Becher, Elise C. MD, MA; Chassin, Mark R. MD, MPP, MPH
Physicians are emerging victorious from the managed care battlefield. Health plans are in various stages of retreat across the country, often with their stock prices plunging. They are reducing or eliminating utilization review, broadening their hospital and physician networks, and easing access to physicians within networks.1,2,3,4 Federal and state legislators continue to debate and enact new laws to further restrict health plans from pursuing the practices deemed most objectionable by patients and physicians.5,6 Having had their more controlling care management tactics resoundingly rejected, health plans are reinventing themselves and emphasizing their role as purveyors of information to consumers about health care options.3 Large public and private employers that previously supported health plans in their efforts to command significant authority over decision making are also reversing course and joining in this effort.7
Although these developments may be welcomed by physicians, celebration of complete victory may be premature. Employers, government, and health plans do not propose to return to the days when they simply paid the increasing costs of all health care interventions ordered by physicians. Instead, they talk about creating a market in health care in which consumers have information about quality and price and make their own choices. The endgame of this strategy may be the defined contribution approach to financing, in which employers provide a fixed amount of subsidy for their employees' health care, and the employees take the rest of the responsibility to find health insurance and pay for care. Once established, this principle would permit employers to limit substantially their exposure to the costs of health care, while putting employees at greater financial risk. Instead of confronting faceless insurance company functionaries on the telephone, physicians may in the future have to contend directly with many more of their patients over the costs of care.
Physicians now enjoy a moment of tactical advantage in the evolution of the struggle for control over health care in the United States. We suggest that the most effective way to capitalize on this—perhaps fleeting—position and to more permanently alter the balance of power in their favor is for physicians to establish strong and visionary leadership in health care quality improvement, with the goal of achieving unprecedented levels of excellence. Such an undertaking, if successful, could place the very essence of health care—defining, measuring, and improving its quality—in the hands of physicians. To succeed in such an endeavor requires understanding the different kinds of quality problems that plague our health care system, the different kinds of errors that lead to them, and how amenable these different kinds of errors may be to different interventions.
In this paper, we delineate a conceptual framework that describes these relationships, and we discuss their implications for conducting effective and durable quality improvement. We illustrate the practical ramifications of physicians' taking a vigorous leadership role in quality improvement by outlining how physicians could engage in this activity in three different settings: a four-or-five-physician primary care practice, a 50-physician multispecialty group, and a 450-bed community hospital.
DEFINING QUALITY PROBLEMS AND ERRORS
The Institute of Medicine's 1990 definition of quality has proved the most useful and durable of all the attempts to capture succinctly the vital elements of quality in health care: “Quality is the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge.”8 The several concepts embodied in this formulation of quality have been described in detail elsewhere.9 Here, it is most important to emphasize that quality is impaired in different ways by the three major classes of quality problems: overuse, underuse, and misuse.10 When patients receive services from which they will not benefit (overuse—e.g., patients who receive antibiotics for colds), the likelihood they will experience good outcomes is reduced because they are exposed to the risk of the adverse effects of these treatments without the possibility of benefit. When patients do not receive services from which they would benefit (underuse—e.g., heart failure patients with systolic dysfunction who are not treated with angiotensin-converting enzyme inhibitors), their likelihood of desired outcomes is reduced because they do not realize the health benefits of these treatments. Finally, when appropriate health services are provided ineptly (misuse—e.g., patients with renal insufficiency who require aminoglycoside antibiotics but receive doses that are not reduced to accommodate their kidney dysfunction), the probability of good outcomes is diminished by the added risk of avoidable complications.
