Discrimination and calibration on external validation of the original prognostic index to the Amsterdam II and SMILE cohorts.

a(p)FAs: (persistent) Frequent Attenders: frequently attending patients during 1 and 3 years, respectively. bConcordance statistics (95% confidence interval). cFemale sex; any medically unexplained symptoms; any psychoactive medication; mean monthly number of prescriptions of antibiotics. dIdeally, the slope should be 1, which indicates perfect calibration of predicted and observed risks. Values <1 indicate overoptimism (shrinkage), that is, high risks are overestimated, while low risks are underestimated. #using a model with shrunken coefficients of the original model (shrinkage coefficient 0.993). Sensitivity, specificity, likelihood ratios and positive and negative predictive values were calculated at the value where their sum was maximal (Q-point of the ROC curve).