The selective impact of a cardiology data bank on physicians' therapeutic recommendations.

We asked the physicians and medical students caring for 60 patients with symptomatic coronary artery disease, immediately after reviewing cardiac catheterization data, to choose medical or surgical therapy and to estimate prognosis one and three years after either therapy. The next day, each participant was given prognostic estimates generated from a large coronary artery disease data bank and again asked to estimate prognosis and choose therapy. Participants unanimously chose medicine for 20 patients (Group I) and surgery for 21 patients (Group III). For 19 patients (Group II), participants were divided on their choice of therapy. After seeing data bank estimates, participants rarely changed recommendations for Group I or Group III, but changed ten percent (9/90, p less than 0.01) of their Group II recommendations. Changes of recommendations by far (9/12, p = 0.02) favored medicine, causing the majority recommendation to change to medicine for two Group II patients. Therapeutic recommendations were guided mostly by pathoanatomy and the chance of improving medical regimens. Computer-generated prognostic data selectively influenced choices among the Group II cases where recommendations had been divided, resulting in changes toward less costly therapy.