Improving Medication Safety Across a Multihospital System

JOURNAL

ON

QUALITY IMPROVEMENT

Team leaders and members often feel that a change has been an improvement and are comfortable in making the change permanent. Yet to spread impact beyond the immediate change, teams must persuade others—and this requires measurement. PATIENT SAFETY

Improving Medication Safety Across a Multihospital System KEN FARBSTEIN, MPP JEANETTE CLOUGH

I

n the June 2000 issue of this journal, Weingart described the architecture of the CareGroup hospital system’s multiyear strategy of becoming the world standard for medication safety.1 This article describes the methods, success factors, barriers, and updated results of our achievements to date. Six tangible, measured improvements in patient safety have occurred; these are described in detail as case studies. CareGroup is a network of six hospitals in eastern Massachusetts—Beth Israel Deaconess Medical Center (BIDMC), from community hospitals that were affiliated with Deaconness Hospital before it merged with Beth Israel (Deaconess Glover, Deaconess Nashoba, Deaconess Waltham, Mount Auburn), and New England Baptist Hospital. The network includes eight affiliated community health

centers and numerous medical groups. CareGroup has a total of 13,000 employees, including 2,000 medical staff.

Ken Farbstein, MPP, is Managing Principal of Melior Consulting Group, Needham, Massachusetts. Jeanette Clough is

Bette Bertini, RN, Julie Bonenfant, RN, Karen Hoikala, RPh, Marie Isham, RN, Loretta Joy, RN, Joseph McLaughlin, RPh, James Mullen, RN, Nancy Miller, RN, Herminia Shermont, RN, and Hope Violette, RPh. Please address requests for reprints to Ken Farbstein, Melior Consulting Group, 166 Lindbergh Ave, Needham, MA 02494-1526; phone 781/444-5525; fax 781/4441396; e-mail [email protected].

President and CEO of Mount Auburn Hospital, Cambridge, Massachusetts. The authors salute the present and past team leaders and coleaders: Anne-Marie Audet, MD, Michael Benari, MD, Copyright © 2001 by the Joint Commission on Accreditation of Healthcare Organizations

VOLUME

27

NUMBER

Background While individual hospitals have improved medication safety, the broad improvement of medication safety across hospital systems has been rare. Hospitals and hospital systems have either used a high-technology approach, a low-technology approach, or a combination of approaches. Some hospitals—for example, Promina Gwinnett Hospital in Lawrenceville, Georgia, and Fairview Hospitals in Minnesota—use a low-technology, fairly inexpensive approach. Promina Gwinett, as mentioned in the guidebook Reducing Adverse Drug Events,2 has tested and implemented most of the 16 best practices

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

123

THE JOINT COMMISSION Article-at-a-Glance Background: The Massachusetts Coalition for the Prevention of Medical Errors and the Institute for Healthcare Improvement have identified 16 best practices to reduce adverse drug events. CareGroup, a network of six hospitals in eastern Massachusetts, multiplied its routine use of these best practices tenfold in the first 18 months of its medication reliability project. Developing the collaborative strategy: Although CareGroup’s long-term plans included technological advances such as clinical order entry, computer systems in the pharmacy, dispensing stations on patient floors, and bedside bar-coding, efforts first focused on manual improvements feasible within a year’s time. A 4-year strategy involves helping the medication reliability team leaders at each hospital to create impressive local results, publicize the results to their colleagues, invite their clinical colleagues to learn to use plan-do-study-act (PDSA) cycles, and have colleagues lead PDSA cycles themselves. At monthly or bimonthly task force meetings, team results are presented and team leaders are given specific assignments for their teams. Case studies: One project reduced the time to blood anticoagulation for heparinized patients. The second dramatically reduced lookalike/soundalike errors. The third improved the safety of patient-controlled analgesia. The fourth reduced coumadin incidents. The fifth improved the education of patients about their medica-

recommended by the Massachusetts Hospital Association and its Massachusetts Coalition for the Prevention of Medical Errors3,4—using preprinted orders for heparin and other hazardous drugs; a policy for rejecting nonstandard orders; computer profiling of patients; standardized doses, packages, and labeling; computerized alerts; patient partnering and selfadministration of medications; pharmacy admixture (rather than mixture by nurses); unit dosing; removal of concentrated potassium chloride (KCl) from nursing floor stock; and elimination to the extent possible of lookalike/soundalike drugs. Fairview has enlarged the authority of the pharmacy and uses forcing functions, pretyped orders (for example, in chemotherapy), and changes in dosing protocols (for example, granting the pharmacist the authority to change dosage in accordance with the patient’s height or weight), among other practices.5

124

tions. The sixth greatly reduced the morning dispensing backlog in the pharmacy. Success factors: Key success factors, in addition to leadership, are the use of data, forcing functions, appropriate pacing, inexpensive practices, and a consultant. The pace needed to implement the best practices overall made it imperative to make many changes rapidly. Often, the team initiated several changes at one time, rather than sequencing changes in successive PDSA cycles. Limitations, barriers, and next directions: CareGroup faces key challenges in measurement and in spreading and deepening the involvement of clinicians, particularly physicians. It lacks an overall, objective measure of medication safety. Spread of the changes made has been incomplete although the adoption of the best practices increased tenfold (from 6 to 60) in 21 months. Two of the case study interventions—in coumadin order sequencing and dedicating a pharmacy technician to order entry—have been implemented at only one site to date, even though the adoption of the change ideas across hospitals is encouraged. The eventual impact of the changes planned for the future, through automated systems such as computerized order entry, is much larger. Considerable progress is anticipated in adoption of best practices; improvement in top-priority areas of each hospital; improved automation and technology in ordering, dispensing, and administering medication; and better reporting.