Research in cognitive and behavioral psychology over the last several decades has probed the nature of the mental processes by which we carry out everyday actions. Studying the commonplace errors that occur in ordinary life has illuminated these mental processes and provided great insight into how they govern human behavior. James Reason has provided what is perhaps the most comprehensive and useful synthesis of this body of work in his landmark book, Human Error.11 Lucian Leape's incisive analysis was one of the earliest attempts to apply these principles to health care.12 A few concepts are most relevant to this discussion. We humans have a limited capacity for focused reasoning that applies knowledge to solve new problems (the “knowledge-based” level of performance). This scarce resource is, therefore, rationed only to those problems that cannot be adequately dealt with in other ways. Many behaviors are guided by other psychological mechanisms that involve recognizing patterns of signals from the outside world and reacting to them by employing previously learned and stored sequences of behaviors (the “rule-based” level). At the lowest attention-requiring level, many actions are governed by pre-programmed instructions that are initiated by a single intention (the “skill-based” level). Thus, if you decide to get up from your chair and walk down the hall to a colleague's office, most of the actions that carry out that plan are conducted at the skill-based level of performance (e.g., standing up, pushing your chair back, stepping away from your desk, walking out the door of your office, etc.). Rule-based actions may also be initiated unconsciously (e.g., seeing that the door to the hallway is closed, you initiate the sequence of actions required to open it). These cognitive strategies act to preserve the scarce resource of the knowledge-based level of performance so that on the way down the hall, you can think about the subject you wish to discuss with your colleague.
Humans strongly prefer to operate at the skill- and rule-based levels. Faced with a specific problem, we will search the environment exhaustively for patterns that match our storehouse of rules before admitting that the problem is one not previously encountered and is one that requires invoking the knowledge-based level of performance for a solution. The acquisition of expertise results in more and more actions' being conducted at the skill- and rule-based levels. As knowledge-based solutions to new problems are developed, tested, and proven, they become rules that can be employed when the same pattern of signals that identifies the problems is recognized in the future.
Errors occur at all three levels of performance. Of key importance to the application of this body of work to health care, research has demonstrated that the likelihood of error is far lower for skill- and rule-based actions than for knowledge-based ones. When one examines all errors, however, the vast majority occur at the skill- and rule-based levels, because so many more actions are conducted at these levels compared with knowledge-based actions.
Reason's definition of error, which was adopted by the Institute of Medicine for its study of errors in medicine, states that error is the failure of a planned action to achieve its objective or the use of a wrong plan to achieve an objective.11,13 Errors occur in three fundamentally different forms: slips, lapses, and mistakes. All of these kinds of errors have different psychological underpinnings and different implications for remediation.
Slips are errors of execution that occur at the skill-based level of performance. They are often easy to observe, because they are errors of commission. Pouring orange juice into a bowl of cereal, adding the same ingredient twice to a recipe, and trying to start a car with a house key are slips.
Lapses are usually failures of memory that also occur at the skill-based level. They are typically errors of omission and more difficult to detect than slips. Walking into a room and failing to recall what you wanted to do there, neglecting to make an extra stop at the market on the way back from work, and failing to alter the wake-up settings on an alarm clock to accommodate a change in routine are lapses. Skill-based activities are monitored from time to time by devoting some attention to checking whether these actions are proceeding as expected.
Slips and lapses occur because of failures of attention at critical times in particular sequences of actions. These failures are of two types: inattention and overattention. Inattention refers to the failure to check on a sequence of actions at the correct time, very often at the point at which one intends to deviate from a familiar routine. Overattention refers to the initiation of an attentional check at the wrong time, commonly before it is needed.
Mistakes occur at both rule- and knowledge-based levels of performance. They are uniformly characterized by the selection of a plan that is inadequate to achieve the intended objective. At the rule-based level, mistakes can occur either because a rule previously proved to work well is selected for the wrong situation or because a bad rule is selected. At the knowledge-based level, the complexity of causation increases dramatically and includes inadequate information about the problem, a variety of biases in interpreting information, and inadequate expertise to solve the problem. Mistakes are generally the most difficult of all errors to detect.
RELATING TYPES OF ERRORS TO QUALITY PROBLEMS
Integrating the body of research from cognitive psychology described above with what we know about health care quality problems yields insights that have important implications for the design and conduct of quality improvement activities. Each class of quality problems is associated with a unique blend of the three different kinds of errors.14 Overuse problems are triggered by mistakes (the selection of wrong plans). They are rarely, if ever, caused by slips or lapses. Sometimes these will be rule-based mistakes, but more often they will be knowledge-based. Prescribing an antibiotic for a viral upper respiratory infection could occur because an inexperienced physician fails to appreciate that a purulent-appearing nasal discharge is by itself insufficient evidence to support a diagnosis of sinusitis. This error is a rule-based mistake—a good rule (“Treat sinusitis with antibiotics”) is misapplied. More often, however, physicians who prescribe antibiotics knowingly for colds are making knowledge-based mistakes. They may, for example, discount the likelihood of an injury due to the antibiotic and be disinclined to expend the effort required to explain to the patient why an antibiotic is unnecessary. Underuse problems can be initiated by mistakes, but they can also be caused by lapses. Because underuse problems represent errors of omission, they are not caused by slips. An influenza immunization for an elderly patient may be omitted because, despite the physician's intention to order it, she is distracted by the complexities of dealing with the patient's hypertension and hyperlipidemia (a lapse). Alternatively, a physician may intentionally omit a routine measles immunization of a child who has a cold, because he believes that a mild viral infection is a contraindication to routine immunizations (a rule-based mistake).
