Analysis of Cardiovascular Patient Data during Preoperative, Operative, and Postoperative Phases

University of Michigan College of Engineering Practicum in Hospital Systems Program and Operations Analysis Analysis of Cardiovascular Patient Data during Preoperative, Operative, and Postoperative Phases Final Report To: Mary Barry, RN Clinical Care Coordinator/Data Manager, Department of Cardiac Surgery Brett Cross, Clinical Information Analyst, Department of Cardiac Surgery Mike Donnelly, Administrator, Department of Cardiac Surgery Donald Likosky, PhD, Section Head, Section of Health Services Research and Quality, Department of Cardiac Surgery From: Kenneth Kokko, Industrial and Operations Engineering Siyu Liu, Industrial and Operations Engineering Andrea McAuliffe, Industrial and Operations Engineering Date: December 11 th, 2012 CC: Mark Van Oyen, Instructor, University of Michigan Andrei Duma, Industrial Engineer, University of Michigan Cardiovascular Center Joshua Pigula, Industrial Engineer, Program and Operations Analysis

Table of Contents 1.0 Executive Summary... 1 1.1 Findings... 2 1.2 Conclusions... 3 1.3 Recommendations... 3 2.0 Introduction... 4 3.0 Background... 4 4.0 Goals and Objectives... 5 5.0 Key Issues... 6 6.0 Project Scope... 6 7.0 Methodology... 6 7.1 Observations... 7 7.1.1 Preoperative Observation... 7 7.1.2 Operative Observation... 7 7.1.3 Postoperative Observation... 7 7.1.4 Clinical Information Analyst Observation... 7 7.2 Data Collection... 8 7.2.1 Preoperative Self-Collection Data... 8 7.2.2 Preoperative Historical Data... 8 7.2.3 Operative Historical Data... 9 7.2.4 Postoperative Historical Data... 9 7.3 Research... 9 7.4 Surveys... 9 8.0 Analysis... 10 9.0 Findings and Conclusions... 10 9.1 Observations Findings and Conclusions... 10 9.2 Data Collection Finding and Conclusions... 11 9.2.1 Preoperative Phase Findings and Conclusions... 12 9.2.2 Operative Findings and Conclusions... 16 9.2.3 Postoperative Phase Findings and Conclusions... 17 9.3 Research Findings and Conclusions... 19 9.4 Survey Findings and Conclusions... 20 9.5 Summary of Conclusions... 22 ii

10.0 Recommendations... 22 10.1 Training Recommendation... 22 10.2 Self-Audit Recommendation... 22 10.3 Monthly Meeting Recommendation... 23 10.4 Online Data From Recommendation... 23 11.0 Expected Impact... 23 Appendix A Preoperative, Operative, Postoperative Patient Data Collection Form... a Appendix B Self-collection Ladder Log... k Appendix C Revised Self-collection Form... l Appendix D Survey Questions... m Appendix E STS Risk Model... p iii

Table of Figures and Tables Figure 1 - High-level Flow Map of Current Patient Data Collection System at the CVC... 1 Figure 2 - Detailed Flow map of Current Patient Data Collection Process in the Dept. of Cardiac Surgery 5 Figure 3 Percentage of Incomplete Patient Data Forms over Total Number of Examined Forms in August, September and October 2012, N=sample sizes... 12 Figure 4 Distribution of Missing Information by Section on Preoperative Patient Data form over Aug, Sept and Oct 2012, N=sample sizes... 13 Figure 5 - Distribution of missing information by sections on pre-operative forms in a descending order of total counts over Aug, Sept and Oct 2012... 14 Figure 6 Total Count of missing information on pre-operative patient data forms from August 2011 to October 2012... 15 Figure 7 Total Count of Missing Information for Sections on the Operative Patient Data Form Containing the Top 55% of Errors for August 2012, September 2012, and October 2012... 16 Figure 8 Total Count of Missing Information for Sections on the Operative Patient Data Form Containing the Top 55% of Errors for August 2012 through October 2012 Combined... 17 Figure 9 - Count of Missing Information by Section on Postoperative Patient Data form over August, September and October 2012... 18 Figure 10 - Distribution of missing information by sections on postoperative forms in a descending order of total counts over Aug, Sept and Oct 2012... 19 Table 1 Summary of Sample Sizes for Preoperative, Operative and Postoperative phases.11 Table 2 Summary results of Self-collection forms for preoperative patient data form 15 Table 3 Comparison of survey responses from PAs and CCCs.21 iv

1.0 Executive Summary The Department of Cardiac Surgery at University of Michigan collects data on patients undergoing cardiac surgical procedures. These data are ultimately submitted to the Society of Thoracic Surgeons Database housed at the Duke Clinical Research Institute at Duke University. However, the Section Head of Health Services Research and Quality is concerned that the data collection process is inefficient. The Section Head expressed interest in reducing process inefficiencies and errors due to the volume of data collected per patient, an increase in patient volume, the limited human resources available and importance of accuracy in the data for the institution. Under the Cardiovascular Center s (CVC) current data collection system no one party takes ownership of the data forms and the resulting errors. Inaccuracies and discrepancies are pushed down stream for the Clinical Information Analyst to handle. An IOE 481 student team from the University of Michigan was asked to examine the current system for inefficiencies and sources of error and identify ways in which the system could be improved. Specifically, the team was tasked with providing recommendations to improve the process efficiency, reliability, and accuracy. Figure 1 below is a brief flow map illustrating the current process: Figure 1 - High-level Flow Map of Current Patient Data Collection System at the CVC Eight tasks were performed to evaluate and improve the CVC s current data collection process. Preoperative Observation. The team shadowed the Physicians Assistants in the 4C unit to understand their process for completing the preoperative data form. Operative Observation. The team observed two surgeries to understand how the operative data form is completed and who is responsible for the various sections. Postoperative Observation. The team observed the Clinical Care Coordinators (CCC) to understand their role and their method for completing the postoperative data form. Clinical Information Analyst Observation. The team observed the Clinical Information Analyst to understand how the data is inputted into the system and how errors upstream affect the analyst s role within the system. Historical Data Collection. The team collected 14 months of historical preoperative data, and 3 months of operative and postoperative data to examine trends and distributions. Physician Assistant Self-Collection. The team created a self-collection form to estimate the amount of time needed to complete the preoperative data form, capture an idea of how often PAs were interrupted while completing the data form and identify the common sources of trouble when completing preoperative data forms. 1

