Speaker
Stein Atle Lie, Research leader at Uni Health & Professor II, Institute for Surgical Sciences, University of Bergen.
Abstract
Most of us will during life have one or more incidents of sickness absence, either self-certified or certified by a physician. The duration (length) of the sick leave events will vary, before the individual returns to work. If sick leave is prolonged (1 year), other types of benefits may be relevant (medical rehabilitation, occupational rehabilitation, temporary disability or permanent disability pension).
There is great variation in how sick leave data are analyzed and presented. Time series analyses, linear models and classical survival models are often used. Survival analyses is often used to look at the duration of single sick leave episodes without take into account that sick leave are recurrent events, and that the distance between sick leave (i.e. how long the person stays at work) also varies.
This presentation will discuss how multi-state models can be used to analyze the time course and shifts between the recurrent events work and different social benefits, and events that are wholly or partially absorbing (death and disability pension). Within the framework of multi-state models we demonstrate simple calculations of transition hazards, transition probabilities and state probabilities, as well as suggestions for regression models for hazards and probabilities.
Examples are taken from sick leave data in the time course after a randomized controlled study of 400 individuals, follow-up of 500 people after stay at a rehabilitation clinic, and predictive factors at military enrollment on the life time course of social security benefits, for 900,000 Norwegian men.