individuals, or any nonfarm injury incurred by a farmer, farmworker, or farm family member
in the course of handling, producing, processing, transporting, or warehousing farm
commodities.” The sample was also categorized based on whether the injury was work-
related or not. An injury was defined as work-related if it occurred at a workplace or during
an activity related to work-function (eg, traveling to a meeting). Of the 113 662 occupational
injuries among rural patients in the trauma registry, there were 3935 (3.4%) agricultural
injuries, 107 728 (94.8%) rural nonagricultural injuries, and 1999 (1.8) missing.
2.2 | Study variables
Patient variables included age, sex, and injury information. Injury variables included
mechanism measured through external cause of injury ICD codes (machinery, transportation,
fall, cut/pierce, struck by/against, and other), type of injury measured through ICD diagnosis
codes (amputation, burn, crushing, dislocation/sprain, fracture, head injury/spinal cord
injury/nerves, internal organ/blood vessels, open wound, other Injury), severity of injury
measured by injury severity score (ISS), length of hospital stay, and hospital charges. ISS is
an anatomically based consensus-driven scoring system that measures injury severity based
on the threat to life in trauma patients. ISS-based ISSs have been validated for predicting
mortality and scores are categorized as mild (ISS = 1–8), moderate (ISS = 9–15), and severe
(ISS = 16+). We further created three age groups: less than 18 years, 18 to 64 years, and 65+
years; and three injury severity groups: minor injuries (ISS = 1–8), moderate injuries (ISS =
9–15), and severe injuries (ISS = 16+).
19
Hospital trauma level based on the American
College of Surgeons levels of I through IV, described above, was collected for each patient.
20
Payer source was the main exposure variable in this analysis. Payer source was collected
through the medical record as part of the trauma registry, and represent payers to which a
claim was submitted. Payer source had 18 categories that were combined into five payer
groups: public (Medicare, Medicare/Medicaid, Medicaid, welfare, other federal government,
other local government, and other state government, CHAMPUS, CHAMPUS/VA); private
(HMO, PPO, self-insured, auto insurance, commercial insurance); uninsured (charity, no
charge, and self pay); other (research fund, teaching fund, victim’s fund, other), and
workers’ compensation as a separate fifth category. Employer-provided health insurance
would be classified as private insurance. An injury hospitalization could be billed to five
separate payer sources. Of the total sample size (113 662), 22.9% and 81.3% had unknown/
missing responses for the first payer and second payer sources, respectively. Because a
missing response on the second payer source could mean there was no other payer, we
limited our analysis to the primary payer source (first payer). However, we used workers’
compensation information from any of the five payer fields to ensure complete information
on workers’ compensation payer. Workers’ compensation was the second payer source for
fewer than 3% of cases.
2.3 | Statistical analysis
The data showed a relatively high percentage of missing data (22.9%) on the main exposure,
primary payer source. Missingness was differential by work-relatedness of injury (main
outcome), patient age, sex, injury severity, type of injury, hospital level, and calendar year,
Missikpode et al.
Page 4
Am J Ind Med
. Author manuscript; available in PMC 2020 January 06.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript