Machine learning-based risk predictive models for depression in patients with diabetes: a systematic review and...

Area of Science:

• Medical Informatics
• Clinical Prediction Models
• Diabetes Mellitus Research

Background:

• Inconsistent findings exist regarding machine learning (ML) models for predicting depression risk in patients with diabetes mellitus (DM).
• This systematic review addresses the need for a comprehensive evaluation of existing ML prediction models' performance, strengths, and limitations.

Purpose of the Study:

• To systematically evaluate the performance and clinical applicability of ML-based depression risk prediction models in patients with DM.
• To provide evidence-based guidance for healthcare professionals in selecting and optimizing depression prediction models.

Main Methods:

• Systematic search of PubMed, Embase, Cochrane Library, and Web of Science databases up to January 2026.
• Included studies were assessed for risk of bias and clinical applicability using PROBAST-AI.
• Pooled area under the receiver operating characteristic curve (AUC) was calculated using a random-effects model.

Main Results:

• 14 studies with 64 ML models were included; all had high risk of bias but high clinical applicability.
• Pooled AUC for best-performing models was 0.822, indicating good predictive performance.
• Deep learning models (AUC=0.802) and general ML models (AUC=0.789) outperformed traditional regression models (AUC=0.765). Logistic regression was most common.

Conclusions:

• ML models demonstrate satisfactory predictive performance for depression risk in DM patients.
• Limitations include small sample sizes and lack of external validation in most studies.
• Future research should focus on improving study design and data processing for enhanced generalizability and clinical stability.