A systematic review of risk prediction models for post-stroke acute kidney injury

Objective To systematically evaluate the risk prediction model for acute kidney injury (AKI) after stroke.Methods Studies on post-stroke AKI risk prediction models from PubMed, Web of Science, Cochrane Library, Embase, CNKI, Wanfang, VIP, and Chinese Biomedical Literature Database were searched from inception to December 23, 2024. The Prediction Model Risk of Bias Assessment Tool (PROBAST) were used to evaluate the bias and applicability of studies, and descriptive methods were used to analyze model characteristics.Results A total of 15 studies were included, including 33 predictive models. 11 studies (73.3%) used Logistic regression models to construct predictive models, 6 studies (40.0%) selected predictive factors based on single factor analysis, 3 studies (20.0%) did not report methods for handling missing data, 14 studies (93.3%) presented predictive models through column charts, risk scoring scales, and regression equations. The common predictive factors included in the models included age, hypertension, serum creatinine, blood urea nitrogen, use of diuretics. 11 studies (73.3%) were internally validated, and 7 studies (46.7%) were externally validated. Among the 33 models, 26 models reported the area under the curve of the receiver operating characteristic curve, and 13 models (39.4%) were evaluated for calibration using calibration curves or Hosmer-Lemeshow goodness of fit tests. All included studies had a high risk of bias, and 11 studies had good applicability.Conclusion The quality of the modeling methodology for AKI risk prediction models after stroke is uneven, and the overall risk of bias is high. In the future, the development quality of prediction models should be further improved by following PROBAST standards and TRIPOD reporting standards.