Parametric hazard rate models for long-term sickness absence.

PURPOSE In research on the time to onset of sickness absence and the duration of sickness absence episodes, Cox proportional hazard models are in common use. However, parametric models are to be preferred when time in itself is considered as independent variable. This study compares parametric hazard rate models for the onset of long-term sickness absence and return to work. METHOD Prospective cohort study on sickness absence with four follow-up years of 53,830 employees working in the private sector in the Netherlands. The time to onset of long-term (>6 weeks) sickness absence and return to work were modelled by parametric hazard rate models. RESULTS The exponential parametric model with a constant hazard rate most accurately described the time to onset of long-term sickness absence. Gompertz-Makeham models with monotonically declining hazard rates best described return to work. CONCLUSIONS Parametric models offer more possibilities than commonly used models for time-dependent processes as sickness absence and return to work. However, the advantages of parametric models above Cox models apply mainly for return to work and less for onset of long-term sickness absence.