The best-known example of a high-technology approach widely spread across a multihospital system is at the U.S. Veterans Administration (VA). The VA hospital system is implementing an internally developed order entry system through its 173 hospitals and plans to add an internally developed bedside barcoding system. An improved system for reporting near-misses is also in use.6–8 The most well-known examples of the hightechnology approach largely centered in a hub of a hospital system have occurred at Brigham and Women’s Hospital in Boston, Wishard Memorial Hospital in Indianapolis, and LDS Hospital in Salt Lake City. These hospitals appear to have greatly reduced adverse drug events by using these expensive systems, whether by eliminating ambiguously interpretable handwritten orders or, as in the case of LDS, by using the clinical decision support system. The Brigham and Women’s

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

JOURNAL

ON

QUALITY IMPROVEMENT

Hospital physician computer order entry system has produced very impressive results; in one study, its implementation was associated with a 55% reduction in medication errors and a 17% reduction in preventable adverse drug reactions.9 The system at Wishard, which has now been used for two million orders, uses clinical decision support logic that in another context was found to frequently and substantially influence physicians’ orders.10–12 At LDS, the Health Evaluation through Logical Processing (HELP) system has elicited impressive results in antibiotic use in the intensive care unit.13 Plans have been made to spread this technology to other hospitals in these hospitals’ systems.

Recap At CareGroup, the initial successes stem from a lowtech approach through six interdisciplinary medication reliability hospital teams. Yet the broader strategy relies on the development of a clinical order entry system, the direct descendant of the Brigham and Women’s Hospital system, and other technologies. Soon after arriving as head of the CareGroup system in July 1998, Dr James Reinertsen set the goal of becoming “the world standard for reliability in medication administration.”14 Knowing the power of a stretch goal, he had selected this goal to inspire broad change in the hospitals’ core functions. In winter 1998 Reinertsen designated Jeanette Clough, chief executive officer (CEO) of Mount Auburn Hospital in Cambridge, Massachusetts, to lead the effort across the six hospitals. Milestones for the first 18 months of the project appear in the chronology shown in Figure 1 (p 126). Reinertsen assigned other systemwide goals to the other CEOs. This project follows the first systemwide collaboration—the successful race to become Y2K compliant— which had been led by Dr Alan Robbins, CEO of New England Baptist Hospital. The medication safety project is expected to take seven to ten more years. Clough convened meetings of an advisory group of 10–12 internal leaders and clinicians, mostly from BIDMC, and since February 1999 has retained Ken Farbstein of Melior Consulting Group as a consultant.

Developing the Collaborative Strategy: Methods Planning the Strategy To become the world standard, CareGroup must become the best at each of three functions: ordering,

VOLUME

27

NUMBER

dispensing, and administering medication. CareGroup must also become the best at reporting errors and developing pertinent safeguards. To accomplish this, changes must occur in the everyday practices and technologies of all physicians, nurses, and pharmacists. As a massively parallel effort is needed, work is under way on each of these tracks simultaneously. While we* have high expectations for the technologies of clinical order entry, computer systems in the pharmacy, dispensing stations on patient floors, and bedside bar-coding, we chose not to wait for technological advances. Rather, we proceeded aggressively on manual improvements feasible within a year’s time. This article focuses on the results of this first wave of improvements. We set the broad outlines of strategy in biweekly meetings, where we developed and updated a fouryear plan. Our strategy has been to work through the team leaders at each hospital—to help the team leaders create impressive local results, publicize them to their colleagues, invite their clinical colleagues to learn to use plan-do-study-act (PDSA) cycles, and have the colleagues lead PDSA cycles themselves. Since February 1999 monthly or bimonthly task force meetings including the leaders of the six medication reliability teams have been held to note and celebrate progress. The team leaders know that we expect them to present their results at each meeting, and this expectation paces the teams’ work. Indeed, specific assignments to the teams at the end of each task force meeting provide direction. Moreover, at each meeting, a scorecard (Figure 2, p 127) is presented and distributed to show each hospital’s progress at adapting the 16 best practices. This has set up a friendly competition among the teams and has kept them accountable to each other. During spring and summer 1999, meetings were held to discuss measurement and cultural and attitudinal issues in ad hoc groups, especially on driving out fear of increased reporting of adverse drug events. We have been developing and are now pilot-testing a Red Alert system to prevent near-misses and adverse drug events from occurring in other hospitals. Because most of the team leaders and participants are nurses and pharmacists, rather than physicians, our * This work is thoroughly collaborative. The words we and our in this article in the context of strategy denote Clough and Farbstein; in the context of work by the teams, we and our denote the team members and Farbstein.

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

125

THE JOINT COMMISSION Project Chronology Jul 1998

Dr James Reinertsen starts as CEO of CareGroup.

Jan 1999

Jeanette Clough, with lead responsibility for improving medication safety, sets assignments for the kick-off meeting of the task force.

Feb 1999

Ken Farbstein is retained as the consultant. Kick-off meeting of the task force.

Mar 1999

Self-assessment by teams of best practices in place.

Apr 1999

The six teams agree to adopt and spread the first set of 16 best practices across the entire network within a year.

May 1999

Dr Lucian Leape speaks at the task force meeting. The first scorecard shows that KCl has been removed from patient floors in all hospitals, making it the first of the best practices in routine use throughout CareGroup.

Jul 1999

IHI two-day workshop takes place.

Sep 1999

Pilot tests are under way at all hospitals.

Oct 1999

Improvement in coumadin safety is publicized via the first e-mail “gong” message. PCA incidents at BIDMC are reduced by 80%.

Jan/Feb 2000

Unit dosing has come into routine use throughout CareGroup.

Nov 2000

Adoptions into routine use of best practices have risen to 60, up from 6 since the task force began 21 months earlier.

Figure 1. Data collection time frames of the case studies are shown in Table 1. CEO, chief executive officer; KCl, potassium chloride; IHI, Institute for Healthcare Improvement; PCA, patient-controlled analgesia; BIDMC, Beth Israel Deaconess Medical Center.

early successes focused on dispensing and administration of medication, rather than on ordering. For example, the best practice of unit dosing was fully adopted throughout CareGroup by January/February 2000. By contrast, efforts by nurse leaders to improve

