A Comparative Study of the Psychometrics and Feasibility of Artificial Intelligence Algorithms Versus the Standard Screening Methods for Oropharyngeal Dysphagia in Older Adults: A Scoping Review.

Abstract

Oropharyngeal dysphagia (OD) is highly prevalent (35.6%-47.4%) in hospitalized older patients but clinical screening is slow and labor-intensive, leading to an 80% underdiagnosis. The aim of this study was to describe and compare the psychometrics and feasibility of current OD clinical screening methods versus artificial intelligence (AI) algorithms for OD. A scoping review was performed using PubMed Central and Web of Science databases (January 2009-August 2025), focusing on psychometric data for OD screening methods in older adults (>65 years) and comparing them with AI-based tests, including the AI Massive Screening for OD (AIMS-OD). Quality of the studies was assessed by using Joanna Briggs Institute Critical Appraisal Checklists (JBICAC). Of 166 articles identified, 137 remained after duplicate removal. Following title and abstract screening, 53 full-text articles were reviewed and 41 met inclusion criteria; 99 clinical screening methods were identified, showing 37-100% sensitivity, 29-97% specificity, 37-100% reliability and required 5.6-10min/patient. AI-based tests demonstrated good overall psychometric performance and reliability, AIMS-OD showed 88.6-94% sensitivity, 42-99.3% specificity, 100% reliability and took only 0.7 seconds/patient without the need of human resources. OD screening methods are accurate but time-consuming, which may limit feasibility and contribute to underdiagnosis. AI tools show comparable psychometric performance, while offering rapid, automated, and highly reliable screening without using human resources. AI-based tools are a feasible and scalable approach to providing early systematic screening of hospitalized older patients, facilitating timely identification and referral for further assessment and potentially reducing OD-related complications.