Better Care and Better Outcomes: Title and subTitle BreakThe Continuing Challenge

Understanding of the links between medical knowledge and clinical practice has evolved considerably over the last 4 decades. Notwithstanding the accelerated pace of medical progress after World War II and early evidence suggesting opportunities to improve the quality of clinical care, few researchers focused on knowledge-behavior relationships during the 1950s and 1960s. It was tacitly assumed that medical advances diffused appropriately into practice after publication in the clinical literature. Perhaps naively, many observers implicitly accepted the concept that physicians could maintain high-quality care by browsing a few journals, attending occasional continuing education events, interacting with their peers, and filtering promotional information from drug and device manufacturers.

By the 1980s, researchers had repeatedly shown that the traditional approach sometimes galvanized dramatic shifts in clinical behavior, but was seldom sufficient to cause consistent and optimal transfer of evidence into practice. Active dissemination of high-quality clinical guidelines was held up initially as a potential solution to this emerging "quality problem." Guideline writers would compile the totality of relevant evidence on a particular condition or technology for practitioners unable to assimilate disparate and conflicting information. Moreover, guideline writers could draw on expert judgment and weigh values and circumstances, thereby helping clinicians navigate the many gray zones of practice¹ in which evidence alone was insufficient (or so the argument went). However, as Lomas² noted, faith in active guideline dissemination presupposed that exposure to better information alone leads to behavior change, and ignored other factors in the practice environment that might affect the clinician's willingness and ability to change.

More recently, there have been 3 complementary spheres of activity aimed at simultaneously changing clinical behaviors and improving quality of care. One set of activities promotes explicit reports on processes and outcomes of care for the general public, referring physicians, and payers. Proponents of these "report cards" argue that external market forces and public pressure are potent catalysts for local quality improvement. The second set of activities, spearheaded most prominently by Berwick,³ has brought the industrial model for continuous quality improvement to health care. Internally driven and based on organizational transformation, this approach demands a fundamental shift in culture, with action by a range of health care professionals, administrators, and support staff. The third wave of activity includes a variety of local and regional initiatives aimed at condition-specific or procedure-specific quality improvement. Leaders of these programs often espouse continuous quality improvement principles, but the implementation tends to be piecemeal and pragmatic.

Along the way, controlled studies^{4 - 5} have shown that knowledge is most likely to change practice if physicians receive salient evidence or information from credible sources and if measures are taken locally to facilitate change. Single-episode educational sessions seldom succeed. Information transfer strategies that do appear effective are outreach visits (academic detailing) and involvement of opinion leaders, educationally influential physicians nominated by local peers as reliable sources for new knowledge. Even intermittent audit and feedback about actual practices has limited impact without mechanisms to ensure clinician buy-in, address local barriers to change, or rectify relevant gaps in clinical knowledge.

This issue of JAMA contains 2 important studies that add to our understanding of how noneconomic interventions can change clinical behaviors and improve clinical care. Both deal with a common and frequently lethal condition: acute myocardial infarction (AMI).

The study by Marciniak and colleagues⁶ is unprecedented in scope and scale, a massive quasi-experimental assessment of the effect of feedback by peer review organizations (PROs) on AMI care in 4 states. The PROs abstracted clinical records of Medicare patients hospitalized for AMI during several months in 1992 and 1993, and PRO physicians provided feedback data on quality indicators to 379 participating hospitals during 1994.

Comparing AMI care in 1992 with results from another chart audit in 1995, the authors found highly significant improvements across a wide range of acute phase and predischarge quality indicators. Because inferences about causation are often invalid when based on before-and-after studies, Marciniak et al⁶ also compared postfeedback patterns of care in the 4 states with a random concurrent sample of Medicare beneficiaries hospitalized for AMI in other states. The results generally favor the 4 target states and are most convincingly significant for prescription of aspirin and β-blockers at discharge.

This quasi-experimental comparative design would have been stronger if the research team had abstracted a baseline sample of records for control states. Nonetheless, the findings strongly suggest that the PRO intervention accelerated positive trends in AMI care. This is even more striking because 29% of the PRO feedback sessions were by telephone or mail.

At first blush, these findings belie the general lessons of controlled studies on clinical behavior change, ie, that neither a single educational session nor a single feedback of data on practice patterns is effective. However, there is a catch. The PROs were known to be acting under contract with the Health Care Financing Administration (HCFA); hospitals were advised of suboptimal results and 73% of the hospitals prepared a concrete response. Hospitals and physicians may have implemented quality improvement plans more assiduously because of HFCA's role and authority in the health care system.

Also in this issue of JAMA, the trial reported by Soumerai and colleagues⁷ provides an excellent test of the effect of local opinion leaders as change agents. The study team randomized 37 community hospitals in Minnesota. They mailed to the medical director in control-group hospitals a booklet showing site-specific data for use of various drugs and procedures as compared with anonymous peer profiles. In intervention hospitals, the authors surveyed clinicians to identify local opinion leaders in AMI care. These influential clinicians met with recognized experts to review relevant evidence, examine their hospitals' baseline performance data, and discuss barriers to change. Thereafter, the opinion leaders worked on their own initiative to improve local care, drawing communications materials and logistical support as needed from the study team.

