Predictors of early retirement and rehabilitation for use in a screening to detect workers in need of rehabilitation

In the German pension insurance sector a screening is being suggested to ensure the participation in rehabilitation measures of workers at risk of health-related early retirement. Screening presupposes empirical indicators of rehabilitation needs. A study is presented that determined predictors of early retirement and rehabilitation from longitudinal data for use in a screening for the selection of workers likely to be in need of rehabilitation. We gathered longitudinal data by conducting a second survey with a cohort for which the first survey had delivered clinical findings, lab. values, medical diagnoses and self reports regarding morbidity, medication, health-related behaviour, family-and occupational-related strains, and sociodemographic information (first survey T0: 1975/76, n = 3.968; second survey T1: 1992/93, n = 28.463). The survey of T1 also comprised inquiries of the pension insurance institutions concerning the retirement and rehabilitation status for pension-insured study subjects (n = 1.794). Based on these subjects, using multi-and bivariate regression analysis, we determined those T0 variables which were significantly related to the events of early retirement (98 cases/357 controls), rehabilitation (127 cases/200 controls) and early retirement or rehabilitation (185 cases/270 controls) in the period T0-T1. The significant T0 variables were subsequently used for the definition of a selection index which measured rehabilitation need by a simple sum score (number of significant T0 variables present). We tested the discriminative power of this index for a subsample of the cohort (cases who retired early or underwent rehabilitation and controls). The index classified 68% of the cases correctly. The sensitivity reached 57% and the specificity 76%. In connection with this result, the long prognostic time interval (up to 17 years) has to be considered. In the case of screening the preselection of workers via the index would occur at the same time as the medical assessment of the actual need for rehabilitation. An earlier study showed that this would raise sensitivity and specifity of an index based on predictors of early retirement substantially. (C) 1998 Chapman & Hall Ltd.