Optimizing Treatment Strategies in the Bipolar Disorder Spectrum With Classical AI Approaches: Systematic Review of Performance, Bias, and Clinical Applicability.

Acute symptomatic response models performed modestly (pooled area under the curve [AUC] 0.68), while imaging improved accuracy (74%-77%).

Relapse and readmission prediction achieved a pooled AUC of 0.71, with digital phenotyping and rule-based methods performing best (AUC 0.85-0.88).

Safety and dose optimization models achieved 85%-97% accuracy.

However, 3 studies were considered at high risk of bias due to small sample sizes associated with disproportionately high-performance estimates.