Objective: To establish whether the Finnish diabetes risk score for predicting the incidence of diabetes (FINDRISK) is also valid in the Netherlands, and to choose cut-off points suitable for the Dutch situation.
Design: . Descriptive.
Method: The FINDRISK was validated in 3 Dutch cohort studies by means of repeated glucose measurements: the Hoorn study (n=5434), the PREVEND study (n=2713) and part of the Maastricht cohort from the MORGEN study (n=863). The predictive value was evaluated using receiver operating characteristic (ROC) analyses. The risk categories were defined on the basis of sensitivity, specificity and positive predictive value.
Results: The predictive value of the FINDRISK was best in the PREVEND cohort (area under the ROC curve 0.77) and was lower for the Hoorn study and the Maastricht cohort (area under the ROC-curve 0.71 for both). The scores were divided into three risk categories: low risk (score lower than 7), slightly increased risk (score 7-9) and increased risk (score so or higher). The percentage of persons with incident diabetes within about 5 years was < 6 in the low risk category, 6-14 in the category with slightly increased risk and 12-26 in the category with increased risk. 16-28% of the Dutch population studied had a score of 10 or higher.
Conclusion: The FINDRISK is a reasonably good predictor for incident diabetes in the Netherlands.