Research. The team interviewed the Quality Collaborative Coordinator for the State of Michigan to gain insight into the audit process, as well as discuss her experiences working with various hospitals within the state, including the CVC. In addition the team spoke with the data manager at St. Joseph Mercy Hospital in Ann Arbor to discuss their data collection system. Surveys. The team created a survey to distribute to the staff involved in completing the patient data forms during preoperative and postoperative phases. The surveys gathered opinions on what is most difficult or frustrating about the PA s or CCC s role and gathered feedback on possible recommendations for the data collection system. 1.1 Findings From the observations, the team found that the staff responsible for completing the data forms does not have knowledge of the system outside their current role or how their work affects the data collection process as a whole. This lack of knowledge reduces the staff s accountability for the collected data. The staff does not feel responsible for the form once it leaves their phase. The Clinical Information Analyst s role has evolved over time to accommodate the increasing amount of missing information pushed downstream. As a result, the analyst has taught himself how to find the missing information to save the time spent sending emails to staff responsible for completing the forms. Inaccurate data also affects the hospital on a larger scale; the state quality collaborative performs audits on the data, which may impact reimbursement to the institution. The Clinical Information Analyst highlights fields when information is missing. By highlighting, he can contact the staff responsible for the data and ask them to complete the missing field. The team relied on the clinical analyst s highlighting marks, to gather the count of missing information on a form. The team discussed the reliability of the highlighting with the Clinical Information Analyst; he told the team he has sent fewer emails and highlighted fewer fields because it takes less time finding the information himself than contacting the other staff. Due to this finding, the team expects the data collected for missing information to be a smaller total count than the actual amount of missing information. The preoperative patient data forms that the team collected, exposed a spike in missing information during the months of August 2011 and August 2012. The team learned new residents join the hospital mid July, and by August are responsible for completing some of the Health and Physical Examination (H&P) information for each patient. The H&P is a main source of information for the preoperative data forms; when the H&P is a less reliable source, the PAs leave more fields blank when completing the data form. During the operative phase, a form had at least one missing information 45% of the time, with the additional work sheets for coronary artery bypass graft surgery (CABG) being the most common source of missing information. There was also a large spike in missing information in August for operative patient data form. For the postoperative phase the most common source of error was in the discharge phase; the CCCs have to read through multiple documents to understand the patient s discharge experience and often find conflicting information on the medical records. When there is conflicting information, the CCCs are more likely to leave a field blank. There was no spike in August for the postoperative phase. The team s research consisted of speaking to various people who have experience with different data collection systems used in the state. The first person the team spoke with was the Quality Collaborative Coordinator for the Michigan Society of Thoracic and Cardiovascular Surgeons 2

Quality Collaborative (MSTCVS-QC). The team learned that the University of Michigan has not performed well recently among the audited sites in the state of Michigan. The student team also spoke with the data manager from St. Joseph Hospital, which has the second best audit score. St. Joseph s system s main difference is their Physician Assistants (PAs) and RNs carry the data cards for their patients with them and update them in real time as they see the patient. The close proximity of the preoperative and postoperative units makes this possible; the number of cases and geography of the University of Michigan s hospital make this unlikely to be a viable solution. The data manager at St. Joseph Hospital believes that if it is not possible to have only a handful of people collecting the data then the group collecting the data must be well trained. St Joseph conducts monthly meetings to review STS definitions and areas they need to improve. The team collected surveys from 11 of the staff at the CVC who are involved in the data collection process; this included 5 PAs and 6 CCCs. All staff indicated that they are adequately trained for the data collection processes and aware of the reasons and importance of collecting patient data. They also feel personally responsible for the completeness and accuracy of the data they collect. All 5 PAs indicated they never seek help when completing forms while 5 of 6 CCCs indicated they seek help at least once. The CCCs involved in the postoperative data indicated Post-Operative Complications the most difficult to fill out. The PAs indicated they have the most trouble filling out Hemodynamics/Catheterization/Echo and Patient Data which validates the historical data. All staff involved indicated they are not the ideal person to fill out the data form. 9 of 11 staff surveyed indicated that an online form would not be easier, but if initiated, auto-population of fields and dual monitors would increase efficiency. 1.2 Conclusions A lack of knowledge and accountability is adversely affecting the ability of the CVC to collect data accurately and efficiently. This lack of accountability and the cultural norms that have been established have become a burden on the department, resulting in a workload shift downstream. Not only are the jobs of the staff involved becoming more time consuming, the inaccuracy of the data collection is reflected in state audits. 1.3 Recommendations The team has five main recommendations as follows: Better training for the staff who collect data, including quarterly training sessions Monthly audits performed by the Data Manager Monthly meetings with involved staff to discuss definition changes and self-audit results The creation of an online data collection form with electronic feeds from other institutional databases. The continuation of this study with a focus on different software options and equipment to increase the accuracy of the data, simplify the data form completion process, and improve employee satisfaction The main focus of the education is better educating the staff on the current STS definitions and the other phases of the data collection system. Understanding how one phase s work affects the others will increase the accountability. The addition of training, monthly meetings, and audits will keep everyone up to date on definition changes and improvement areas. Leveraging existing institutional databases to populate the STS forms likely will reduce user-inputting errors, and create a more efficient system. 3

2.0 Introduction The Department of Cardiac Surgery at University of Michigan collects data on patients who are undergoing cardiac surgical procedures. Due to the extensive amount of data collected, the department believes that the current patient data collection system is burdensome, inefficient, unreliable and at times inaccurate. The Section Head of Health Services Research and Quality asked an IOE481 student team from University of Michigan to evaluate the current system through observation and data collection and provide recommendations for an improved system. Based on the observations and data collection, the team identified key areas within the current system to address, and recommended changes for improvement. The project is complete and this report presents the student team s data collection, analysis methods, findings and conclusions, and recommendations to improve the patient data collection system. 3.0 Background The Cardiac Surgery Department currently collects data from over 900 procedures performed annually on adult patients, and reports the data to the national database housed by the Society of Thoracic Surgeons (STS). The current data collection forms contain hundreds of fields. The data elements contained on these forms change over time, yet include the following general areas: patient and disease characteristics, method of the operation, and postoperative outcomes. A local cardiac surgeon is the head of the STS Adult Cardiac Surgery Database, and also the head of the Michigan Society of Thoracic and Cardiovascular Surgery Quality Collaborative (MSTCVS-QC). These STS data are used for quality assurance, quality improvement, public reporting, and research. A Clinical Information Analyst transfers the information from the paper-based patient data forms to an online database called Lumedx, a certified 3 rd party vendor, to generate and submit the STS data form. The forms are also used for generating surgeon operative notes, submitting referral letters, billing patients, as well as measuring and improving the quality of patient care. The patient s surgical acuity may be classified into one of the following categories: elective, urgent or emergent (see Appendix A). For each category, the current patient data collection system consists of three main phases of a patient s care: preoperative care, operative care, and postoperative care. Each phase involves many personnel and paper forms. Physician Assistants at the Cardiovascular Center Intensive Care Unit (CVICU) or on the 4C unit initiate the data recording process by completing preoperative patient data forms. Surgeons and Perfusionists in the operating room continue the process by completing operative patient data forms to record patient data during surgery. Clinical Care Coordinators (CCCs) then complete postoperative patient data forms after a patient is discharged from the hospital. All three sets of forms are collected by the clinical information analyst and entered into Lumedx. Within 30 days, more follow-up notes are taken by CCCs and entered by the Clinical Information Analyst into Lumedx. A more detailed flow map may be found below in Figure 2. 4