By contrast, misuse problems are caused by all three kinds of errors. A physician may fail to cannulate a peripheral vein with an intravenous catheter because he exerts too much force and the catheter travels through the vein (a slip), leading to an increased risk of hematoma. A nurse may fail to deliver a medication at the time ordered because she is distracted by an urgent problem with another patient (a lapse). A physician may fail to adjust the dosage of an aminoglycoside antibiotic correctly for a patient with renal insufficiency because he does not know that such an adjustment is required (a knowledge-based mistake) or because he applies the wrong formula for making the adjustment (a rule-based mistake).
IMPLICATIONS FOR IMPROVEMENT
It is important to understand the error types that lie behind particular quality problems, because the individual error types call for different approaches to amelioration or prevention, and some are more easily remedied than others. We believe that some of the observed variability in the effectiveness of different intervention strategies may be due to a lack of appreciation of this set of relationships. Errors attributable to slips and lapses are most amenable to improvement because they represent failures of execution in which appropriate, intended plans are not successfully completed. When interventions focus on achieving the aims intended by physicians or other clinicians, they will be more readily accepted. Many efforts to reduce errors in medication ordering involve checking orders against acceptable dosages for specific drugs. This approach to improvement can detect many slips in which, for example, a physician writes an order for 10 mg when he really means 1 mg. Correcting this kind of error will meet with little resistance because there will be no disagreement that an error has occurred. Because slips can occur in a wide variety of different ways, however, ascertaining that they have occurred may be more challenging than correcting the ones that are discovered.
Identifying and correcting lapses will be similarly noncontroversial. Reminder systems, whether automated or manual, should be highly effective interventions for errors of this type. The design of such systems should be informed by the insights from cognitive psychology noted earlier. Lapses can occur either because of failures to check on the progress of specific processes or because of ill-timed checks (i.e., either too early or too late to avoid the error). Thus, the timing of reminders to physicians should be carefully considered so they take place at the point in the process of care most opportune for preventing the omission of needed tests or treatments.
When mistakes are the subject of improvement, the road may be bumpier. Identifying and fixing rule-based mistakes will require achieving agreement that a particular rule is faulty or misapplied. Thus, the physician who believes that a simple cold is a contraindication to administering routine immunizations to a child must be persuaded that this “rule” is faulty, because it is inconsistent with the best scientific evidence. We physicians employ dozens of rules in the daily practice of patient care. In this penchant, we are making the best use of one of the most human of intellectual faculties, the ability to recognize complex patterns and to activate appropriate response pathways. As noted above, however, the distinctive measure of expertise in human endeavor is the enlarged repertoire of appropriate and effective rules at the command of the expert compared with the novice. Viewed from this perspective, the whole movement to create evidence-based practice guidelines in medicine can be seen as an attempt to systematically improve the “rules” that govern our clinical practice by codifying and disseminating those used by experts. The use of an opinion-leader strategy would seem ideally suited for addressing this kind of mistake.
Rule-based mistakes that arise from the misapplication of “a good rule” require a different approach. The physician described above who treated simple rhinitis with antibiotics was employing a good rule (“Treat sinusitis with antibiotics”) in the wrong situation. The problem requiring correction here is not the rule but the recognition of the appropriate pattern of symptoms and signs that should trigger its application. An opinion-leader intervention that focused too narrowly on the delineation of appropriate antibiotic regimens for sinusitis would likely fail to correct this kind of error.
