Background: Lockdowns and stay-at-home orders announced internationally for COVID-19 have led to physical and social distancing, with reports of many individuals experiencing social isolation (SI) and loneliness. Although the emergency declaration in Japan was declared as a “mild” lockdown requested by the government without penalties for violations, the lockdown measures, including SI, had several influences on people’s lives and mental health as in other countries. Furthermore, Japan declared a state of emergency multiple times; thus, it is necessary to examine the influence of the transition of SI caused by repeated emergency declarations and the deterioration of mental health associated with these changes.
Objective: This study longitudinally investigated the transition of SI and its related factors during the mild lockdown under 2 declared states of emergency in Japan and analyzed psychosocial characteristics by extracting clusters where people with specific transition patterns of SI predominated.
Methods: We collected data on 7893 inhabitants (3694 [46.8%] women, 49.6 [SD 13.7] years old) living in the 7 prefectures where the initial emergency declaration was applied. The investigations took place online in the final phase of the first and second states of emergency: phase 1 (between May 11 and 12, 2020) and phase 2 (between February 24 and 28, 2021). Nonparametric Bayesian coclustering was used to visualize the exhaustive interaction structure between the transition pattern of SI and the psychosocial variables.
Results: There were no improvements in social networks and loneliness between the 2 phases, although psychological distress significantly improved and depression slightly decreased. Overall, 3868 (49%) of the 7893 participants remained socially isolated through phases 1 and 2, and 947 (12%) were socially isolated in phase 2, even though they were not socially isolated in phase 1. More participants experienced persistent SI in unmarried, childless, and low-household-income groups. The persistent-SI group had fewer cohabitants than other transition pattern groups. The nonparametric Bayesian coclustering results showed that most clusters, including participants without SI throughout phases 1 and 2, had healthy behaviors, more interactions, good relationships, and less loneliness and psychological stress. Furthermore, the cluster in which relationships deteriorated in phase 1 recovered in phase 2. Comparatively, the clusters with SI throughout phases 1 and 2 were divided into clusters with increased loneliness and psychological stress; clusters were close to participants’ average scores in this study. The clusters with increased loneliness and psychological stress were notable for deteriorating relationships and less online interaction.
Conclusions: This study revealed the actual state of transition of SI and related psychological, social, and behavioral factors under repeated declarations of a state of emergency. These results should help construct intervention methods that fit individual characteristics of people in SI during a pandemic.