Hazard functions to describe patterns of new and recurrent sick leave episodes for different diagnoses

Abstract

Objectives This study aims to identify the hazard functions that describe the occurrence patterns of new and recurrent sick leave (SL) episodes for mental, respiratory, and musculoskeletal diagnoses. Methods The data come from a cohort of workers in the Hospital das Clínicas da Universidade Federal de Minas Gerais, Brazil, including all employees working ≥20 hours per week, whose first employment relation with the hospital started between 1 January 2000 and 31 December 2007 (N=1579). We created 15 samples corresponding to combinations of diagnoses causing SL and the number of previous episodes already suffered. We fitted Weibull, log-normal, and log-logistic models by resampling and selected the model having the lowest Akaike information criterion in the greatest number of resamples. Results Differences were observed in the probability distributions associated with the process generating a SL. Diagnosis showed important differences in terms of risk intensity: mental episodes were the least frequent. There were differences in risk intensity and shape of the function over time depending on the episode number, particularly between the first episode and recurrences. In addition, these differences varied by diagnosis. Conclusions In most of the samples analyzed, we identified a mixture of distributions, implying a need to revise the statistical methods of analysis for SL occurrence with the aim of obtaining consistent estimates of the risk and the associated factors.