Relationships between In-person and Digital Social Interactions with Family and Friends and Loneliness among Older Adults in Japan: A Cross-Sectional Analysis of the 2024 JACSIS

Study Design and Population

This cross-sectional study used data from the 2024 Japan COVID-19 and Society Internet Survey (JACSIS), an ongoing, large-scale, web-based, self-administered questionnaire survey conducted by Rakuten Insight, Inc. The JACSIS project was launched in August 2020 and has been conducted annually to examine the social and behavioral impacts of the COVID-19 pandemic in Japan.24) Participants are recruited from a large online panel, with sampling quotas based on age, sex, and prefecture applied to approximate the national population distribution. However, the panel does not constitute a strict probability sample.

The 2024 survey was conducted from December 2, 2024, to January 16, 2025, and collected information on participants’ demographic characteristics, lifestyle factors, health status, and social relationships. For the present analysis, we restricted the sample to adults aged ≥65 years. Among 28,000 respondents aged 18–84 years, 6,786 older adults met the age criterion and had complete data on communication frequency, loneliness, and all covariates. Since our primary aim was to examine associations rather than estimate population prevalence, analyses were conducted using unweighted data.

Exposure Variables

The exposure variables in this study were the frequency and mode of social interaction, capturing both in-person and technology-based communication. Participants reported how often they engaged in various types of communication with non-cohabiting family members and friends or acquaintances.

Social interaction was assessed using JACSIS survey items asking, “During the past month, how often did you engage in the following activities?” In this study, social interaction included in-person meetings, text messaging (including email, chat, and LINE), audio-only telephone calls, and video calls with non-cohabiting family members and friends. Response options ranged from “none,” “once per month,” “2–3 times per month,” “once per week,” “2–3 times per week,” “4–5 times per week,” to “almost every day.”

To ensure consistent interpretation across modes, frequency categories were recategorized into three groups: none, monthly (1–3 times per month), and weekly or more (≥1 time per week). Each communication mode was coded separately for family and friends to examine distinct interaction patterns by relationship type. In all analyses, “none” served as the reference category. Additionally, communication frequency was treated as a nominal variable, and dummy variables were created for the “monthly” and “weekly or more” categories.

Outcome Variable

The outcome variable of this study was loneliness, assessed using the three-item short form of the Japanese version of the UCLA Loneliness Scale. The original UCLA Loneliness Scale (Version 3) was developed by Russell,25) its Japanese version was validated by Masuda et al.,26) and the three-item short form of the Japanese version was subsequently validated by Arimoto and Tadaka.27)

To evaluate the reliability of the three-item scale in the present study, we calculated Cronbach’s alpha. A previous study of Japanese mothers reported an alpha coefficient of 0.790.27) In our dataset, the alpha coefficient was 0.891, which exceeds the commonly accepted threshold of 0.700, indicating adequate internal consistency.

Loneliness was measured by asking participants how often they had experienced the following during the past 30 days: (1) feeling a lack of companionship, (2) feeling left out, and (3) feeling isolated from others. Participants responded using four options: 1 (always), 2 (sometimes), 3 (rarely), and 4 (never), with higher scores indicating greater loneliness. To align the direction of the scores with the construct, we transformed the data by subtracting the raw total score from 15. In the following analyses, higher scores reflect higher levels of loneliness.

To classify individuals as “lonely” or “not lonely” for analytical purposes, loneliness scores were dichotomized. In the primary analysis, participants scoring 4 or higher (≥4) were classified as “lonely” (coded as 1), while those scoring 3 were classified as “not lonely” (coded as 0). The threshold of 4 was selected because it represented the median score of the study population, ensuring sufficient sample size for each group while allowing for a broad screening of individuals experiencing even mild levels of loneliness. Furthermore, to examine the robustness of the results against variations in the cut-off point, we conducted a sensitivity analysis using the third quartile score of 6 as an alternative threshold. In this analysis, participants scoring 7 or higher (≥7) were classified as “lonely” (coded as 1), and those scoring 6 or lower (≤6) as “not lonely” (coded as 0).

Covariates

We adjusted for demographic, household, health-related, lifestyle, and socioeconomic factors as potential confounders. Demographic variables included age (continuous) and sex (male/female). Educational attainment was coded into four categories: high school or less (reference); vocational school or junior college; university or graduate school; and other/do not know. Household factors were assessed using household size, categorized into living alone (reference), two-person, three-person, or four-or-more–person households. Household annual income was categorized as <5 million Japanese Yen (JPY) (reference), 5–10 million JPY, ≥10 million JPY, and “do not know/prefer not to answer.” This latter category was retained as a separate classification in all regression models to minimize data loss. Health-related factors included physician-diagnosed chronic conditions, coded as 1 for having at least one condition and 0 for none.

