The Model for Improvement

Heartland Medical Group was a primary care practice of 12 family physicians, 7 internists, 8 adult nurse practitioners (NPs), and 1 obstetrician who cared for women with high-risk pregnancies. Heartland had 1 clinic site in a town of 12,000 people. It served people from this town as well as several smaller towns nearby. The 19 physicians and 8 NPs in Heartland were divided into 3 groups with separate front desk staff and separate medical assistant staff. In other words, there were 3 teams of clinicians and support staff. In addition, a group of registered nurses (RNs) and administrative staff supported all 3 teams. Heartland used an electronic health record (EHR) to which it had access as a result of an affiliation agreement with the hospital where the physicians also practiced.

Public reporting of medical group performance was becoming a reality in the state where Heartland was located. A project had been underway for 3 years to introduce the idea. Medical groups’ performance in preventive care and chronic disease care had been measured and reported back to the medical groups but not more widely. All medical groups knew that the quality of care measurements would soon be reported on a public website, starting in 1-2 years.

One component of chronic care being measured was care for patients with asthma. The 2 measures reported were the percentage of patients who had high scores on a self-administered questionnaire about their asthma symptoms (a high score indicating that the patient had few symptoms) and the percentage who reported that they had been hospitalized for asthma during the past year or had visited an emergency department (ED) twice (or more) for asthma. In the most recent report, Heartland had scored poorly relative to other medical groups on both measures. Only 20% of their patients had high scores on the self-administered questionnaire, and 14% of their patients had been hospitalized during the past year or had 2 more ED visits. The physicians and NPs wanted to improve these scores.

With the approval of Heartland’s Clinical Services Committee (the day-to-day operations committee), a quality improvement (QI) team was formed to serve all 3 primary care teams. The QI team consisted of a physician from one primary care team, an NP from a second team, another NP from the third team, an operations manager, a medical assistant, an RN, and a staff member from the hospital’s EHR support group. The general aim of the QI team was to improve Heartland’s scores on the asthma measures so that Heartland could “hold its head up with pride” in the company of the other medical groups in the state.

The team met for the first time to review its charge and get to work.

Like the story about surgical safety training in Chapter 14, this story is recurring throughout the United States as clinics and hospitals become motivated to improve their performance. In many cases, the motivation of the clinics and hospitals stems from their concern for their reputations. The concerns are being generated by public reporting of how well patients fare under their care. Sometimes insurance company payments are also tied to meeting certain thresholds on quality measures although not in the case of Heartland and its asthma care. There was no question that Heartland wanted to achieve better results for its patients with asthma. What they sought was a pathway to better results. What approach did they use?

The traditional method for Heartland was the Model for Improvement, which is shown in Figure 16–1 (Langley et al, 2009, pp. 15-25). The Model for Improvement is one interpretation of the CQI approach to process improvement. It derives from the work of Deming and Shewhart, who originated the cycle of Plan-Do-Study-Act or PDSA (Langley et al, 2009, p. 465). The Model also includes 3 preliminary questions to be answered before beginning cycles of PDSA. In the United States, the Model for Improvement has become very widely used. It was devised and has been taught by Associates in Process Improvement (API), Austin, TX, and is sometimes referred to as the API Improvement Model. The model has been disseminated and promoted by the Institute for Healthcare Improvement, Boston, MA.

The first step in the Model for Improvement is to answer the question “What are we trying to accomplish?” In other words, the first step is to state as clearly as possible the aim of the improvement project or endeavor. General statements of aim are not useful without further specification because they do not lend themselves to careful determination of whether progress is being made. For example, in the case of Heartland’s asthma care project, the team might have decided that they were trying to accomplish an improvement in their care of patients with asthma. Although this statement is, of course, correct, it would have afforded the team little direction and little basis for knowing whether they were succeeding. To some people at Heartland, accomplishment of this aim might have meant that asthma education should be improved. To others, it might have meant that more patients should be treated with medication having long-term effects, for example inhaled corticosteroids (beclomethasone, fluticasone, and similar drugs delivered with an inhaler). Still others might have been thinking of improving the assessment of the severity of patients’ disease. Any of these actions might have been appropriate; but in the end, how would the QI team know with confidence whether they were making progress toward their goal of improving asthma care?

As it happened, the Heartland QI team was not tempted to settle for a general statement of its aim because the organization measuring their performance had in essence assigned them a twofold aim: (1) to reduce the burden of symptoms in their patients with asthma, and (2) to reduce their asthma patients’ needs to obtain care for their symptoms in the inpatient portion of a hospital or in an ED. The statement of these 2 aims provided an adequate answer to the question of what the team was trying to accomplish. These 2 aims are not the only defensible aims that the team could have chosen. For example, they might have settled on symptom reduction as one aim but decided that improvement of actual breathing capacity (speed of airflow into and out of the lungs) would be their second aim, to be measured by devices used by the patients at home or in the clinic. But the team was content with the 2 aims stipulated by the reporting organization.

