WCLC 2025 - Posters & ePosters-P2.15 .10 Improving Patient-Reported Outcome Symptom Monitoring Using Bayesian Networks

Pdf Summary

This study aimed to optimize patient-reported outcome (PRO) symptom monitoring in lung cancer patients by improving the alert system using Bayesian networks. The SYMPRO-Lung multicenter randomized trial involved 244 lung cancer patients starting new treatments, who completed weekly online symptom reports for up to one year. Alerts were generated when symptoms crossed predefined thresholds, with patients able to mark some alerts as unnecessary, highlighting a need to reduce false alarms.<br /><br />A Noisy-OR Bayesian network model was developed to decrease unnecessary alerts without missing severe symptoms. Using data from 7,185 weekly symptom reports, the updated algorithm maintained sensitivity by not missing any severe symptoms while reducing alerts from 1,325 (18%) to 1,281 (18%), with necessary alerts decreasing slightly from 60% to 56%.<br /><br />A full Bayesian network was created to identify determinants of alarming versus non-alarming symptoms, taking into account patient baseline characteristics. Results showed that undergoing treatment (probability 0.7) and impaired emotional functioning (probability 0.4) were significant predictors for symptom reports. Patients reporting alarming situations were more likely to have comorbidities, disease progression, and emotional impairment. For example, impaired emotional functioning had probabilities of 0.57 in alarming versus 0.29 in non-alarming situations, and having two or more comorbidities was 0.37 versus 0.17. <br /><br />The study concludes that the Noisy-OR network effectively reduces unnecessary alerts while maintaining symptom detection accuracy. Recognizing emotional functioning and comorbidities as key factors aids in better symptom interpretation. This optimized alert algorithm supports the broader implementation of PRO symptom monitoring in clinical practice, potentially improving patient quality of life during lung cancer treatment.

Asset Subtitle

Cindy Verberkt

Meta Tag

Speaker Cindy Verberkt

Topic Multidisciplinary Care: Nursing, Allied Health and Palliative Care

Keywords

patient-reported outcome

symptom monitoring

lung cancer

Bayesian networks

SYMPRO-Lung trial

Noisy-OR model

alert system optimization

emotional functioning

comorbidities

clinical implementation