Lifestyle factors included smoking, alcohol use, and physical activity (walking, running, and sports). Smoking and alcohol use were dichotomized as 0 for none and 1 for any history. Physical activity variables were coded as 0 for no activity and 1 for any activity.

Socioeconomic status was assessed using current employment status, coded as 0 for retirees, homemakers, and unemployed individuals and 1 for participants who were currently employed. All binary variables were coded such that 1 indicated the presence of the characteristic and 0 indicated its absence.

These covariates were selected because demographic, household, health-related, lifestyle, and socioeconomic factors have been shown to influence social interaction patterns and loneliness among older adults in previous research28-30) and may confound the associations of interest in this study.

Statistical Analysis

First, we described sample characteristics by reporting medians and ranges for continuous variables, and counts and percentages for categorical variables, for the lonely and non-lonely groups defined by two cut-off points (scores ≥4 and ≥7). Differences between groups were examined using Wilcoxon rank-sum tests for continuous variables and chi-square tests for categorical variables.

To examine the associations between communication modes and loneliness, separate binary logistic regression models were fitted for each mode (in-person, telephone, text messaging, and video) and relationship type (non-cohabiting family and friends). The “no contact” category served as the reference group in all analyses. We performed these regression analyses using the primary cut-off for loneliness (score ≥4) and subsequently repeated them using the alternative cut-off (score ≥7) as a sensitivity analysis.

Multivariable models were adjusted for age, sex, household size, comorbidity, smoking, alcohol use, physical activity, and employment status. Adjusted odds ratios (aORs) and 95% confidence intervals (CIs) were estimated.

All analyses were conducted using JMP® Student Edition 18.2.1 (SAS Institute Inc., Cary, NC, USA), and two-sided p-values <0.050 were considered statistically significant.

Ethical Considerations

The study was reviewed and approved by the Shimonoseki City University Research Ethics Committee (Approval No. 2509–0623) and the Tohoku University Graduate School of Medicine Ethics Committee (Approval No. 2024-1-1035). All participants provided informed consent online prior to participation. The survey data were anonymized by the survey company before being provided to the researchers.

Participant Characteristics

Table 1 summarizes the characteristics of the study population (n=6,786). The median age was 72 years (range 65–84), and 50.4% were male. When applying the cut-off of 4, 3,732 participants (55.0%) were classified as lonely (score ≥4). When applying the cut-off of 7, 1,534 participants (22.6%) were classified as lonely.

Table 1.

Sample characteristics by loneliness status (UCLA Loneliness Scale) (n=6,786)

Comparisons between the lonely (score ≥4) and non-lonely (score 3) groups revealed significant differences in household, socioeconomic, and health-related factors. Participants classified as lonely using the cut-off of 4 were significantly more likely to live in one-person households (22.4% and 15.7%, respectively; p<0.0001), have a household annual income of less than 5 million JPY (59.3% and 51.0%, respectively; p<0.0001), and have at least one chronic condition (62.3% and 50.5%, respectively; p<0.0001). They were also significantly less likely to be currently employed (24.6% and 30.7%, respectively; p<0.0001) and had higher rates of smoking history (52.3% and 47.2%, respectively; p<0.0001). No significant differences in age, sex, and educational attainment were observed between the groups at this threshold. These trends were generally consistent using the cut-off of 7, although significant age differences emerged in the higher-threshold analysis (p<0.0001).

Regarding social interaction, participants classified as lonely (score ≥4) reported considerably lower frequencies of communication compared to non-lonely participants across most modes and relationship types. Specifically, weekly interactions via in-person meetings, text messaging, and telephone calls were significantly less frequent among the lonely group for both family and friends (all p<0.0001). For video calls, significant differences were observed for family members at the cut-offs of 4 and 7 (p=0.039 and p=0.006, respectively). In contrast, for friends, a significant difference was found only when using the cut-off of 7 (p=0.004), whereas no significant difference was observed with the cut-off of 4 (p=0.392).

Logistic Regression Results

Table 2 presents the adjusted associations between communication frequency and loneliness. The results are described below by communication mode, primarily focusing on the analysis using the cut-off of 4. In these analyses, the “no contact” category for each communication mode was used as the reference group.

Table 2.

Non-adjusted and adjusted odds ratios for the association between communication frequency and loneliness (n=6,786)

For family members, in-person meetings weekly or more were significantly associated with lower odds of loneliness (aOR=0.764, 95% CI 0.664–0.879), whereas monthly meetings were not. For friends, both monthly and weekly meetings were strongly associated with lower odds of loneliness. The association was stronger for weekly meetings (aOR=0.435, 95% CI 0.377–0.500) than for monthly meetings (aOR=0.621, 95% CI 0.549–0.701).

