Overview

• Moffitt Cancer Center tested Bayesian networks combining clinical records, patient-reported outcomes, and Fitbit-derived heart rate and sleep metrics.
• In the cohort of 58 patients on systemic therapy for non–small cell lung cancer, the combined-data models better distinguished high- from low-risk patients than clinical data alone.
• The approach emphasizes interpretability, letting clinicians see how symptom reports, sleep quality, and lab values contribute to predicted risk.
• Authors say such risk stratification could enable earlier interventions to reduce treatment-related urgent care visits and hospitalizations.
• The study, published in JCO Clinical Cancer Informatics with NIH support, was single-center with a modest sample and will be validated in larger multi-institutional cohorts.