Capture-recapture estimates of workplace injury rates.
Atlanta, GA: U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health, R01-OH-007596, 2007 Jan; :1-18
This study used the two most widespread and comprehensive sources of nonfatal occupational injury and illness reporting to evaluate completeness of reporting and factors that affect differential reporting. These sources are state workers' compensation (WC) data and the Survey of Occupational Injuries and Illnesses (SOII) conducted annually by the Bureau of Labor Statistics (BLS). For 1998-2001, we compared reported injuries from these two sources by matching individual lost-time injuries and illnesses in 6 states: MN, NM, OR, WA, WI, and WV. We used probabilistic matching methods to link data from the two sources. We then used capture-recapture analysis for each of these states to develop improved measures of nonfatal injury incidence. Logistic regression was used to account for differential capture by employer, injury, and worker characteristics. This helped to identify factors associated with underreporting. Then, we compared the degree to which the two data systems capture occupational injuries and illnesses in the these states. Assuming that biased estimates of underreporting toward zero (i.e., reporting from each source is independent), we found substantial underreporting in both WC and the BLS survey, with the exception of reporting of lost-time cases in WA and WV. The state median of estimated underreporting in WC was 25%, and the median in the BLS survey was about 40%. The range in WC was from 5% to 37%, and the range in the BLS survey was from 23% to 47%. We tested the sensitivity of these estimates to the assumption of source independence. Assuming that the odds of a case being reported to a WC system are 5 times higher if it is reported to the BLS than if it is not, then the state median of estimated underreporting in WC was 45%, and the median in the BLS survey was about 50%. We also found that simply counting the cases that appear in either system substantially improves the overall completeness of reporting in most states. On the assumption of source independence, the minimum ascertainment in any state was 86%. An important limitation of this study is that injuries or illnesses that employers are unaware of will be absent from both the WC and BLS survey data. This may happen because workers are concerned about the stigma of claiming an injury due to concern about employer retaliation, or because both worker and employer are unaware of the work-relatedness of the condition. This makes capture-recapture analysis particularly inappropriate for determining underreporting of chronic occupational diseases.
Surveillance-programs; Injuries; Statistical-analysis; Analytical-methods; Analytical-processes; Quantitative-analysis; Workplace-monitoring; Workplace-studies; Disease-incidence; Epidemiology
Leslie I. Boden, Ph.D., Department of Environmental Health, Boston University School of Public Health, 715 Albany Street, Boston, MA 02118
Final Grant Report
NTIS Accession No.
National Institute for Occupational Safety and Health
Boston University Medical Campus