Takemi Program in International Health
The Takemi Program in International Health seeks to improve health and health systems around the world by welcoming mid-career health professionals and scholars to the Harvard T. H. Chan School of Public Health to conduct path breaking research and develop their leadership skills.
665 Huntington Avenue, Bldg. 1, Room 1210
Boston, MA 021151, USA
Heterogeneous Associations Between Virtual Communication and Loneliness:
A Machine Learning Approach
Daisuke Kato,1,2,3 MD, PhD; Ichiro Kawachi,4 MD, PhD; Atsushi Nakagomi,2 MD, PhD; Katsunori Kondo,2,5 MD, PhD; Koichiro Shiba, PhD 6
¹ Takemi Program in International Health, Harvard T.H. Chan School of Public Health, Boston, MA
² Center for Preventive Medical Sciences, Chiba University, Chiba, Japan
³ Department of Family Medicine, Mie University Graduate School of Medicine, Japan
⁴ Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health, Boston, MA
⁵ Institute for Health Economics and Policy, Japan
⁶ Department of Epidemiology, Boston University School of Public Health, Boston, MA
Introduction
Background: Loneliness affects approximately 25% of the population globally and is exacerbated in older adults by social isolation and mobility limitations.
Role of Virtual Communication (VC): Digital tools such as email, social media, and video calls are proposed to alleviate loneliness, with their importance highlighted during COVID-19.
Research Gap: VC’s impact on loneliness is mixed—some studies show benefits, others find little effect. Factors like digital literacy and social networks may affect outcomes, suggesting the need for more nuanced approaches to understand when and for whom virtual communication is most effective.
Objectives
Estimate Average Treatment Effect (ATE) and Conditional Average Treatment Effects (CATE) of virtual communication on loneliness, using causal forests with Japan Gerontological Evaluation Study (JAGES) data. Identify subgroups most likely to benefit from virtual communication interventions, informing targeted approaches to reduce loneliness in older adults.
Methods
Study Design & Data
- JAGES baseline (2019) and follow up (2022).
- Adults ≥65 years from multiple municipalities; final sample: n=4,807.
Key Measures
- Exposure (Virtual Communication, VC): Internet communication with family, friends, or acquaintances (e.g., email, LINE, Zoom) ≥2–3 times/week.
- Outcome (Loneliness): Three-item UCLA Loneliness Scale (score 3–9).
- Covariates (42 variables): Age, gender, education, income, health status, social engagement, etc.
Statistical Analysis
- ATE: Targeted Maximum Likelihood Estimation (TMLE) comparing loneliness in VC vs. non-VC groups.
- CATE: Causal forests (generalized random forest, GRF) to detect subgroup differences and interactions.
- Validation: Best Linear Projection (BLP), calibration by quartiles (Q1 Q4), and covariate distribution checks.
- Missing Data: Imputed via MissForest (random forest–based algorithm).
Software: R (version 3.6.0)
Discussion
Key findings
- VC modestly reduced loneliness (ATE: -0.136, 95% CI: -0.194 to -0.078).
- Effects varied—greatest benefits in socially active, less educated women.
- Limited effects in highly educated men with fewer social interactions.
Implications
- Small loneliness reduction is still meaningful for mental and physical health.
- VC works best as a supplement to existing social interactions.
- Accessible, user-friendly technology is crucial for older adults.
- Socially isolated individuals may need additional support beyond digital tools.
Conclusions
- VC helps reduce loneliness, but effectiveness varies.
- Personalized interventions based on social and digital engagement are needed.
- Bridging the digital divide is key to maximizing VC benefits.
References
Nguyen, M. H., et al. (2020). Changes in Digital Communication During the COVID-19 Global Pandemic: Implications for Digital Inequality and Future Research. Social Media + Society, 6(3), 2056305120948255.
Shiba, K., et al. (2023). Associations of online religious participation during COVID-19 lockdown with subsequent health and well-being among UK adults. Psychological Medicine, 53(9), 3887–3896.
Surkalim, D. L., et al. (2022). The prevalence of loneliness across 113 countries: Systematic review and meta-analysis. BMJ, 376, e067068.
Acknowledgements
This project was conducted with the support of the Takemi Program in International Health at the Harvard T.H. Chan School of Public Health. This study used data from JAGES (the Japan Gerontological Evaluation Study), supported by Grant-in-Aid for Scientific Research and various funding sources from Japanese institutions.