Assessing the Propensity for Presenteeism with Sickness Absence Data

Presenteeism occurs when an employee attends work while sick or unwell. It is a major Human Resource and organizational issue: in addition to productivity losses, presenteeism is believed to increase sickness absence and decrease self-rated health. However, by its very nature, presenteeism cannot be monitored in the same manner as sickness absence. We show how the probability of presenteeism can be estimated from simple absence data by means of a zero-inflated binomial regression analysis (ZINB). The approach is validated on a Danish data set that contains self-reported sickness absence and presenteeism, whereas causality and reliability are verified by conducting Monte-Carlo simulations. The objective of paper was to explore how far the traditional but costly tool used to assess presenteeism behaviour, a questionnaire, could advantageously be replaced by a statistical approach that relies on easily available information on sickness. We show that the ZINB model captures presenteeism well via the inflation process and delivers insight on both absenteeism and presenteeism. Using Monte Carlo simulations, we further highlight that the model can be used to compute a global indicator, propensity for presenteeism, even when important assumptions are violated.