JMIR Publications

ABSTRACT

Background:

Frailty is defined as a clinical state of increased vulnerability due to age-associated decline of an individual’s physical function resulting in increased morbidity and mortality when exposed to acute stressors. Early identification and management can reverse frail individuals to robust.

Objective:

Therefore, we propose an approach for early diagnosis of frailty in community-dwelling elderly individuals in Thailand using a machine learning model generated from individual characteristics and anthropometric data.

Methods:

The datasets of 2692 community-dwelling Thai old adults in Lampang from 2016 and 2017 were used for model development and internal validation. The derived models were externally validated by the dataset of community-dwelling old adults in Chiang Mai, 2021. The machine learning algorithms implemented in this study include the K-Nearest Neighbors (KNN) algorithm, Random Forest ML algorithms (RF), Multi-layer Perceptron Artificial Neural Network (MLP), Logistic Regression (LR) models, Gradient Boosting Classifier (GBC), Linear Support Vector Classifier (SVC).

Results:

The discrimination performance of externally validated models was LR (mean AUC 0.75, CI 0.71-0.78), KNN (mean AUC 0.54, CI 0.51-0.57), RF (mean AUC 0.74, CI 0.71-0.78), MLP (mean AUC 0.54, CI 0.51-0.57), GBC (mean AUC 0.73, CI 0.57-0.63) and SVC (mean AUC 0.73, CI 0.70-0.77). LR and MLP were well-calibrated to the expected probability of external validation dataset.

Conclusions:

Our findings showed that our models have potential to be utilized as a screening tool in Thai community-dwelling older persons to identify frail individuals who require early intervention to become physically robust.