Drug choice as a problem-solving process.

Abstract

A model of the drug prescribing process, which incorporates prescribers' personal values about treatment outcomes and beliefs about treatment effects, was tested under actual clinical conditions. Forty physicians were given two fictional case histories and six disguised case histories of patients whom they had recently treated for hypertension or maturity-onset diabetes mellitus. The physicians completed questionnaires based on each case history that measured 1) the beliefs about the probability that seven treatment-related outcomes would result from the prescribing of several alternative treatments and 2) the values placed on each outcome. The physicians were also asked, in an open-ended question, how they would treat the patient described in the case. The 40 physicians proposed 172 drug treatments that corresponded to treatment alternatives for which beliefs about treatment effects had been measured. The model correctly predicted 1) prescribing intent in 81% of hypertension cases and in 87% of the diabetes cases and 2) actual prescribing in 76% of hypertension cases and in 70% of the diabetes cases, significantly more than would be expected at random (P less than 0.01). The prescribing model appears useful for predicting drug choices for the outpatient treatment of hypertension and diabetes by resident physicians.