Evaluating global epidemiology of type 2 diabetes mellitus among the working-age population: A 60-year study by interpretable machine learning framework.

Aims: This study aimed to characterise the epidemiology in the trends of type 2 diabetes mellitus (T2DM) burden among working-age population (WAP) from 1990 to 2021 and to forecast future patterns up to 2050 using interpretable machine learning, with an emphasis on providing key drivers for public health strategies.

Materials and methods: Granular data were extracted from the Global Burden of Disease 2021 study. We built the eXtreme Gradient Boosting (XGBoost) models to predict T2DM burden in WAP and clarified the driving factors using SHapley Additive exPlanations (SHAP) to enhance interpretability.

Results: Global incident cases in WAP are projected to increase five-fold from 6.33 million (1990) to 32.38 million (2050), with North Africa/Middle East showing the fastest growth (estimated annual percentage change, EAPC = 3.45). Notably, the traditional inverse relationship between socio-demographic index and T2DM burden is reversing (p < 0.01), with high-income regions like the UK facing accelerating age-standardised DALY rates by 2050 (EAPC = 1.43). SHAP analysis identified age as the predominant contributor, while high blood glucose, BMI, and air pollution remained consistently influential. High temperature and alcohol consumption emerged as increasingly significant factors.

Conclusions: Global T2DM burden among the WAP shows alarming trends, particularly in ageing societies where labour forces are highly valued. Tailored WAP guidelines may reduce T2DM burden worldwide, but stronger efforts on obesity and climate change are essential for effective T2DM management.