And yet stating the twofold aim was not enough to enable the team to chart its progress. The team needed to answer the Model’s second question, “How will we know that a change is an improvement?” In other words, for each aim, they needed to have a well-defined measure that would enable them to quantify their progress and know whether the changes they were making in the clinic were lessening their patients’ symptom burdens and decreasing the patients’ needs for emergency care. Again, the measurement question was defined for them by the reporting organization, which had adopted detailed specifications for the measurement of symptom burden and use of emergency care. Briefly, the specifications started with a definition of the population of asthma patients in Heartland’s practice. This definition was needed for the measurement of progress on both aims. Next, the use of a specific questionnaire was called for, enabling a symptom score to be obtained for each patient indicating the patient’s level of symptoms during the week prior to the date when the patient completed the questionnaire. The proportion of symptom-free patients was then defined as the number of patients with a high score (above a specified threshold) on the questionnaire divided by the number of asthma patients. The use of emergency services was to be determined by using hospital and ED records as reflected in billing statements sent to insurance companies. The proportion of patients needing emergency care was defined as the number of patients with a hospitalization or 2 ED visits (or more) for asthma during the past year divided by the number of asthma patients. Additional details were also spelled out in the specifications. Again, various team members might have preferred different measures of attainment of the aims. For example, they might have preferred to use a different questionnaire or to define use of emergency services so that the number of ED visits in the definition was 3 instead of 2. But the team was satisfied enough with the definitions used by the reporting organization.

The next step in the Model for Improvement is to answer the question “What change can we make that will result in improvement?” This question is open to misinterpretation because it provides no indication of how an idea for a change is to be obtained. It might be tempting for a QI team to seize upon the first idea that is voiced by anyone on the team. It would be imprudent to proceed with the first idea offered or with someone’s pet idea not subjected to any evaluative thinking. But it also would be wasteful to delay taking action while the possibilities for changes are investigated at great length. Some systematic approach to generating ideas for changes and evaluating them is appropriate at this stage in using the Model.

It is always useful to inquire what other organizations have done to solve the problem at hand. If the problem is a clinical problem, sometimes the inquiry can be done by searching the literature, using MEDLINE or an alternative database. However, it is often faster and more productive to inquire of other medical groups or hospitals that are facing or have faced the same challenge.

The other approach commonly employed to generate ideas is to use one or another of the tools for creativity presented in Chapter 10, for example, some version of team brainstorming. Alternatively, a QI team can ask 2 or 3 of its members to confer and present a few possibilities to the whole team for consideration. This approach assures that possible process changes receive some critical appraisal even before they are presented to the team and that further debate and critique occurs in the whole team—provided that the small group of 2 or 3 is asked to present some well-considered possibilities and not what they have firmly concluded are the best change ideas.

The QI team at Heartland proceeded first by gathering some background information that could be collected quickly. First, they wanted to know how many asthma patients Heartland cared for. They learned from a review of their electronic records that there were 2521 asthma patients in the practice. They suspected that the practice actually had more asthma patients, but 2521 was the number that could be verified by using their data. Because the approach to assessment and treatment is different for children and adults, they also determined how many patients were in each age group.

Their next step was to generate a list of changes that they might make to improve care and achieve their aims. They generated their list of possibilities as a group, not by asking a sub-group to meet separately and present a list to the whole group. The team used a version of the nominal group technique, discussed in Chapter 10. Every member of the team was asked to write down 3 ideas for changes that could be made. The chair of the team then asked team members one at a time to present an idea. She continued calling on team members around the table until all of the group’s ideas were listed on a flip chart. The list consisted of 26 possible changes, some of them overlapping. Among the suggested changes were:

• Assure that every clinic visit by an asthma patient includes a formal assessment of the patient’s asthma symptoms.

• Contact asthma patients who have not returned to the clinic for 6 months and ask them to return for a visit.

• Assure that inhaled corticosteroids are prescribed for all asthma patients whose symptoms are not well controlled by short-acting inhaled medications (β₂-agonists).

• Track ED use and contact patients seen in an ED for asthma symptoms to assure that they come to the clinic within 1 week of the ED visit.

Some of the change ideas came from reading done by QI team members, and some were solely the results of individual insight and creativity. Other ideas were taken from other medical groups in the state, whose efforts to improve asthma care were known by some of the QI team members.

After a list of possible changes was generated, the next step was a discussion of criteria to use in choosing among the possibilities. The criteria might have been devised before the list of possible changes was generated, but it is usually easier for teams to deal with the criteria after they have a concrete list of possible changes in front of them. Criteria often considered in making choices among possible changes are:

• The change under consideration has a reasonable likelihood of success.

