To evaluate the effectiveness and implementation challenges of AI-augmented swallowing rehabilitation for adults with oropharyngeal dysphagia, highlighting the significance of AI in addressing these challenges.
Approach:
Key Findings:
AI-augmented interventions produce short-term gains in functional oral intake and physiological measures, but effects diminish after cessation, indicating a need for ongoing support.
Adherence to AI interventions declines sharply without clinician supervision, emphasizing the importance of human oversight.
Digital literacy, cognitive impairment, and interface usability are significant barriers to implementation, particularly for populations with the greatest need, which could hinder overall effectiveness.
AI algorithm performance is rated at very low certainty, with validation primarily in healthy volunteers, limiting the applicability of findings to broader populations.
Interpretation:
The findings suggest a need for pragmatic trials in supervised post-stroke rehabilitation, while highlighting insufficient evidence for other aetiologies, unsupervised settings, and the necessity for research involving diverse populations.
Limitations:
Evidence for sustained outcomes and effectiveness in diverse populations remains inadequate, which is critical given the therapist shortage.
Validation of AI systems is largely confined to healthy volunteers, limiting generalizability and applicability to clinical settings.
Conclusion:
The study underscores the necessity for further research to address the gaps in evidence and implementation challenges in dysphagia rehabilitation, particularly the role of AI in enhancing rehabilitation outcomes.
