JMIR Publications

ABSTRACT

Background:

The two most commonly used methods to identify frailty are the frailty phenotype and the frailty index. However, these two methods have limitations in clinical application. Furthermore, methods for measuring frailty have not yet been standardized.

Objective:

We aimed to develop and validate a classification model for predicting frailty status using vocal biomarkers in community-dwelling older adults based on voice recordings obtained from the picture description task (PDT).

Methods:

We recruited 127 participants aged ≥50 years. Data was collected via a short form of the Comprehensive Geriatric Assessment scale and voice recording with a tablet device during the Korean version of the PDT. Three artificial intelligence (AI) models, SpeechAI (speech data only), DemoAI (demographic data only), and DemoSpeechAI (combining both data types), were designed according to the input type to predict the frailty status.

Results:

In the comparison between the SpeechAI and DemoAI models, the SpeechAI (area under the curve [AUC] 0.89, 95% confidence interval [CI] 0.86-0.92) model showed superior performance to the DemoAI model (AUC 0.74, 95% CI 0.73-0.75) in AUC values (t=8.705, P<.001), and the DemoSpeechAI model (AUC 0.93, 95% CI 0.89-0.97) was superior to the DemoAI model in AUC values (t=7.978, P<.001). However, there was no significant difference between the SpeechAI and DemoSpeechAI models (t=1.057, P=.35) The DemoSpeechAI model, which jointly analyzed acoustic features and demographics, outperformed the models that used only speech or demographics on the metrics.

Conclusions:

The results confirmed that the AI model developed through a deep neural network and supervised learning, enhanced with self-supervised learning, accurately predicted the frailty status. Clinical Trial: This study was registered with the Clinical Research Information Service (Internet) in 2021 (KCT0006708 , available from: https://cris.nih.go.kr/).