Development of a Machine Learning Model to Predict Therapeutic Responses in Laryngopharyngeal Reflux Disease
View abstract on PubMed
Summary
This summary is machine-generated.Machine learning models can predict treatment success in laryngopharyngeal reflux disease (LPRD). Key factors influencing response include proximal reflux time and weak acid exposure, aiding in personalized LPRD management.
Area Of Science
- Otolaryngology
- Gastroenterology
- Computational Medicine
Background
- Laryngopharyngeal reflux disease (LPRD) presents a significant clinical challenge, necessitating effective treatment strategies.
- Understanding factors influencing therapeutic response is crucial for optimizing patient outcomes in LPRD management.
Purpose Of The Study
- To develop and validate machine learning models for predicting therapeutic responses in LPRD.
- To identify key demographic and diagnostic parameters that influence treatment outcomes in LPRD.
Main Methods
- Retrospective analysis of LPRD patients treated with proton pump inhibitors, with response defined as a ≥50% reduction in reflux symptom index score.
- Application of four machine learning models (logistic regression, random forest, support vector machine, gradient boosting) using demographic and 24-hour multichannel intraluminal impedance (MII)-pH monitoring data.
- Internal and external validation of model performance using independent datasets.
Main Results
- All four machine learning models demonstrated comparable predictive performance for therapeutic response in LPRD.
- The logistic regression model achieved the highest accuracy (82.98% internal, 84.91% external) and F1 scores (88.24% internal, 86.21% external).
- Proximal total reflux time and weak acid time were identified as significant factors influencing treatment response.
Conclusions
- Machine learning models offer a valuable tool for predicting therapeutic responses in LPRD patients.
- Identifying key influencing factors like reflux time and acidity can refine LPRD management strategies.
- Integration of machine learning into clinical practice holds promise for improving LPRD treatment outcomes.