JMIR Publications

ABSTRACT

Background:

Although geriatric depression is prevalent, it is difficult to diagnose depression in older adults who live alone. Diagnosis using self-reporting instruments is limited in measuring depressive symptoms of older adults in a community setting. Ecological momentary assessment (EMA) using wearable devices could be used to collect additional data to classify depression groups.

Objective:

The aim of our study was to develop a machine learning algorithm to predict classification of depression groups among older adults living alone. We focused on utilizing diverse data collected through a survey, actiwatch, and EMA report related to depression.

Methods:

The prediction model using machine learning was developed in four steps: (1) data collection, (2) data processing and representation, (3) data modeling (feature engineering and selection), and (4) training and validation to test the prediction model. Older adults (N=47) living alone in community settings completed an EMA to report depressive moods four times a day for two weeks. Participants wore an actiwatch, which measured their activity and ambient light exposure every 30 seconds for two weeks. At baseline and the end of the two-week observation, depressive symptoms were assessed using the Korean versions of the Short Geriatric Depression Scale (SGDS-K) and the Hamilton Depression Rating Scale (K-HDRS). Conventional classification based on binary logistic regression was built and compared to four machine learning models (the logit model, decision tree, boosted tree, and random forest models).

Results:

Based on the SGDS-K and K-HDRS, 38% of the participants (18/47) were classified into the depression group. They reported significantly lower scores of normal mood and physical activity and higher levels of white and red, green, and blue (RGB) light exposures at different degrees of various 4-hour time frames (all P values < .05). Sleep efficiency was chosen for modeling through feature selection. Comparing diverse combinations of the selected variables, daily mean EMA score, daily mean activity level, white and RGB light at 16-20 pm exposure, and daily sleep efficiency were selected for modeling. Conventional classification based on binary logistic regression had a good model fit (accuracy: 0.705; precision: 0.770; specificity: 0.859; and area under receiver operating characteristic curve: 0.754). Among the four machine learning models, the logit model had the best fit compared to the others (accuracy: 0.910; precision: 0.929; specificity: 0.940; and area under receiver operating characteristic curve: 0.960).

Conclusions:

This study provides preliminary evidence for developing a machine learning program to predict the classification of depression groups in older adults living alone. Clinicians should consider using this method to identify under-diagnosed subgroups and monitor daily progression regarding treatment or therapeutic intervention in the community setting. Furthermore, more efforts are needed for researchers and clinicians to diversify data collection methods using by survey, EMA, and a sensor.