Similar to in-person meetings, text messaging with family was significantly associated with lower odds of loneliness, but only at the weekly or more frequent frequency (aOR=0.747, 95% CI: 0.659–0.846). For friends, both monthly (aOR=0.881, 95% CI 0.778–0.997) and weekly interactions (aOR=0.534, 95% CI 0.470–0.608) were significantly associated with reduced loneliness.

Regarding audio-only telephone calls, weekly interaction with family members was significantly associated with lower odds of loneliness (aOR=0.624, 95% CI 0.543–0.718), while monthly interaction was not. For friends, significantly lower odds were observed for both monthly (aOR=0.734, 95% CI 0.659–0.819) and weekly calls (aOR=0.489, 95% CI 0.419–0.570).

In the analysis using a cut-off of 4, video calls showed no statistically significant association with loneliness for either family or friends, regardless of frequency.

Sensitivity Analysis (cut-off of 7)

When using the stricter cut-off of 7, the associations generally became stronger, with lower odds ratios observed across all modes. Unlike the primary analysis, where family contact required weekly frequency to be significant, this analysis showed that even monthly contact with family, including in-person meetings (aOR=0.656, 95% CI 0.576–0.746) and phone calls (aOR=0.773, 95% CI 0.679–0.880), was significantly associated with lower odds of severe loneliness. Notably, video calls, which were not significant in the primary analysis, showed significant protective associations in this sensitivity analysis: weekly video calls with family (aOR=0.736, 95% CI 0.547–0.991) and monthly video calls with friends (aOR=0.656, 95% CI 0.470–0.917) were significantly associated with lower odds of loneliness. For other modes, the pattern remained consistent: friends showed remarkable protective effects even with monthly contact, whereas family required weekly contact to show strong associations.

Differences by Relationship Type

The finding that interactions with friends were associated with lower loneliness, even at lower frequencies such as monthly, compared to family interactions may be explained by differences in these relationships. This observation is consistent with the perspective of socioemotional selectivity theory, which suggests that older adults tend to prioritize relationships that offer emotional satisfaction.31) In other words, friendships retained in later life are likely to represent emotionally valuable relationships that remain after many non-voluntary connections have been pruned.

Friendships are typically voluntary relationships based on shared interests, and older adults have been reported to experience more positive emotions during interactions with friends than with family.32,33) Consequently, interactions with friends via digital tools or face-to-face visits may provide psychological satisfaction even at lower frequencies. This suggests why even monthly contact was associated with reduced loneliness.

In contrast, family relationships often possess characteristics of obligatory relationships.34) Interactions with family may involve instrumental support or administrative matters in addition to emotional exchange. Although the quality of contact is more important for well-being than its quantity,35) the emotional density per interaction might be lower than that of friendships, given the obligatory aspects of family interactions. Therefore, more frequent contact of weekly or more, regardless of whether digital or face-to-face, may be necessary to maintain a sense of connection associated with lower levels of loneliness.

Accordingly, the results suggest the importance of maintaining and promoting connections with friends, as well as with family. Given that interactions with friends were associated with lower loneliness even at relatively low frequencies, facilitating participation in hobby groups or local communities may be a beneficial approach.

Comparison of Communication Modes

Regarding communication modes, consistent associations with lower loneliness were observed for interactions using text messaging and audio calls, in addition to face-to-face interactions.

Previous studies have reported inconsistent findings regarding text messaging. Some studies indicate that increased ICT use or frequent social messaging is associated with higher levels of loneliness.16,20) However, the results of this study are consistent with findings that text-based interactions with close others are associated with lower loneliness.18,36) Although the associations of ICT interaction vary greatly depending on the relationship and purpose of use, text messaging likely had a positive effect in this study because the focus was limited to interactions with family and friends, serving as a means to maintain existing social bonds.

Similarly, findings regarding audio calls are mixed. While some reports associate higher telephone frequency with lower loneliness,18) others suggest that increased loneliness may prompt greater telephone use.37) The protective association observed for audio calls in this study suggests that telephone contact may play a supplementary role in situations where face-to-face interaction is scarce.38)

Remote contact is unlikely to be a perfect substitute for face-to-face interaction, and digital means alone may not fully compensate for emotional connection.17,39) Nevertheless, the results of this study suggest that even simple tools such as text messaging and audio calls can serve as important channels for reducing loneliness when used continuously within close relationships.

Conversely, the results regarding video calls were complex. Although no significant association was observed in the analysis using the standard definition of loneliness with a cut-off of 4, a protective association was suggested when using the definition for severe loneliness with a cut-off of 7.

