Risk prediction models for acute kidney injury in patients with acute pancreatitis: a systematic review.

Abstract

Background: Acute kidney injury (AKI) is a life-threatening complication of acute pancreatitis (AP), significantly increasing mortality and extending hospital stay. Early identification of high-risk patients is crucial for timely intervention yet remains clinically challenging. Although several prediction models have been developed, their methodological quality, predictive performance, and clinical applicability vary greatly and lack comprehensive evaluation. A systematic synthesis of existing evidence is therefore imperative to inform clinical practice and guide future research.

Methods: Two reviewers examined Chinese databases (including CNKI, VIP, SinoMed, and Wanfang), along with PubMed, Web of Science, the Cochrane Library, Embase, and Scopus, to identify risk prediction models for AKI in patients with AP published from inception to December 2025. Two reviewers independently reviewed the literature, retrieved relevant information, and evaluated the risk of bias using the PROBAST tool. The main performance measures of the models, including AUC, calibration, validation method, and predictors, were retrieved.

Results: Our review assessed 20 studies with 9481 patients. The prevalence of AKI varied between 13.8% and 72.8%. The reported AUCs of the prediction models ranged from 0.735 to 0.920. All 20 studies were assessed as having high overall risk of bias according to the PROBAST tool, primarily owing to retrospective study designs, lack of external validation, and incomplete model development processes.

Conclusions: Existing AKI prediction models for AP patients show moderate to good apparent discrimination but are based on studies with uniformly high risk of bias. No model can be recommended for routine clinical use without prospective external validation.