Patient-Centered Outcome Assessment May Lead to Different Conclusions and Different Treatment Decisions
Published Online: June 20, 2014
Robert M. Kaplan, PhD, chief science officer, Agency for Healthcare Research and Quality
Discussions of how to measure health are often characterized by 2 themes: first, illness and premature death are undesirable, so at least 1 component of health is avoidance of serious illness and mortality; and second, the effects of illness and disability on everyday functioning and quality of life are important considerations as well.1 Disease and disability would typically disrupt the usual activities of daily living. For example, cancer or heart disease may shorten life expectancy and can reduce a person’s capacity to engage in meaningful life activities in the years prior to death. Even relatively minor illnesses can have disruptive effects. A common cold, for example, can interfere with social and work activities, but symptoms usually improve in a relatively short time. However, chronic illnesses, such as recurrent low back pain, can result in permanent disruption of enjoyable life activities.2 A comprehensive conceptualization of wellness, therefore, must consider risk of death, reduced quality of life, and duration of health states.1,3,4
Over the last few decades, the Agency for Healthcare Research and Quality (AHRQ), clinicians, policy makers, and patient advocates have demonstrated a growing interest in measuring patient-reported outcomes. Most illnesses are now evaluated in terms of their effects on usual life activities. For example, modern medicine applies common laboratory tests, such as the blood chemistry panel, to assess wellness. Although these tests are clinically instructive, they are not always correlated with outcomes from the perspective of the patient, nor are they direct measures of the disease process. We often refer to the tests as “surrogate” markers because they serve as proxy measures of the disease process. Sometimes the surrogate markers do not correlate with either life expectancy or outcomes when viewed from the patient’s perspective.5
Many clinical studies now apply standardized measures that address a patient’s quality of life. Figure 1 summarizes the number of scientific papers identified in PubMed between 1972 and 2012 that included the “quality of life” key-word phrase. In 1972, PubMed did not identify any papers with this phrase; but by 2012, more than 11,000 articles containing the phrase were identified. This trend has continued in recent years, with a 78% increase in the 5 years between 2007 and 2012. During the last 4 decades, many new quality-of-life tools have become available. These tools allow for a more sophisticated analysis of patient-reported outcomes specific to a variety of illnesses such as cancer,6 diabetes,7 and heart disease.8
Identifying quality of life as an important outcome was perhaps first highlighted in the 1988 Shattuck Lecture by Paul M. Ellwood Jr, MD.9 Dr Ellwood advocated for what he referred to as “a technology of patient experience.” In contrast to managing symptoms, Dr Ellwood emphasized the importance of managing patient outcomes. He saw medical care that factored in quality of life as relying on 4 techniques:
1) Developing standards and guidelines that match treatments with patient desires
2) Measuring patient well-being and functioning
3) Using normative data to interpret patient outcomes within the context of other people
4) Disseminating information in ways that could affect decision makers
This approach puts the patient at the center of healthcare and uses patient-centered reports to offer guidance and perspective for clinical care. More recently, use of these methods has come to be known as patient-centered outcomes research (PCOR). The use of PCOR is well represented in the Affordable Care Act and in the Patient-Centered Outcomes Research Institute (PCORI). At the center of PCOR is the measurement of outcomes from the patient perspective.
Although PCOR may seem uncontroversial, the PCOR perspective often leads to different conclusions than more traditional biomedical research. PCOR gives preference to outcomes with reference to only 2 central measures: length of life and quality of life.10,11 The central premise of PCOR is that the goal of medicine and public health is to lengthen human life and/or improve its quality during the years that people survive. This perspective argues that physiological measures are important only if they relate to life duration or life quality. Blood pressure, for example, is a meaningful biological measure because it is highly predictive of early death or disability associated with myocardial infarction (MI) or stroke. Other measures less clearly relate to the twin objectives of improved life quality or lengthened life expectancy. Catecholamine variations in response to acute stress, for example, are less clearly related to the objectives and outcomes researchers focus on.
Another perspective arising from PCOR is the focus on all-cause mortality as opposed to disease-specific mortality.12 A variety of large clinical trials have demonstrated reductions in 1 cause of death but have shown compensatory increases in other causes of death.13 Trials on screening mammography, for example, frequently show that breast cancer screening leads to a reduction in breast cancer mortality. Yet the same trials often fail to show that breast cancer screening increases overall life expectancy.14 Although breast cancer deaths might be reduced, other causes of death are increased.15
Another example of this conundrum is illustrated by the Physicians’ Health Study. In this landmark study, approximately 22,000 physicians were randomly assigned to take either 325 mg of aspirin every other day or placebo. When the data were first analyzed, significantly fewer physicians in the low-dose aspirin group had died of MI than those in the placebo group. However, considering all causes of cardiovascular death, the number of physicians who had died was exactly the same in the aspirin and placebo groups (Figure 2). All of the deaths occurred during the study period and all were considered premature deaths. Aspirin may have changed what was recorded on the death certificate, but it did not extend participants’ life expectancy.16 Considering the specific cause of death (MI) would lead to the conclusion that using aspirin as a primary preventive was highly effective. From the outcomes perspective, though, aspirin had no effect. From the perspective of the patient, we would argue that people and families are more concerned about the person’s vital status and less concerned about a specific cause of death.
PDF is available on the last page.