Risk prediction models for incident type 2 diabetes: A scoping review and an update to a systematic review

Early identification of high-risk individuals is key for type 2 diabetes prevention. This study combines a scoping review of historical risk prediction models with a systematic review update of recent models (2019-2026) to map methodological evolution, geographical trends, and reporting quality. PubMed was searched for existing systematic reviews (1993-January 2026) for the scoping review, and for new primary studies (November 2019-January 2026) to update prior evidence. Original English-language regression-based prediction models for incident type 2 diabetes in adults were included; validation studies, machine-learning models, and non-original publications were excluded. Data were synthesised narratively and descriptively, with quality assessed using CHARMS, TRIPOD, and PROBAST. The scoping review synthesised 50 primary studies (median n = 5,217; follow-up 7 years); internal validation was reported in 84.0% versus 16.0% external, and AUC in 90.0%. The systematic update identified 20 new studies (median n = 8,879; follow-up 8.7 years); internal and external validation were respectively reported in 55.0% and 50.0%, calibration in 75.0%, and AUC in 75.0%. Risk of bias remained high in 85.0% of updated studies. Discrimination reporting is now standard, but external validation, calibration, and risk-of-bias reporting remain inconsistent. Future models should prioritise external validation and evaluation in diverse populations before clinical implementation.