Bayesian statistics as applied to hypertension diagnosis

Abstract

This paper deals with the following hypertension diagnoses: essential hypertension and five types of secondary hypertension: fibrodysplasic renal artery stenosis, atheromatous renal artery stenosis, Conn's syndrome, renal cystic disease, and pheochromocytoma. Only blood pressures, general information and general biochemical data are taken into account. Nineteen items were finally selected, by statistical investigation of experimental data, as being both discriminative and independent. The marginal density distributions of every item, and then joint density distribution functions were determined within six types of hypertension. The frequency of a given hypertension type within the hypertensive patients was used as prior probability of this state. The loss matrix was established by medical arguments. The expected loss corresponding to six possible decisions could thus be calculated for all cases. Both the ratio of secondary hypertensions that could be inferred from our set of data (not including the results of complementary tests) and that of correct "essential" hypertension diagnosis proved to be satisfactory.