Validity evaluation of recently published diabetes risk scoring models in a general Chinese population

The study aimed to assess the validity of some recently published diabetes risk scoring models in a general Chinese population. In 2007, there was a re-examination of 711 individuals who were originally examined in 1992. Since 24 individuals had diabetes in 1992, 687 individuals were available for analysis. Validity was assessed with area under the receiver operating characteristic curve (AROC), and we assessed seven prospective and four cross-sectional models. When applied to our population, AROCs tended to be higher in Asian models than in non-Asian models (average AROCs 0.694±0.034 vs. 0.667±0.040, p=0.258), and those tended to be higher in prospective models than in cross-sectional models (average AROCs 0.695±0.028 vs. 0.652±0.042, p=0.072). A prospective model from Taiwan performed best (AROC 0.749; 95% CI 0.691-0.807). In conclusion, diabetes risk scoring models could not always be generalized from one population to another before validation. Asian models might be more suitable for Asian populations than non-Asian models, and prospective models might be more suitable for predicting future diabetes than cross-sectional models. When applied to our population, a prospective model from Taiwan performed best, and widespread application might be considered in the population.