Inter-method agreement of self-reports compared to register data collection for return to work and sickness absence: a systematic review and meta-analysis

To determine the inter-method agreement of self-reported versus register data of 'sickness absence' (SA) and 'return- to-work' (RtW) outcome measurements.We conducted a systematic review and a meta-analysis of studies reporting mean differences and sensitivity and specificity for self-report versus register data and an inductive analysis of the self-report question formulations. An information specialist searched Medline, Embase, PsycINFO for studies published from inception to November 2022. Screening and data extraction was done by two authors independently.Twenty-three studies were included of which eighteen with an overall high risk of bias. Self-reports had a pooled mean difference of 1.84 sickness absence days (95% CI 0.26 to 3.41, I2 98%, 18 studies, 38716 participants) Compared to registries which varied among studies from 204 more to 17 days less. The median average sick leave in studies in the self-report group was 8 days (IQR 4 -23 days). Being absent from work measured with self-report had a sensitivity of 0.83 (0.76-0.88 95% CI) and a specificity of 0.92 (0.88 -0.94 95% CI) compared to registry data. The high heterogeneity amongst the studies could not be explained by recall time, gender, register type, prospective or retrospective self-reports, health problem, sickness absence at baseline or risk of bias. Studies lacked standard outcome reporting, had unclearly formulated questions in self-reports and there was little information on the registers' quality.Current self-reports may differ from register-based absence data but in an inconsistent way. Due to inconsistency and high risk of bias the evidence is judged to be of very low certainty. Further research is needed to develop clear standard questions which can be used for sickness absence and return to work self-reports. Quality of registers needs to be better evaluated. Percentage positive and negative agreement, mean differences and 2x2 tables should be reported for studies investigating agreement between sickness absence and return to work outcome measures.