Variation in medical practice: implications for the hospital epidemiologist.

Why do two physicians faced with similar patients make different decisions about the care their patients should receive? This area of health services research has been energized by recent concerns about rising healthcare costs. In 1982, Wennberg and Gittelsohn'1 described the great variation in surgical procedure rates throughout the northeastern United States during the 1970s. Population rates of hysterectomy and prostatectomy varied fourfold between local service areas; tonsilectomy rates varied sixfold. Differences in patient populations could not explain the variation in rates. The authors concluded that a large portion of the variability was attributable to the personal preferences of the surgeons practicing in the various communities. Similar differences in procedure rates are easy to replicate and are evident at several levels of aggregation. On larger scales, regional and international differences can be documented easily.2 On a smaller scale, practice variation within individual hospitals is frequently evident. Dr. David Eddy, Professor of Health Policy and Management at Duke University, has studied the conditions that allow two physicians to treat the same condition in dramatically different ways.3 The physician's decision to recommend a course of treatment is the outcome of a decision-making process. The idealized process has several elements. First, the physician gathers evidence relevant to the patient's complaint. Using his training and experience, the physician judges the evidence to arrive at a diagnosis. After the diagnosis is made, the physician weighs the advantages and disadvantages of various treatment options according to the expected benefit for the patient. Some variation is to be expected, and in fact, is essential. The physician's primary obligation is to his patient. Each patient has a different set of social and h alth needs to which the physician must respond. However, for each step in the decision-making process, physician-to-physician variation not attributable to patient differences can be documented.4 Dr. Eddy believes that it is the uncertainty surrounding the true health risks and benefits associated with various treatment modalities that account for much of the variability in medical practice. For example, 140 urologists were asked to estimate the chance that a 65-year-old man with moderate symptoms of benign prostatic hypertrophy would develop complete uriary retention in the subsequent five years. Their estimates ranged from 10% to 80%. Cardiovascular surgeons were asked to estimate the probability of replacement valve failure for xenografts and mechanical valves. The estimates of failure rates ranged from 3% to 95% for xenografts and from 0% to 50% for mechanical valves.3