Background Short and long term sickness absence (SA) are due to different causes, therefore require different approaches to intervention. In research on SA, incidence had usually been analysed separately from the duration of episodes.

Objective Describe the usefulness of the analysis of SA incidence by segmentation of duration of episodes versus non-segmented analysis.

Methods We compared overall incidence of non-work related SA and incidence of SA segmented by duration (short duration (≤ 15 days), medium (16–30, 31–90) and long (> 90)). We used data from a health insurance company (“mutua”) in Spain during 2011 (230,332 episodes, 752,906 workers). The incidence rate was calculated according to age groups (<25, 25–34, 35–44, 45–54, ≥55). We computed rate ratios, with their 95% confidence interval (95% CI), to quantify linear trend (RR-T) and association (RR) of SA between age groups.

Results The overall incidence of SA showed a steady decline with age (RR-T = 0.92, 95% CI = 0.91–0.92) with an incidence of 36.4% in <25 years versus 29.3% in ≥55 years (RR = 1.24, 95% CI = 1.22–1.27). The overall excess incidence observed in young is produced by short duration (RR-T = 0.82, 95% CI = 0.82–0 82; RR = 1.98, 95% CI = 1.93–2.03). In addition, the incidence of long term SA changed this pattern (RR-T = 1.32, 95% CI = 1.30–1.33), being more frequent in ≥55 years relative to the youngest (5.5% versus 2.1%, RTI = 3.19, 95% CI = 2.97–3.43).

Conclusions The analysis of indicators of SA segmented by duration of episodes provides more precise image than that obtained from global data. Information on the type of episodes of IT according to duration provides valuable information for SA intervention.