Knowledge-based mistakes will be the most difficult of all errors to approach. These errors necessarily involve more complex and subtle data. One may not even be able to achieve agreement that an error was made. In these situations, a more collaborative and deliberative approach to improvement is required, one that seeks a variety of different perspectives and is prepared to conclude that simple solutions may not be possible. An example of an intervention that might be suitable in situations like this is the multidisciplinary case conference. Such conferences are employed on a regular basis to evaluate patients' suitability for invasive cardiac procedures or to discuss the most effective treatment regimens for patients with breast cancer. Because different error types are likely to respond to different interventions, a careful analysis of the kinds of errors that underlie a particular quality problem in a particular setting will be important to the creation of the most effective intervention strategy. This step is particularly important when dealing with underuse and misuse problems, where the greatest variety of error types occur. For example, in approaching an underuse problem, such as the failure to prescribe beta blockers to survivors of acute myocardial infarctions (MIs), it is vital to understand the extents to which lapses, rule-based mistakes, and knowledge-based mistakes contribute to causing the problem. If lapses constitute a major cause, reminder systems will be most effective. Some physicians may believe that a history of cigarette smoking contraindicates the use of beta blockers (a rule-based mistake). Reminding these physicians to prescribe beta blockers for MI survivors will have little effect in correcting this error. An opinion-leader intervention might be more successful in persuading them that smoking does not contraindicate this treatment. Finally, some patients' histories may call for more individualized decisions because it is not clear how the evidence from the studies of the efficacy of beta blockers applies to them. For example, this therapy has not been studied in patients with acute MIs treated successfully with thrombolytic agents. Deciding on the appropriate use of beta blockers in patients like these will likely require a consensus process that brings together clinicians from all the relevant disciplines to apply their clinical judgment to extrapolate from the available evidence.
PHYSICIANS' RESPONSIBILITIES FOR IMPROVING QUALITY
So how should individual physicians go about engaging in quality improvement? Many physicians address the desire to provide high-quality care by making a personal commitment to excellence and striving to do their best for every individual patient during each telephone call, office visit, or hospital stay. This professionalism lies at the core of what it means to be a physician and is an indispensable motivation for quality improvement. It is a necessary component of an effective approach, but insufficient by itself. If the medical profession is to succeed in taking the leading role in improving quality of care, we suggest that the overall effort must be characterized by several guiding principles.
First, physicians must take ownership of all aspects of the problem, developing a full-spectrum quality improvement effort, one that addresses important overuse, underuse, and misuse issues. Such an effort will focus on clinical quality topics but should also include critical aspects of how patients experience their care. For example, how well informed patients are about their illnesses and treatment options can have important effects on their adherence to what are often complex and burdensome regimens of diagnosis and management. In some situations it may not be possible to improve the processes of clinical care unless improvements in communication with patients are undertaken as part of the intervention.
Second, measurement of important, precisely defined parameters of quality must be the hallmark of this effort—both to guide improvement interventions and to document impact to a variety of stakeholders.
Third, we must endeavor to improve both individual performances and systems of care. Successful system improvements will complement exceptional individual performances, but we should not as individuals abandon our personal commitment to excellence in the expectation that system improvements by themselves will solve all our problems. Finally, all physicians must get into the game. No matter how large or small the practice, no matter what the organizational make-up or the reimbursement mixture, no matter how technologically advanced or antiquated the information systems are, the science about what works and what does not to improve health outcomes for patients is the same. The ways in which physicians may prioritize and attack quality problems may differ depending upon the scopes and sizes of their organizations, but the need to initiate an organized effort is constant across all settings.