Figure 2 - Detailed Flow map of Current Patient Data Collection Process in the Dept. of Cardiac Surgery According to the Clinical Information Analyst, the most common errors and inefficiencies include missing fields, inaccurate information on the forms, or handwriting discrepancies. These errors and inefficiencies delay the time between the raw data recording and entry into Lumedx. 4.0 Goals and Objectives The goals of this project were to evaluate and suggest changes to improve the efficiency, reliability, and employee satisfaction of the data collection system while reducing redundancies and inaccuracies in the system. With the information collected from observations, historical data collection, research and surveys, the team developed recommendations and necessary resources to: Identify the inefficiencies between the processes of data collection and reporting Recommend methods to reduce the identified inefficiencies Recommend methods to minimize the amount of missing information of the data reporting system Identify material and other resources necessary to implement recommendations and improve the data collection process 5

5.0 Key Issues The following key issues drove the need for this project: Hundreds of data points are collected for each patient The data collection process is time and user-intensive, resulting in incomplete and inaccurate data Incomplete or inaccurate information delays the required data input into the STS database The department relies on accurate and timely data for billing patients, submitting operative notes, submitting referral letters, and measuring and improving the quality of care Staff along the continuum (data collection and inputting) are frustrated with the amount of time and paper the current system uses 6.0 Project Scope This project focused on the data collection process. During the preoperative phase the team studied only the urgent patient population in the Department of Cardiac Surgery. An urgent case refers to a hospitalized patient who needs surgery prior to discharge from the hospital; the patient is not stable enough to go home and then return for surgery. For the operative and postoperative phases, the patient data collection processes are centralized and completed using the same process for elective, urgent, and emergent cases. Therefore, the team collected operative and postoperative data on all three classes of acuity. The team studied only the Adult Cardiac Surgery Data Collection Form Version 2.73. Observations took place in the OR office, the CVC OR, the CVC Intensive Care Unit (ICU), the University Hospital 4C unit and the CCC s offices on the 5th floor of the CVC. The team did not directly analyze the care and clinical outcomes. Only the paperwork and its completion process were studied. 7.0 Methodology The team gathered information and data via four methods: observations, data collection, research, and surveys. First, observations for the PAs were done in the 4C unit. Perfusionists and surgeons were observed in the operating rooms of the CVC. The CCC s were observed at their offices on the 5th floor of the CVC. Second, data was collected in two ways, self-collection and historical data collection. The PA s on 4C used self-collection to collect real time data and the student team examined patient data forms from all three phases to collect historical data. Third, the student team conducted research to learn about patient data collection systems used in the past at the CVC as well as learn about best practices from a hospital with one of the best data audit scores within the MSTCVS-QC. Fourth, surveys were sent to the PAs and CCCs to gather qualitative information reflecting the current system. The rest of the methods section will discuss the details about the observations, data collection, research and surveys. 6

7.1 Observations The team completed four types of observations to understand the patient data collection system. The data collection system across each phase (preoperative, operative, and postoperative) of the patient s index admission was observed. The team also observed the process of inputting the data by the Clinical Information Analyst through the data entry screens into Lumedx. Observations helped the team make recommendations regarding improving data collection methods. 7.1.1 Preoperative Observation First, the team observed the preoperative phase in the CVICU and 4C. Since CVICU only completes an urgent patient data form 1 to 2 times a month, the team decided to observe elsewhere to gather enough data point and focused observations only on the 4C unit. The 4C unit completes about 20 urgent case forms a month. After observing for 8 hours over 1 week, the team found the patient data form to be completed at very different times throughout the day depending on the PA s schedule; the team observed only 6 data forms being completed. To improve the efficiency of data collection during the limited project timeline, the team created and implemented a self-collection form, which is described in Data Collection (section 7.2.1). 7.1.2 Operative Observation Second, the team observed the operative phase in the operating rooms of the CVC. The team observed 2 surgical cases, about 5 hours long each, to understand the roles of the perfusionist and surgeon during the completion of the operative patient data form. The team found that perfusionists and surgeons either complete the data forms during the surgery or immediately before leaving the operating room. The team discussed with the Section Head whether more cases should be observed. Due to the standardization of the process, the team and Section Head agreed it was not necessary to observe more surgical cases. 7.1.3 Postoperative Observation Third, the team observed the postoperative phase in the CCCs offices. The team shadowed the CCCs for 8 hours over 2 weeks. Observations revealed each CCC is responsible for completing an entire patient data form on her own. Due to the interruptions of a regular workday, many of the CCCs prefer to complete the forms at home or when fewer interruptions are expected. The team did not see a need to implement self-collection with the CCCs. The team found that the data collection method relatively consistent across all CCCs. 7.1.4 Clinical Information Analyst Observation Fourth, the team observed the Clinical Information Analyst in his Cardiac Surgery office at the CVC. The team observed the patient data forms from the three phases being entered into Lumedx. During this process, if an error is observed, the analyst highlights the missing information and attempts to locate the missing information in various clinical databases. If the analyst is unable to find the information, an email is sent to the person responsible for filling out the patient data form. If after 24 hours a response isn t received, another email is sent to both the person responsible and his/her supervisor. 7

