JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) models show potential for enhancing the prediction of treatment response to cognitive behavioral therapy (CBT) in patients with anxiety disorders.

Objective:

This systematic review and meta-analysis aim to quantify the overall performance of AI models in predicting treatment response following CBT for anxiety disorders and to examine how data sources, algorithmic approaches, and diagnostic subtypes influence predictive performance.

Methods:

We conducted a systematic literature search in PubMed, Embase, Web of Science, Cochrane Library, and PsycINFO up to August 2025. The analysis included studies that developed or validated AI models for predicting CBT outcomes. Predictive performance metrics including sensitivity, specificity, accuracy, and area under the curve (AUC) were extracted and pooled.

Results:

Eleven studies (n=38,563 patients) were included in the analysis. The pooled sensitivity of AI models was 0.64 (95% CI: 0.58–0.71), specificity was 0.61 (95% CI: 0.53–0.69), accuracy was 0.68 (95% CI: 0.64–0.71), and the summary area under the curve (AUC) was 0.67 (95% CI: 0.63–0.71), indicating moderate overall predictive performance. Models utilizing multimodal data demonstrated the strongest performance, achieving a sensitivity of 0.70 and an AUC of 0.76. Furthermore, predictions were most accurate for generalized anxiety disorder and panic disorder compared to other anxiety disorder subtypes.

Conclusions:

AI shows promising potential for predicting CBT responses in anxiety disorders, with multimodal approaches demonstrating superior performance. However, clinical application remains limited by study heterogeneity, small sample sizes in some studies, and insufficient external validation. Future research should prioritize large, prospective, multi-center studies with standardized protocols to enhance the generalizability and clinical utility of these models. Clinical Trial: Registration ID: CRD420251137096