126

safety outside their area of formal responsibility—for example, in ordering—have not yet been fully implemented. We have used the energy and enthusiasm of the leaders who have stepped forward on their own, and have had them lead changes in areas within their responsibilities. Network leaders in information systems and executive levels have focused on developing an electronic order entry system as the key change in ordering, freeing the teams to focus on improving the dispensing and administration of medication. Leaders Given the broad scope of the medication reliability project, we need numerous leaders. We needed to recruit them as we proceed. Initially, each CEO selected the hospital’s medication reliability team leader. One team had the pharmacy director and vice president of patient care services assume the leadership from the risk manager. In two teams, the leader moved to another hospital, and a member of the team was made its leader. Dedicated time is a factor when involving practicing, active staff physicians. While it is desirable, physician leadership at the hospital team level is not indispensable early in a broad change effort like this one. Training As the teams formed in spring 1999, the consultant trained them in rapid-cycle techniques during their meetings. For example, when the consultant realized that the team would benefit from quickly listing hypotheses about why medications were most often missing, the nominal group technique (NGT) was summarized in two minutes, sticky notes were distributed, and the team performed its first NGT. In the meeting wrap-up, the team members were offered several ways to use NGT in the future and were asked to consider using it as the opportunity arose. In July 1999 the teams’ knowledge of these techniques was reinforced by a two-day workshop organized and delivered by the Institute for Healthcare Improvement (Boston). This workshop served as a key vehicle for training team leaders and participants. In addition, Lucian Leape, a national authority on medical errors, spoke at a day-long symposium in May 1999. Recruiting the Teams In establishing teams, we asked the CEO of each hospital to select a team leader and participants. We

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

VOLUME

27 piloting

Hosp D

Hosp E

NUMBER

Legend

Not yet implemented

Nonpunitive error reporting

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations x Equivalent in place; won’t implement

24-hour-a-day pharmacy services available

Critical pathways for complicated care

Maximum 12-hour shifts for all nursing personnel

ER condition management protocols

Warning labels for lookalike/soundalike drugs

Pharmacist on unit, at rounds

next Next pilot test

piloting Began pilot test

Already in routine use in part of hospital

piloting

Throughout entire hospital

piloting

next

piloting

next

Hosp F

ON

Pharmacy-based admixture of IV solutions

Unit dosing

Computerized drug profiling

Limited numbers of dosages and pump types for IV solutions

Colored patient allergy wristbands

Patient partnering

Removed KCl concentrates from all units

piloting

Hosp C

piloting ICU

piloting

Hosp B

Enforcement of standardized prescribing

piloting

Hosp A

Weight-based heparin protocols

Chemotherapy protocols and chemo preprinted orders

Best Practice

JOURNAL QUALITY IMPROVEMENT

Scorecard: Spread of Best Practices Across Inpatients as of November 14, 2000

Figure 2. At each task force meeting, the scorecard is presented and distributed to show each hospital’s progress at adapting the 16 best practices.

Nonpunitive error reporting is shown but is not included in the count, since it is in the second list of best practices, which are considered feasible in 1–3 years. ICU, intensive care unit; KCl, potassium chloride; IV, intravenous. © 2000 CareGroup and Melior Consulting Group. Permission to copy this scorecard is granted if the source is identified.

127

THE JOINT COMMISSION did this to preserve, rather than undercut, the internal leadership of each hospital. Nursing leaders, pharmacists, and risk managers constituted most of the teams’ members. Physicians attend most of the meetings of four of the teams; physicians participate from time to time, as needed, in the other two teams. Four teams had eight members, one had four, and one had five (including the leader but not the consultant). The two teams performing the most PDSA cycles were the smallest. Larger teams were more likely to implement changes without pilottesting them. This is consistent with the consultant’s experience—small teams tend to more readily pilottest ideas, and larger teams tend to digress into areas outside their control, such as hospital policy. When this happens, the larger number of members and agendas, the limited time for meetings, and the often collegial style of leading meetings can often prevent the team from beginning or reviewing PDSA cycles. Cultural Issues Each of the six hospitals has a unique culture; size and academic mission mark the key differences in cultures. As a very large academic medical center, affiliated with Harvard Medical School, BIDMC—representing the merger of Beth Israel and Deaconess hospitals—has a mission that is national in scope. It provides an internationally recognized model of patient-centered nursing care. The risk in a large academic medical center is that a new goal, even one deemed critical by the CEO, could have to compete with other institutional goals—and so we have adapted our approach accordingly. The six hospital medication reliability teams are urged to raise awareness by showing the intense, emotionally provocative “Beyond Blame” 10- to 12-minute videotape prepared by Bridge Medical, Inc (Solana Beach, Calif). Yet BIDMC’s nursing leaders had already been promoting a culture of safety. As part of this work, they had been surveying clinicians about their mindfulness of the possibility for medication errors. Initially they preferred to continue raising awareness through the survey, and we accommodated their preference. (Indeed, more recently, we have encouraged the other hospitals to conduct similar surveys.) Subsequently, BIDMC has been showing and discussing the videotape. This is one example of accommodating and capitalizing on cultural differences across hospitals.

128

Kick-Off Clough asked each hospital to designate a team leader and form a medication reliability team. The kick-off meeting of the six teams occurred February 8, 1999. For the next meeting of this task force, we asked each team leader to formally present a description and an analysis of an actual error that had occurred within his or her hospital. We also asked teams for their definitions of error to establish clear baseline language across the network. Each hospital also presented a detailed flow chart of the steps in providing medication for a patient. Finally, teams presented assessments of the current status of their use of best practices, using the criteria set forth by the Massachusetts Coalition for the Prevention of Medical Errors. Identifying and Spreading Best Practices. In late 1998 three lists of best practices were compiled to improve medication safety, drawing primarily on recommendations from the Massachusetts Coalition for the Prevention of Medical Errors3,4 and the Institute for Healthcare Improvement.1 The first list consisted of 16 best practices feasible within 1 year; the second list, of six best practices feasible in 1–3 years; and the third, the six best practices whose implementation was thought to require more than 3 years. At the April 1999 task force meeting, the six teams agreed to adopt and spread the first set of 16 best practices across the entire network within 1 year. This, rather than a specific numerical reduction in adverse drug events, was the goal. We are proud of the team leaders for their commitment to this difficult responsibility. We asked each team to inventory its best practices. We developed the scorecard to track progress networkwide. On the initial self-assessment, we compiled the first scorecard in May 1999. This scorecard showed that KCl had already been removed from the patient floors in all six hospitals, making this the only best practice that had been fully adopted across the network at that point. For the teams at the three largest hospitals, the consultant was often asked for advice, either during team meetings or in private meetings with team leaders. When data on errors and the results of changes would be useful, the consultant prepares run charts, Pareto graphs, and other data displays. The consultant aided two other teams in their efforts. Team leaders and members often feel that a change has been an

