Bayesian calculations for the determination of the etiology of pleuritic chest pain in young adults in a teaching hospital

We have studied use of the preceding program for the differential diagnosis of young patients presenting with pleuritic chest pain. We have evaluated the effects of roundoff approximations on computational accuracy and have compared the diagnoses made with Bayesian calculations with actual clinical diagnoses. Although the effects of roundoff were small, the computed diagnoses agreed with the discharge diagnoses only 73% of the time. Moreover, the correct diagnosis of pulmonary embolism was made in only 45% of the cases. By introducing a “no decision” region for patients whose maximum Bayesian calculation was less than 0.55 or 0.65, the total number of correct diagnoses fell but a larger number of patients with pulmonary embolism was potentially identified. (April 1978)