Genetic improvement of dairy cattle health using producer-recorded data is feasible. Estimates of heritability are typically low, indicating that genetic progress will be slow. Health improvement may also be possible through incorporation of environmental and managerial aspects into herd health programs. The objective of this research was to use the more than 1,100 regularly recorded herd characteristics to benchmark cow health status. These data were combined with producer-recorded health event data. Non-parametric models including support vector machines and random forests were used to predict and benchmark cow health status. Random forest models attained the highest accuracy for predicting health status in all health categories when evaluated by 10-fold cross validation. Accuracy of prediction (SD) of random forest models ranged from 0.87 (0.06) to 0.93 (0.001). Results of these analyses indicate that non-parametric algorithms, specifically random forest, can be used to accurately identify cows likely to experience a health event of interest. Further development of predictive models into herd management programs will continue to improve dairy health.