Machine learning models using step count data of the two preceding weeks were able to predict hospitalization in the upcoming week for patients during systemic therapy, but not other adverse events.