A population-based risk algorithm for the development of diabetes: development and validation of the Diabetes Population Risk Tool (DPoRT)

Background National estimates of the upcoming diabetes epidemic are needed to understand the distribution of diabetes risk in the population and to inform health policy. Objective To create and validate a population-based risk prediction tool for incident diabetes using commonly collected national survey data. Methods With the use of a cohort design that links baseline risk factors to a validated population-based diabetes registry, a model (Diabetes Population Risk Tool (DPoRT)) was developed to predict 9-year risk for diabetes. The probability of developing diabetes was modelled using sex-specific Weibull survival functions for people >20 years of age without diabetes (N=19 861). The model was validated in two external cohorts in Ontario (N=26 465) and Manitoba (N=9899). Predictive accuracy and model performance were assessed by comparing observed diabetes rates with predicted estimates. Discrimination and calibration were measured using a C statistic and Hosmer-Lemeshow chi(2) statistic (chi(2)(H-L)). Results Predictive factors included were body mass index, age, ethnicity, hypertension, immigrant status, smoking, education status and heart disease. DPoRT showed good discrimination (C=0.77-0.80) and calibration (chi(2)(H-L) <20) in both external validation cohorts. Conclusions This algorithm can be used to estimate diabetes incidence and quantify the effect of interventions using routinely collected survey data.