Prediction of long-term and frequent sickness absence using company data

<span class="paragraphSection"><div class="boxTitle">Abstract</div><div class="boxTitle">Background</div>More insight into predictive factors is needed to identify employees at risk for future sickness absence. Companies register potentially relevant information regarding sickness absence in their human resources and work schedule administration.<div class="boxTitle">Aims</div>To investigate which combination of administrative company data best predicts long-term and frequent sickness absence in airline employees.<div class="boxTitle">Methods</div>Socio-demographic and work-related variables between 2005 and 2008 were retrieved from the administrative data of an airline company. Logistic regression analyses were used to build prediction models for long-term (>42 consecutive days) and frequent (more than three episodes) sickness absence in 2009. Both models were internally validated.<div class="boxTitle">Results</div>Data on 7652 employees were available for analysis. Long-term sickness absence was predicted by a combination of higher age, recent pregnancy, having a parking permit, having ‘aggravated working conditions’ and previous sickness absence. Recent marriage appeared to reduce the risk. Frequent sickness absence was predicted by being single, not having children of 16 years and older, not having a company parking permit, no shift work, having a job with special operational requirements and previous sickness absence. The long-term and frequent sickness absence models had a discriminative ability of 0.72 and 0.73, and an explained variance of 10.9 and 14.2%, respectively.<div class="boxTitle">Conclusions</div>The results show that it is possible to compose prediction models for employees at risk of sickness absence using only administrative company data. However, as the explained variance was low, additional factors should be identified to predict risk of future sickness absence.</span>

http://ift.tt/2pwnRZW