The Value Of Bayesian Discriminant Analysis In Diagnosis Of Primary Aldosteronism

Objective：To explore the value of Bayesian discriminant analysis inclinical diagnosis of primary aldosteronism(PA).Methods： Three hundred and twenty-one suspected cases of PA wereenrolled and divided into two groups: PA group and essential hypertensiongroup. Bayesian probability discriminant analysis was used to get theoriginal classification function coefficients and posterior probability of PA.Then, best cut-off point was obtained according to the receiver operatingcharacteristic (ROC) curves. Finally the value of Bayesian discriminantanalysis in diagnosis of PA were compared with that of upright plasmaaldosterone/renin ratio (ARR).Results: When the posterior probability is0.28, the accuracy of Bayesiananalysis in diagnosis of PA reach to the highest level, sensitivity andspecificity are0.87,0.70respectively.Upright ARR take43.45as cut-offpoint reaches the highest sensitivity and specificity,respectively0.88,0.76.Area under the curve could rank the posterior probability of PAsimilarly as upright ARR do. When the diagnosis of PA meet both uprightARR>43.45and posterior probability>0.28, the specificity of the diagnosiscan up to0.87; however take meet one of them as diagnosis condition, the diagnostic sensitivity of PA can be improved to0.96.Conclusion: The accuracy of Bayesian discriminant function is similar toupright ARR. Their combination can be used in clinical diagnosis of PA.PAshould be considered when reached one of them, and the possibility of PAmay be improved when meet both of them without further confirmatorytest,while if fail to reach both of them, PA can be excluded directly.