Predictive model for severe intraventricular hemorrhage risk in preterm infants: a systematic review and meta-analysis

Objective: Risk prediction models offer a potential approach for early identification of severe intraventricular hemorrhage (SIVH) in preterm infants, yet their clinical applicability and methodological quality remain uncertain. This systematic review aimed to identify existing SIVH prediction models in preterm infants, evaluate their performance, and assess their risk of bias and clinical applicability. Methods: We systematically searched PubMed, Web of Science, Embase, CINAHL, MEDLINE, SinoMed, CNKI, and Wan-Fang databases for relevant studies up to September 30, 2025. Data extraction followed the CHARMS framework, while risk of bias and applicability were assessed using PROBAST. Meta-regression explored heterogeneity sources. The review is registered with PROSPERO (CRD42023486813). Results: From 13,311 initially retrieved studies, 16 prediction models were included. A meta-analysis of 7 models yielded a pooled AUC of 0.805 (95% CI: 0.756-0.853). However, all studies exhibited high risk of bias, primarily in the analysis domain, with frequent shortcomings in handling of missing data (93.75%), use of univariable analysis for predictor selection (62.50%), inadequate calibration assessment (68.75%), and non-robust internal validation (50.00%). Methodologically rigorous models demonstrated better performance. Conclusion: Current SIVH prediction models show promise but require methodological improvements. Future efforts should prioritize prospective designs, optimized predictor selection, enhanced external validation, and better calibration to improve clinical utility. Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/view/CRD42023486813, PROSPERO CRD42023486813.