Using a Neural Network Analysis to Assess Stressors in the Farming Community.

In the 1980s and 1990s, with decreasing numbers of full-time farmers and adverse economic conditions, chronic stress was common in farmers, and remains so today. A neural network was implemented to conduct an in-depth analysis of stress risk factors. Two Colorado farm samples (1992-1997) were combined (n = 1501) and divided into training and test samples. The outcome, stress, was measured using seven stress-related items from the Center for Epidemiologic Studies-Depression Scale. The initial model contained 32 predictors. Mean squared error and model fit parameters were used to identify the best fitting model in the training data. Upon testing for reproducibility, the test data mirrored the training data results with 20 predictors. The results highlight the importance of health, debt, and pesticide-related illness in increasing the risk of stress. Farmers whose primary occupation was farming had lower stress levels than those who worked off the farm. Neural networks reflect how the brain processes signals from its environment and algorithms allow the neurons "to learn". This approach handled correlated data and gave greater insight into stress than previous approaches. It revealed how important providing health care access and reducing farm injuries are to reducing farm stress.