7.2 Data Collection After observations were complete, the team collected data on the patient data forms for the preoperative, operative, and postoperative phases. To supplement the data collected through observations, the team implemented a self-collection form used by the PAs during the preoperative phase. Historical data was collected for all three phases. 7.2.1 Preoperative Self-Collection Data Initially, the team developed a self-collection form, called a ladder log, for the PAs to fill out while completing the preoperative patient data forms (Appendix B). The log was designed to capture: (1) the amount of time necessary to complete a data form and (2) whether or not the data form was completed all at once. The PAs marked on the ladder log when they started and stopped working on the data form. The PAs wrote down the data form number and a task code in between the start and stop lines, which described if the data form was completed with or without interruptions. After five days of self-collection, the team only received two partially completed ladder logs. The team concluded the ladder log was too cumbersome for the PAs, which contributed to the low number of responses. After ending the ladder log, the team re-evaluated what information the self-collection needed to capture. The team decided self-collection needed to be used to mark areas on the data forms that PAs had difficulty completing. The new self-collection form (Appendix C) allowed the team to decide if improvements must be focused on specific sections of the patient data form. The revised self-collection form was smaller and simpler to complete than the ladder logs, having checkboxes for most fields. The revised self-collection form was printed on a quarter sheet of paper, which was stapled to the front of the patient data forms. When the PAs completed a patient data form, they first marked their initials on the attached self-collection form. Once they are finished with the patient data form, they write an approximation of how long the patient data form took them to complete and checked any boxes for sections which they had trouble completing. The PAs self-collected for 13 days. The checkboxes on the self-collection form correspond to the sections on the patient data forms, which are: Patient Data Risk Factors Incidence Previous C Cardiac Status Pre-Op Medication Hemodynamics/Catheterization/Echo 7.2.2 Preoperative Historical Data The team examined 251 preoperative patient data forms that had already been entered into Lumedx by the Clinical Information Analyst from August 2011 to October 2012. The team only looked at data forms of urgent cases. The Clinical Information Analyst highlighted areas missing information, which was how the team determined when errors occurred. The team recorded the sections of the forms with missing information. 8

7.2.3 Operative Historical Data The team examined 201 operative patient data forms of elective, urgent, and emergent cases. These forms were from August 2012 to October 2012 and had already been entered into Lumedx by the Clinical Information Analyst. The operative forms are already broken into sections of data. The team used these sections when taking notes on when and where a data form was missing information. 7.2.4 Postoperative Historical Data The team examined 202 postoperative data forms from August 2012 to October 2012 that had already been entered into Lumedx. The data forms were for elective, urgent, and emergent cases. Missing information was identified in the same method as the preoperative and operative data forms. The team recorded the data sections on the form that were missing information. The team spoke to the CCCs to determine how to split the data form into sections. 7.3 Research The team contacted the Quality Collaborative Coordinator from the MSTCVS-QC and a nurse data manager at St. Joseph Hospital in Ann Arbor, MI, to learn about data collection processes used at other institutions. The Quality Collaborative Coordinator conducts audits for the MSTCVS-QC and has worked at the University of Michigan CVC for three months to help with the postoperative phase of the patient data forms. The coordinator has also experienced changes to the data collection process and is familiar with the different versions of the CVC s data collection system. The team discussed the Quality Collaborative Coordinator s experiences while working at the University of Michigan. The Quality Collaborative Coordinator also described the differences between various hospitals within the MSTCVS-QC, and reasons different data collection systems lead to better audit scores than others. The second source the team contacted was St. Joseph Mercy Hospital as recommended by the Section Head. The hospital has the second best audit score in the state and two of the CCCs worked at St. Joseph before coming to the University of Michigan Hospital. The team asked a data manager from St. Joseph about the hospital s data collection process and advantages and disadvantages the St. Joseph data collection process. 7.4 Surveys After completing observations, data collection, and research, the team wanted to collect employees comments and suggestions on the current patient data collection system. The team developed a survey using the Michigan survey developer, Qualtrics. The survey was distributed to the staff that completes the patient data forms during the preoperative and postoperative phases, including all eight PAs on 4C and all eight CCCs at the Department of Cardiac Surgery. The surveys gathered opinions on which sections of the patient data forms the clinical teams find most difficult or frustrating to complete. The survey also gathered staff comments on current system performance and suggestions on potential improvements. The student team collected 11 survey responses. The survey questions can be found in Appendix D. 9

8.0 Analysis The team analyzed the historical data of missing information from all the data sections of the collected preoperative, operative and postoperative patient data forms. The data was separated and analyzed by sections for each of the three different phases forms using Microsoft Excel. The team looked at the percentage of forms with at least one piece of missing information. The counts of missing information for each section was also studied based on monthly trends as well as total counts. The team used the self-collection form to compare sections the staff found difficult to complete and the sections with the most amount of missing information, as identified from the historical data collection. The self-collection forms also illustrated how long it takes to complete a form and how interruptions affect the data form completion. The team utilized the survey results to collect opinions and comments from PAs and CCCs to validate the frustration and difficulties expressed from the conversations during team s observation and data collection period. 9.0 Findings and Conclusions The following findings are organized the same way as the methods sections of the report, observations, data collection, research, and surveys. Each section discusses the qualitative and quantitative findings and conclusions for the project. 9.1 Observations Findings and Conclusions Through observations, the student team learned how the form is completed during the various stages of patient care. Different staff are responsible for different patient data forms, spread out through different phases of the index admission. The team was told by the staff that PAs, Perfusionists, CCCs, and the Clinical Information Analyst do not have any knowledge of the other staff s roles. The lack of downstream or upstream knowledge affects the quality of the patient data forms. The staff does not feel responsible for the patient data form after it leaves their phase. After observing the Clinical Information Analyst search for missing information, a cultural norm was identified. This cultural norm pushes information that cannot be found down stream, eventually leaving the Clinical Information Analyst responsible for the information on the patient data forms. While observing the Clinical Information Analyst, the team found the analyst s role has evolved over time to accommodate the increasing amounts of missing information he must complete. The team observed the analyst search through a patient s records, looking for missing information, multiple times. The Clinical Information Analyst has unofficially trained himself to look for patient data instead of sending increasing numbers of emails to the staff that are responsible for the initial completion of the data form. Searching for information increases the time it takes the analyst to enter one data form into Lumedx, and the likelihood for inaccurate information (as many fields require a clinical expertise). The student team s observations revealed multiple people completing the same work during various phases of the data form. The staff does not have any knowledge of the data collection 10

system outside their own phase. These observations highlighted the large amount of rework and the lack of transparency within the data collection system. 9.2 Data Collection Finding and Conclusions The team collected historical data from August 2012 to October 2012 for preoperative phase. The team examined preoperative data forms of urgent cases only and examined an average number of 20 data forms per month. For the operative and postoperative phases, the team examined patient data forms for all three types of cases (elective, urgent and emergent) and collected an average of 84 data forms over August, September and October 2012. First, the team looked at the number of patient data forms with missing information for each phase of patient care for August 2012, September 2012, and October 2012. Table 1 below presents the sample sizes of preoperative, operative and postoperative data forms over August, September and October 2012 respectively. Table 1 Summary of Sample Sizes for Preoperative, Operative and Postoperative phases Aug-12 Sep-12 Oct-12 Total Percentage Incomplete Total Percentage Incomplete Total Percentage Incomplete Preoperative 21 81% 20 20% 16 31% Operative 95 46% 75 43% 31 42% Postoperative 94 5% 76 14% 32 34% Due to the different number of forms for each phase, the team studied the percentages of forms with missing information. Figure 3 compares the percent of forms with missing information for each phase during August 2012, September 2012 and October 2012. 11

