Post-stroke depression risk prediction models in stroke patients: A systematic review

Abstract

Background

Post-stroke depression (PSD) is a severe complication that significantly affects patients' prognosis. It is widely believed that risk prediction models can be employed to identify patients early and develop intervention measures. Although the number of PSD prediction models has gradually increased, the quality and applicability of these models remain unclear.

Objective

This study aims to systematically review the published research on risk prediction models for PSD.

Methods

A computer search was conducted in databases including CNKI, WanFang Data, VIP, CBM, PubMed, EMbase, Web of Science, CINAHL, and The Cochrane Library, collecting studies on PSD risk prediction models. The search time frame spanned from the establishment of these databases to March 1, 2025. Two researchers independently screened the literature, extracted data, and assessed the risk of bias for the included studies before performing a qualitative systematic review.

Results

A total of 12 studies were included, comprising 13 risk prediction models. The area under the curve (AUC) or C-index of these models ranged from 0.726 to 0.928. The risk of bias assessment indicated that all the included models were at high risk, with three models demonstrating poor applicability. The most commonly included predictors in the models were, in order: Barthel Index, NIHSS score, age, hypertension, and education level.

Conclusion

Overall, the predictive performance of PSD risk prediction models is promising. However, limitations remain that require further optimization, including issues related to data sources, study design, and data processing. Future research should prioritize the external validation of existing prediction models or the development of higher-quality models with enhanced applicability.