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

JOURNAL

ON

QUALITY IMPROVEMENT

improvement and are comfortable in making the change permanent. Yet to spread impact beyond the immediate change, teams must persuade others. Measurements make this possible. Indeed, this is the most powerful use of the measurement of medication safety: It persuades others to join the effort. The team leaders were strongly encouraged but not required to use the PDSA method for rapid-cycle change, and most did so. Two teams used the worksheet “Building Knowledge for Improvement,” developed by Tom Nolan (Associates in Process Improvement, Silver Spring, Md)15 or the consultant’s adaptation, or used run charts to show data both before and after. Such run charts represent PDSA projects’ gold standard for showing improvement. The other teams generally used the PDSA method but developed baseline measurements after changes had begun and resisted pleas to use the worksheet. All the successful projects, by definition, measured results. The consultant guided the teams through the PDSA method. Preferably, this starts with the aim and proceeds through questions about baseline data, possible changes, selection of a single change, planning of a pilot test, knowledge gained upon the early assessment of results and follow-up actions, and overall lessons—questions 1–8 in the Nolan worksheet. More typically, the process was less sequential, though it had an eddying, flowing logic of its own. For example, since the hectic morning pace in the pharmacy could be problematic, the pharmacy team at a community hospital began addressing the pharmacy backlog with a discussion of possible changes (question 3) and reached agreement on the aim and baseline data for the problem. Desiring to guide the group without blocking its enthusiastic flow, the consultant stated the aim (answering question 1) while listing the changes and adding a possible change, but he refrained from insisting that the team consider baseline data and ways to ascertain whether a change had been an improvement (question 2). Rather, the consultant ensured that the team listed a broad set of possible changes, and then had it identify a single change. Then, he provided a one-page planning template (to provide a structured answer to question 4), some of whose questions the team had already answered. The consultant had them answer the remaining questions on the planning template. As it completed the planning template, the team described

VOLUME

27

NUMBER

how it would know if the change had been an improvement (question 2). The team then ran the pilot test, and guided by the consultant, answered the worksheets’ final three questions, in the preferred order. In some cases the smaller community hospitals have preferred to “just do it” rather than use the PDSA approach. There is less bureaucracy and there are fewer committees, and organizational change is less complex. For certain changes that the smaller hospital teams deem almost certain to constitute improvements, they prefer to form a policy hospitalwide rather than temporarily run two processes in parallel during a pilot test (Figure 3, p 130). In general, the steps that limit the rate at which practices are adopted are, for the easily implemented best practices, the time and attention of senior leaders at each hospital. The rate-limiting steps for the more difficult best practices involve the teams’ mastery of PDSA. When the study phase of the PDSA cycle showed that a change has been an improvement, the teams sought to spread the change throughout the hospital. To reach hospitalwide change, they generally used the formal authority structure to make policies. P&T (pharmacy and therapeutics) committees approved hospitalwide policies on the basis of team leaders’ presentations on the evidence of successful PDSA changes. A case in point is the development of preprinted orders for heparin at Mount Auburn Hospital, a regional community teaching hospital. Surprisingly, the medication reliability team used informal networks of influence only minimally in their six successes. We have encouraged the spread of the best practices networkwide, across the six hospitals. Team leaders routinely describe their accomplishments in detail at the task force meetings. When measurement reveals a tangible improvement in the safety of care, an e-mail message is sent to all members, with a request that they publicize it throughout their hospitals and explicitly consider its applicability for them locally. These efforts have resulted in the use of several specific practices at additional hospitals, for example, spread of a weight-based herapin protocol. To make this systemwide adoption more systematic, a pilot test of a meta-level change across the six-hospital system is under way.

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

129

THE JOINT COMMISSION Methods Used for Rapid-Cycle Change Results Certain

Complexity of Implementation

Uncertain

Easy

Just do it

PDSA preferred (not applicable for best practices)

Difficult

Plan-do-study-act

PDSA preferred (not applicable for best practices)

Figure 3. The team leaders’ decisions about the best methods include these factors. PDSA, plan-do-study-act.

Results: Measurement Methods As the teams proceeded in adopting best practices, we sought to measure both the number of best practices adopted and the local outcomes of change. With 16 best practices and six hospitals, ultimately we will have 96 local adoptions (not including nonpunitive error reporting). The 60 adoptions to date represent a tenfold increase in the number of adoptions from the baseline of 6 early successes. Figure 4 (p 131) shows the growth in the adoptions by month. At first, the scoring on the scorecard was necessarily subjective. Recently, however, operational definitions and measures for each have been developed. (They are available for free download at www.meliorconsulting.com.) This has provided more specific goals for the teams and more precise knowledge of and confidence in the extent of improvements. We also sought to measure the local results of changes. The case studies in this article describe the first six projects for which data became available. Additional improvements should become known as measurement efforts grow. Three of these six changes were measured from incident reports, one was measured electronically from laboratory results, one was measured by audit and observation, and one was measured by timing task completion. Table 1 (p 132) shows the time frames for data collection for each of the six early successes. The results for each are discussed in the case studies.

■ Faster therapeutic anticoagulation for patients receiving heparin; ■ Fewer lookalike/soundalike errors; ■ Fewer patient-controlled analgesia (PCA) administration adverse events; ■ Safer administration of coumadin; ■ Improvement in giving patients information on their medications; and ■ Improved processing of the morning backlog in dispensing medication in the pharmacy. Case studies of the six improvement projects now follow.* Three of the projects resulted in reductions in the number of medication error incidents; the other three were primarily measured in other ways. More specifically, we knew that the lookalike/soundalike, PCA, and coumadin changes were improvements because there were fewer incidents of those types, even while the overall number of incidents was generally increasing. We knew that the heparin changes had made an improvement because lab test results showed that more patients reached a therapeutic level of anticoagulation. We knew by observation and audit that patients were better informed at New England Baptist Hospital. We knew by measuring times of day that the pharmacy backlog had been reduced at Deaconess Waltham. At the November 2000 task force meeting, the Nashoba team announced that with its use of allergy wristbands, 479 consecutive days had elapsed without allergy incidents. It also announced a threefold reduction in heparin incidents from fiscal year 1999

Case Studies Spread of best practices is shown by data regarding six tangible, measured improvements in the quality of care:

130

* Readers may obtain additional details on the case studies by contacting the consultant or the key contact listed for each case study.

