Liver function test indices-based prediction model for post-stroke depression: a multicenter, retrospective study.

ABSTRACT

BACKGROUND: Post-stroke depression (PSD) was one of the most prevalent and serious neuropsychiatric effects after stroke. Nevertheless, the association between liver function test indices and PSD remains elusive, and there is a lack of effective prediction tools. The purpose of this study was to explore the relationship between the liver function test indices and PSD, and construct a prediction model for PSD. METHODS: All patients were selected from seven medical institutions of Chongqing Medical University from 2015 to 2021. Variables including demographic characteristics and liver function test indices were collected from the hospital electronic medical record system. Univariate analysis, least absolute shrinkage and selection operator (LASSO) and logistic regression analysis were used to screen the predictors. Subsequently, logistic regression, random forest (RF), extreme gradient boosting (XGBoost), gradient boosting decision tree (GBDT), categorical boosting (CatBoost) and support vector machine (SVM) were adopted to build the prediction model. Furthermore, a series of evaluation indicators such as area under curve (AUC), sensitivity, specificity, F1 were used to assess the performance of the prediction model. RESULTS: A total of 464 PSD and 1621 stroke patients met the inclusion criteria. Six liver function test items, namely AST, ALT, TBA, TBil, TP, ALB/GLB, were closely associated with PSD, and included for the construction of the prediction model. In the test set, logistic regression model owns the AUC of 0.697. Compared with the other four machine learning models, the GBDT model has the best predictive performance (F1 = 0.498, AUC = 0.761) and was chosen to establish the prediction tool. CONCLUSIONS: The prediction model constructed using these six predictors with GBDT algorithm displayed a promising prediction ability, which could be used for the participating hospital units or individuals by mobile phone or computer.