350 – Administrative Data, Record Linkage, and Latent Class Models
Comparing Injury Data from Administrative and Survey Sources: Methodological Issues
Nicole Nestoriak
Bureau of Labor Statistics
Brooks Pierce
Bureau of Labor Statistics
The Bureau of Labor Statistics' Survey of Occupational Injuries and Illnesses (SOII) is an important source of information for workplace injuries. Recent work comparing the SOII to Workers' Compensation (WC) administrative data concludes that the SOII undercounts injury and illness cases. Because the SOII is a sample, case by case comparisons between the two sources must distinguish between WC cases which are missing from the SOII because of sampling and cases which are missing because of underreporting. Previous research made this distinction by identifying SOII sampled establishments within WC data. This approach requires accurate employer information in WC data and subjective analysis of an establishment match. As an alternative, after matching SOII and WC data at the case level, we estimate the number of linked cases for the population by applying survey weights to the linked cases in the SOII sample. This allows us to estimate as a residual the number of WC cases missing in SOII, without having to directly identify SOII establishments in WC data. We describe the relative merits of this approach, and provide an alternative measure of the SOII undercount using Kentucky data.
