Patient-level simulation models, mainly developed in Western populations, capture complex interactions between risk factors and complications to predict the long-term effectiveness and cost-effectiveness of novel treatments and identify high-risk subgroups for personalized care. However, incidence of outcomes varies significantly by ethnicity and region. We used high-quality, patient-level register data to develop the Chinese Diabetes Outcomes Model (CDOM) for predicting incident and recurrent events in type 2 diabetes (T2D).
The CDOM was developed using the prospective Hong Kong Diabetes Register (HKDR) cohort (n = 21,453; median follow-up duration, 7.9 years; 166,433 patient-years). It was externally validated with a retrospective territory-wide cohort of Chinese patients with T2D attending Hong Kong publicly funded diabetes centers and community clinics (n = 176,120; follow-up duration, 7.2 years; 953,523 patient-years).
The CDOM predicted first and recurrent events with satisfactory performance during internal (C-statistic = 0.740–0.941) and external (C-statistic = 0.758–0.932) validation after calibration. The respective C-statistic values for cancer were 0.664 and 0.661. Subgroup analysis showed consistent performance during internal (C-statistic = 0.632–0.953) and external (C-statistic = 0.598–0.953) validation after calibration.
The CDOM, developed using comprehensive register data with long-term follow-up, is a robust tool for predicting long-term outcomes in Chinese patients with T2D. The model enables the identification of patient subgroups to augment study design and develop tailored novel treatment strategies, inform policy, and guide practice to improve cost-effectiveness of diabetes care.
This article contains supplementary material online at https://doi.org/10.2337/figshare.28250639.