So practically speaking, how would it work? How should physicians adopt quality improvement as part of the routine of their professional lives? Consider first a primary care practice comprising four or five internists. The first step the group should take is to agree upon initial priorities for quality improvement. For example, the physicians might decide to focus on specific problems of overuse in acute care (using antibiotics to treat colds or uncomplicated bronchitis) or underuse of preventive care for elderly patients (such as the receipt of vaccines). The group should identify the relevant scientific literature and any applicable guideline statements establishing best practices, and then select one or a few specific measures that can be assessed and tracked in the outpatient setting. One easy and convenient place to start is the National Guidelines Clearinghouse's user-friendly Web site (www.guidelines.gov). A wide variety of evidence-based guidelines are readily accessible in this searchable database. Most specialty societies are represented, and many of the guideline documents contain succinct evidence summaries in addition to their clinical recommendations. For instance, if the group of internists wants to focus on immunizations, the specific measures selected might be: (1) the percentage of patients aged 65 or older who have ever received pneumococcal vaccine15; and (2) the percentage of patients aged 50 or older who received influenza vaccine for the most recent influenza season.16 To identify the population for study, if possible, the physicians should use the practice's information systems to generate a list of all patients aged 50 or older. Some sophisticated systems will even be able to produce (using billing records) lists of patients who received influenza vaccine during the most recent flu season. The physicians should identify a sizable but manageable number of patients' records to review. In this case, approximately 50 records of patients 65 years old or older (to assess receipt of both pneumococcal and influenza vaccines) and an additional 50 records of those aged 50–64 (for influenza only) should suffice. If the group has adequate computer software and prowess, it can automate selection of a random sample from among the entire group of patients. If the information systems are capable of identifying which patients were influenza vaccine recipients, then approximately equal samples of vaccine recipients and non-recipients should be randomly selected. If the available computer hardware and software are inadequate to select the sample, the names for review can literally be drawn out of a hat. Should the information systems be inadequate to produce any kind of automated patient list, the physicians can do the study on a prospective basis, including all patients aged 50 or older when they present to the office for appointments.
The physicians should create a simple paper check list, including spaces to record basic demographic information about each patient (age, sex, race, etc.), whether the patient ever received pneumococcal vaccine (for those patients aged 65 or older), whether the patient had received influenza vaccine during the most recent year (for those patients aged 50 or older), and whether there were any medical problems, social factors, or logistic issues (such as number or timing of visits for the year) that might have impacted the receipt of vaccines. The physicians or nurses in the practice should then review the medical records and complete the paper checklists for the sampled patients. Clerical personnel can enter the information into a simple computerized database or perform these simple tabulations by hand. The group can then determine how the practice is doing on the two selected measures and can look to see whether there are any identifiable factors that are associated with greater or lesser likelihood of vaccine delivery.
In conducting any quality improvement project, following completion of the initial data collection, the physicians should discuss whether the data point to any common types of errors (slips, lapses, rule-based or knowledge-based mistakes), thereby suggesting what types of interventions may be most likely to improve practice. In this case, the data might show that the influenza vaccination rate for the previous year was only 42%. Further review might demonstrate that 90% of patients who presented for office visits for any reason from September through January were vaccinated, but that only 22% of patients who did not visit the office during that period received the vaccine. In discussing these results, the physicians will recognize that the data suggest that lapses are the cause for the majority of the vaccination failures. When patients do visit the practice during the months leading up to flu season, they are vaccinated at high rates. Clearly the physicians in this practice are generally following the correct “rule” (give influenza vaccine to all patients over age 49 who do not have contraindications).16 However, the practice is not doing a satisfactory job of reaching out to patients who do not come in for another reason during the vaccination season and fail to remember on their own to make appointments for the shot. Therefore, the group might want to focus initial improvement efforts on working harder to remind patients to come in during that time period.
On the other hand, data from another practice with an overall influenza vaccination rate of 42% might reveal that most of the patients did make at least one office visit during the months of September through January. Further, the vaccination rate might be 90% among the patients aged 65 or older, but only 22% among the patients aged 50–64. In this case, lapses are not likely to be the cause of a substantial portion of the problem. The group might find upon further discussion that two of its physicians believe that the recently revised recommendation to vaccinate all patients aged 50 or older (rather than 65 or older) is overzealous, and that not all patients in this age group should be vaccinated. In this case, the measured underuse of vaccine would be largely attributable to rule-based mistakes, rather than lapses. The physicians are not forgetting to give the vaccine; they are intentionally withholding it. The group would need to carefully review the literature supporting the change in the recommendations and attempt to come to consensus on desired practice. After reviewing the literature, the group might agree to follow the recommendation to vaccinate all patients aged 50 or older, or it might decide to compromise and select out a subset of higher-risk patients (those with chronic illnesses that make them more likely to suffer complications with influenza infection) in the younger age group.
Projects such as these are not grandiose; the resources available for quality improvement within a small primary care group are limited. The existing clerical staff must devote their time to pulling records and entering data into a database, and the medical staff have to identify the projects, determine best practices, select the measures, develop the checklists, review the medical records, and analyze the results of the findings. Furthermore, plans must be made to remeasure performance periodically after interventions are implemented to ensure that improvements are maintained. Despite the relatively small scale, however, the projects are nonetheless meaningful. They will translate into better outcomes for patients because they are tied to interventions that have been proven to improve health.
