BACKGROUND: Currently, there are over 2,000 known work-related hazards that can lead to numerous occupational respiratory diseases (ORD). ORD surveillance is a crucial means to identify diseases, exposures, and emerging hazards, and ultimately informs occupational disease prevention efforts. Previously, Washington State (WA) conducted surveillance for work-related asthma (WRA) by identifying cases through a basic keyword search on the WA workers’ compensation (WC) Report of Industrial Injury or Occupational Disease form (ROIID). The two aims of this study were to (1) enhance the current WRA surveillance method and compare a new case-capture algorithm to the previous strategy and (2) create and evaluate surveillance methods for additional ORDs including silicosis, Legionnaires’ disease, and Coccidioidomycosis.
METHODS: We used WA’s WC data to identify potential ORD cases from 2011-2015. Claimant information, diagnosis (ICD) codes, and illness narratives were extracted from the ROIID. Diagnosis ICD codes were also extracted from medical billing records. We created individual algorithms for each ORD disease using claimant information, keyword searches, respiratory ICD codes, and Occupational Injury and Illness Classification (OIICS) codes. Cases identified in the final disease algorithms were manually reviewed to determine the predictive positive value and compared to case estimates for each ORD in WA.
RESULTS: A total of 44,377 claims were identified as having either a keyword (15%), an ICD code (81%), and/or an OIICS code (33%) indicating occupational respiratory disease. The enhanced asthma-specific algorithm included additional keywords, common misspellings, and OIICS nature codes. The enhanced WRA algorithm captured 98% of the cases previously identified and identified a total of 1,174 potential cases. Manual review confirmed the enhanced WRA algorithm had a positive predictive value of 57.5%. The silicosis algorithm identified 44 cases of potential silicosis, including 5 that had diagnostic evidence consistent with silicosis. Additional variables including county, occupation, and common preliminary diagnoses were essential to the Coccidioidomycosis and Legionnaires’ algorithms in order to reflect those disease profiles in WA.
CONCLUSIONS: Components of each algorithm varied greatly based on disease characteristics and manual review will remain an integral process to case identification. While manual review can be time intensive, accurate case counts for each disease provides a needed baseline and will be vital to WA’s long-term ORD surveillance efforts. The iterative algorithm development process will help inform the creation of additional algorithms for less common ORDs to better describe their morbidity and mortality for Washington workers.