Table 5 Predicting the propensity to being susceptible versus immune to any sickness absence (zero-inflation part) and the duration of sickness absence, if susceptible (negative binomial part)
 Zero-inflated part (ZI) Negative binomial part (NB) Coefficient OR (95 CI) Coefficient MR (95 CI) Baseline (odds for ZI, mean for NB) 0.29 0.75 (0.41 to 1.37) 1.60 4.93 (3.45 to 7.04) Male blue collar 1.99 7.30 (4.72 to 11.30) 0.67 1.95 (1.45 to 2.62) Male white collar (ref) 0 1 0 1 Female white collar 1.42 4.13 (2.20 to 7.76) 0.02 1.02 (0.68 to 1.54) Age (years) 1829 0.16 1.17 (0.48 to 2.86) 0.05 1.05 (0.72 to 1.52) 3034 0.21 1.24 (0.52 to 2.93) 0.04 0.96 (0.65 to 1.40) 3539 (ref) 0 1 0 1 4044 0.22 0.81 (0.40 to 1.63) 0.15 0.86 (0.62 to 1.20) 4549 0.68 0.51 (0.26 to 1.01) 0.01 1.01 (0.71 to 1.43) 5054 0.64 0.53 (0.26 to 1.06) 0.09 0.91 (0.65 to 1.27) 5561 0.68 0.51 (0.26 to 1.01) 0.16 1.18 (0.84 to 1.66) Health problems None (ref) 0 1 0 1 One 0.24 1.27 (0.76 to 2.12) 0.62 1.87 (1.44 to 2.42) Two or more 0.92 2.51 (1.35 to 4.68) 1.23 3.41 (2.64 to 4.40)
• Estimated model coefficients, odds ratios (OR) and mean ratios (MR) with 95 CI from fitting a zero-inflated negative binomial regression Model 1 including age, gender, occupational grade and the presence of self-reported health problems as covariates.

• The estimate of the dispersion parameter was 0.56.

• Zero-inflated part refers to the model component for predicting membership to the subpopulation A with high propensity to zero absence, and Negative binomial part to the component predicting the days on sick leave among the susceptible subpopulation B. To facilitate interpretation, for the zero-inflation part we have shown the odds ratios associated with the complementary propensity to having any sickness absencethat is, inclusion in subpopulation B.