Issue Index
Steps Toward Predicting Hospitalization During Chemoradiotherapy
A new machine learning (ML) model that analyzes the number of daily steps taken by patients being treated with chemoradiotherapy can help identify their level of risk for unplanned hospitalization. This potentially practice-changing tool that combines artificial intelligence with wearable devices was developed through a collaboration between Julian Hong, MD, MS, and Nitin Ohri, MD, MS. They recently published results of their joint research in “Machine Learning–Based Prediction of Hospitalization During Chemoradiotherapy With Daily Step Counts” in JAMA Oncology.The study, led by Izzy Friesner, used data from three prospective trials of activity monitoring during chemoradiotherapy that were conducted by Dr. Ohri and his colleagues at Montefiore Einstein Comprehensive Cancer Center, in the Bronx, NY. The patients had a variety of cancers and were provided with the wearable devices, which they wore for throughout their courses of curative-intent concurrent chemoradiotherapy.
Ms. Friesner and Dr. Hong led the development of the ML models to predict hospitalization. They compared the performance of ML models based on step counts only, clinical features (performance status, age, sex, etc.) only, and step counts and clinical features combined. The model using step counts only was the best at predicting patients’ risk of unplanned hospitalization, although the combined model was similar. Both models that included step counts were better than the model that relied on clinical features alone, indicating that activity metrics could dramatically enhance clinical care during chemoradiotherapy. Soon, ML approaches using step counts could be used to direct supportive care to reduce unplanned hospitalizations during chemoradiotherapy.
This new article is the latest outcome from Dr. Hong and Dr. Ohri’s collaboration, which has its roots in ROI grants awarded independently to each investigator. The support served as a catalyst for their ongoing partnership that combines technologies that are transforming health care and promises to improve outcomes for cancer patients.