• Compared with other possible changes, the change is likely to have a greater impact in achieving the project’s aim(s).

• The healthcare practice or hospital has the resources necessary to make the change.

• There is evidence that the change was effective when it was used in other similar settings.

The QI team’s next step was to discuss how the possible changes in the list compared with one another when they are judged against the criteria. Some teams go so far as devising a matrix to make explicit the scores for each possible change with respect to each of the criteria, with the scores usually assessed only roughly as “high,” “medium,” or “low.” Sometimes different possible changes are combined, and some possibilities may be dropped at this point if they are deemed by consensus to be unpromising.

Finally, the QI team chose a change to test. Sometimes more than one change is chosen for testing simultaneously. The final choice is usually made by consensus. Alternatively, there are several voting procedures that can be used, as explained in Chapter 10. One popular voting method is multivoting (Scholtes et al, 2003, pp. 3-20 to 3-21). Team members multivote by voting for 2-5 of their top choices, thus avoiding an artificial forced choice by each member of a “top choice.” This procedure recognizes that preferences among possibilities usually are not black-and-white but differ like shades of gray. The votes are tallied as usual, and the top choices for the team are those with the highest numbers of votes. The team may choose to test the change that received the highest number of votes, or it may choose the top 2 or 3 changes to test.

The steps described here, from posing the question “What change can we make that will result in improvement?” to choosing a change to test, may appear time consuming. However, provided that team members have done some homework in advance to inform their discussion, the entire sequence can usually be completed in a single meeting of 90 minutes or so.

The Heartland asthma QI team first chose to test whether performance could be improved by reaching out to patients who had not had a clinic visit for 6 months. The medical assistant on the QI team plus one of the NPs proceeded to identify which of the 2521 patients with asthma had not been seen in the clinic for 6 months. They discovered that only a tiny proportion of the 2521 patients had not been seen, and so this approach was abandoned. However, there was benefit for the team in that they learned that the low performance scores for Heartland were not due to poor follow-up of patients. The clinicians were seeing the patients on a timely basis, but they needed to change what they were doing during the visits.

The next change to be pursued was to assure that inhaled corticosteroids were prescribed for all patients whose symptoms were not well controlled by short-acting inhaled medications. The medical assistant and RN worked with the EHR expert to obtain a list of all asthma patients who had been prescribed medication for their symptoms but not corticosteroids (or more potent medication). They examined a sample of these patients’ records and found that at the most recent visit it was usually not possible to determine whether the patient’s symptoms were controlled or not. Formal assessments of symptom severity were generally not done or, if done, were not recorded in the record. The team decided to test whether improvement could be achieved by increasing the frequency of both routine formal assessment of symptoms and prescription of inhaled corticosteroids when needed.

At this point, the QI team had reached the beginning of the PDSA cycle in Figure 16–1. The next step was to Plan, that is, plan the change. They decided to add a prompt to the EHR that would remind the medical assistant to have all patients with asthma complete the customary symptom questionnaire at every visit, construing the administration of the questionnaire in the same way that they regarded measuring a patient’s blood pressure, that is, as a routine element in every visit for patients with asthma. At the same time, they planned to promote the prescription of inhaled corticosteroids for appropriate patients, to be identified by their poor scores on the symptom questionnaire. This promotion would be done by academic detailing, that is, by having 2 highly interested physicians initiate conversations with all other clinicians in the medical group to remind them about the appropriate prescription of inhaled corticosteroids (Soumerai and Avorn, 1990). In addition, the QI team decided to distribute short e-mail updates on the improvement project to all clinicians and medical assistants at Heartland.

The next step in the Model for Improvement is Do, that is, make the change in a limited or test environment. The QI team decided to try this 2-pronged change (using symptom questionnaires and academic detailing) in just one of the 3 teams in the clinic. The EHR prompt was added and the academic detailing was carried out within the test setting within a few days. (Meanwhile, the e-mail updates on the project were distributed every 2 weeks to those on all 3 clinical teams.)

