Predictive models for dysphagia in older adults: a systematic review and meta-analysis

Area of Science:

• Gerontology
• Clinical Medicine
• Biostatistics

Background:

• Dysphagia (difficulty swallowing) is prevalent in older adults, often linked to aging and neurodegenerative diseases.
• It significantly impacts quality of life, leading to malnutrition, frailty, and aspiration pneumonia.
• Early detection of dysphagia risk is challenging, with existing predictive models showing variable performance.

Purpose of the Study:

• To systematically review and assess the performance of predictive models for identifying dysphagia risk in the elderly population.
• To identify common predictors associated with dysphagia in older adults.

Main Methods:

• A comprehensive literature search was performed across multiple databases (e.g., PubMed, Web of Science, CNKI) up to September 15, 2025.
• Seventeen studies met the inclusion criteria, with a meta-analysis conducted on pooled data.
• Risk of bias and external validation were assessed for included studies.

Main Results:

• The pooled AUC from meta-analysis was 0.82 (95% CI: 0.77-0.88), indicating moderate predictive ability.
• Commonly identified predictors include advanced age, smoking, reduced self-care, polypharmacy, frailty, malnutrition, cognitive impairment, and poor oral health.
• Most studies exhibited a high risk of bias, and only two models underwent external validation.

Conclusions:

• Predictive modeling for dysphagia in older adults is an emerging field with significant methodological limitations.
• Insufficient external validation and high bias risk hinder the clinical utility of current models.
• Future research requires adherence to PROBAST standards and large-scale, multicenter validation for improved reliability.