90% 80% 70% Percentages 60% 50% 40% 30% 20% Pre-operative, N=57 Operative, N=201 Postoperative, N=202 10% 0% Aug-12 Sep-12 Oct-12 Figure 3 Percentage of Incomplete Patient Data Forms over Total Number of Examined Forms in August, September and October 2012, N=sample sizes As seen in the Figure 3, operative forms are incomplete 45% of the time. The percentage of forms during the preoperative forms is the least consistent with a large spike seen in August 2012. The postoperative form is usually the most reliable form in terms of completeness, having the lowest percentage of forms with missing information every month but October. The team then broke down the historical data by sections on pre-operative, operative and postoperative patient data forms, to gain insights on the distribution of missing information varied by sections on each form. 9.2.1 Preoperative Phase Findings and Conclusions The team categorized the missing information from the preoperative phase into the same sections as on the patient data form. The team measured the amount of missing information in each section for August, September and October 2012; the results can be seen in Figure 4 below. 12

Count of Missing Information 50 45 40 35 30 25 20 15 10 5 0 Aug-12, N=21 Sep-12, N=20 Oct-12, N=16 Figure 4 Distribution of Missing Information by Section on Preoperative Patient Data form over Aug, Sept and Oct 2012, N=sample sizes Figure 4 shows September to have 2 or less pieces of missing information per section, while October has a range of 0 to 10 pieces of missing information per section. The team believes the October data is the most accurate because the Clinical Information Analyst knew about the student project and tried to highlight every time he found a field missing information. In October 2012, 31% forms had missing information. August 2012 shows more counts of missing information for most sections than September and October 2012. The higher total counts per field August 2012 coincides with the greater percentage of forms with missing information found in Figure 3. After looking at the sections by month, the team looked at the distribution of missing information for all three months together. Figure 5 illustrates which sections have the most counts of missing information over the three months. 13

Count of Missing Information 60 50 40 30 20 10 0 Figure 5 - Distribution of missing information by sections on pre-operative forms in a descending order of total counts over Aug, Sept and Oct 2012, sample size = 47 data forms As figure 5 illustrates, Hemodynamics, Catheterization, Echo and Patient Data are the two sections with most missing information. Both of these sections rely on a patient s most recent lab results. The team learned that the fields are often left blank while waiting for the most recent laboratory values to be completed. If the PA does not remember to complete the sections before the patient is taken to surgery, the responsibility moves downstream to the Clinical Information Analyst. The analyst must either email the staff or search through the patient s records to find the missing information. Given the spike in August 2012 the team decided to collect data back to August of 2011 to determine if there was a pattern to the data, this is shown below in Figure 6. 14

Count of Missing Information 100 80 60 40 20 0 Figure 6 Total Count of missing information on pre-operative patient data forms from August 2011 to October 2012, sample size = 251 data forms The figure above indicates a cyclical behavior with missing information on the data forms. A discussion with the PAs on 4C and the client indicated an addition of new residents at the end of July of each year might affect the preoperative patient data. The PAs retrieve information for the patient data forms from History and Physical reports which are sometimes completed by the new residents. Differences in syntax or missing information will cause the PAs to have difficulties completing the patient data forms. The team also analyzed how the preoperative data would affect the STS Risk Model, which is a list of factors used to calculate the patient s estimated risk of morbidity and mortality. When information on the preoperative patient data form is missing, the risk calculation will be negatively influenced. Through interviews, the team learned that the surgeons do not use the formal STS risk models when calculating a patient s mortality rate. Often they use past experiences and knowledge to calculate a general mortality risk. The student team believes that if fewer fields are missing on the data form, the STS Risk Model may be used more frequently and accurately, which may result in improvements in informed consent and care. A sample of STS Risk Model sheet may be found in Appendix E. The team implemented a self-collection form for two weeks, to collect information on how PAs complete the preoperative patient data form and in which section(s) they have the most difficulty completing. The team collected 11 completed self-collection forms by the end of data collection period. Table 2 summarizes results gathered from the self-collection forms. Table 2 Summary results of Self-collection forms for preoperative patient data form Total Forms Number of times more than 1 person Number of times a PA had trouble with a section on the data form completed a form Patient Data Hemodynamics/Catheterization/Echo 11 2 3 1 5 Number of Interruption 15

The team compared the self-collection results to the historical data collection. Patient Data and Hemodynamics/Catheterization/Echo were the sections the PAs had the most trouble finding patient information for. These results validate the historical data collection; Patient Data and Hemodynamics/Catheterization/Echo had the highest count of missing information as found from historical data collection. 9.2.2 Operative Findings and Conclusions After the preoperative data was analyzed the team looked at the operative patient data. The operative data forms have over a hundred data points. To avoid separating the operative form into over a hundred sections, the team only analyzed the sections on the form that made up the most frequent 55% of missing information. Figure 7 shows the count of missing information on sections of the operative form for August to October 2012. Count of Missing Information 20 18 16 14 12 10 8 6 4 2 0 Aug-12, N=97 Sept-12, N=74 Oct-12, N=30 Figure 7 Total Count of Missing Information for Sections on the Operative Patient Data Form Containing the Top 55% of Errors for August 2012, September 2012, and October 2012 Figure 7 also shows an increase in August s missing information, similar to the preoperative patient data form. September 2012 and October 2012 follow a similar trend. The additional CABG Worksheets that are used for CABG (Coronary Artery Bypass Graft) surgeries have the largest amount of missing information. The team was informed that the CABG sheet was newly added to the operative data form within the past year. The patient data from the preoperative and operative forms are used to generate the operative note, which is entered into a patient s medical record after surgery. When information is missing, the Clinical Information Analyst has to search for the information, which increases the time to generate an operative note. 16

