Risk Prediction Performance of Blood Biomarkers for Bipolar Disorder With Psychotic Symptoms

Objective  To investigate biological markers associated with psychotic symptoms in patients with bipolar disorder (BD) based on electronic medical records of patients, and to develop an interpretable risk prediction model that supports the identification of high-risk individuals and that facilitates decision-making for providing clinical intervention in a timely manner.

Methods  A total of 2352 patients diagnosed with BD and admitted to West China Hospital, Sichuan University were enrolled using the electronic medical records system of the hospital. The participants were divided into two subgroups, the bipolar disorder depression (BDD) group and the bipolar disorder mania (BDM) group. The logistic regression algorithm was used to train and validate the prediction model, and interpretability methods were used to analyze the contribution of each feature to individuals and the effect of the features on specific target prediction decisions.

Results  The logistic regression model demonstrated robust predictive performance across the BD, BDD, and BDM cohorts, with areas under the curve (AUC) of the receiver operating characteristic curves always exceeding 81.6%. The core predictive features included platelet distribution width (PDW), fibrinogen (FIB), platelet large cell ratio (P-LCR), activated partial thromboplastin time (APTT), prothrombin time (PT), and triglyceride (TG). The logistic regression model exhibited strong interpretability and was combined with nomograms for intuitive risk quantification and individualized prediction.

Conclusion  The logistic regression model enables rapid and simple screening of BD patients with psychotic symptoms. Distinct patterns of changes observed in blood biomarkers of BDD and BDM subgroups enrich the understanding of the underlying pathophysiological mechanisms and highlight the importance of considering subtypes in the intervention and management of patients.