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

JOURNAL

ON

QUALITY IMPROVEMENT

Growth in Adoption of Best Practices Total number of Best Practices in Routine Use Throughout Hospitals

60 52

55

56

55

58

60

40

40

35

36

26

20

18

21

22

21

22

12 6 0

0

0 May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 1999 2000

Figure 4. The figure shows the growth in the change adoptions by month among the six CareGroup hospitals. Decreases in Sep 99 to Aug–Sep

2000 and February reflect the fact that after reporting “hospitalwide, routine use,” a hospital team infrequently discovers that a best practice has not been fully implemented; the scorecard is revised until hospitalwide, routine use is achieved. The counting of adoptions was suspended from Jan 2000 to Feb 2000 while operational definitions of the best practices were drafted.

to fiscal year 2000, coinciding with the use of a weight-based heparin protocol (key contacts: James Mullen, RN, and Joe McLaughlin, RPh). Case Study 1. Faster Therapeutic Anticoagulation of Patients Receiving Heparin at Mount Auburn Hospital Context. Administration of heparin requires careful titration so that blood thinning is appropriate to treat the clinical condition. In this process the patient’s blood may become over- or undercoagulated. The goal is to reach and sustain the proper therapeutic range of anticoagulation in a timely and efficient way. The medication reliability task force had identified heparin as a high-risk drug whose use is known from the literature to be prone to error, as it requires titration and frequent recalculation of dosages.2 The team’s goal was to help patients receiving heparin achieve a rapid therapeutic level of anticoagulation and ensure maximum safety. Changes. The team created a new dosing schedule and revised order form (for example, the form now orders all prothrombin time tests [PTTs]) for the new protocol; reduced the heparin bolus dose and infusion rates; and improved the nurses’ documentation form. Measuring method. Chart review occurred before and after the changes.

VOLUME

27

NUMBER

Analysis of 100 charts as a baseline showed that in the first 3 days of heparin use, 44% of patients ever reached a PTT within the therapeutic range of 50–80. The fraction of patients who ever achieved a therapeutic level within the first 3 days of heparin use more than doubled, to 93% (n pre = 100; n post = 29 [for this and all results shown below]; p <.001, based on a one-tailed z-test).16 The follow-up sample size was economically small. Charts are being reviewed to ascertain a second post sample to ensure that the improvement has persisted more than a year after the changes occurred, in summer 1999. The fraction of patients who reached the therapeutic anticoagulation range by the end of the first day doubled, from 16% to 31%. The degree of anticoagulation was measured more often after the changes (4.9 times per patient versus 3.3 times), and the fraction of all these measurements that revealed out-of-range (that is, nontherapeutic) levels decreased by a third, from 69% of all measured PTTs to 47%, for an improvement of 32%. On obtaining the internal hospital committees’ approval, the team has trained clinicians and the protocol is in routine use throughout the hospital. The team has continued to run numerous additional PDSA cycles, and most recently it has developed and spread Results.

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

131

THE JOINT COMMISSION Table 1. Time Frames for Data Collection*

Project

Time studied before change

Time studied after change

Heparin

Aug 1999

Nov 1999

Lookalikes/soundalikes

Feb 1, 1999–Jul 31, 1999

Aug 1, 1999–May 2000

PCA

Jun 7, 1997–Mar 15, 1999

Mar 16, 1999–Nov 2000

Coumadin

Apr 1999

May 1999–Nov 2000

Patient partnering

Jul 1999

Sep 1999 and Mar 2000

Dispensing times

Jan–Mar 2000

Apr 2000

* PCA, patient-controlled analgesia.

hospitalwide the use of separate order forms for heparin for each of three different patient conditions: coronary syndrome, stroke, and deep vein thrombosis (DVT). Success factors. Factors included leadership by an influential physician on the P&T committee, who chaired the monthly adverse drug events subcommittee, and generation of clever and resourceful changes—for example, printing in the pharmacy of the initial batch of colored forms to enable their use to begin even before the print shop delivered them. Spread. Four of the six hospitals have since both adopted weight-based heparin protocols and documented clinically and statistically significant declines in the time to therapeutic level of anticoagulation. Key contact. The key contact was Michael Benari, MD. Case Study 2. Fewer Lookalike/Soundalike Errors at New England Baptist Hospital Context. Four incidents had occurred during 6 months because drug names looked alike or sounded alike. Changes. To increase the awareness of the drugs used within the institution, a contest was held to see which patient care area could identify the most lookalike/soundalike drugs. A list of approximately 100 drugs was developed and posted in each medication room. Drugs that look alike that were next to each other in the Pyxis Medstation® (Pyxis Corp; San Diego), in various medication kits, or on the CPR (cardiopulmonary resuscitation) cart were either rearranged or removed, or the drugs were purchased from a different manufacturer with different packaging. For a few high-risk drugs, the pharmacy affixed a warning label to the medication.

132

For each of the soundalike drugs, the pharmacy programmed the indications for the drug into the Pharmacy Drug Data File, which interfaces with the Pyxis Medstation medication profile. When the nurse calls up a drug that is on the soundalike list, the indication of the drug appears on the Pyxis medication profile screen. For instance, if the nurse requests quinine for a patient, the message will indicate that this is for leg cramps. The nurse can then correlate the indication with the patient’s symptoms before administering the drug and avoid confusion between quinine and quinidine, which is used for cardiac arrhythmia. Measuring method. The measuring method was a count of incidents before and after the changes. Results. The hospital has seen a decrease in the number of lookalike/soundalike errors and has not seen any repeat incidents during the past year. Success factors. Success factors included creativity in using a contest and a range of different changes in both the pharmacy and on patient floors. Spread. One other hospital has used these techniques and now routinely meets our criteria for preventive labeling of lookalike/soundalike medications. Key contact. The key contact was Karen Hoikala, RPh. Case Study 3. Fewer PCA Administration Adverse Events at BIDMC Context. Several incidents occurred related to equipment failures, use of the infusion pump, calculation errors, dispensing errors, and so on. Some of the incidents were due to the ordering, dispensing, and administration phases of the medication process.