The team also looked at the counts of missing information for the top 55% of sections for all months combined, as seen in Figure 8 below. 30 25 20 15 10 5 0 24 22 14 11 11 11 10 9 Figure 8 Total Count of Missing Information for Sections on the Operative Patient Data Form Containing the Top 55% of Errors for August 2012 through October 2012 Combined The results from Figure 8 confirm the student team s analysis from Figure 7. As stated before, the operative forms contain hundreds of data points and when fields are missing, increasing the amount of time it takes the Clinical Information Analyst to enter the data into Lumedx. 9.2.3 Postoperative Phase Findings and Conclusions The team collected and analyzed the postoperative data in a similar way as the preoperative and operative. The student team spoke to the CCCs to understand how the postoperative patient data form can be separated into sections. The count of missing information for each section, by month is seen in Figure 9 below. 17

16 Count of Missing Information 14 12 10 8 6 4 2 Aug-12, N=94 Sep-12, N=76 Oct-12, N=32 0 1 (supportive devices) 2 (extubation) 3 (post-op testings) 4 (Discharge) Post-op Complications Figure 9 - Count of Missing Information by Section on Postoperative Patient Data form over August, September and October 2012 Figure 9 illustrates a consistent count of missing information between months. The CCCs data collection system does not show the same spike in August 2012 as seen in the preoperative data. The CCCs had additional help completing the data forms during August, and the additional staff was very well trained on the definitions for STS data collection. The extra help resulted in lower counts of missing information for the postoperative phase during August. The data follows a similar trend for each section across all three months, except for the discharge section. The distribution of missing information can be seen in Figure 10 below. 18

30 25 24 20 15 10 5 7 5 3 2 0 4 (Discharge) 3 (post-op testings) Figure 10 - Distribution of missing information by sections on postoperative forms in a descending order of total counts over Aug, Sept and Oct 2012 Section 4 (Discharge) has three times the amount of missing information as the next highest section. The team spoke to the CCCs to learn more about potential causes of this spike. The CCCs report that when they are completing the discharge section, they often have to read through multiple documents to understand the patients discharge experience. These documents can sometimes have conflicting information. After completing the historical data analysis and findings, the team spoke with a Quality Collaborative Coordinator and a data manager from St. Joseph Hospital to learn more about the CVCs data collection system as well as another hospital s system. 9.3 Research Findings and Conclusions 2 (extubation) 1 (supportive devices) Post-op Complications After speaking to a Quality Collaborative Coordinator from the MSTCVS-QC, the team learned about different data collection systems used throughout the state. According to the Coordinator, the University of Michigan Hospital System (UMHS) has not performed well recently in its audits by MSTCVS-QC. The Coordinator explained poor audit scores are common for hospitals as large as the UMHS. Smaller hospitals have fewer people trained and responsible for collecting data. At a smaller hospital, the preoperative and postoperative care units are closer together; the data forms are completed within the same area for all three phases. The CVC does more cases than smaller hospitals like St. Joseph and has more than three areas where the patient data forms are completed for cardiac surgery patients, and therefore requires more staff for the data collection process. The number of cases as well as the geography of the hospital is not something the CVC can change; however, communication between phases and accountability for the information may be improved. The coordinator also talked about self-auditing. If the CVC were to randomly audit a few data forms from the three phases each month, the department would have a better understanding of which sections on the data forms need to be improved. 19

The Quality Collaborative Coordinator and the Section Head encouraged the team to contact St. Joseph Hospital because St. Joseph has one of the best audit scores in the state. The student team spoke to the data manager from St. Joseph Hospital. At St. Joseph, the PAs and RNs are responsible for the patient data forms of the patients they round on during their shift. The patient data cards are carried by the staff throughout the entire day and updated in real time after a PA or RN sees a patient. The preoperative and postoperative units at this hospital are located near to each other, which allows the same group of staff to complete the patient data forms for both phases. St. Joseph has a little more than half of the CVC s caseload. Due to the fewer cases, the two data collection systems are not directly comparable. However, the team did learn two important techniques to improve the data collection system. The data manager believed that if a hospital cannot have only a few people collecting data for the patient data forms, then the group that is collecting data must be well trained. Uneducated collectors will not understand the STS definitions and will complete the data forms incorrectly. St. Joseph meets monthly to discuss new STS definitions and review old definitions. They also talk about areas they need to focus on to improve the quality of the data forms. After speaking to both the MSTCVS-QC Quality Collaborative Coordinator and the Data Manager, the student team created a survey to gather employee opinions about the current system and possible areas of improvement. 9.4 Survey Findings and Conclusions The team collected eleven responses out of sixteen total survey recipients: five responses from the PAs at 4C unit, and six responses from the CCCs. The survey asked about the staff s perspective on current patient data collection process, as well as their opinions on inefficiencies and potential improvements. There are two versions of survey questions tailored according to the different roles. A list of findings from the survey responses are: Among the eleven responses, only one CCC is a new hire (<1 year experience) All of the respondents believe they are aware of the reasons and importance of collecting patient data The respondents expressed personal responsibility for the completeness and accuracy of the information on the patient data form The respondents expressed adequate training for the data collection processes Eight of the respondents indicated that they refer to STS definition sheet more than once when completing one patient data form All five PA respondents expressed that they never seek help from others when completing the forms Five out of six CCC respondents stated they do seek help once or twice when completing one patient data form Hemodynamics/Catheterization/Echo is the most difficult section to complete during the preoperative phase, followed by Cardiac Status, Previous CV Interventions and Patient information with equal rankings 20

Post-op complications is the most difficult section to complete during the postoperative phase, followed by Second section (extubation time), and then Third section (post-op testing) Three PA respondents stated that the Clinical Information Analyst is the best person to complete the pre-operative patient data form since the task is better matched with their (the clinical information analyst) role responsibilities Two PA respondents stated that someone else should be specifically trained to take over this patient data collection job since it s time consuming and should have another clerical role Four of the CCC respondents expressed that the PAs and nurses who directly take care of patients are better to complete the postoperative form since the PAs and nurses have better access to the required information (real time data) Two CCC respondents expressed that the Clinical Information Analyst would be the best person to complete the postoperative patient data form since the task is better matched with their role responsibilities Nine of the respondents do not think an online data from would be easier to complete than a paper form Auto-population of some information from the patient s record online is the most preferable improvement feature, followed by dual-screen at workstation and training workshop Most of the respondents believe an average of 32% of information on the patient data form can be auto-populated, based on the existing data from patient s online record 56% of time the staff need to search for information from other s written notes To gain more insights on how two different groups of respondents view the current patient data collection, the team broke down the findings from responses into Table 3 below: Table 3 Comparison of survey responses from PAs and CCCs PA(5 respondents) CCC (6 respondents) Refer to STS definition sheet 2 times 1-2 times Seek help from others Never 1-2times Suggested another role Clerical role Inpatient team Online form easier No 1/3 Yes Amount of data can be auto-populated ~40% ~30% Amount of data from written notes >50% >50% Auto-population preferable Yes Yes The team studied the survey results and discussed the comments and suggestions from PAs and CCCs. The team then concluded following major insights: 1. The PAs preference of an extra person trained only for the patient data form completion reflects the PAs heavy workload. The reflection aligns with the goal to reduce any unnecessary workload. 2. Despite resistance to an online patient data form, half of responses suggest autopopulation of some existing data would ease the data form completion process. 21

