Participants.

In this study, we included 2,664 older adults aged 65–79 years living in town A in Hokkaido, Japan, as of 17 May 2021. The cutoff age was 79 years because the percentage of individuals certified as requiring assistance with daily activities is higher for those aged ≥80 years. Additionally, declining physical function may be associated with reduced social participation, which may influence loneliness, prompting the exclusion of individuals >79 years old from the study. Overall, 2,516 participants were enrolled after excluding those certified as requiring assistance in daily activities (n = 120), those who were hospitalized for ≥3 months (n = 27), or those residing at a special address (n = 1).

Survey.

The study description, questionnaire, and consent form were mailed to the participants, and the questionnaires were collected via return envelopes. The study purpose and research ethics were explained to the participants using the research instructions; participation was voluntary, and written informed consent was obtained. The survey was conducted between May 24 and June 11, 2021. This study was approved by the Ethical Review Committee of the Faculty of Health Sciences, Hokkaido University (No. 21−7). All responses were anonymized prior to analysis.

Target area.

Many municipalities in Hokkaido are experiencing depopulation and aging populations, with a significant percentage (21.7%) of older adults living alone [20]. This depopulation has made it challenging to provide quantitative services, which is one potential measure to address loneliness, in many areas due to inadequate public transportation. In the target area of this study, the only available public transportation is by bus, which is not convenient and makes it difficult to deliver sufficient quantitative services. Additionally, the aging rate for individuals aged ≥65 years in the target area is 37.7% [21], which is comparable to nearly half of the municipalities in Japan. Thus, the target area reflects trends seen in many aging municipalities that may benefit from measures aimed at addressing loneliness, where providing quantitative services is difficult due to depopulation.

Survey items.

We used a 59-item self-administered questionnaire that assessed basic attributes, loneliness, social participation, life space, self-evaluation of ability, and depressive symptoms.

Basic attributes.

Six basic attributes were evaluated: age, self-reported gender, household members, exercise days (i.e., the amount of physical activity), smoking, and drinking. The participants were asked about the number of family members they live with: “1. one person,” “2. two people,” “3. three people,” “4. four people,” “5. five or more people.” Smoking frequency was assessed with the following options: “1. never,” “2. used to smoke but stopped,” “3. sometimes,” “4. daily.” Drinking frequency over the past year was categorized as follows: “1. do not drink (cannot drink),” “2. used to drink but stopped,” “3. less than 12 times a year,” “4. 1–3 days a month,” “5. 1–2 days a week,” “6. 3–4 days a week,” “7. 5–6 days a week,” “8. daily.” Higher scores for each item indicate an increase in household members, exercise days, smoking frequency, and drinking frequency.

Loneliness.

Loneliness was assessed using the Ando, Osada, and Kodama Loneliness Scale, which comprises 10 items with dichotomous responses: “1. Yes” and “2. No.” The total score ranges from 0 to 10, with higher scores indicating a greater sense of loneliness. The reported mean score is 1.46 ± 2.27 for individuals aged 60–69 years and 1.58 ± 2.36 for those aged 70–79 years. This scale does not have a designated cutoff value for loneliness severity [22].

Social participation.

Social participation was evaluated based on activities generally performed by older adults in the Japanese social context [23,24]. Recognizing that the nature and significance of social participation vary across social contexts, we identified activities that are representative of the everyday experiences of older adults in Japan. Respondents were asked about the frequency of participation in eight types of social participation: volunteering, group exercise, hobbies, study/education circles, care prevention classes, senior citizen club activities (i.e., local recreational clubs for older adults), neighborhood association activities, and work.

Volunteering referred to unpaid public service activities. Group exercise included structured physical activities performed with others, such as community gymnastics. Hobbies referred to leisure pursuits such as gardening, crafts, or music. Study/education circles were informal peer-based gatherings for shared learning. Care prevention classes were community-based health promotion programs aimed at maintaining independence. Senior citizen club activities involved recreational or cultural events organized for older adults. Neighborhood association activities were community-based engagements, including participation in residents’ associations, neighborhood meetings, and local events such as seasonal festivals or group clean-up efforts. These activities are often rooted in mutual support and collective responsibility and are characteristic of Japanese local communities. Work included both paid and unpaid labor, such as part-time jobs, self-employment, volunteer work, or helping in family businesses.

Each activity was rated on a six-point frequency scale: “four or more times a week,” “two to three times a week,” “once a week,” “one to three times a month,” “several times a year,” and “not participating.” Each step on the six-point scale was scored in order, with frequent participants receiving a score of 6 and non-participants a score of 1.

Life space.

The extent of life space was assessed using the Life-Space Assessment [25], a tool to evaluate mobility in the spatial extent of an individual’s life, allowing the examination of usual mobility patterns in the month before the assessment. A Japanese version of the Life-Space Assessment has been developed by Harada et al. [26]. It categorizes life-space into five levels, which range from the room where a person sleeps to areas beyond their town: “1. Other rooms of your home besides the room where you sleep,” “2. An area outside your home such as your porch, deck, patio, hallway (of an apartment building), garage, yard, or driveway,” “3. Places in your neighborhood other than your own yard or apartment building,” “4. Places outside your neighborhood but within your town,” “5. Places outside your town.” For each life-space level, participants are asked how often they traveled to that area (“1. less than once a week,” “2. 1–3 times each week,” “3. 4–6 times each week,” “4. daily) and whether they needed assistance from another person or an assistive device (“1. yes” and “2. no”). The LSA scores range from 0 (indicating that the individual is totally room-bound) to 120 (indicating that the individual travels out of town every day without assistance).