The next step in the Model is to Study, that is, find out what has happened as a consequence of instituting the test change. Some years ago, this step in the PDSA cycle was called “Check,” and the cycle is still sometimes called the PDCA cycle. The first part of this step is to check to see whether the planned change was implemented as intended. The second part is to inquire whether the change had its intended effect. At the end of 2 weeks, the QI team assessed its test of the change by simply inquiring of those working in the test setting, first, whether the EHR prompts were appearing on the computer screens as intended. Second, they asked whether the medical assistants were administering the symptom questionnaires and whether the clinicians were using the questionnaires to decide whether to prescribe inhaled corticosteroids. Everyone on the QI team presumed that the reports about the function of the EHR would be reliable but that the medical assistants and clinicians were likely to exaggerate how well they were following through with the planned change—not because they deliberately misreported their actions but because of optimism or selective memory. But, at this stage, a subjective check on progress was sufficient. A more thorough investigation came at 6 weeks when 3 members of the QI team reviewed a sample of records for visits by asthma patients occurring after the start of the test of change. There had been 105 visits by patients with a diagnosis of asthma. One hundred visits were reviewed. For 94 visits, a symptom questionnaire score was recorded in the record. For 14 of these 94 visits, the patient’s symptom score was poor and the patient had not been prescribed inhaled corticosteroids previously. In 13 of these 14 cases, the clinician prescribed an inhaled corticosteroid at the time of the visit. (In the one case that was a failure, the physician said with a wink that she had “lost my head”—made a mistake.) The test of the change was judged a success.

The final step in the PDSA cycle is Act. In the Model, to act means to take broader and more long-lasting action. In the case of the Heartland project, acting meant extending to the other 2 clinical teams the EHR prompts, routine use of the symptom questionnaire, and routine prescription of inhaled corticosteroids to patients whose symptoms were poorly controlled with short-acting inhaled medication.

After 4 more months, the QI team did a full audit of its performance on the 2 measures that would soon be reported publicly. The audit showed that both the symptom burden and the use of emergency services by asthma patients had improved—but not enough for those at Heartland to be satisfied. This result, in turn, raised the question of returning to the PDSA cycle to consider more possibilities for process changes, to choose 1 or 2 more changes to test, to plan them, and to continue on through the cycle again. As in this case, the quest for quality usually continues indefinitely.

This story of Heartland Medical Group’s pursuit of improved performance in caring for its patients with asthma illustrates the Model for Improvement. In other medical groups, the story would have been different. The range of process changes considered might have been different. The changes first chosen for testing might have been different, resulting in different details during the Study phase of the first turn of the cycle. But the steps traversed would have been the same.

There is a good deal more to using the Model for Improvement than the events described in Heartland’s story. For example, no mention was made of the use of statistics, which would help determine whether the apparent improvement in quality scores was real and not simply a result of random fluctuation. The basic statistical tools for analyzing process improvement scores are run charts and control charts. A run chart is a simple plot of measurements over time with no statistical calculation results included in the chart except for the average of the measurements (Brassard and Ritter, 2008, pp. 124-127). A control chart includes results of statistical calculations that permit large, real changes to be distinguished from random fluctuations (Montgomery, 2009, pp. 179-343). A third type of chart for tracking process changes, commonly used in manufacturing and service industries but seldom used in health care, is the cumulative sum chart or cusum chart. A cusum chart includes more complex calculation results that permit small but persistent, real changes to be distinguished from random fluctuations (Hawkins and Olwell, 1998; Montgomery, 2009, pp. 400-419).

In the Heartland story, there is also no mention of the many diagrammatic tools that are used in process improvement, for example, process maps, cause and effect diagrams, and project management charts (Nelson et al, 2007, pp. 296-307, 313-320, 362-368). Finally, there is no discussion in the story of the organizational supports that are needed for a team to be able to succeed consistently in improving its processes. These supports include an organizational culture that nurtures perpetual improvement and construes clinical work as encompassing not only the care of patients but also work to improve the processes of patient care. These topics are covered in works by Deming (1986), Ohno (1988), Harry and Schroeder (2000), and Nelson and colleagues (2011).

Some of the organizational supports needed for effective process improvement are the same as those needed for teams to succeed in general. Chapter 18 deals with the roles and responsibilities of senior leaders in supporting teams. Some of the necessary supporting actions aid process improvement.

OTHER APPROACHES TO PROCESS IMPROVEMENT

As mentioned earlier, the Heartland QI team could have used a different step-by-step process improvement method to improve the care it provided for Heartland’s patients with asthma. Two other choices are described here briefly. These methods are the only other methods that currently receive frequent attention in the healthcare literature. They are less commonly used in the United States than the Model for Improvement.

Lean Production

The Lean Production approach to process improvement focuses on eliminating waste or “fat” in processes, hence the name. This approach is often called the Toyota Production System (TPS) because it was developed in the Toyota Motor Company beginning in about 1950. The system is a fully developed method for the leadership and management of any enterprise, and it includes all aspects of performance improvement. Embedded within the system is a methodical approach to process improvement in particular (Liker, 2004).

Forms of Waste

In the Lean approach to improvement, processes are improved by ridding them of 7 different types of waste (Ohno, 1998, pp. 19-20). The types of waste, listed in Table 16–1, are:

Stay updated, free articles. Join our Telegram channel

Join