Theory and Measurement in the Study of Medical Practice Variation

Abstract

<p>Variations in the rate of use of health care services (a.k.a. medical practice variations) have been described in the literature for over eighty years. The literature suggests three general sources of variation: patients, physicians, and environment. The relative influence of these sources and the specific mechanisms that produce observed variations are not well understood. This thesis presents four studies that identify and examine methodological issues that preclude our ability to understand the variation phenomenon.</p> <p>It is commonly believed that the physician is in part responsible for observed variation in health care services use. However, determining the influence of the physician in this regard is problematic, as it is difficult to isolate the effect of the physician from that of the patient and environment (including available resources). The first study presented in this thesis suggests there is meaningful variation in treatment recommendations between physicians working in a common environment, even after controlling for important patient clinical characteristics. Next, I present an experiment that suggests that factors related to the patient’s unique social context might influence how the physician intends to manage a patient’s care. As variations studies do not measure or adjust for social context, this might indicate an important methodological limitation of those studies if indeed context is an important (and justifiable) determinant of what care the patient will receive.</p> <p>Not all variation is necessarily bad. The third study I present explores how previous researchers discriminate between warranted and unwarranted variation. This study indicates that few researchers explicitly do so, and that a clear, consistent, and functional definition of unwarranted variation is lacking – a feature that potentially limits the interpretation of study results. The final study argues that traditional methods for examining regional variations are inadequate for informing health care managers because they examine variation in health care service use rather than needs.</p>