Decision tree distinguish affective disorder diagnosis from psychotic disorder diagnosis with clinical and lab factors

Abstract

BackgroundAffective symptoms usually occur at the same time of psychotic symptoms. An effective predictive method would help the differential diagnosis at an early stage of the mental disorder. The purpose of the study was to establish a predictive model by using laboratory indexes and clinical factors to improve the diagnostic accuracy. MethodsSubjects were patients diagnosed with psychiatric disorders with affective and/or psychotic symptoms. Two patient samples were collected in the study (n = 309) With three classification methods (logistic regression, decision tree, and discriminant analysis), we established the models and verified the models. ResultsSeven predictors were found to be significant to distinguish the affective disorder diagnosis from the psychotic disorder diagnosis in all three methods, the 7 factors were Activities of daily living, direct bilirubin, apolipoproteinA1, lactic dehydrogenase, creatinine, monocyte count and interleukin-8. The decision tree outperformed the other 2 methods in area under the receiver operating characteristic curve, and also had the highest percentage of correctly classification. ConclusionWe established a predictive model that included activities of daily living, biochemical, and immune indicators. In addition, the model established by the decision tree method had the highest predictive power, which provided a reliable basis for future clinical work. Our work would help make diagnosis more accurate at an early stage of the disorder.