Objective To systematically search and evaluate predictive models for neonatal hypoglycemia risk in infants born to mothers with gestational diabetes mellitus, providing a reference for neonatal clinical workers to choose or develop suitable risk assessment tools. Methods A systematic search was conducted in both Chinese and English databases, including PubMed, Web of Science, The Cochrane Library, Embase, the China BioMedical Literature Database (CBM), the VIP Database, China National Knowledge Infrastructure (CNKI), and Wanfang Database, with the search period set from the inception of each database to November 12, 2024. Two trained researchers independently performed literature screening and data extraction. The prediction model risk of bias assessment tool (PROBAST) was used to assess the risk of bias and applicability of the included studies. Results A total of 13 studies on risk prediction models for neonatal hypoglycemia in infants born to mothers with gestational diabetes mellitus were included, encompassing 14 prediction models. The number of predictors ranged from 2 to 16, with the most common predictors being preterm birth, maternal age, mode of delivery, glycemic control during pregnancy, and neonatal birth weight. The area under the curve (AUC) of receiver operating characteristic curve for the included models ranged from 0.580 to 0.887. The quality assessment results for all 13 studies indicated a "high risk of bias" and "low concern regarding applicability." Conclusions Current evidence suggests that the overall predictive performance of risk prediction models for neonatal hypoglycemia in infants born to mothers with gestational diabetes mellitus is moderate, with a high risk of bias but low concern regarding applicability. The overall quality still has considerable room for improvement. Future model development studies should follow the recommendations in the PROBAST tool and the CHARMS checklist to reduce the risk of bias, and should conduct multi-center external validation to enhance the generalizability and reliability of these models across diverse clinical settings.
2026, Vol. 14 
