Table 3

Performance of sickness absence (SA) prediction models
Development setting Validation setting
Fixed coefficients Re-estimated coefficients Gender inclusive SRH exclusive
SA days model
Regression coefficients (SEa)
Age −0.016 (0.015) −0.016 (0.015) −0.016 (0.014) 0.004 (0.014) −0.001 (0.014)
Prior SA 0.007 (0.001) 0.007 (0.001) 0.003 (0.002) 0.003 (0.002) 0.004 (0.002)
Self-rated health −0.718 (0.244) −0.718 (0.244) −0.356 (0.170) −0.349 (0.173) not included
Gender not included not included not included 0.699 (0.269) not included
Predictive performance
Nagelkerke’s pseudo R2 0.12 0.03 0.03 0.05 0.02
Discrimination (AUCb) 0.73 0.65 0.68 0.68 0.65
Calibration (slope) 0.94 0.89 0.87 0.86 0.86
SA episodes model
Regression coefficients (SEa)
Age −0.043 (0.016) −0.043 (0.016) −0.039 (0.015) 0.008 (0.015) 0.005 (0.015)
Prior SA 0.472 (0.070) 0.472 (0.070) 0.465 (0.067) 0.473 (0.068) 0.477 (0.065)
Self-rated health −0.715 (0.255) −0.715 (0.255) −0.190 (0.185) −0.187 (0.188) not included
Gender not included not included not included 0.463 (0.256 not included
Predictive performance
Nagelkerke’s pseudo R2 0.32 0.18 0.21 0.22 0.21
Discrimination (AUCb) 0.83 0.76 0.78 0.78 0.77
Calibration (slope) 0.98 0.96 0.98 0.95 0.98

a standard error; barea under the receiver operating characteristic curve.

The table shows the regression coefficients and performance measures in a development sample of 535 health care workers and the current validation sample of 593 office workers.

Roelen et al.

Roelen et al. BMC Public Health 2013 13:105   doi:10.1186/1471-2458-13-105

Open Data