5
Case Study Vanderbilt’s Journey Toward System-Supported Practice1

William W. Stead


This discussion focuses on how to shift from expert-based practice to system-supported practice. To begin, one important observation is that if a unit or institution performs each of seven practices 90 percent of the time, the probability that it will perform all seven for an individual patient is only 48 percent. Health care providers and practitioners often do not realize this, however, because they report successes when they achieve 90 percent on individual practices. Overall, they may not achieve the desired clinical outcomes.

A shift from expert-based practice to expert-managed, system-supported practice can increase the level of success. A case example is the work done at Vanderbilt University for the past two years to improve ventilator management. This example, described below, clarifies how a clinical unit can deal with two challenges. Challenge I involves translating evidence into standard practice (Stead and Starmer, 2008). Challenge II is creating something that works, like closed-loop control, while keeping people in the loop (Stead et al., 2008).

In expert-based practice, experts are supposed to bring knowledge and technical skills, assimilate data, make wise decisions, and do what

1

This chapter is based on the author’s presentation and responses to questions during a plenary session of the NAE-IOM workshop on Harnessing Operational Systems Engineering to Improve Traumatic Brain Injury Care in the Military Health System on June 11, 2008. The author would like to thank NAE staff member Jessica Buono for her assistance in preparing this material for publication.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 69
5 Case Study vanderbilt’s Journey Toward System-Supported Practice1 William W. Stead This discussion focuses on how to shift from expert-based practice to system-supported practice. To begin, one important observation is that if a unit or institution performs each of seven practices 90 percent of the time, the probability that it will perform all seven for an indi- vidual patient is only 48 percent. Health care providers and practitioners often do not realize this, however, because they report successes when they achieve 90 percent on individual practices. Overall, they may not achieve the desired clinical outcomes. A shift from expert-based practice to expert-managed, system- supported practice can increase the level of success. A case example is the work done at Vanderbilt University for the past two years to improve ventilator management. This example, described below, clarifies how a clinical unit can deal with two challenges. Challenge I involves translating evidence into standard practice (Stead and Starmer, 2008). Challenge II is creating something that works, like closed-loop control, while keeping people in the loop (Stead et al., 2008). In expert-based practice, experts are supposed to bring knowledge and technical skills, assimilate data, make wise decisions, and do what This chapter is based on the author’s presentation and responses to questions during a 1 plenary session of the NAE-IOM workshop on Harnessing Operational Systems Engineering to Improve Traumatic Brain Injury Care in the Military Health System on June 11, 2008. The author would like to thank NAE staff member Jessica Buono for her assistance in preparing this material for publication. 

OCR for page 69
0 SyStEmS ENgiNEEriNg to imProvE trAumAtiC BrAiN iNjury CArE is necessary to ensure that their decisions are carried out. This practice is built around the expertise of the physician, and the performance of the system depends on the performance of this individual. Disagree- ment is expected among experts, and the performance of the system is no better than the performance of the individual expert. Typically it is worse, and experts are responsible for recognizing and learning from their mistakes. In comparison, the idea behind system-supported practice focuses on the system’s performance; teams of people, a well defined process, and information technology tools work in concert to produce desired results consistently. People bring compassion and judgment, the process brings simplification and standardization, and information technology reduces dependence on memory and forces action when needed. Collectively, the goal is to ensure the desired performance every time, and, if this fails, each failure becomes an immediate and iterative improvement. Figure 5-1 depicts this systems approach to health care, which joins system devel- opment to cycles of system-supported practice. The left-hand side represents iterative cycles of system development. First, a high-priority population is selected and defined, such as ventilator patients. The term population is used to imply a condition that needs to be managed to Individualize and Act • Assess • Plan • Order Pick Population •Risk •Cost System-Supported •Variability Practice Evidence System • Research Development • Guidelines • Practice database Monitor and Correct Process Patient • Sentinel events Status Workflow • Process outcomes Results • People’s roles • Clinical outcomes Trends • Process • Technology tools FIguRE 5-1 Systems approach to care. Figure 5-1.eps

