Unsupervised machine learning can assign older adult patients to low, moderate, or high prediagnosis symptom severity categories, allowing healthcare providers to identify patients at risk for hospitalization or death even before initiating treatment, researchers reported in JAMA Network Open.