Big Data/Machine Learning/AI
Digital Solutions for Loneliness: Exploring Heterogeneous Effects of Virtual Communication Using Machine Learning DAISUKE KATO* DAISUKE KATO Ichiro Kawachi Atsushi Nakagomi Katsunori Kondo Koichiro Shiba
Background:
Loneliness among older adults negatively affects health and quality of life, increasing risks of depression, cognitive decline, and mortality. Evidence suggests loneliness’s impact varies by demographic and socioeconomic (SES) factors. This study examined whether internet-based virtual communication reduces loneliness and explored potential group differences.
Methods:
Data were analyzed from the Japan Gerontological Evaluation Study (JAGES) using 2019 and 2022 surveys. Participants included 4,807 adults aged 65+ who were independent in daily activities (ADL). The primary exposure was virtual communication (e.g., email, Zoom) between 2021 and 2022, and the outcome was loneliness, measured by the UCLA Loneliness Scale. Covariates included demographic and health factors from 2019, such as age, sex, marital status, baseline health, and in-person interaction frequency.
The average treatment effect (ATE) of virtual communication on loneliness was estimated using Targeted Maximum Likelihood Estimation (TMLE). Causal Forest analysis assessed variations across demographic and SES factors.
Results:
Of 4,807 participants, 2,152 engaged in virtual communication, and 2,655 did not. Virtual communication was associated with reduced loneliness scores (ATE: -0.120, 95% CI: -0.213, -0.026). Causal forest analysis showed 86.3% of participants using virtual communication showed improved loneliness scores. Significant heterogeneity was observed (p = 0.006). The Best Linear Predictor estimate was -0.12 (SE = 0.05, p = 0.012), suggesting factors like age and social activity levels influenced effects. The mean forest effect (-0.08, p < 0.001) supported the overall association, while the differential forest effect (0.12, p < 0.001) highlighted group differences.
Conclusion:
Virtual communication is associated with reduced loneliness in older adults, with variations across groups. These findings underscore the potential of digital technologies to address loneliness in aging societies.