Smartphone app agreed with expert frailty ratings in heart failure
The app had a weighted agreement score of 0.78 against expert frailty ratings in patients aged 75 or older with heart failure.
Multicenter
Limit: Agreement with expert ratings was evaluated; clinical outcomes were not measured
Edited by Boris Djordjevic · AgingResearch.News
Across six hospitals, a prospective validation study enrolled 462 patients aged 75 years or older with symptomatic chronic heart failure. Participants averaged 81.8 years of age, and 55.4% were male. Patients walked an L-shaped, 4-meter course while an iPhone camera recorded video. An artificial intelligence application assigned ratings on the Clinical Frailty Scale. Eleven trained cardiologists used the videos and daily living data to reach a consensus reference rating for each patient.
In the full analysis set of 442 patients, agreement between the application and expert ratings showed a weighted score of 0.78, exceeding the predefined threshold. Ratings differed by two or more grades in 0.5% of patients. In a reproducibility test using two consecutive measurements, agreement was 0.89. The system maintained consistent agreement across age, sex, and heart failure severity.
Why it matters
Frailty is linked to worse outcomes in people with heart failure. The findings address whether automated tools can provide more consistent frailty assessments than traditional tests that rely on subjective clinician judgment.
Caveats
The study evaluated agreement with expert ratings rather than clinical outcomes. The findings may not apply to people younger than 75 or those without symptomatic chronic heart failure.
- Ai-based smartphone frailty assessment system
- Machine learning
- Heart failure
- Clinical Frailty Scale
- AI-derived Clinical Frailty Scale
- Humans