OCR for page 69
 CASE StuDy: vANDErBilt’S jourNEy achieve a consistent outcome. Although only one population is worked with at a time, a patient can be in multiple populations. After selecting a population, an evidence base is gathered related to the target population, including literature on clinical research and consensus practice guidelines. Each consensus guideline is decomposed into a column of a table with a row for each practice it recommends. In the case of ventilator management at Vanderbilt, the table had a column for the University Hospital Consortium (UHC), one for the Institute of Healthcare Improvement (IHI), and one for the Centers for Disease Control and Prevention. The completed table displayed the variability in available evidence. A small group of subject-matter experts (two to six individuals) was then assembled to build a straw-person set of standard practices and an idealized process for executing them (Figure 5-2), as well as methods of measuring performance of the practice. A common fact base was compiled by documenting actual performance against the straw-person set of practices. This is the starting point for a model. At this point, the focus changes from a common fact base to cross- enterprise agreement. In this instance, the focus was on intensive care units. Executives, medical, and nursing leadership from each unit, subject-matter experts, and key support personnel came together for an intensive day of cross- enterprise design. They used the common fact base developed from the straw person to identify problematic practices, processes, and mea- sures. Then they broke into groups to work out designs for alternative solutions, reported their results, and agreed on revisions to the straw person. The design was limited to changes the group believed could be implemented in 45 days or less. It was also agreed that there would be a standard set of practices used throughout the entire medical center. For ventilation management, this meant these practices would be used from the neonatal intensive care unit through the burn and trauma units. These constraints eliminated many “blue sky” suggestions. Through iterative cycles, evidence was translated into actionable standard practices. Take, for example, stress-ulcer prophylaxis. The UHC recommendation stated that mechanically ventilated patients are at high risk of developing gastrointestinal (GI) bleeding and should receive stress-ulcer prophylaxis, unless medically contraindicated. The IHI recommendation stated that stress ulcerations are the most com- mon cause of GI bleeding in ICU patients and that the presence of GI bleeding is associated with a five-fold increase in mortality compared

OCR for page 69
 Implementation Shift/daily assessment of bundle Real-time dashboard for all Orders of bundle compliance ventilator patients in all ICUs (with electronic pop-up (with alerts to unit managers) 1. Physician orders reminders for deviations) vent mode and settings Documentation in HED as checklist 1. Elevate head of bed 2. Place sign on bed 2. Physician orders 3. Document in HED ns “Standard Ventilator notes Practices” Stress-Ulcer Prophylaxis Wean Pt comes to Pt placed on Oral, TPN, IV Extubate unit on vent vent in ICU DVT Prophylaxis Resp therapy RT sets vent Pharm &/or Mechanical validates settings setting Oral Care 1. Brush teeth Standard Ventilator Practices Orders 2. Hypopharyngeal suctioning 1. Elevate head of bed 30–45 degrees 3. Swabbing Yes No Contraindications Y 2. Stress-ulcer prophylaxis Nurse given Yes No Contraindications Sedation mgmt target RASS Pt at or Pt meets Is patient Continue Pass SBT 3. DVT Prophylaxis Y N and above target criteria for agitated or Goal directed to monitor ? Yes No Contraindications documents RASS? SBT in pain? RASS vs vacation score 4. Oral Care Yes No Contraindications N Y N 5. Sedation management with daily interruptions Over sedated/ Resume @ ½ Adjust Resume (RASS vs vacation) do sedation original rate sedation ventilation 6. Weaning/spontaneous breathing trial vacation (SBT) Orange Alert: Red Alert: Super Red Alert: Yellow Alert: Patient started on Patient on vent. Multiple patients on Vent. order placed vent. without without all vent. without all without “Standard “Standard Ventilator “Standard Ventilator “Standard Ventilator Ventilator Practices Practices Orders” in Practice Orders” Practice Orders” Orders” in place place documented in HED documented in HED FIguRE 5-2 Supporting practice (process). Figure 5-2.eps landscape

OCR for page 69
 CASE StuDy: vANDErBilt’S jourNEy with ICU patients without bleeding. Therefore, prophylaxis is recom- mended. However, neither UHC nor IHI explicitly defined stress-ulcer prophylaxis, contraindications, etc. Table 5-1 shows how Vanderbilt eliminated the guesswork by outlining exactly which practice standard should be implemented in specific situations. If parts of a recommended practice would be performed by different people at different times, they were subdivided into a standard practice for each role or time. For example, a recommendation for oral care was divided into three elements, including oral swabs, tooth brushing, and hypopharyngeal suctioning. Together, the elements of each standard- ized practice make this approach actionable and permit more focused accountability and performance measurement. A major breakthrough in the development of system-supported practice at Vanderbilt was the process-control dashboard (Figure 5-3), which shows the entire team the status of a patient against their plan as a set of red, yellow, or green lights. There is a line for each patient and a column for each element of the standardized practice. Each cell represents one action that must be taken by a member of the team. No actions are lumped together. A green light means everything is in the expected state according to that plan. A yellow light means action must be taken, but there is still some time to do so. A red light means an outcome is out of line with the desired performance and action must be taken immediately. By making the display available in real time, the clinical team can tell what is right and wrong about the practices or documentation and help the system improve. Defining the system and outlining differences between evidence and standard practices are key aspects in the development of system-supported practice. The second aspect of system-supported practice is closed-loop con- trol, which means the output of the system feeds back directly to change TABLE 5-1 Standard Practice Table for Stress-Ulcer Prophylaxis VUMC Practice Standard: Gastric Access Gross Blood Y N H2 blocker per tube Y Y Proton pump inhibitor per tube N N H2 blocker IV or TPN N Y Proton pump inhibitor IV

