Towards Transforming Maternal Health in Pakistan Through Equitable, Context-Aware AI for Early Detection of Gestational Diabetes Mellitus (GDM)
Oxford Open Digital Health · 13.11.2026
DOI: 10.1093/oodh/oqag017 ↗Kurz gefasst
Fachartikel in der Zeitschrift Oxford Open Digital Health, online erschienen am 13.11.2026. Verfasst von Ihtesham ul Islam und 8 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Themen: Diabetes mellitus
Abstract (Originalsprache)
This study presents AI-GDMP, an AI-enabled risk prediction tool developed to support early detection of Gestational Diabetes Mellitus (GDM) among pregnant women attending a tertiary hospital in Islamabad, Pakistan. GDM diagnosis relies on mid-pregnancy oral glucose tolerance testing (OGTT) performed between 24– 28 weeks of gestation. By focusing on the first trimester and utilising non-invasive, low-cost features, AI-GDMP offers a potentially equitable and accessible approach to early high/low-risk classification, particularly valuable in resource-constrained antenatal settings. By flagging high-risk women early in pregnancy, the tool opens a wider window for lifestyle modification, nutritional counselling, and closer follow-up before overt hyperglycaemia develops. The AI model was trained and validated on retrospective data from a JCI-accredited tertiary care hospital in Islamabad (2015–2024): eligible OGTT-linked pregnancies were digitized and screened (n = 3,748), yielding a final analytic cohort of 3,102 pregnancies (1,352 GDM; 1,750 non-GDM) with a stratified 80/20 train–test split (n = 2,481/621). A structured feature engineering pipeline reduced 93 candidate variables to 49 early-pregnancy predictors, with additional reduced feature sets (29 and 25 predictors) derived using Lasso- based selection. Among multiple machine learning approaches, the Random Forest classifier delivered the most promising results on the independent hold-out test set, achieving an AUC-ROC of 0.88, an accuracy of 77.9%, and a recall of 75.9%. Data collection followed ethical protocols, including IRB approval and anonymized patient records. The trained-model was integrated into a clinical software environment, enabling real-time high/low-risk output through a physician dashboard and provides a foundation for prospective validation and equity-focused evaluation in Phase 2.
Zitieren
APA
Islam, I. U., Iltaf, N., Afzal, H., Nawaz, S., Hussain, Z. T., Zahid, T., Zia, U., Mehmood, K., & Dadan, S. (2026). Towards Transforming Maternal Health in Pakistan Through Equitable, Context-Aware AI for Early Detection of Gestational Diabetes Mellitus (GDM). Oxford Open Digital Health. https://doi.org/10.1093/oodh/oqag017
Vancouver
Islam IU, Iltaf N, Afzal H, Nawaz S, Hussain ZT, Zahid T, et al. Towards Transforming Maternal Health in Pakistan Through Equitable, Context-Aware AI for Early Detection of Gestational Diabetes Mellitus (GDM). Oxford Open Digital Health. 2026. doi:10.1093/oodh/oqag017
BibTeX
@article{Islam2026,
author = {Islam, Ihtesham ul and Iltaf, Naima and Afzal, Hammad and Nawaz, Shahnaz and Hussain, Zuha Tassawar and Zahid, Tehreem and Zia, Usman and Mehmood, Khawir and Dadan, Sundus},
title = {Towards Transforming Maternal Health in Pakistan Through Equitable, Context-Aware AI for Early Detection of Gestational Diabetes Mellitus (GDM)},
journal = {Oxford Open Digital Health},
year = {2026},
doi = {10.1093/oodh/oqag017}
}
EUVIMED