Innovative Machine Learning Approach Predicts Urgent Care Needs in Lung Cancer Patients

Researchers at the H. Lee Moffitt Cancer Center & Research Institute have developed advanced machine learning models to predict urgent care visits among patients undergoing treatment for non-small cell lung cancer (NSCLC). Published in JCO Clinical Cancer Informatics in September 2025, this study highlights how integrating patient-reported outcomes and wearable sensor data can enhance predictive accuracy.

The study involved 58 patients monitored with Fitbit devices and surveyed through questionnaires. The researchers used explainable Bayesian networks to build predictive models that incorporate clinical data, symptom reports, sleep quality, heart rate, and other wearable metrics. The models demonstrated a significant improvement over traditional clinical data-based predictions, effectively distinguishing patients at high risk for treatment-related toxicities requiring urgent care.

"By combining real-time symptom reporting with continuous wearable device monitoring, we can identify at-risk patients earlier," explained lead researcher Dr. Brian D. Gonzalez. "This approach enables healthcare providers to intervene proactively, possibly preventing hospitalizations and improving patient experiences."

The findings suggest that such multidimensional data integration into machine learning models holds promise for personalized cancer care. The models' transparency, facilitated by explainable AI techniques, helps clinicians trust and understand the predictions, which is crucial for clinical decision-making. Future research aims to expand these models to include molecular data and validate findings across larger, multi-institutional cohorts.

While preliminary, this innovative approach underscores the potential of combining patient-generated data and machine learning to improve outcomes for lung cancer patients during systemic therapy.