Capture-Recapture Estimates of the Undercount of Workplace Injuries and Illnesses: Sensitivity Analysis

BackgroundIn recent years, capture-recapture methods have been applied to state Survey of Occupational Injuries and Illnesses (SOII) and workers' compensation data, but uncertainties arise because of population differences between the two sources, likely source dependence, and concerns about the accuracy of case linkage. MethodsWe linked SOII and workers' compensation records for California and used capture-recapture methods to estimate the proportion of injuries and illnesses involving at least 4 days away from work captured by each source. We then did a sensitivity analysis. ResultsAssuming source independence, estimates of the proportion of injuries and illnesses involving at least 4 days away from work captured by the SOII varied from 42.4% to 49.0%, while workers' compensation estimates were between 76.9% and 77.6%. Re-estimating SOII capture rates assuming source dependence (OR=3) reduced capture estimates substantially. ConclusionsEstimated capture rates remained low after changing several of the underlying assumptions. Positive source dependence had the greatest impact. Am. J. Ind. Med. 57:1090-1099, 2014. (c) 2013 Wiley Periodicals, Inc.