Top Line: It’s not uncommon for patients receiving radiation to wind up in the emergency department or hospital.
The Study: SHIELD-RT was a prospective study at Duke that asked two very interesting questions. First, how well can machine learning identify patients at high risk of acute care (ED visit or hospitalization) during a course of radiation? Second, does a more intensive, bi-weekly treatment management schedule reduce the rate of acute care? The machine learning algorithm was trained to identify patients at >10% risk of acute care during radiation therapy based on pre-treatment data and treatment plan. A third of treatment courses were identified as high-risk, and those patients were randomized to standard once-weekly treatment management visits or experimental twice-weekly visits. Interestingly, the rates of concurrent chemo, IMRT use, and treatment duration were comparable between the low- and high-risk cohorts. GI and CNS malignancies were highly represented in the high-risk group, though. Among high-risk patients, the intensive bi-weekly visits resulted in a 10% absolute and 45% relative reduction in the number of patients requiring acute care (22 → 12%) during radiation. In comparison, patients identified as low-risk had a <3% rate of acute care utilization. Most acute care visits were related to neurological or nutritional complications.TBLs: 1) Machine learning effectively identified patients at high- and low-risk of acute care utilization, and 2) twice-weekly treatment management visits reduced acute care utilization among those at high-risk. | Hong, J Clin Oncol 2020