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

JOURNAL

ON

QUALITY IMPROVEMENT

Changes. The team made the following changes to

reduce adverse drug events and improve patients’ satisfaction with pain relief: revised the pain management flow sheet; developed flow sheet guidelines; developed a skill check list for the Baxter PCA II pump (Baxter; Deerfield, Ill); standardized the type of PCA pump used throughout the hospital; revised PCA order forms; developed equianalgesic reference charts; developed drug calculation reference; educated all clinicians about practice changes; and developed a PCA resource manual for each unit. Measuring method. The measuring method was a count of incidents before and after the changes. Results. A run chart (of mean time between failures) of PCA incidents from June 1997 through March 15, 1999, showed that incidents occurred every 13.8 days, on average, before the changes. In the subsequent 612 days after the changes, only 25 incidents occurred, that is, 1 incident every 24.4 days. The sustained improvement since the changes is 77%. Success factors. Extensive consensus-building across both campuses of the newly merged academic medical center was achieved. Spread. All five other hospitals have since met our criteria that denote the routine use of a limited number of pump types. Key contact. The key contact was Nancy Miller, RN, MS. Case Study 4. Safer Administration of Coumadin at Deaconess Glover Hospital Context. Three coumadin-related adverse events involving two different patients had recently occurred, and additional doses may well have also been missed in the preceding months. Changes. The team changed the standard coumadin administration time from 6 PM to 2 PM to eliminate required hand-offs between day and evening RNs and attending and covering physicians. It also revised the order sheet to more clearly portray the series of coumadin orders for a given patient in light of the PT times. (A copy of the revised order sheet is available from the authors upon request.) Measuring method. The measuring method was a count of incidents before and after the changes. Results. In the 19 months after the changes were made in April 1999, one error of this type occurred. Spread. None of the other hospitals have yet documented improvements in this area.

VOLUME

27

NUMBER

Key contact. The key contacts were Julie Bonen-

fant, RN and Ron Staley, RPh. Case Study 5. Education of Patients About Their Medications at New England Baptist Hospital Context. On administration of medication, nurses did not consistently recite the name of the drug and the rationale for its use to the patient. Without this procedure in place, patients were unable to detect and prevent possible errors with their medications and ask appropriate questions. Changes. The nursing procedure was changed to include reciting the name of the drug and its rationale before administering it to patients. Measuring method. The measuring method involved observation and audits. Results. The frequency of the nurse’s verification with the patient of his or her individual medications increased from 79% to nearly 100%, and then it was measured again at more than 90%, according to observations outside the patient room and audits. The frequency with which nurses stated the rationale for the use of the medication rose slightly, from 63% to 75%, and then substantially more on remeasurement, to more than 90%. Spread. Four of the other five hospitals now meet the criteria for routine use of patient partnering techniques. Key contact. The key contact was Karen Hoikala, RPh. Case Study 6. More Expeditious Processing of the Morning Dispensing Backlog at Deaconess Waltham Hospital Context. The pharmacy was closed from 9 PM to 7 AM, so on many mornings dispensing was hectic and error prone, given the need to process the backlog from the previous evening’s orders. This environment had been identified as a root cause of an adverse drug event. Changes. A dedicated pharmacist was assigned to exclusively handle the entry of orders into the Meditech information system (Westwood, Mass), and the other pharmacist handled all other tasks until the morning backlog was completed. This function rotates from day to day. Measuring method. The measuring method was timing the completion of each morning’s backlog.

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

133

THE JOINT COMMISSION The hospital saw a reduction of 45 minutes, or 50%, in the time needed to process the backlog. In other words, instead of the hectic period lasting 90 minutes, from the pharmacy’s opening at 7 AM until 8:30 AM, this vulnerable window of time was halved, to 45 minutes. The pace was calmer, reducing the morning crunch and the likelihood of error. Spread. The best practice may be adopted elsewhere depending on pharmacy order entry processes. Key contact. The key contact was Arthur Dahl, RPh. Results.

Appropriate Pacing The pace needed to implement the best practices overall made it imperative to make many changes rapidly. Often, a team initiated several changes at one time, rather than sequencing changes in successive PDSA cycles. In fact, five of the six teams used multiple simultaneous changes. This made it difficult or impossible to ascertain the effects of individual changes. If a change had been ineffective, therefore, it was difficult to know that from objective data and to stop the change. The team leaders, however, were not troubled by this, and the multiple changes remain in effect.

Success Factors Key success factors, in addition to leadership, are the use of data, forcing functions, appropriate pacing, inexpensive changes, a consultant, and reporting. Data Most teams made little use of data until prompted to document the improvement that had occurred. Education was needed on ways to collect and display the data. This education often took place after changes had been made; indeed, baseline data were usually compiled after the changes were made, with our help. Now seeing the usefulness of the measurements, the team leaders are developing and using baseline measures earlier in their PDSA cycles. Forcing Functions Many changes in the health care industry rely on education or in-service training to prevent errors—or ultimately on memory, which is inherently unreliable. As Leape explains, more lasting improvement accrues through the use of forcing functions built into the process during redesign.17 Five of the six case study teams used forcing functions. The order forms, for example, embed stat PTT orders as a default rather than requiring the physician to remember to write in “stat PTT test” each time. As another example, alerts about maximum doses for particular drugs now appear automatically on a pharmacist’s entry of an order into the Meditech computer. A key goal is to capitalize on the capability for similar forcing functions via automation of the medication process. And to improve coumadin administration, moving the standard administration time from 6 PM to 2 PM eliminates the need for hand-offs between clinicians on different shifts.

