Risk prediction scores for mortality, cerebrovascular, and heart disease among Chinese people with type 2 diabetes

Abstract

Context: Risk scores for cardiovascular and mortality outcomes have not been commonly applied in Chinese populations.

Objective: To develop and externally validate a set of parsimonious risk scores (HKU-SG) to predict the risk of mortality, cerebrovascular disease and ischemic heart disease among Chinese people with type 2 diabetes; and compare HKU-SG risk scores to other existing ones.

Design: Retrospective population-based cohorts drawn from Hong Kong Hospital Authority health records from 2006 to 2014 for development and Singapore Ministry of Health records from 2008 to 2016 for validation. Separate 5-year risk scores were derived using Cox proportional hazards models for each outcome.

Setting: Study participants were adults with type 2 diabetes aged 20 years or over, consisting of 678,750 participants from Hong Kong and 386,425 participants from Singapore.

Main Outcome: Measures Performance was evaluated by discrimination (Harrell’s c-index), and calibration plots comparing predicted against observed risks.

Results: All models had fair external discrimination. Among the risk scores for the diabetes population, ethnic-specific risk scores (HKU-SG and JADE) performed better than UKPDS and RECODe models. External validation of the HKU-SG risk scores for mortality, cerebrovascular disease and ischemic heart disease had corresponding c-indices of 0.778, 0.695, and 0.644. The HKU-SG models appeared well calibrated on visual plots, with predicted risks closely matching observed risks.

Conclusions: The HKU-SG risk scores were developed and externally validated in two large Chinese population-based cohorts. The parsimonious use of clinical predictors compared to previous risk scores could allow wider implementation of risk estimation in diverse Chinese settings.