JMIR Publications

ABSTRACT

Background:

Agricultural injury surveillance remains challenging because existing data systems do not comprehensively capture nonfatal injuries, particularly among self-employed workers, unpaid family members, and other populations involved in agricultural activities. Emergency medical services (EMS) records represent a potentially valuable but underutilized source of surveillance data. Integrating EMS records with temporal and spatial statistical methods may improve the identification of injury patterns and geographic areas with elevated injury burden.

Objective:

This study aimed to develop and apply a statewide EMS-based surveillance framework to characterize agricultural injuries, examine temporal patterns, estimate small-area agricultural injury rates, and identify areas with elevated injury burden in Florida.

Methods:

A retrospective surveillance study was conducted using Florida Emergency Medical Services Tracking and Reporting System (EMSTARS) records from 2016 through 2023. Agricultural injuries were identified using International Classification of Diseases, Tenth Revision (ICD-10) external cause of injury and incident location codes. Generalized additive models (GAMs) were used to evaluate seasonal variation and long-term temporal trends in monthly injury counts. To estimate agricultural workforce populations at the ZIP Code Tabulation Area (ZCTA) level, county-level agricultural worker counts from the 2022 Census of Agriculture were redistributed using a dasymetric approach informed by the National Land Cover Database classifications and a negative binomial GAM. A Bayesian spatial negative binomial model incorporating Besag-York- Mollié 2 (BYM2) random-effects was then used to estimate small-area injury rates and identify ZCTAs with at least 95% posterior probability of exceeding the statewide background injury rate.

Results:

A total of 1133 agricultural injury incidents requiring EMS response were identified during the study period. Most injured patients were male (n=852, 72.8%) and 85 (7.5%) were younger than 18 years. Contact with agricultural machinery was the most frequently recorded external cause of injury (n=541, 47,7%). Temporal analysis identified a recurring seasonal pattern, with expected monthly injury counts approximately 60% higher during the June-July peak than during the December minimum. Dasymetric redistribution produced ZCTA-level agricultural workforce estimates, with higher concentrations in south-central Florida near Lake Okeechobee. Bayesian spatial analysis identified 94 ZCTAs with elevated agricultural injury rates relative to the statewide background, with the largest concentration occurring in south-central Florida.

Conclusions:

Statewide EMS records can be integrated with temporal modeling, land cover-based workforce estimation, and Bayesian spatial analysis to support agricultural injury surveillance beyond descriptive case identification. This framework enables the identification of seasonal patterns and geographic areas with elevated injury rates while accounting for estimated populations at risk and spatial variability.