The number of women choosing agriculture as an occupation is increasing. Agriculture is dangerous work, and women are at risk of serious injury, but the research on injuries in females is sparse. Women perform different types of farmwork and have different exposures than men. Studies have not assessed injury in a large group of female agricultural operators. In this study, we used XGBoost, a machine learning algorithm, and logistic regression to examine 17 factors hypothesized to be associated with injury in 1529 farm and ranch women. The sample was split into a training group of 1070, and the results were replicated in a test group of 459. The model accuracy was 88%. We compared the results of XGBoost to those of the logistic regression models and computed odds ratios to estimate effect sizes. We found that the two methods generally agreed. XGBoost identified the total number of musculoskeletal symptoms, age, sleep deprivation, high work-related stress, and exposure to respiratory irritants as being important to injury. The multivariate logistic regression model identified higher income, higher stress, younger age, and number of musculoskeletal symptoms as being significantly associated with injury. The analysis highlights the importance of musculoskeletal disorders and work strain to injury in women.
Read full abstract