Measuring Hospital Quality: Title and subTitle BreakWhat Physicians Do? How Patients Fare? Or Both?

Just 2 decades ago, most US citizens presumed they received high-quality care, and neither patients nor physicians were much concerned with quality measurement. The landscape has changed. Lapses in quality of care provided by the US health care system—the most expensive health care system in the world—are now widely recognized.^{1 - 2} The Institute of Medicine's report of 2001 proclaimed a chasm between how the US health care system currently performs and its ideal¹ ; since then, more recent data show persistent deficiencies in care.² There is little doubt of the ample opportunities to improve the health care Americans receive.

With ongoing concerns about deficiencies in care, the need for valid and broadscale measures of quality is clear. However, despite much work, there is no consensus on which aspects of quality to measure: should the focus be “processes” of care, such as whether physicians or hospitals consistently provide known efficacious therapies, or should the focus be clinical “outcomes” such as inpatient mortality rates? Each approach has its appeal and drawbacks.

Measuring process of care is appealing because it can be based on strong scientific evidence, it can be relatively easily assessed and compared, and the processes considered are arguably within clinicians' direct control. If a patient with acute myocardial infarction (AMI) is admitted to a hospital service, it is relatively straightforward to determine whether the admitting physician prescribes aspirin. Generally, patient mix (that is, whether one physician's patients are sicker than those of others) is a minor factor; with few exceptions, all patients with AMI should be prescribed aspirin. In addition, contraindications are incorporated into the definition of the eligible population.

Although appealing, process measures have limitations: only a fraction of care, even for well-studied conditions such as AMI, is evidence-based, and clinical care often relies on nuanced physician judgments that do not lend to the easy development of measures. Many clinical decisions that all would consider “good care” are therefore not included in process measures. Skeptics worry that by focusing on some treatments but not others, clinicians will “practice to the test”³ ; ie, some physicians will provide aspirin but neglect other aspects of care, ultimately producing worse care. Furthermore, process measures are less transparent to patients, and publicizing these measures is less likely to engage patients in quality improvement.

Outcome measures are also appealing. Patients clearly value outcomes, and the goal of clinical care is to improve outcomes. Hence, measuring mortality rates after AMI seems not only reasonable but perhaps more consistent with what patients, policymakers, and payers are seeking. However, outcomes have at least 2 important limitations. They are heavily confounded by patient mix (if one physician's patients are sicker in a way that is not included in a risk adjustment instrument, those patients may have an increased mortality rate that is not a reflection of the care that physician provides). There are also issues of statistical power: even poor practitioners may have mortality rates that are not statistically worse than excellent practitioners when sample sizes are small.

The study by Bradley et al⁴ in this issue of JAMA and a recent study by Peterson et al,⁵ also published in JAMA, shed light on this debate. Both studies examined how processes of care relate to outcomes for patients hospitalized for an AMI. Using different data sets, the 2 studies found that hospitals that perform well on process measures for AMI have lower AMI mortality rates, although the strength of this association differed. Peterson et al found a robust relationship, whereas Bradley et al found that only 6% of the variation in mortality was explained by variations in the process measures they examined. Neither study suggests that the examined measures are unimportant, and in fact, randomized trials have shown these treatments to improve patient outcomes.^{6 - 7} The findings of Peterson et al imply that the process measures are a good proxy for other clinical activities that also likely improve patient outcomes. The findings of Bradley et al suggest otherwise and imply that additional measures need to be developed. Differences in the findings may be due to alternative methodological approaches to handling transferred patients.

Whatever the reasons for the differences, the takeaway message from these studies is that both types of quality measures— processes and outcomes—are important. While process measures examine how effectively hospitals provide specific types of evidence-based care, outcome measures help keep organizations focused on the real goal: maximizing patient health. This is part of the reason the Medicare Payment Advisory Committee has recommended incorporation of 30-day mortality into the battery of measures Medicare uses for assessing hospital quality.⁸ Some states, such as California, have been posting hospital-specific, 30-day mortality for common medical conditions for many years.⁹

Another message from these studies is the need for improvement in measuring and enhancing the quality of care Americans receive. Although the current process measures are a good start, they are too few in number and only capture a small subset of clinical care. Furthermore, prior work has demonstrated that high performance in one area of care does not necessarily guarantee high performance in other areas.¹⁰ Therefore, a greater breadth and depth of process measures are needed and should capture patients admitted for other clinical conditions, such as gastrointestinal diseases, renal disease, and many other conditions. Processes should be examined across the spectrum of care from admission to discharge, and robust data should be collected on outcomes. Ongoing improvements in risk adjustment will help level the playing field and avoid penalizing physicians and hospitals who care for sicker patients. Although data on mortality are helpful in certain conditions, other outcome measures, such as change in functional status or quality of life, would be helpful.

In addition, measures for other aspects of hospital care, such as patient experience, efficiency, and equity are needed. Data on patients' satisfaction with care, such as whether clinicians communicate effectively and respond to patients' needs, are forthcoming. The development of the Hospital Consumer Assessment of Health Providers and Systems survey, which is being field tested and will become publicly available in 2007, should allow consumers, payers, and policymakers to identify hospitals that provide patient-centered care.¹¹ Good measures of efficiency, the ability of providers to deliver high-quality care while using fewer resources, are in early development. Unfortunately, little attention is paid to developing robust measures of equity, the notion that variations in care be clinically appropriate. Given the tremendous inappropriate variation in care that Americans receive,¹² ensuring high-quality care for all Americans, including racial and ethnic minorities, should be prominent in the quality agenda. Of course, even with comprehensive quality measurement, making care better has its own challenges. Well-performed interventions to improve quality often fail to demonstrate a sizable impact^{13 - 14} and suggest that a sustained, multifaceted approach may be needed to make care better.

Of note, the study by Bradley et al used the National Registry of Myocardial Infarction (NRMI) database to examine care for AMI, which is an important reminder that much of the progress in quality measurement and improvement has been made in just a few clinical areas. The NRMI is one of the few examples of successful efforts in the private sector to collect data on the quality of care. Funded by a pharmaceutical company, the NRMI has gathered information about treatment of MI for more than 15 years and from more than 1600 hospitals.¹⁵ The data are timely and relevant and provide key insights about the care received by tens of thousands of Americans who have an MI each year. The existence of databases like the NRMI and the leading efforts of the American Heart Association, the American College of Cardiology, and others in supporting studies and endorsing quality improvement efforts have lead to greater adherence to guidelines in care of patients with AMI.^{16 - 17} These improvements have likely contributed to the dramatic reduction in mortality from coronary heart disease.^{18 - 19} Although nascent efforts are under way in oncology, renal disease, and other disciplines, they pale in comparison to the strides made in improving cardiac care. Many more thousands of lives might well be saved if other specialty societies and private companies follow these models.

Although the US health care system is now committed to quality measurement and the public reporting of such data, debates will continue about what to measure, who collects the data, and what to report publicly. More information is needed on processes and outcomes across a large number of conditions for hospitals, physician practices, and other health care settings and practitioners. Much of these data are on their way, led by major payers such as Medicare and coalitions of employers²⁰ who want greater accountability for the care they purchase and to stimulate improvements in quality of care. In the most expensive health care system in the world, patients and physicians should expect nothing less.