134

Inexpensive Changes Perhaps most notably, the changes have been inexpensive. None required capital expenditures. Indeed, only one change—standardization of PCA pumps—required any outlay whatsoever for purchase or lease. This thrift, in fact, is built into the PDSA method. Team leaders have to select changes that are readily feasible, which generally rules out buying something through an often lengthy purchasing process. Surprisingly, in this era of cost consciousness, this valuable feature of the PDSA method is little recognized. Consultant The consultant mentors the team leaders. The hospitals’ team leaders and members have one-to-one consultations in person and on the phone to discuss tactics and barriers. To encourage the leaders’ ownership, the consultant does not lead the teams. Rather, the consultant takes a more active role in leadership of a team initially, for example, in leading the team through a PDSA cycle, and as the leader becomes more adept at it, the leader takes on that responsibility. The leader and consultant stay in touch by phone, fax, and e-mail, and they swap educational materials. Individual team leaders and team members need to think and act differently. The consultant needs to consider and adapt to the unique personality and preferences of each individual to build a trusting advisory relationship. In the work reported here, the consultant conducted an informal assessment of the fit between each leader’s styles and preferences and his own, using a typology developed by James Grant.* For example, for * James Grant, PhD, head of the Theravision Institute, communications with author [K.F.], 1999–2000.

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

JOURNAL

ON

QUALITY IMPROVEMENT

leaders with a flowing, consensus-building style, he communicates with active listening, empathizing, and responding to their feelings. For more pragmatic, practical leaders, he emphasizes orderliness, structure, facts, and data. He also accommodates the leaders’ individual interests. For example, one research-minded team leader, pursuing her dissertation, is especially interested in areas of overlap between the team’s needs and her thesis. A physician leader is interested in broadening his knowledge of improvement concepts and techniques as a P&T chairman while reducing adverse drug events. In addition to the PDSA model, the consultant uses two other models of social and individual psychology to guide and stimulate change. Rogers’s classic analogy of agricultural extension agents in promoting the use of new agricultural techniques provides one model for coaching and epidemiological diffusion.18 The consultant uses this model, for example, in publicizing locally tested innovations such as a heparin protocol to other hospital teams. The second model considers the readiness to change of team members and leaders, as recommended by Prochaska et al in Changing for Good, which provides a theory for the progressive stages of individuals’ behavioral changes.19 Reporting To date, given the inherent limitations of reporting systems, we have emphasized improvement more than reporting. Beginning in 1997, Risk Management Foundation (Cambridge, Mass) tabulated data, including reported medication errors, related to incident reports across the CareGroup network. More recently, the task force has been using a CareGroupwide intranet Executive Information System (EIS), which allows users to generate histograms and run charts by type and level of medication error. CEOs have presented these data at board meetings. Use of the EIS is not yet a habit among team members, however, and few of them have capitalized on the capability to see these data online. Given that incident reports are likely to constitute a very small fraction of adverse drug events,20 we and the five teams are trying to raise reporting rates. Five of the hospitals have raised their incident reporting rates to a much higher and sustained level, which suggests that a difficult cultural change—nonpunitive error reporting—is well under way.

VOLUME

27

NUMBER

CareGroup is excited about the results of confidential peer interviews of frontline providers, being pioneered by Dr Saul Weingart and colleagues at BIDMC.21 During a 3-month study, the hospital incident reporting system identified 58 incidents on a medical unit (19 involved medications). In the same period, physicians reported 110 incidents (12 involved medications) through confidential peer interviews. A single incident (a fall with minor injury) was identified by both methods. A comparison of this method with the more conventional incident reporting appears in Table 2 (p 136). Because this form of reporting has been particularly useful in stimulating clinicians’ energy for change, Weingart is broadening its use.

Limitations, Barriers, and Next Directions CareGroup is early in this multiyear effort, and while proud of the improvements to date, it is acutely aware of the daunting tasks yet to be done. Assessment of the interventions was limited insofar as it generally involved before–after designs, so that temporal factors could have affected the results. For example, perhaps other simultaneous changes during a pilot test, not known to the team, could have caused the improvement at hand. CareGroup faces key challenges in measurement and in spreading and deepening the involvement of clinicians, particularly of physicians. Broader and better measures are important for marking progress; measures are a call to action and a celebration of action. They show that systemic change is needed and then that systemic change has been possible. It is particularly important to use measurements to bring physicians into this effort. Measurements should also reveal success stories and practical tips so as to provide descriptions of exactly how other physicians have contributed to successes. CareGroup lacks an overall, objective measure of medication safety. We attempt to mitigate this in three ways. First, we use multiple data sources, especially automatic, local, unobtrusive measures where they are available—for example, PTT test times. Second, we solicit the advice and approval of other experts. Finally, we continue to improve our measures. However, these measures are not, and cannot become, comprehensive. With the exception of a handful of particularly motivated physicians, most have watched from afar. A

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

135

THE JOINT COMMISSION Table 2. Comparison of Incident Reporting and Voluntary Structured Confidential Interviewing*

Incident reporting

Structured confidential iInterviewing

How evoked

Pushed up

Pulled

Scheduling

Squeezed into busy day

Scheduled

Data collection mechanism

Gathered by typing into computer

Gathered by specially trained clinician interviewer

Threshold

Requires decision to generate, and sitting down at computer

Requires “yes” in answer to a question and an oral description

Disposition

To risk managers for corrective action

To team leaders for tabulation and PDSA improvement

Resultant number of reports

Low

High

* PDSA, plan-do-study-act.

frustrating case in point lies in the enforcement of standardized prescribing, for example, getting doctors to write units instead of the abbreviation U, which can resemble a zero and cause a tenfold overdose. The process for spreading enthusiasm among doctors has relied on the unvoiced expectation that they would discuss progress with their peers. Physicians are now being enrolled through a more thoughtful process, with the help of Dr Michael Benari, the head of a P&T committee, and Dr Saul Weingart. Dedicated and reimbursed time of physicians will be crucial. CareGroup also needs physician leadership for the best practices involving protocols and guidelines—for example, for emergency room patients and other patients needing complex care. Physician leadership is also needed for two especially difficult best practices: enforcing standardized prescribing and minimizing the use of long physician work shifts. When CareGroup agreed in spring 1999 to pursue those best practices, the goal was to require all clinicians to work no longer than 12-hour shifts. In early 2000, in framing operational definitions for all 16 best practices, CareGroup retreated to require the 12-hour shift for only nurses and pharmacists and delayed the requirement for physicians. As safety experts in the airline and nuclear safety industries realized as they decreased error rates,22–23 it is critical to aggressively track and reduce near-misses as well as adverse events that reach the patient. CareGroup aims to proactively and systematically identify and measure the holes in the safety nets. Individuals, rather than teams, will perform the bulk of the operationalization and collection of measures. Three of