A larger group, such as a 50-person multispecialty practice, can conduct the same kinds of programs, with the same steps in the process, but should be able to benefit from economies of scale. Whereas a small primary care practice will generally not be able to have any staff entirely dedicated to quality improvement, a large multispecialty group should be able to have one or two employees dedicated entirely to coordinating and running the quality improvement endeavors for the group. Additionally, large multispecialty group practices are more likely to have automated patient registration and other information systems at their disposal, and often employ dedicated information technology staff. Such groups should be able to do less of the project planning and implementation with paper, pen, and calculator and to conduct larger and more sophisticated quality improvement projects. Having practitioners with expertise in fields of general medicine as well as many subspecialties significantly broadens the clinical areas from which opportunities for improvement can be selected. Generalists and specialists in the group should meet to discuss and agree upon priorities for quality improvement, and should solicit input from non-physician clinical staff (nurses, technicians, physicians' aides).
Selecting specific projects to undertake also requires recognizing political reality; payment systems in health care do not currently recognize and reward quality. To the contrary, improving quality often results in decreased revenue. For instance, in a setting with predominantly fee-for-service payments, decreasing the number of inappropriate hysterectomies would provide a clear quality benefit for patients, but would decrease revenue for the physicians. It might be necessary to start with projects that are chosen primarily because—in addition to their potential to improve outcomes for patients—they are likely to be either revenue-neutral or revenue-increasing for the physicians. Therefore, a 50-person multispecialty group operating in a largely fee-for-service environment might choose to start out with a portfolio of projects that would include (1) increasing the proportion of acute MI survivors leaving the hospital on beta-blockers; (2) increasing the proportion of hypertensive patients in the practice with diastolic blood pressures maintained under 90 mm Hg, and (3) increasing the proportion of anticoagulated patients with international normalized ratios (INRs) in the therapeutic range. As in the case of the smaller primary care practice, conducting the quality measurement requires reviewing the literature to determine best practices, selecting specific quality measures, constructing data-collection tools, and collecting the data. Once the measurement data are available, the group should attempt to determine the types of errors (slips, lapses, rule-based mistakes, or knowledge-based mistakes) that are responsible for the greatest differences between desired and actual practice.
Consider, for example, a group that decides to look at their success in prescribing beta-blockers to patients following MI.17 They may find that many of the patients who are not prescribed beta-blockers are discharged from the hospital in the afternoon, when the hospital floors are the busiest. The physicians are likely to conclude that these errors are caused by lapses. Probably at least partially because of the busy time of day, the physicians may forget about the beta-blockers. Alternatively, the physicians or nurses may neglect to make sure that the prescriptions get handed over to the patients at a time when they can be safely packed away in a place where they will be found and filled once the patient gets home. Decreasing the frequency of such lapses would likely be best accomplished through the use of one or more reminder systems. The practice might create and use a checklist that includes a few crucial items to be considered when discharging post-MI patients. Prescribing beta-blockers (in the absence of absolute contraindications) would be on that list. This practice may also want to consider implementing a routine telephone call to post-MI patients on the day following discharge to make sure that the patient has made it home with the appropriate prescriptions.
Another practice conducting the same quality improvement project might discover that two physicians (out of a total of 12 internists and cardiologists) have cared for the majority of the post-MI patients discharged without beta-blockers. In discussing the reasons for the differences in practice among the various members of the group, the physicians might discover that these two doctors are mistaken about the contraindications to prescribing the medication. One of them may believe that non—insulin-dependent diabetes is a contraindication, and the other may believe that mild chronic obstructive pulmonary disease is a contraindication. The errors in prescribing made by these physicians are not lapses. Rather, they are rule-based mistakes. Therefore, reminding these physicians to prescribe beta-blockers without exploring the reasons for their misperceptions about the contraindications would be very unlikely to succeed in changing their behaviors. To improve the practice, the group would discuss the literature on contraindications (including recent recommendations suggesting that there are few, if any17) and work toward a consensus on desired prescribing patterns.
