This research examines how arbitrators consider accommodations for employees with physical and mental illnesses. Unlike other recent research on the subject, the authors specifically and purposely draw their sample from recent US arbitration cases—2015 to 2018, n = 209. Additionally, using content analysis software, NVivo, the case characteristics were autocoded, and the case outcomes were manually coded. Using logistic regression, the following model was developed to predict the odds of case outcomes: disability, injury, discrimination, retaliation, absence and reinstatement. The Cox and Snell and Nagelkerke analysis indicates that our model accounts for approximately 15.6 to 21.5 per cent of the variance, with 33.3 per cent of the individual and split arbitration cases outcomes and 91.2 per cent for organisation arbitration cases correctly predicted. The model predicts 71.2 per cent of the cases.