by Gregory D. Stevens, PhD, MHS

A common refrain among clinicians and researchers in evaluating health care is the need for “objective” over “subjective” data.  A major presumption is that objective data (usually things like lab results, clinical records, or billing data) tend to be free of human bias and that “subjective” data (e.g., patient satisfaction) are filled with it. The debate over what data will be used to evaluate performance is heightened by the fact that new payment mechanisms that reward “value” in care are arriving.

In some ways, this simple divide is useful because it makes us consider how human and other potential biases in data have major consequences for our research.  But the line between objective and subjective may not be nearly as bright as we might think, and prioritizing the “objective” data may blind us to the value in what people experience and what they say. If health care is designed to benefit patients, then it makes little sense to exclude the patient perspective in evaluating that care.

A related topic was recently raised in a pair of thoughtful perspectives on person-centered care published in Medical Care. The pieces reflected on a paradigm shift that is underway toward care that is more embracing of patient preferences.  Lines, Lepore, and Wiener highlighted, for example, how this abandons the old “medical model” where patients would simply comply with the decisions of providers.  And moving toward measuring the patient perspective is certainly a corollary to this.

This issue is particularly relevant in primary care research.

Compared to other settings of care, primary care tends to involve fewer procedures that easily lend themselves to fairly objective bean counting.  Here, much of the care is composed of advice and support to help patients stay healthy, change behaviors, cope with anxiety, and navigate complicated care routines for chronic problems.  In such instances, there are nuances to care that are not in any medical records, and it is likely that patients are best situated to tell us how these services meet their needs.

Moreover, billing data (viewed as objective since we arguably closely track what we pay for) has been mostly replaced in primary care by managed care encounter data. Such data reporting has been partly uncoupled from payments (due to capitation and global payments used in primary care, particularly in places like California that have had very high managed care penetration [Berenson and Rich, 2010; Rosenthal et al., 2001]) and sometimes dismissed in importance by physicians.  As such, encounter data that might have been considered a close second to billing data is sometimes viewed as incomplete or inaccurate in primary care.

Patient-reported care and measures of patient satisfaction are certainly complicated by many factors.  These include differing expectations for care and differing abilities to understand and recall information.  But even medical records and encounter data are filled with assessments that are subject to differences in provider expectations.  

For example, in my own recent work to gauge developmental screening behaviors of pediatricians, I gave 20 providers an example case of a 9-month child who is not yet sitting or crawling (a typical red flag for developmental delay) and asked physicians about their next steps.  I expected to hear various assessments and referrals, but 4 of the physicians assessed this as completely normal.  The variations in expectations for developmental progress undoubtedly lead to very different medical records.

Good researchers know that there are limitations to every analytic approach, and this is no less true in choices of measurement.  But one step to improving primary care research is to challenge the automatic assumption that what patients tell us is of less value (and more highly biased) than what is recorded in other sources.