In organizations larger and more complex than a multispecialty group practice, the economies of scale may continue to provide additional benefits for quality improvement, but the process of launching a program becomes more difficult. Consider, for instance, a 450-bed community hospital. Unlike both small and large group practices, the physicians who provide care to hospital patients do not control the governance of the organization. The hospital may have resources already dedicated to some quality-related activities, but the physicians in separate clinical settings may not. The numbers of clinical specialties and treatment arenas available for improvements are great, but the organizational structure is likely to pose significant obstacles to success. The various clinical departments are led by different individuals. Non-physician caregivers (nurses, physician assistants, nursing aides, technicians, and social workers) and other support staff (patient registrars, billing staff, etc.) are commonly under the auspices of entirely different organizational structures from those of the physicians. The medical staff may have limited input into the administrative workings of the areas in which they deliver care. Services such as labs and pharmacy are also under separate leadership. The presence of so many potential stakeholders is, on the one hand, exciting because of the extensive possibilities it creates, but the reality is that there are generally as many competing agendas as there are constituencies, and the complex relationships among the various heads of departments and services make setting common priorities a challenge at best. Getting the leaders of these separate groups to agree in principle that quality improvement is a top priority is not difficult. Turning that agreement into a set of specific goals and objectives and a willingness to commit staff time and other resources to a systematic effort to measure and improve care on a series of quality measures is much harder.
Physicians will need to take it upon themselves, therefore, to collaborate with colleagues to be the driving force for quality improvement programs. They will need to set the goals, secure the funds, and oversee the work. To set priorities, an internist at a community hospital could work with the chief of medicine to bring together general internists and specialists to discuss quality problems affecting patients at the hospital and agree upon a list of high-priority improvement projects. The group might decide, for instance, that it wants to begin by increasing the percentage of heparinized inpatients reaching therapeutic activated partial thromboplastin time (aPTT) levels within 24 hours. A second priority might be to decrease overuse of cephalosporin antibiotics in inpatients to reduce the prevalence of resistant organisms.18 Another might be to ensure that all patients having breast-conserving surgery for early-stage breast cancer receive a course of radiation therapy thereafter. Conducting the actual projects will involve processes very similar to those described in the group practice example earlier (i.e., confirm best practices, select specific quality measures, construct data-collection tools, collect data, compare results to desired performance, evaluate reasons for deviation from desired performance, plan and implement interventions, remeasure). The key to transforming successful measurement into sustainable improvement lies, again, in determining the types of errors responsible for deviations from desired practice, the goal being to attack the problems with the solutions most likely to succeed. As we have suggested, automated systems or double checks are likely to work well to remedy slips. Reminder systems seem particularly well suited to lapses. Decreasing mistakes is a more complicated endeavor. Rule-based mistakes may be best addressed in occasional meetings designed to reach consensus on best practices— what the appropriate rule is for any particular issue, and who the patients to whom the rule should be applied are.
In some instances, application of rules may not be possible, however. Individual patients may be dissimilar enough that there are no readily accessible rules to apply (requiring knowledge-based thinking and leading to knowledge-based mistakes). Consider, for instance, further treatment for early-stage breast cancer after breast-conserving surgery in an 85-year-old woman with chronic obstructive pulmonary disease who had a very small, low-grade lesion. A meta-analysis of many large randomized controlled trials suggests that all patients should receive radiation therapy following breast-conserving surgery.19 But the trials included in this meta-analysis did not enroll many patients like this one. If a quality improvement project looking at this issue finds that it is questionable whether the “rule” (follow breast-conserving surgery with a course of radiation therapy) applies to a substantial percentage of the women who did not receive treatment, then the best solution may be to institute an ongoing series of case conferences where such patients can be discussed by all of the practitioners with relevant expertise before the individual treatment decisions are made. If such conferences become a routine feature of breast cancer care at the hospital, then all of the physicians can feel comfortable knowing that the most difficult cases have been considered and discussed by the clinicians best equipped to help make decisions about their care.
