JMIR Publications

ABSTRACT

Background:

With dementia cases substantially increasing worldwide, adequate treatment tools are urgently needed. Considering the accessibility of cognitive interventions, non-pharmacological treatments such as digital cognitive training are gaining popularity. Past studies specifically focus on cognitive training itself which requires digital literacy and easy to use. To enhance digital literacy and facilitate ease of use, it is possible to incorporate a social aspect, provide rewards for app usage, and utilize AI conversational agents.

Objective:

We developed a 12-week digital cognitive training called ‘Care & Cure’, which consists of a chatbot service called ‘Saemi talk’ and a group chat service called ‘Our Town’. A field study was conducted, to answering two research questions. The primary objective is to determine whether ‘Care&Cure’ program can enhance cognitive function in older adults. The secondary objective is to determine if ‘Our Town,’ which acts as a catalyst to increase ‘Saemi talk’ usage and interaction between AI agent and the user, improves the social aspect element.

Methods:

A total of 133 participants (age range: 51–83 years; mean age: 64.75 years) who had not been diagnosed with dementia were recruited. All participants received the ‘Care & Cure’ program for three months. The primary outcomes were changes in Korean-Mini Mental State Examination-version 2 (K-MMSE 2) and Cognitive Impairment Screening Test (CIST) scores after 3 months. and the secondary outcomes were the participants' log data, degrees of social support (MOS-SSS), depression scores (SGDS-K), and engagement scores (TWEETS).

Results:

Overall, the Care & Cure intervention improved cognitive function, as measured by the K-MMSE pre-post score (Hedge’s g = 0.26, p < .001) and CIST pre-post scores (Hedge’s g = 0.35, p < .001). In secondary results, emotional/informational support showed the highest difference before and after use (t = -6.509, p < .001) with a moderate effect size (0.70; Hedge’s g). Willingness to participate was significantly different between the factors of cognition (t = 2.159, p < .05) with a weak effect size (0.43; Hedge’s g) and affect (t = 2.008, p < .05) with a moderate effect size (0.51; Hedge’s g). However, the behaviors did not show significant differences between pre- and post-use. Additionally, depression showed a significant difference between pre- and post-use (t = 3.093, p < .01) with a weak effect size (0.21; Hedge’s g). In addition, both the active user’s group and non-active user’s group showed a significant increase in CIST scores after treatment (paired t-test; p < 0.001 for active user’s group; p < 0.05, active user’s group). In addition, individuals who used ‘Our Town’ tended to use ‘Saemi talk’ more frequently (p < .001). And, those who used ‘Saemi talk’ more frequently had higher social support scores (p < .001), and individuals with higher social support tended to have higher K-MMSE scores (p < .05).

Conclusions:

Care & Cure successfully improved cognitive function, participation, social support, and emotional relief. We found that improving social support among participants was a key strategy for improving cognitive function.