An explanatory model of medical practice variation: a physician resource demand perspective

Abstract

Practice style variation, or variation in the manner in which physicians treat patients with a similar disease condition, has been the focus of attention for many years. The research agenda is further intensified by the unrealistic assumption that by reducing variation, quality will be improved, costs will be reduced, or both. There is a wealth of literature that identifies differences in health care use of many kinds, in apparently similar communities. Attempts have been made by many scholars to identify the determinants of variation in terms of differences in the population characteristics (e.g. age, sex, insurance, etc.) and geographical characteristics (e.g. distance to provider, number of physicians, number of hospital beds, etc.). When significant differences in use rates prevail after controlling for differences in population characteristics, it is often attributed to 'uncertainty', or the fact that there is no consensus on what constitutes the optimum treatment process. It is suggested by this literature that the greatest variation can be found in the circumstances where there is the most 'uncertainty'. In this work, a physician resource demand model is proposed in which it is suggested that, during the diagnosis and treatment process, physicians demand resources consistent with the clinical needs of the patients, modified by the intervening forces under which they practice. These intervening forces, or constraints, are categorized as patient agency constraints, organizational constraints and environmental constraints, which are characterized as 'induced variation'. It is suggested that when all of the variables that constitute these constraints are identified, the remaining variance represents 'innate variance', or practice style differences. It is further suggested that the more completely this model is specified, the more likely area differences will be attenuated and the smaller will be the residual variance.