Virtual Booth

Type Your Abs

Cerebrovascular disease (CeVD) is a major cause of death and disability among people with Type 2 Diabetes Mellitus (T2DM) in Malaysia. However, most existing risk tools were developed in Western populations and often need complex computing to use. The I-Stroke Score was developed as a simple, point based and AI-calibrated risk prediction tool for CeVD among T2DM patients, designed for routine use in Malaysian primary care.

The score was developed and validated using the Pulau Pinang National Diabetes Registry (2019 to 2024), involving 20,338 patients across three temporal phases. Logistic regression with backward elimination produced an eight predictor score covering age, sex, diabetes duration, body mass index, hypertension, chronic kidney disease stage, ischaemic heart disease and district. Patients are grouped into Low, Medium and High risk tiers. Six machine learning algorithms, including Random Forest, XGBoost and CatBoost, were benchmarked against the score.

The score achieved AUC values of 0.681 at development, 0.663 at temporal internal validation and 0.630 at temporal external validation, with a consistent rise in observed CeVD prevalence across risk tiers. No machine learning model significantly outperformed the score. When calibration drifted in the latest cohort, AI-based recalibration using Platt scaling restored good calibration (Hosmer Lemeshow p = 0.261) without changing the points table.

The tool is deployed as a free web calculator (zuhairyusoff.github.io/IStrokeCalculator) and is copyright registered with MyIPO (CRLY2025W02431). The I-Stroke Score offers a transparent, low cost and updatable approach to stroke risk stratification that can be adapted to other health systems.