3. More than half of the CCCs suggest the inpatient team has better access to the real time patient information, specifically discharge information. The CCCs currently have to search for 56% of the data from notes written by the inpatient team. It would contribute to overall efficiency if the staff that complete the notes also complete corresponding fields on the postoperative patient data form. 9.5 Summary of Conclusions The team summarized the findings draw from observation, data collection, research interviews and survey into 4 major conclusions: 1. A large amount of rework exist the data collection processes, which lead to the inefficiency and frustration among staff 2. All three phases of the data forms reveal missing information, the distribution of missing information of some phases are less consistent than others over different time periods 3. The fact that staff frequently refer to the STS definition sheet or seek help from others when completing a data form results in the increase of completion time 4. Observations and interviews illustrated a lack of standardization and transparency across three phases which leads to rework, unreliable data and common frustration 10.0 Recommendations After completing the methods and analysis, the team has developed four main recommendations. The recommendations consist of training, self-audits, staff meetings, and an online data collection form. 10.1 Training Recommendation First, better training should be available for the staff who collect data. Quarterly training sessions will help employees refresh their understanding of the STS definitions and help standardize the process for completing the patient data forms. Employees will be trained in the same manner and inconsistencies can be minimized. M-Learning modules should be created to teach STS definition changes and best practices. Best practices include knowing which data is the source of truth if there are inconsistencies in the system and also knowing what procedure to follow or who to ask for help when a field cannot be completed. A M-Learning module will also ensure that each data collector has been exposed to training modules, and must achieve a certain level of competency. 10.2 Self-Audit Recommendation For the second recommendation, the Data Manager and Clinical Information Analyst should perform monthly self-audits on the data forms. Ten forms could be randomly selected from the 22

prior month and checked for missing information and data accuracy. Self-audits will help the department monitor its performance and understand which areas of the data forms should be focused on for improvements. 10.3 Monthly Meeting Recommendation Third, the CVC should hold monthly meetings for all of the staff involved in STS data collection. The meetings will provide the opportunity to discuss what procedures are considered best practices, data collection issues found during self-audits, and ways each phase of the process affects downstream work. The meetings can be informal and held as quick huddles for the teams on each unit, or they can be more formal and held in a library with every unit gathered together, depending on the issues that need to be discussed. Monthly meetings will encourage ownership of the data and also help standardize the collection process. The results from the self-audits should be presented at monthly meetings; discussing the problems of the data collection system (with the people involved) will increase accountability to the collection system. 10.4 Online Data From Recommendation The fourth recommendation has two parts. First, the team believes it will be beneficial to create an online data form that extracts information from existing institutional databases. Extraction will reduce manual searching for information by the staff. When a mouse is hovered over a field on such an online form, the definition could appear, eliminating the need for the staff to search through multiple pages of the STS definition packet. The online form could also prevent the staff from submitting an incomplete form by alerting the user of missing fields and requiring they be completed before submission. For the second part of this recommendation, the student team recommends continuing the study to learn more about software options available for patient data collection. The team suggests investing in dual monitors, to help staff eliminate the need to switch between windows frequently. More studies must also be done to determine which fields on the data forms can be auto-populated. 11.0 Expected Impact In conclusion, the team expects the following results from the recommendations. Training sessions, M-Learning modules and monthly meetings would increase the standardization of the data collection process. With a more standardized process and better training, the rework done by the Clinical Information Analyst can be reduced by at least 50%. Using an online data form would also reduce the amount of rework and improve the efficiency of the process by at least 30%. Conducting self-audits and holding monthly meetings will increase the ownership of the data, resulting in a more reliable data collection process. Finally, with more reliable data, models such as the STS Risk Model can be used more frequently and reliably. If risk models are more accurate the quality of patient care will undoubtedly improve. 23

Appendix A Preoperative, Operative, Postoperative Patient Data Collection Form Patient Name: Preoperative CPI: Surgeon: Race White Black Asian Am.Indian/Alaskan Nat Native Hawaiian/Pacific Islander Other Hispanic, Latino, or Spanish Ethnicity Referring Cardiologist/City: Referring PCP/City: Admit Source: Elective Admission Emergency Department Other Transfer in from another acute care facility (Name of Hospital ) WBC HCT PLT INR Tbili Albumin A1c Creat HIT antibodies Not Drawn Risk Factors Incidence Cigarette Smoker 1 yr First CV surg (1st op) First Re op CV surg (2nd op) Current Smoker 2 wks Second Re op (3rd op) Third Re op (4th op) Other Tobacco Use 1yr Family CAD Fourth or more Re op (5th or more) Family Aortic dz Previous CV Interventions Diabetes If Yes Previous Cardiac Interventions If Yes Diabetes control None Diet Oral Insulin Other CABG Dyslipidemia Valve If Yes Renal Failure Prev AV Replacement Dialysis Prev AV Repair Hypertension Prev MV Replacement Infectious Endocarditis Prev MV Repair Type Treated Active Prev TV Replacement Culture Negative Staph aureus Strep species Prev TV Repair Coag neg staph Enterococcus Fungal Other Prev PV Replace/Repair Chronic Lung Disease No Mild Moderate Severe Prev AV Balloon Valvuloplasty PFT FEV1 % DLCO% No DLCO Prev MV Balloon Valvuloplasty ABG on Room Air If Yes po2 pco2 Prev Transcatheter Valve Replacement Home O2 Prev Perc Valve Repair Inhaler/Bronchodilator Date known If Yes Date Sleep apnea If No Estimate # Months Liver Disease Immunocompromised Oth Card Surg If Yes Arrhyth. Surg Yes No PAD Congenital Unresponsive state Syncope ICD Cerebrovasc Disease PPM If Yes CVA If Yes 0-2 wks > 2 wks PCI If Yes TIA PCI during this admit Yes UM Yes OSH No Carotid Stenosis > 79% None Right Left Both If Yes Why Surgery? PCI Complication Right 80-99% 100% Left 80-99% 100% PCI Failure Hx CEA/stent PCI/ CABG Hybrid Illicit Drug Use PCI Stent If Yes Bare Metal Unknown ETOH Use <=1 drink/wk 2-7 drink/wk >=8 drink/wk Drug-eluting Pneumonia No Recent Remote PCI Interval <=6 hr > 6 hr Mediastinal Rad Cancer within 5 yr Type Oth CV Proc Five Meter Walk T1 sec T2 sec T3 sec Connective Tissue Dz Transplant Other a

