Background: The Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) was commissioned by the government to evaluate healthcare interventions to manage individuals on, or at risk of, long-term sickness absence. Inclusion was restricted to studies evaluating intervention effects on return-to-work (RTW).
Objectives: The aim of this substudy was to describe how the effects on RTW were synthesized by a synthesis without meta-analysis.
Methods: A systematic literature review was performed to identify well-performed randomised controlled trials investigating intervention effects on RTW.
Results: Most of the 68 included unique studies based their RTW outcomes on sickness absence register data. Despite this, there was considerable heterogeneity in how RTW was defined, operationalized, estimated and presented in the publications. Furthermore, it was common that authors presented several RTW outcomes, using different models on the same data. Meta-analysis was not considered feasible; thus, a synthesis without meta-analysis was performed including all reported RTW results. For each study, the number of statistically significant results in favour of the intervention or the control was first assessed. Thereafter, experts in the field evaluated each study result’s potential clinical significance, based on the magnitude of statistically significant effects. Experts then formulated summarized results, weighing the evidence of favourable effects against the evidence of nonfavourable or nonpresent effects, to determine whether the overall results suggested directional effects. The GRADE domains were considered for summarized results, and uncertainty arose mainly because of risk of bias and imprecision. The certainty of results was assessed as being very low to low, and results were presented according to GRADE’s recommendation of informative statements. The advantages of this method of synthesis include the comprehensiveness when summarizing all results in the included studies. The disadvantages involve loss of precision, affecting the possibility to summarize the magnitude of effects with associated uncertainty.
Conclusions: The considerable heterogeneity in results reporting effects of interventions aiming at RTW might be handled by performing synthesis without meta-analysis. However, it is desirable that the results of RTW interventions are reported in a manner that allows meta-analysis to be performed. The new core outcome set in the field might facilitate this.