Risk of bias and applicability of prediction models for outcomes of peritoneal dialysis-related peritonitis: a PROBAST systematic review

Background Peritoneal dialysis (PD)-related peritonitis frequently causes hospitalization, catheter loss, hemodialysis transfer, and death. Existing reviews have focused on peritonitis incidence rather than post-diagnosis outcomes. We systematically reviewed prediction models for post-peritonitis outcomes and assessed their risk of bias and applicability using PROBAST. Methods PubMed, EMBASE, Web of Science, Scopus, Cochrane Library, CNKI, Wanfang, and VIP were searched from inception to January 30, 2026, with an updated search on March 15, 2026. We included studies developing or externally validating multivariable models for adverse outcomes after peritonitis diagnosis in adults on maintenance PD. Two reviewers independently screened records, extracted data, and applied PROBAST across four domains for both risk of bias and applicability. External validation was classified under a prespecified four-tier framework; calibration completeness required slope and intercept. Results Seven studies reporting seven prediction models met the inclusion criteria, comprising 3,821 peritonitis episodes in six studies (one study did not report its sample size). No model achieved an overall low risk of bias. One model underwent limited independent external validation. Two models were rated low risk in the participants, predictors, and outcome domains but unclear overall owing to incomplete calibration reporting, unclear missing-data handling, and lack of explicit overfitting assessment; applicability concerns were identified in the analysis domain for all models. The remaining five models were rated high risk overall. No study among the included model development/validation studies performed prospective clinical impact assessment. Conclusions Currently, no model has sufficient evidence for routine clinical implementation or screening; under strict supervision, local calibration, and a clear research purpose, some models may serve as research tools. Trial Registration PROSPERO CRD4202654321.