Cardiac Status Prior MI If Yes MI When <=6Hr >6 but < 24 Hr 1-7 Days 8-21 Days > 21 days Anginal Class No Angina CCA I CCA II CCA III CCA IV CHF within 2 weeks If Yes NYHA 1 2 3 4 (please mark yes to CHF for all VAD/Transplant cases) Prior CHF Cardiac Presentation No Symptoms/Angina Symptoms but no ischemia Cardiogenic Shock Stable Angina USA Non-STEMI STEMI Resuscitation w/in 1 hr Arrhythmia None Remote > 30 days Recent w/in 30 days VT/VF Sick Sinus 2nd Deg HB 3rd Deg HB AF/Flutter VAD in place? Implanted at U-M OSH VAD implant date / / Previous VAD indication: Bridge to Tx Bridge to recovery Destination Paroxysmal Cont/Persistent Post cardiotomy ventricular failure Device Malfunction End of life Type LVAD RVAD BiVAD TAH Device name Pre Op Medication Beta Blocker 24 hr Contraindicated AceI/ARB 48 hr IV Nitrates 24 hr Anticoagulant 48 hr If Yes Heparin (unfractionated) Heparin (LMW) Thrombin Inhibitor Other Antiarrhythmics 24 hr Coumadin 24 hr Inotropes 48 hr Steroids 24 hr Aspirin 5 days Lipid Lowering 24 hr If Yes Statin Non-statin Both ADP Inhibitor 5 days If Yes ADP Inhibitor D/C # days before surgery Antiplatelet 5 days Glycoprotein Iib/IIIa 24 hr If Yes Abciximab (ReoPro) Tirofiban (Aggrastat) Eptifibatide (integrilin) Thrombolytics 48 hr Hemodynamics/Cath/Echo Cath done Date / / No. of Dz Cor Vessels None 1 2 3 (LM> 50%=2) Left Main >= 50% Proximal LAD >= 70% EF Done EF Method LV gram Nuc Med Estimate Echo MRI/ CT Other LV Systolic Dimension (LVSD) (LVIDs) (mm) LV End Diastolic Dimension (LVEDD)(LVIDd) (mm) PA Systolic during Cath PA Systolic Pressure Valve Disease If Yes Valve Stenosis Mean Grad Valve area Aortic Mitral Tricuspid Pulmonic Insufficiency None Trace Mild Moderate Severe PreOp Complete PreOp Entered By/Date By/Date b

c

d

e

f

g

h

IABP inserted Post op If Yes Reason Hemo instability PTCA support USA Prophylatic Date IABP Removed Catheter Assist Device If Yes Impella Tandem Heart Other ECMO When inserted: PreOp IntraOp PostOp Reason inserted: Post Operative Date Device Removed If Yes Hemo Instability CPB wean failure PCI Failure Other Initiated PreOp IntraOp PostOp Reason Cardiac Failure Resp. Failure Hypothermia Rescue/ Salvage Blood Usage (Data entry) RBC FFP Cryo Platelet Highest Creatinine ICU Admit Date/Time Readmit ICU Readmit to ICU D/C ICU Date/Time D/C ICU Ext in OR Date/Time Date/Time Initial Ext Date/Time Reintub Dt/Time Ext Dt/Time Post Op Echo If Yes Insufficiency None Trace Mild Moderate Severe AV MV TV Post Op EF EF Cardiac Enzymes If Yes Peak CKMB Peak Troponin I Peak Troponin T 12 lead EKG Not Performed No Significant Changes New Pathological Q Waves or LBBB Imaging Study Not Performed Angio evidence new thrombosis or occlusion of graft or native coronary Imaging evidence of new loss of viable myocardium No evidence of new myocardial injury Discharge Date Discharge Disposition: Home Oth. Hosp ECF/TCU/Rehab Cardiac Rehab N/A Death Hospice Perm. Nursing Home Smoking Cessation N/A Cardiac Neuro Renal Antibiotics D/C 48 hr N/A Vascular Infection Pulm Valvular Unknown Other Discharge Medications Death location ADP Inhibitor Antiarrhythmic Aspirin Ace Inhibitor or ARB Beta Blocker Lipid Lowering Coumadin Direct Thrombin Inhibitors Contraindicated Contraindicated No, Not Indicated Contraindicated Contraindicated Statin Non-statin Both Other i

Post Operative Events/Complications Complication Yes Complication Yes Operative Pulmonary Re Op Bleeding/Tamponade Prolonged Vent > 24 hr If Yes Bleed Timing Acute Late Pneumonia Re Op Valvular Dysfunction Venous Thromboembolism Re Op Graft Occlusion If Yes PE Re Op Other Cardiac DVT Re Op Other Non cardiac Pleural Effusion w/drainage Open Chest/ planned delayed closure Renal Sternotomy Issue (ANY) Renal Failure (3x preop or creat >=4 or New HD) If Yes Sternal instability/dehiscence Yes No If Yes Dialysis Required Yes No Infection Dialysis after D/C Yes No Surgical Site Infection Ultra Filtration If Yes Superficial Sternal Vascular Deep Sternal Iliac Femoral Dissection Mediastinitis If Yes Acute Limb ischemia Diagnosis Date Other Left Open with packing Rhythm requiring Device Wound Vac PPM ICD PPM/ICD Muscle Flap Cardiac Arrest Omental Flap Anticoagulant Event Thoracotomy Tamponade (no OR) Harvest/Cannulation GI Event (specifiy) Wound w/ packing (not primary) Wound w/ Wound Vac (not primary) Multi system Failure Sepsis Afib If Yes Positive Blood Culture Yes No Aortic Dissection Neurologic Laryngeal nerve injury Stroke Phrenic nerve injury TIA Other (specify) Encephalopathy/Coma If Yes Anoxic Embolic Drug Paralysis Metabolic Bleed Other If Yes Type: Perm Transient PostOp Complete PostOp Entered By/Date By/Date j

Appendix B Self-collection Ladder Log k

Appendix C Revised Self-collection Form l

Appendix D Survey Questions m

n

o

Appendix E STS Risk Model p

q

r