Re: Amelsvoort et al. "Approaches for predicting long-term sickness absence".

We would like to thank Van Amelsvoort et al (1) for the interest in our study (2) and take the opportunity to clarify here that none of the workers were sick-listed when they participated in the baseline health survey. We mentioned in the abstract that incident (ie, not prevalent) long-term sickness absence was retrieved from an occupational health register (2). Our explanation of how to interpret the area under the receiver operating characteristic (ROC) curve as measure of discrimination between workers with and without long-term sickness absence might have given the impression that the study population was a mix of workers with and without sickness absence.