OCR for page 69
 SyStEmS ENgiNEEriNg to imProvE trAumAtiC BrAiN iNjury CArE FIguRE 5-3 Process-controlFigure 5-3.eps real-time feedback for supporting dashboard showing practices. Source: Stead and Starmer, 2008. Reproduced with permission from the bitmap Vanderbilt University Medical Center. Copyright 2007 by the Vanderbilt University Medical Center. the inputs, thereby changing system performance and bringing it back under control (see Figure 5-4). A simple example of this is the interaction between a thermostat and a furnace to control room temperature. There are many inputs to the room temperature; the furnace is just one of them. The thermostat is set to a desired temperature with a targeted control limit for how far above or below the desired temperature it can go. When the thermostat senses the temperature is falling below the control limit, it calls for heat, the fur- nace turns on, the temperature rises, and the thermostat approaches the upper-control limit and turns off the heat. Because it keeps a record of its immediate past performance, if it overshoots the next time, the thermostat turns off a little earlier until the temperature is within the control limits. If someone opens a window and changes the inputs to the system, the thermostat adapts to the change without reprogramming. The desired performance is achieved without programming complex interactions

OCR for page 69
 CASE StuDy: vANDErBilt’S jourNEy Collective Collective Inputs Control Inputs Control Outputs Outputs Figure 5-4.eps FIguRE 5-4 Closed-loop control. Source: Stead et al., 2008. Reprinted with permis- sion of the American Clinical and Climatological Association. among inputs or modifying the program as inputs change. This is what is needed in health care. To achieve this, there must be agreement on an end-to-end plan of actions. Real-time measurement is also necessary to show what is hap- pening and to display the instant status of the patient against the plan. The human team can then become the effecter mechanism that acts on input and allows for human judgment to override the system when the need arises. Figure 5-5 shows a model for adapting the practice for closed-loop control. However, health care situations are often too complex for an end-to- end plan at the level of detail for an order or a prescription. Therefore, Vanderbilt developed a set of nested plans at different levels of specificity. Tier 1 is the highest level and has the least specificity. It outlines broadly Status Plan Performance • Objectives • Actions • Process • Results • Steps • Measures FIguRE 5-5 A model for adapting practice for closed-loop control. Source: Stead et al., 2008. Reprinted with permission of the American Clinical and Climatological Association. Figure 5-5.eps

OCR for page 69
 SyStEmS ENgiNEEriNg to imProvE trAumAtiC BrAiN iNjury CArE applicable objectives and describes what is to be accomplished, rather than how to accomplish it. In the case of ventilators, the objective was to minimize the time a patient was on mechanical ventilation and, at the same time, minimize complications. Table 5-2 summarizes the highest level plan for ventilator management. Tier 2 adds more specific plans and outlines how work is to be done to accomplish the Tier 1 objectives, including who should do what, how they should do it, and how it will be measured. Tier 2 is the level built into the cross-process control dashboards and tuned to the patient popu- lation and provider capabilities. Tier 2 must fit into the context of the Tier 1 objectives agreed to by stakeholders. Table 5-3 shows the Tier 2 plan for the objective of minimizing time on mechanical ventilation. Tier 3 includes enough detail for caregivers to flex the plans in the tier above to the individual patient and to day-to-day variations in the facility’s medication formulary, etc. Tier 3 is the level repre- sented in order sets. Table 5-4 shows the Tier 3 plan for stress-ulcer prophylaxis. Overall, with these three tiers, the goal is to develop a modularized end-to-end plan that makes the complexity of any com- ponent manageable. The time line for the Vanderbilt initiative is shown in Figure 5-6. All of the preparatory work was done in January 2007, and the design was completed in February 2007. By March, the order sets and education had been implemented. The process-control dashboard was developed and available to the team by May 2007, so, as they made rounds, they could see what was actually going on and what the dashboard showed as the status of their patient in comparison to their plan. They could TABLE 5-2 Ventilator Management: Highest Level Plans (Tier 1) Objectives Process Steps Measures Minimize time on • Avoid over-sedation • Risk-adjusted time on mechanical ventilation • Wean as rapidly as mechanical ventilation possible • Unplanned extubation rate •➢ Failed extubation rate Minimize complications • Pneumonia prophylaxis • Incidence per 1,000 ➢ • Stress-ulcer prophylaxis ventilator days ➢ • DVT prophylaxis SOURCE: Stead et al., 2008. Reprinted with permission of the American Clinical and Climatological Association.