These complex results reflect the inconsistency reported in previous studies regarding the effectiveness of video calls. While some studies report their utility for maintaining connections with family in care settings,40) evidence regarding their effect on loneliness reduction is limited,41) and surveys during the pandemic also found no clear association with family or friends.21) The lack of association at the standard cut-off in this study is consistent with these findings. Video calls among older adults are strongly linked not only to technical proficiency but also to frailty and the presence of caregivers.42) Therefore, for the general older population, the technical stress associated with complex operations43) and barriers to use may have offset the benefits of communication.

However, the analysis focusing on severe loneliness revealed a protective association for video calls. This aligns with reports that video call use is associated with lower loneliness among socially isolated older adults.44,45) According to media richness theory, rich media containing visual cues are suitable for conveying ambiguous and complex messages.46) Studies applying this theory to support older adults suggest that individuals experiencing severe loneliness found text or audio alone insufficient and required the strong social presence provided by seeing the partner’s face via video calls.47) Therefore, for these individuals, video calls functioned as an essential compensatory mechanism for face-to-face interactions.

Consequently, the findings suggest that supporting digital utilization involves selecting tools tailored to individual needs and skills. Text messaging and audio calls appear less burdensome for many older adults and offer meaningful benefits in reducing loneliness. Therefore, uniformly recommending the acquisition of advanced technologies like video calls might not be necessary. Instead, video call support seems most relevant for individuals requiring richer communication, such as those with severe loneliness or physical constraints limiting face-to-face interaction.

Practical and Research Implications

Regarding practical implications, our findings suggest that public health interventions should prioritize maintaining diverse communication channels tailored to the individual's social context and technological proficiency. In municipal health settings, the identified thresholds for loneliness could serve as screening criteria to prioritize limited administrative resources. For the general older population, supporting the use of accessible tools like text messaging and audio calls may be more feasible and effective for maintaining social connections than uniformly promoting advanced technologies. Conversely, resource-intensive support, such as providing digital devices and training for video calls, appears most relevant for individuals with severe loneliness, for whom visual presence may serve as a critical compensatory mechanism.

As for research implications, future studies should employ longitudinal designs to further clarify the nature of these associations. Specifically, stratified analyses focusing on older adults living alone are necessary, as their reliance on digital communication and its potential role in relating to loneliness may differ from those living with others. Furthermore, future research should consider the concurrent use of multiple communication modes. Such approach would provide a more nuanced understanding of how various digital and face-to-face interactions independently or synergistically relate to loneliness in the post-pandemic era.

Limitations

Despite the importance of the findings presented herein, this study has some limitations. The cross-sectional design precludes causal inferences, as the observed associations between communication modes or frequency and loneliness do not imply causality. It is possible that individuals with higher loneliness selectively engage in certain types of communication, and unmeasured confounding cannot be ruled out. Furthermore, the web-based survey restricts participation to those with digital access, potentially underrepresenting older adults without such access and limiting generalizability. Engagement in digital modes, particularly video calls, may also depend on ICT literacy, social support, and health status, which could introduce selection bias. The generalizability of the results is also bounded by specific Japanese cultural norms and the timing of the survey.

Regarding the measurement of loneliness, the UCLA Loneliness Scale was analyzed using dichotomous cut-offs rather than as a continuous measure. While this approach was chosen to enhance the interpretability of findings for public health practice and to broadly explore the potential of various communication modes, it may result in information loss or misclassification. Given that loneliness is a graded construct, the data-driven nature of these thresholds should be considered when interpreting the robustness of the associations. Moreover, the findings may reflect associations concentrated at specific thresholds rather than providing a comprehensive view of how social interactions relate to loneliness across its entire distribution. Additionally, the use of a sample-specific median restricts direct comparability with other populations.

In terms of interaction characteristics, this study assessed frequency but not quality or content, limiting a comprehensive understanding of social connectedness. Moreover, as each communication mode was analyzed separately, the potential confounding effects of concurrent use of multiple methods were not fully accounted for. Since many older adults likely engage in various modes simultaneously, the independent association of each specific mode should be interpreted with caution. The analysis also did not incorporate social network size, such as the number of individuals with whom participants maintain contact, due to data unavailability. Since network size may be significantly associated with loneliness, its potential role as a confounding factor cannot be ruled out. Finally, self-reported data may be subject to recall bias, and despite extensive adjustments, residual confounding from unadjusted or unmeasured factors, including social network characteristics, may persist.

Conclusion

This study demonstrates that the association between communication frequency and loneliness varies considerably by relationship type and mode. Interactions with friends showed a protective association with loneliness, even at lower frequencies, compared to interactions with family. Regarding digital modes, text messaging and audio calls showed consistent associations with lower levels of loneliness, whereas video calls were less associated with reduced loneliness, although they appeared beneficial for severe loneliness. Preventing social isolation among older adults requires a flexible combination of face-to-face and digital interactions, as well as respecting the diverse ways each individual prefers to connect.