Discriminant Analysis of Occupational Performance Characteristics in Patients with Bipolar Depression and Healthy Individuals

In this retrospective study, occupational performance characteristics were compared between 21 patients with BD and 43 healthy control participants during a single, standardized artistic activity-based occupational therapy session at Kyorin University Hospital. Characteristics showing significant group differences were entered into multivariable logistic regression analysis, adjusting for age and gender, and into discriminant analysis using the stepwise method.

Model performance was evaluated using receiver operating characteristic (ROC) curve analysis and leave-one-out cross-validation (LOOCV). Results: Among 65 identified occupational performance characteristics, 12 differed significantly between groups. The final discriminant model identified two characteristics reduced tendency to “Ask questions and consult” and greater “Difficulty in understanding procedures” as contributing to group discrimination (Wilks’ λ = 0.630, χ² = 28.188, df = 2, p < 0.001).

The overall classification accuracy was 78.1%, which was identical in LOOCV. ROC curve analysis demonstrated moderate discriminative accuracy (AUC = 0.829, 95% CI = 0.718–0.941, p < 0.001; sensitivity 74.4%, specificity 85.7%). Conclusion: Two observable occupational performance characteristics reduced help-seeking behavior and difficulty understanding procedures distinguished patients with BD from healthy control participants with moderate accuracy.

These findings may reflect executive and information-processing vulnerabilities associated with BD and support occupational therapy as a source of clinically meaningful behavioral information that complements conventional symptom-based assessment.</p>