Validation of Existing Diabetes Risk Models in a Swedish Population

Several type 2 diabetes risk prediction models have been developed but how it will perform in a Swedish population is not known. No diabetes risk prediction model is routinely used in Sweden. The aim of this study is therefore, to validate 11 non-invasive models and to evaluate the performance to predict incident type 2 diabetes in a Swedish population. A population-based cohort from the Västerbotten Intervention programme (VIP) from 1990 to 2020 will be the validation sample. Incident type 2 diabetes within 10-years of follow-up, will be determined by oral glucose tolerance test or through self-reports. A self-administered questionnaire is completed, and anthropometric, clinical, and biochemical measures are obtained at each of the health screening visits. In the statistical analysis the overall performance of the models will be compared using the Brier score. In addition. discrimination and calibration of all the models will be evaluated. Recalibration of models will be done.

Inclusion criteria

• At least one visit within the Västerbotten Intervention program

Exclusion criteria

• Prevalent diabetes at first visit defined by a fasting capillary plasma glucose >7mmol/L, a 2-hour capillary plasma glucose of ≥12.2 mmol/L or self-reported history of diabetes