The unexpected influence of physician attributes on clinical decisions: results of an experiment

This experiment was designed to determine: (1) whether patient attributes (specifically a patient's age, gender, race, and socioeconomic status) independently influence clinical decision-making; and (2) whether physician characteristics alone (such as their gender, age, race, and medical specialty), or in combination with patient attributes, influence medical decision-making.

An experiment was conducted in which 16 (= 2(4)) videotapes portraying patient-physician encounters for two medical conditions (polymyalgia rheumatica (PMR) and depression) were randomly assigned to physicians for viewing. Each video presented a combination of four patient attributes (65 years or 80 years of age; male or female; black or white; blue or white collar occupation). Steps were taken to enhance external validity. One hundred twenty-eight eligible physicians were sampled from the northeastern United States, with numbers balanced across 16 (= 2(4)) strata generated from the following characteristics (male or female; < 15 or > or = 15 years since graduation; black or white; internists or family practitioners). The outcomes studied were: 1) the most likely diagnosis; 2) level of certainty adhering to that diagnosis; and 3) the number of tests that would be ordered.

Patient attributes (namely age, race, gender, and socioeconomic status) had no influence on the three outcomes studied (the most likely diagnosis, the level of certainty, and test ordering behavior). This was consistent across the two medical conditions portrayed (PMR and depression). In contrast, characteristics of physicians (namely their medical specialty, race, and age) interactively influenced medical decision-making.

Epidemiologically important patient attributes (which Bayesian decision theorists hold should be influential) had no effect on medical decision-making for the two conditions, while clinically extraneous physician characteristics (which should not be influential) had a statistically significant effect. The validity of idealized theoretical approaches to medical decision making and the usefulness of further observational approaches are discussed.