MACHINE learning can help predict urgent care visits in patients with non–small cell lung cancer receiving systemic therapy. Researchers found accuracy improves when wearable and patient-reported data are included.
Patients receiving systemic therapy for NSCLC often experience treatment-related toxicities that lead to emergency or urgent care visits, negatively affecting outcomes and straining healthcare resources. To address this, researchers tested the use of Bayesian Network models, incorporating multidimensional data to anticipate when patients may require urgent care during the first 60 days of treatment.
In the study, 58 patients with NSCLC completed PROMIS-57 quality-of-life surveys and wore Fitbit devices to provide continuous patient-generated health data. Demographic and clinical information was extracted from medical records, and machine learning algorithms were applied to develop predictive models.
Initial models using only demographic and clinical data achieved moderate predictive accuracy, with area under the curve (AUC) values of 0.72 before systemic therapy and 0.81 during treatment. When patient-reported outcomes and wearable sensor data were added, performance improved substantially. The final model reached an AUC of 0.86, representing a significant gain in predictive accuracy.
These findings demonstrate that combining clinical characteristics with real-time data from patients and wearables can provide more robust insights into healthcare utilization patterns. The explainable Bayesian Network approach also helps clinicians understand the interactive factors that influence urgent care visits, offering opportunities for proactive intervention.
The study underscores the potential for integrating multidimensional data into predictive models to improve supportive care for patients with lung cancer. By identifying patients at higher risk of urgent care visits early in therapy, clinicians may be able to intervene before toxicities escalate, reducing unplanned healthcare use and improving the quality of cancer care delivery.
Reference: Gonzalez BD et al. Using Bayesian Networks to Predict Urgent Care Visits in Patients Receiving Systemic Therapy for Non–Small Cell Lung Cancer. JCO Clin Cancer Inform. 2025;9.