Self-evaluation of ability.

Various studies have reported that self-evaluation of usefulness and ability is associated with life satisfaction and loneliness [17,18]. Higher levels of self-evaluation of ability have been associated with increased social participation and may reduce feelings of loneliness. Therefore, we assessed self-evaluation of ability by examining memory decline, which is common in older age. Subjective memory was assessed using the Prospective Retrospective Memory Questionnaire [27], a self-report scale measuring everyday memory failures related to retrospective and prospective memory.

Depressive symptoms.

The depressive symptoms were evaluated using the Geriatric Depression Scale (GDS)-15 [28]. It consists of 15 yes-or-no questions that assess various aspects of mood and emotional well-being over the preceding week, making it suitable for quick evaluations in clinical settings. For the GDS-15, scores range from 0 to 15, with higher scores indicating greater severity of depressive symptoms. A Japanese version was developed by Sugishita et al. [29].

Statistical analysis.

To examine exploratory models of the determinants of loneliness by gender, we conducted analyses referring to the method proposed by Mitchell H. Katz [30]. The primary purpose of this study was to appropriately evaluate the independent effects of the types of social participation on loneliness. Therefore, we deemed it necessary to rely on simple correlations and examined the activities that are independently associated with loneliness after controlling for confounding factors.

Based on Katz’s approach, we used a liberal threshold of p < .25 in the correlation analysis to broadly retain candidate variables. This was to avoid overlooking variables that might appear weak in simple correlations but could become significant after adjusting for other factors. Since the types of social participation are partially interrelated, eliminating them solely based on the presence or absence of bivariate correlations would risk excluding truly important types. Moreover, because the sample size in this study was small (N = 200), including variables that had no association at all would increase noise and inflate standard errors.

We also considered the approach of incorporating all variables into the model at once (enter method); however, types of social participation with small effects may be absorbed by other variables and judged as statistically insignificant. As a result, there is a risk of misinterpreting truly associated variables as having no effect. For these reasons, we did not adopt the forced entry method in this study. Instead, we used a method that first identified a broad range of candidates through correlation analysis and then input them hierarchically into the multiple regression analysis.

Depression is a proximal psychological factor that can be a result and a cause of loneliness [15,16], and it is highly likely to function as a confounder and a mediator. Therefore, including it in the early stages might lead to overadjustment and an unjust underestimation of the effects of social participation, thus introducing depression only in the final step.

The specific procedure used was as follows:

1. (1). For loneliness and ordinal-scale variables (household members, smoking, drinking, volunteer activity, group exercise, hobbies, study/education circles, care prevention classes, senior citizens club, neighborhood association activities, and work), we conducted Spearman’s rank correlation analysis, and for interval-scale variables (age, number of exercise days, self-evaluation of ability, life-space, and depression), we conducted Pearson’s product-moment correlation analysis.
2. (2). Items with p ≤ .25 in the correlation analysis were included as independent variables, and loneliness as the dependent variable in multiple regression analysis.
3. (3). Independent variables with p ≤ .25 in the multiple regression analysis were included as Step 1, and depression was added as Step 2. Hierarchical multiple regression analysis was conducted; this was used as the final model. The validity of the model was evaluated by testing the amount of change in R².

Regarding missing data, diagnostics did not reject the hypothesis that the missing pattern was missing completely at random (p = 0.166). For loneliness, the main outcome of this study, listwise deletion was applied to ensure accuracy and interpretability of the results. Participants with missing data on loneliness were excluded, as the dependent variable must be present for inclusion in the regression model. For other variables, single imputation via regression with Bayesian estimation under a normality assumption was applied. This method was chosen in consideration of the exploratory nature of the study and the relatively large number of predictors, to minimize potential bias and preserve sample size. All statistical analyses were performed using IBM SPSS Statistics version 28 (IBM Corp., Armonk, NY).

Group exercise and loneliness.

Hierarchical multiple regression analysis revealed that group exercise significantly influenced loneliness in men. Although previous studies have demonstrated that physical activity helps prevent loneliness in older adults [36,37], our study provides new insight by showing that group-based physical activity has a protective effect specifically among men aged 65 and older.

Group exercise not only provides opportunities for social interaction but may also improve psychological states [38]. For example, low self-efficacy is associated with loneliness, while men tend to have stronger physical self-efficacy, which has been shown to predict reduced loneliness [39]. Group physical activity may enhance self-efficacy through cooperation and a sense of accomplishment, which may contribute to reduced loneliness. In addition, physical activity has been reported to alleviate anxiety about loneliness and death in older men [40].

Our findings suggest that promoting participation in group physical activity, rather than physical activity alone, may be particularly relevant to reducing loneliness among men.