Application of data linkage techniques to Pacific Northwest commercial fishing injury and fatality data

Abstract Background Commercial fishing consistently has among the highest workforce injury and fatality rates in United States. Data related to commercial incidents are routinely collected by multiple organizations which do not currently coordinate or automatically link data. Each data set potential generate a more complete picture inform prevention efforts. Our objective was examine utility of using statistical linkage methods incident when personally identifiable information is available. Methods In this feasibility study, we identified true matches discrepancies between de-identified sets Python Record Linkage Toolkit. Four from Oregon Washington were linked: Fishing Incident Database, Vessel Casualty Nonfatal Injuries Trauma Registry. The each covered different date ranges within 2000–2017, containing 458, 524, 184, 11 cases respectively. Several classifiers evaluated. Results Naïve-Bayes classifier returned number these small sets. A total 41 8 close identified, 29 determined be duplicates. addition, highlighted 4 records that Washington. optimum match parameters date, state, vessel official number, people on board. Conclusions Statistical enables accurate, routine matching for such as those fishing. It provides needed improve accuracy existing records. also expanding sharpening details individual support occupational safety research.