Use of Capture-Recapture Analysis to Estimate the Number of Agricultural Fatalities and Severe Agricultural Injuries in Indiana

Abstract

ABSTRACT Conducting surveillance of agricultural injuries and fatalities in the United States has been an ongoing challenge, with many cases falling outside the criteria of national and local surveillance systems. In this research, capture-recapture analysis was used to estimate the number of fatal agricultural injuries in Indiana between 2016 and 2020. A limited analysis of non-fatal injuries is also provided. This analysis was possible because of two publicly available datasets containing incident descriptions with sufficient detail for case matching. The first dataset consisted of summary lists of fatal and nonfatal agricultural injuries in Indiana published in annual agricultural fatality reports produced by the Purdue Extension. The second data source was AgInjuryNews, which gathers reports of agricultural injuries and fatalities published in news media and other publicly available sources. Results of the capture-recapture analysis estimate that, every year in Indiana, the Purdue Extension misses 18% of fatal incidents and AgInjuryNews misses approximately 60%. AgInjuryNews identifies approximately 3 fatal incidents per year that are missed by Purdue Extension. Analysis of nonfatal incidents was limited by the fact that both data sources only included nonfatal injuries that were extremely severe and/or connected to a fatality. The Purdue Extension is estimated to miss 22% and AgInjuryNews is estimated to miss 25% of nonfatal agricultural injuries meeting that narrow definition. While capture-recapture analysis only provides estimates of true injury rates, the results provide evidence that Purdue Extension’s surveillance captures most agricultural fatalities in the state. AgInjuryNews has been able to identify cases missed by Purdue, and this research takes an important step forward in quantifying how media reports found in this data source differ from extension surveillance. This research also highlights the continuing limitations in the surveillance non-fatal injuries and the ways in which publicly available data can aid researchers in filling gaps in surveillance.