The trial's scope was somewhat narrow, focusing only on drug therapy within 24 hours of a patient's first contact with a study hospital. Unfortunately, the number of eligible elderly patients was too small in the control arm for a meaningful assessment of the comparative change in thrombolysis use. However, in the intervention group, the overall baseline use rate was already 73%, similar to the 68% use rate among ideal thrombolysis and primary angioplasty candidates reported in the study by Marciniak et al.⁶ Both teams reasonably question how much higher these rates can be pushed. Soumerai et al⁷ also aimed to reduce unnecessary use of lidocaine prophylaxis among patients of all ages. Intriguingly, the median absolute decline in inappropriate lidocaine prophylaxis was similar for both intervention and control-group hospitals. On the other hand, for 2 other primary outcomes, use of aspirin and use of β-blockers among eligible elderly patients, intervention hospitals showed significant improvements compared with control-group hospitals.

The findings of Soumerai et al,⁷ together with an earlier randomized trial of opinion leaders in obstetrical care,⁸ illustrate the potential impact of cultivating and supporting respected clinicians who can act as "quality champions." These educationally influential clinicians may not be physicians in managerial or formal leadership positions. Accordingly, administrators must try to understand the sociology of local practice before they launch quality improvement initiatives.

Several other lessons and directions for research can be inferred from these 2 important studies. The first, drawn from the control groups, is that contemporary practice changes continuously in positive directions, presumably as a function of ongoing diffusion of evidence from the literature of medicine, dissemination of guidelines and care maps, promotion of drugs and devices by industry, and countless ad hoc local initiatives to improve care. Therefore, readers should take a skeptical view of any uncontrolled study claiming that a particular intervention improves processes or outcomes of care.

Second, these studies highlight the need for better health information systems. Both research teams relied on retrospective chart abstraction methods, with all of their attendant costs and potential inaccuracies. Whereas Marciniak et al⁶ were able to abstract most of the designated hospital charts, Gurwitz et al have reported elsewhere that during the baseline period, 11% of patients with a potential AMI were excluded from the study because of inability to access the medical records.⁹ Electronic patient records and registries of important clinical variables would help facilitate quality improvement activities.

Third, neither study addressed some of the unanticipated effects of the interventions. On the positive side, it is not known whether these interventions also improved the use of potentially lifesaving drugs for nonelderly patients. On the negative side, neither study addressed whether there was any increase in the inappropriate use of thrombolytic drugs for individuals without confirmed AMI.

Fourth, neither study addressed costs in relation to outcomes, but this issue must be squarely faced in future research on quality-enhancing strategies. Large investments to improve processes of care may not always lead to commensurate improvements in clinical outcomes. Clinical trials show that many modern treatments must be administered to scores of patients to avert 1 adverse event or prolong 1 life. Unlike a trial in which outcomes for nearly complete adherence to a given therapy are measured against placebo controls, quality-enhancing strategies start from a baseline of partial adherence and may lead to marginal changes.¹⁰ For example, based on placebo-controlled trials of long-term β-blockade after AMI, 100 medium-risk patients must be treated for a year to prevent 1 death.¹¹ Marciniak et al⁶ found that, compared with control states, intervention states used 8% more β-blockers at discharge for ideal patients. In effect, this means involving 1200 patients in a quality-improvement initiative to save 1 life.

Interventions targeting several areas of suboptimal treatment simultaneously will multiply these effects, but the net impact may remain small. Soumerai et al⁷ did not attempt to measure changes in clinical outcomes. Because of limited statistical power, Marciniak et al⁶ showed no significant mortality advantage for intervention states compared with control states using abstracted clinical records. Using 1995 and 1996 claims data, they found significant survival advantages that ranged up to 0.9% at the l-year follow-up interval. However, the postintervention survival advantages were not deflated for preintervention differences in mortality among states, and the comparison is potentially confounded by many other factors. Postintervention l-year mortality among Medicare patients in 1995 remained around 30%, highlighting that for AMI, major improvements in outcomes will be possible only with new and better therapies. Those caveats aside, the effects of the PRO program may be underestimated because HCFA already had launched a similar PRO intervention in control states during 1995. Above all, even small improvements in AMI survival translate into hundreds, if not thousands, of saved lives.

In sum, these 2 studies on AMI care have helped move the goalposts in our understanding of how to make good practices even better. We still know too little about which practice improvement strategies are most effective and efficient in particular contexts or clinical conditions. There is much to learn about why physicians readily adopt particular practices but reject others, and how the life cycle of useless practices can be shortened. But what is already very clear is that the quality of medical care can be measurably improved by knowledge-based interventions. A firm foundation now exists for an active agenda of research and implementation that speaks to physicians' professionalism and their patients' needs, and does not answer first to economic imperatives.