# Interpretable machine learning for accessible dysphagia screening and staging in older adults

## TL;DR

This study developed interpretable machine learning models to screen and stage dysphagia in older adults, achieving high accuracy and creating a web app for real-time clinical use.

## Contribution

The novel contribution is the development of interpretable ML models for dysphagia screening and staging, validated across multiple centers and implemented in a web application.

## Key findings

- CatBoost achieved 0.914 AUC for binary classification of dysphagia.
- Neural networks achieved 0.884 AUC for multiclass classification.
- A web application was developed to support real-time screening and stratification.

## Abstract

Dysphagia in older adults causes serious complications, and efficient and scalable screenings are needed. This prospective multicenter study developed interpretable machine learning (ML) models for the early identification and staging of dysphagia. Nine ML models were built using the clinical data from 1,235 patients and externally validated on 720 patients. All patients were older adults from seven Suzhou hospitals whose dysphagia was confirmed via videofluoroscopic swallowing studies. Features were selected via random forest, and model interpretability was analyzed with SHapley Additive exPlanations (SHAP). The CatBoost model achieved an area under the receiver operating characteristic curve (AUC) of 0.914 for binary classification, while neural network gave AUC 0.884 for multiclass classification. External validation confirmed robustness (binary AUC, 0.909 and multiclass macro-AUC, 0.860). SHAP identified ten core features—oral/pharyngeal function influenced all stages, and masticatory/phonatory features acted selectively. A web application was created accordingly to facilitate real-time screening and stratify dysphagia patients.

•Nine machine learning models were compared in triaging geriatric dysphagia•The models used multi-domain data and were interpretable•Both the binary and multiclass classifiers relied on only ten core features•A web app with SHAP visualizations was developed to support clinical screening

Nine machine learning models were compared in triaging geriatric dysphagia

The models used multi-domain data and were interpretable

Both the binary and multiclass classifiers relied on only ten core features

A web app with SHAP visualizations was developed to support clinical screening

Gastroenterology; Health sciences; Internal medicine; Medical specialty; Medicine