136

these data collection cycles—gathering data on pharmacies’ intercepts of unclear or inaccurate orders, allergy alarms, and drug–drug interaction alarms— have already been performed. Generally, in quality improvement it is highly advantageous to improve the vital few problem areas.24 These improvements are most likely when the key actors and those with formal authority for a process lead the effort. Hence, medication ordering, the most common source of adverse drug events,25 will offer a fertile area for improvement, as soon as physicians who order medication can be equipped with fail-safe methods for order entry and revision. Spread of the specific changes made in the case studies, as opposed to the spread of the best practices themselves, has been limited. Two of the case study interventions—in coumadin order sequencing and dedicating a pharmacy technician to order entry— have been implemented at only one site to date. As more complex changes are developed and proven, particularly at the larger hospitals, it will be helpful to tap the power of informal influence networks.18 Team leaders have been urged to identify the early adopters among department or nursing floor heads and to begin using them to spread changes that had been successfully tested in the adopters’ own units. Success with tapping informal influence networks, however, has been modest; the teams have preferred to continue working with their original change leaders. We strongly encourage the adoption of the change ideas across hospitals. Yet given innate differences among hospitals, not every change is appropriate for widespread use.

MARCH 2001 JOURNAL Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

JOURNAL

ON

QUALITY IMPROVEMENT

While the changes are all positive developments, the eventual impact of the changes planned for the future, through automated systems such as computerized order entry, is much larger. The consultant and information technology staff are now gathering preliminary data on the extent of improvement due to automatic alarms in the pharmacy computer systems—the earliest of our automated systems. We anticipate continued progress for five reasons. First, our teams have acquired skill with PDSA. Second, they are more facile with, and less fearful of, measurement. Third, the positive effects of most of these improvements should remain in force, since five of them rely on forcing functions. Fourth, the reporting of near-misses has been validated and is now being spread. Fifth, the aggressive development of key technological changes is continuing. Yet the changes we seek are vastly and inherently broader than any team leader’s ability to

control them, and we must involve every physician, nurse, and pharmacist in them. Our team leaders owe much of their own professional success to their attention to detail, close management, orderliness, and ability to control complex processes. The key leadership tasks before us require the delegation of many tasks— that is, the release of the very control that the team leaders have used to great effect. Their understandable reluctance to offload certain tasks to others may limit our efforts to spread changes more deeply. We anticipate much progress on six simultaneous fronts: adoption of best practices; improvement in top-priority areas of each hospital; improvement in automation and technology in ordering, dispensing, and administering medication; and better reporting. In each of these individual six areas we hope that CareGroup is, within a year, among the safest hospital systems in the United States. J

References 1. Weingart S: Making medication safety a strategic organizational priority. Jt Comm J Qual Improv 26:341–348, 2000. 2. Leape L, et al: Reducing Adverse Drug Events. Boston: Institute for Healthcare Improvement, 1998. 3. Kirle LE, et al: The Massachusetts Coalition for the Prevention of Medical Errors. Proceedings of Enhancing Patient Safety and Reducing Errors in Health Care. Rancho Mirage, CA, Nov 8–10, 1998. 4. See www.mhalink.org. 5. Pharmacists key in a low-tech plan to reduce medication errors. Briefings on Patient Safety, 1(4): 1, 4, 12, 2000. 6. Larson L: Ending the culture of blame: A look at why medical errors happen and what needs to change. Trustee 53(2):6–10, 2000. 7. Charatan F: Senators introduce bill to improve patient safety. BMJ 320:465, 2000. 8. Ayers T: Computers aid

VOLUME

27

NUMBER

patient safety. Science 282:2295, Dec 18, 1998. 9. Bates D, et al: Effect of physician computer order entry and a team intervention on prevention of serious medication errors. JAMA 280:1311ff, 1998. 10. The press release about the 2,000,000th order can be found at www.medicine.indiana.edu/news_ releases/archive_00/physician_ computer_00.htm. 11. McDonald C: Protocolbased computer reminders, the quality of care and the nonperfectability of man. N Engl J Med 295:1351–1355, 1976. 12. Overhage J, et al: A randomized trial of “corollary orders” to prevent orders of omission. J Am Med Inform Assoc 4:364–375, 1997. 13. Evans RS, et al: A computerassisted management program for antibiotics and other anti-infective agents. N Engl J Med 338:232–238, 1998.

14. Pham A: Regrouping, with care: With experience, a touch of zeal, and a belief in joyousness, CEO takes on tough task. Boston Globe, Apr 18, 1999, p F1. 15. Langley G, et al: The Improvement Guide: A Practical Approach to Enhancing Organizational Performance. San Francisco: Jossey-Bass, 1996. 16. Matlack W: Statistics for Public Policy and Management. North Scituate, Massachusetts: Duxbury Press, 1980, p 290. 17. Leape L: Error in medicine. JAMA 272:1851–1857, 1994. 18. Rogers E: Diffusion of Innovations 4th ed. New York: The Free Press, 1995. 19. Prochaska JO, Norcross JC, DiClemente CC: Changing for Good. New York: William Morrow, 1994. 20. Cullen DJ, et al: The incident reporting system does not detect adverse drug events: A prob-

lem for quality improvement. Jt Comm J Qual Improv 21:541–552, 1995. 21. Weingart S, et al: Confidential clinician-reported surveillance of adverse events among medical inpatients. J Gen Intern Med 15: 470–477, 2000. 22. Barach P, Small S: Reporting and preventing medical mishaps: Lessons from non-medical near miss reporting systems. BMJ 320: 759–763, 2000. 23. Rees J: Hostages of Each Other: The Transformation of Nuclear Safety Since Three Mile Island. Chicago: University of Chicago Press, 1988 (see especially the appendix). 24. Juran JM: Juran on Quality by Design. New York: The Free Press, 1992, pp 59–61. 25. Bates D, et al: Incidence of adverse drug events and potential adverse drug events. JAMA 274: 29–34, 1995.

3

Copyright 2001 Joint Commission on Accreditation of Healthcare Organizations

137