Introduction Clinical usefulness of a risk function is a new concept that should be considered beyond the traditional checking of discrimination and calibration for performance of a model. Framingham risk score (FRS) for cardiovascular disease is a widely used one which has been validated in different countries but its clinical usefulness has been neglected.
Methods We checked discrimination of FRS and so its calibration and clinical usefulness before and after recalibration in a population based cohort, Tehran lipid and glucose study, of 2640 men and 3584 women aged 30–74 years. To check clinical usefulness, we used decision curve analysis (DCA) and calculated net benefit of treatment for patients with ≥20% of 10 year probability of disease according to FRS model.
Results The area under the curve for FRS model, was 0.794 and 0.838 for men and women respectively. The original model had a poor calibration but got a good one after recalibration (Hosmer-Lemeshow χ2 statistic of 16.8 for men and 18.4 for women). Based on DCA, FRS was clinically useful in cut points of 10%–30%, as threshold probability of disease that a patient should be treated, before and after recalibration. The net benefit of model to treat patients at cut point of 20% did not differ significantly before and after recalibration in both men and women (p>0.3 based on bootstrap resampling).
Conclusion Original FRS has a good discrimination and poor calibration in Iran but considering clinical usefulness, it can be used even without recalibration.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.