Population-based diagnostic prediction scores to identify undiagnosed type 2 diabetes mellitus in low- and middle-income countries in South and East Asia and the Pacific: A systematic Review and meta-analysis.

INTRODUCTION: Undiagnosed diabetes prevalence is high in low- and middle-income countries (LMICs) in South Asia, East Asia & Pacific (SAEA&P). Non-invasive scores may be helpful for population screening to identify people for blood glucose testing.

AIMS: To assess population-based diagnostic prediction scores' performance to identify undiagnosed type 2 diabetes mellitus (T2D) in adults in SAEA&P LMICs.

METHODS: A systematic search was conducted in PubMed, OVID, and Cochrane Library in July 2025, followed by hand searches. We included studies that developed or validated T2D diagnostic prediction scores in adult general populations without known T2D from LMICs in SAEA&P. A bivariate diagnostic accuracy meta-analysis was performed for scores with multiple validations across different countries in studies without a high risk of bias. Discriminative performance was measured based on the Area Under Curve (AUC). This review was registered with PROSPERO (CRD42024584501).

RESULTS: We identified 9651 articles and assessed 85 full-text articles for eligibility. Thirty articles were included in the review: eleven that developed and 19 that validated population-based T2D diagnostic prediction scores across eight countries. The meta-analyses of Finnish Diabetes Risk Score (FINDRISC) variations and of the Indian risk score show fair pooled discriminative performance for both scores (AUC = 0.735 and AUC = 0.704, respectively). Scores adapted from those developed in Europe had higher specificity but lower sensitivity than those developed in SAEA&P. No articles were considered to have a low risk of bias.

CONCLUSIONS: Some T2D diagnostic prediction scores could be useful in SAEA&P LMICs; however, robust external validation in local settings is required before implementation.