In the case of the community hospital, after deciding upon the proposed projects, the group leading the quality effort will need to create a strategy for convincing hospital administration to fund the quality measurement and improvement. The most likely route to success will be to demonstrate how the proposed quality improvement projects will improve the financial situation of the hospital. Hospital administrators are likely to be receptive to proposals that will increase the likelihood of good outcomes for patients and simultaneously improve the hospital's financial standing; the overall quality improvement program may well need to be initiated with a series of such projects. The group should create a business plan that details the necessary hospital investment and the expected return on investment associated with the proposed endeavors. In most cases, some of the projects on the priority list will naturally be good candidates for revenue enhancement, and others will not. In the examples given above, the cephalosporin project may decrease antibiotic costs (if less expensive drugs are substituted for the third-generation cephalosporins) as well as patient care costs (by reducing nosocomial infections caused by resistant organisms). The breast cancer project, however, will not enhance revenue unless the hospital provides radiation services, and it is unclear what impact the heparin project would have. The physicians would therefore be wise to begin by seeking to convince hospital administration to fund the effort to decrease overuse of cephalosporins. While conducting this project (and others like it) with financial support from the hospital, the physicians should be able to acquire equipment (such as computers) that they can then use to support the high-priority work that is unlikely to result in financial gain. Additionally, the positive results of the initial series of projects might eventually convince administrators to support revenue-neutral or revenue-negative work to a limited degree.
PROSPECTS AND PITFALLS
We have described a conceptual framework that delineates the relationships among the three different kinds of quality problems and the errors that trigger them. We suggest that interventions to improve quality are not likely to succeed unless they are carefully targeted at the particular kinds of errors underlying the quality problems one wishes to ameliorate. Further, we have outlined how physicians in three very different practice settings can take reasonable and practical steps now to establish effective quality improvement programs.
Power relationships in health care are currently in flux. Managed care companies and large employers are retreating from the aggressive postures they assumed as recently as a few years ago. Government is unlikely to employ regulation in an attempt to increase its control over the health care delivery system. Yet, consumers are increasingly concerned about quality, and the costs of health care are again rising faster than costs in other sectors of the economy.
We suggest that now is an opportune time for physicians and the organizations they direct or guide to take the leadership role in health care quality improvement. In doing so, physicians can regain much of the autonomy over the practice of medicine previously lost to government and to managed care. But capitalizing on this opportunity effectively will require changes in many traditional approaches to quality. Physicians in all practice settings will need to conduct quality improvement as an integral part of patient care, rather than seeing it as a burdensome, externally imposed task that is irrelevant to the core goals of their practices. By focusing improvement activities on carefully selected aspects of patient care, physicians will be able to achieve results rapidly—results that will convey substantial health benefits to patients. Major reductions in the use of antibiotics for colds, increases in the use of inhaled steroids in patients with asthma, reductions in prescribing errors, and a host of other clinically meaningful objectives are within the reach of most, if not all, physicians' practices.
Physician leadership in quality will be defined also by accountability. Physicians should seize the initiative for leading in the effort to remedy all three kinds of quality problems. It is not enough for physicians to be vocal about the need to provide effective care to all who could benefit (i.e., solve underuse problems). If physicians aspire to regain legitimacy as responsible stewards for the vast resources our society expends on health care, we must surely also undertake to lead the effort to eliminate the harmful and wasteful effects of overuse and misuse.9,20,21 Leading this movement does not mean that physicians must bear the burden of solving all of the problems alone. But it does mean speaking clearly, loudly, and often about the need to deal with all three kinds of problems, and it means devoting significant time, effort, and resources to developing the strategies, tools, and methods to improve. This is both an individual and a collective responsibility. Physicians must set examples by working to improve the care they deliver personally, the care they deliver with the colleagues with whom they practice, and the systems that surround them. Accurate predictions of future power shifts in health care have been rare in recent years. It is possible that a consumer-led quality revolution in health care, mediated in part by widespread access to unprecedented information about hospitals and doctors on the Internet, will fizzle.22,23,24,25 It is possible that employer-led efforts to provide consumers with information about quality to drive a market-based competition on quality will be ineffective.7 It is possible that recent initiatives to pay bonuses to physicians who meet standards of quality set by managed care companies will fail to spread widely.26 So, if we physicians do little or nothing to define and improve quality on our own terms, we may be no worse off several years hence that if we had made the commitments outlined here. On the other hand, what are we waiting for? The evidence of the magnitude of our problems is overwhelming. Do we wish to solve these problems by defining for ourselves and our patients what excellence in quality is and how it may be achieved consistently, or do we wish to cede that responsibility to others? The choice may be ours to make for now but not, perhaps, for much longer.