OCR for page 69
 CASE StuDy: vANDErBilt’S jourNEy TABLE 5-3 Ventilator Management Tier 2 Plans Objectives Process Steps Measures Goal-directed sedation • Physician order for • Ordered value, date, individual target RASS and time q12hr • Charted value, date, • Nurse assessment of and time actual RASS q4hr Rapid weaning • Rapid weaning protocol • Screen pass/fail, date, • If >24 hr daily screen for and time trial readiness by RT • Trial pass/fail, date, • If pass, spontaneous and time breathing trial within 6 hrs ➢SOURCE: Stead et al., 2008. Reprinted with permission of the American Clinical and Climatological Association. TABLE 5-4 Ventilator Management Tier 3 Plans Objectives Process Steps Measures Stress-ulcer prophylaxis • + tube – blood => • Use or non-use of Pepcid 20 mg PT order set q12 hr • Tube +/- • + tube + blood => • TPN +/- Prevacid 30 mg PT • Drug order q24 hr • Administration record • – tube – blood => Pepcid 40 mg /c TPN q24 hr ➢ • – tube + blood => Pepcid 20 mg IV q12 hr SOURCE: Stead et al., 2008. Reprinted with permission of the American Clinical and Climatological Association. easily report whether a practice shown as red was applicable, whether documentation was wrong, or whether the algorithm used to calculate the color for the dashboard was misleading. As a result, the process began to improve itself. By September, the team felt that all of the processes were basically accurate, and they made it a high priority in every round of patient

OCR for page 69
 SyStEmS ENgiNEEriNg to imProvE trAumAtiC BrAiN iNjury CArE February 2007 Bundle design shop March 2007 Order set revisions & nursing education implemented May 2007 Dashboard as Screensavers June/July 2007 Dashboard refinements (resptherapy, oral care) August 2007 Reports available October 2007 “Improvement opportunity” reports available Nov -December 2007 Emphasis on consistent execution of complete bundle for each patient Jan -March 2008 Nurse charting improvements March 2008 Renewed nursing education + educational materials FIguRE 5-6 Time line for the Vanderbilt initiative. contact to turn red cells into green cells on the process-control dash- board for each patient. Data received at the end of April showed a 50-percent reduction in the incidence of ventilator-acquired pneumo- nia in Vanderbilt Adult Hospital compared to the previous six-month period. This means 11 more people left the hospital, and 1.6 million dollars was saved in health care costs. In addition, with process improve- ment, Vanderbilt was performing all of the practices for 70 percent of patients and individual practices for more than 90 percent of patients. A few key practices are highlighted in this model. The first is avoid- ing the desire to get everything right for everyone all the time. This is the wrong approach. The objective is to get something that is workable, deliver it, then constantly modify it so it will be self-correcting and self- sustaining. Table 5-5 outlines the rapid iterative cycles in the system- supported approach necessary to achieve a workable model. The second key practice is to have more than one metric for success. The approach at Vanderbilt is measurement-driven based on agreement TABLE 5-5 Rapid Iterative Cycles in the System-Supported Approach Phase Goal # People Duration Initiation Focus 2–4 1–2 hrs Pre-work Approach 6–10 2–6 wks Design shop Agreement 30–60 1–2 days Development Components 10–20 3–12 wks Pilot System 30–100 2–6 wks Rollout Dissemination > 100 4–6 wks Improvements Performance > 100 Continuous

OCR for page 69
 CASE StuDy: vANDErBilt’S jourNEy on a standard set of practices and an iterative improvement process tar- geted at execution with 100 percent reproducibility. If continued mea- surement of a practice shows the desired improvement, it continues. If not, it is changed. Rather than relying on one metric, a set of indicators is used; each indicator has an explicit definition consisting of the process or system, the indicator, the actor, and the timing of the process. Overall, system-supported practice is a combination of people, processes, and technology. It cannot be achieved by inserting informa- tion technology into our current way of working. The important thing is to focus on what must be improved rather than on external measures; use measurement-driven, iterative cycles to create self-correction and sustained improvements; use a common fact base to encourage agree- ment; set a target of 100 percent performance for the set of practices appropriate to a patient; and combine people, processes, and technology to achieve desired results. Another factor critical to the success of this initiative is that the institution and the support staff must approach the project with a collective will for process improvement. REFERENCES Stead, W.W., and J.M. Starmer. 2008. Beyond Expert-Based Practice. Pp. 94–105 in Evidence-Based Medicine and the Changing Nature of Health Care. 2007 IOM Annual Meeting Summary. Washington, D.C.: National Academies Press. Stead, W.W., N.R. Patel, and J.M. Starmer. 2008. Closing the loop in practice to assure the desired performance. Transactions of the American Clinical and Climatological Asso- ciation 119: 185–195.

OCR for page 69