[Validity of risk scores to predict type 2 diabetes in the Dutch population]

To validate risk scores to predict occurrence of type 2 diabetes in the Dutch population.

Prospective cohort study.

Twelve basic risk scores and 13 extensive risk scores with biomarkers were used to predict the risk of developing type 2 diabetes during 7.5 years in a prospective cohort of 38,379 Dutch men and women. Occurrence of diabetes was documented through repeated questionnaires and validated against medical records. The capacity of the risk scores to correctly identify those at high risk of developing diabetes was determined using the C-statistic. The capacity of the risk scores to correctly quantify the absolute risk of diabetes was determined by testing the difference between the predicted and observed risk in the population.

The capacity of basic risk scores to identify those at high risk of diabetes was good, with C-statistics ranging from 0.74 (95%-CI: 0.73-0.75) to 0.84 (0.82-0.85). The extended risk scores were very capable of identifying those at high risk of diabetes, with C-statistics ranging from 0.81 (0.80-0.83) to 0.93 (0.92-0.94). Most risk scores, however, were unable to correctly quantify the absolute risk of diabetes; the risk was usually overestimated. Only the basic KORA model correctly quantified the risk in this Dutch population.

In the Dutch population, risk scores to predict the occurrence of type 2 diabetes are very capable of identifying those at high risk. Extension with biomarkers improves this capacity. Quantification of the absolute risk of diabetes was insufficient in most risk scores.