Comparisons of self-reported and register data on sickness absence among public employees in Sweden

Self-reported assessments of sickness absence are often performed in epidemiological studies. The objective of this study was to compare the number of sick-leave days according to self-reported data over 12 months with data from the employer's register for the same period. An additional aim was to ascertain whether the self-reported information and the recorded data would show equivalent associations with self-reported general health.

The study was based on a cohort of 4869 municipal employees in Sweden, about 80% women, who answered a questionnaire in 2001-2. The responses provided by the employees included information on number of sick-leave days and self-rated health. Data on sick-leave days, occupation and age were derived from the employers' computerised registers. The questionnaire information on sick-leave days was compared with the corresponding information retrieved from the employer register by means of calculating sensitivity and specificity, using the employers' data as the "gold standard".

The annual number of sick-leave days was lower according to the self-reported information than to the register data. For women the agreement between the two sickness absence measures for no sick-leave days, 1-7 days and >/=28 days were 74%, 72% and 67%, respectively. The sensitivity of questionnaire versus register information regarding any self-reported sick-leave day was 91% and the specificity was 74%. Sensitivity and specificity for sickness absence >/=28 days were 67% and 98%, respectively. The results for men were similar to those for women. Self-reported and recorded sickness absence were both associated with self-rated health. The odds ratios were 7.27 and 8.25, for subjects with >/=28 recorded and self-reported number of sick-leave days respectively, compared to subjects with no sickness absence.

Good agreement was found between self-reported and register information on sickness absence. Self-reported data on sickness absence may be useful in common epidemiological applications.