagegrp EDSAT * Patagegrp EDREC * Patagegrp Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) Sum of Squares df Mean Square F Sig E E E E E E E E E E E E E E E E E E Note: Unit Patient level unit of analysis

169 161 Analysis of Variance Between Groups: Patient Age Group ANOVA Table ANSWER * patgender EXPLAIN * patgender TRUST * patgender RESPECT * patgender COURTESY * patgender AVAILABILITY * patgender DRNURSEWK * patgender EDSAT * patgender EDREC * patgender Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups Between Groups Within Groups (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) (Combined) Sum of Squares df Mean Square F Sig E E E E E E E E E E E E E E E E E E Note: Unit Patient level unit of analysis

170 162 Appendix J. Patient Satisfaction Principal Component Analysis Inter-Item Correlation Matrix ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK ANSWER EXPLAIN TRUST RESPECT COURTESY AVAILABILITY DRNURSEWK Note: Unit Patient level unit of analysis PCA with All Items Component Variance Explained Initial Eigenvalues Extraction Sums of Squared Loadings % of Variance Cumulative % % of Variance Cumulative % Extraction Method: Principal Component Analysis.