Machine Learning Models Predicting Hospital Admissions During Chemotherapy Utilising Longitudinal Symptom Severity Reports and Patient-Reported Outcome Measures
- Zuzanna Wójcik 1,2, Vania Dimitrova 2, Lorraine Warrington 3, Galina Velikova 3,4, Kate Absolom 3,5, Samuel D Relton 5
- Zuzanna Wójcik 1,2, Vania Dimitrova 2, Lorraine Warrington 3
- 1UKRI Centre for Doctoral Training in Artificial Intelligence for Medical Diagnosis and Care, University of Leeds, Leeds, UK.
- 2School of Computer Science, University of Leeds, UK.
- 3Leeds Institute of Medical Research, St James's University Hospital, Leeds, UK.
- 4Leeds Cancer Centre, Leeds Teaching Hospitals NHS Trust, Leeds, UK.
- 5Leeds Institute of Health Sciences, Leeds, UK.
- 0UKRI Centre for Doctoral Training in Artificial Intelligence for Medical Diagnosis and Care, University of Leeds, Leeds, UK.
Related Experiment Videos
Contact us if these videos are not relevant.
Contact us if these videos are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.Machine learning models using patient-reported symptom data can predict chemotherapy-related hospital admissions. Incorporating patient-reported outcome measures (PROMs) enhanced model accuracy for better patient care.
Area Of Science
- Oncology
- Health Informatics
- Machine Learning in Medicine
Background
- Chemotherapy toxicity frequently causes acute hospital admissions, straining healthcare resources and affecting patient quality of life.
- Existing predictive models for emergency admissions often overlook crucial patient-reported data.
Purpose Of The Study
- To develop and evaluate machine learning (ML) models for predicting hospital admissions risk during chemotherapy using longitudinal symptom data.
- To assess the impact of patient-reported outcome measures (PROMs) on the performance of these ML models.
- To compare the predictive accuracy of models with and without PROMs.
Main Methods
- Longitudinally collected patient symptom severity reports were utilized.
- Four machine learning models, including Random Forest and Extreme Gradient Boosting, were trained and tested.
- Model performance was evaluated for predicting overall hospital admissions and short-term (14-day) admissions risk.
- A comparative analysis was conducted between models developed with and without PROMs.
Main Results
- Random Forest and Extreme Gradient Boosting models demonstrated excellent predictive performance for hospital admissions (balanced accuracy, recall, specificity > 0.9).
- Short-term (14-day) admissions risk prediction accuracy was found to be poor.
- The inclusion of PROMs significantly improved the overall performance of the machine learning models.
Conclusions
- Longitudinal collection and analysis of symptom severity reports and PROMs are crucial for understanding chemotherapy toxicity.
- These data-driven insights can enhance the prediction of emergency admissions, informing clinical decision-making and patient management.
- Integrating PROMs into ML models offers a promising approach to proactively identify patients at risk of chemotherapy-related hospitalizations.
Related Experiment Videos
Contact us if these videos are not relevant.
Contact us if these videos are not relevant.