This study used a cross-sectional, survey-based research design with a nonrandomized sampling method. The study design, data collection, and reporting followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for cross-sectional studies [33].

Participants were Croatian citizens of both genders, aged 65 years and older, living in Croatia. To define older adults, we used a cutoff age of 65 years [34]. The sample comprised 2053 participants.

As shown in Table 1, the sample was mostly composed of older adults, with a majority being female, residing in urban areas, and either married or widowed. More specifically, the median age of the participants was 74 (IQR 69-81) years. Regarding gender distribution, women were more represented in the sample (n=1324, 64.49%) than men (n=729, 35.51%). Most participants lived in urban areas (n=1585, 76.57%), while a smaller proportion lived in rural areas (n=468, 22.61%). In terms of marital status, about half of the participants were married (n=1000, 48.71%). A significant proportion were widowed for over 1 year (n=745, 36.29%), while smaller percentages were recently widowed (less than 3 months: n=65, 3.17%; 3 months to 1 year: n=56, 2.73%). In addition, 97 (4.73%) of participants had never been married, and 90 (4.38%) were divorced.

Participants were recruited from community settings using a snowball sampling technique [35,36]. Initially, health students from 6 Croatian cities (Zagreb, Split, Rijeka, Zadar, Dubrovnik, and Osijek), each guided by a university professor, were instructed to identify 6 individuals aged 65 years and older among their relatives, ensuring inclusion of both sexes. Eligible participants were community-dwelling older adults who were able to understand the study purpose and complete the questionnaire independently. Each recruited participant was then asked to invite up to 6 peers from their own social circles (friends or relatives) who met the same criteria. Data collection continued until the target sample size (n=2053) was achieved, calculated using the Cochrane formula for continuous data [37].

Before participation, potential respondents were informed about the study aims, the voluntary nature of participation, and the guarantee of anonymity and confidentiality. Anonymity of the respondents was ensured by placing each completed questionnaire in a sealed envelope. After filling out the questionnaires, respondents returned the sealed envelopes to the student. The students then delivered the envelopes to their supervising professor, who subsequently forwarded them to the principal investigator. The research was conducted between April and October 2021 in nonclinical, community-based environments (primarily private households).

The following measures were used to collect data and address the research objectives:

Descriptive measures were first used to summarize demographic data, with frequencies and percentages calculated for categorical variables such as gender, marital status, time spent alone, socializing with others, and place of residence. Means and SDs were used for continuous variables such as age, loneliness, psychological distress, and self-efficacy. Hierarchical regression analysis was conducted in a stepwise manner to assess the contribution of various predictors to the dependent variables of loneliness and psychological distress. The predictors included demographic factors, psychological factors, and social behaviors during the lockdown. Model fit was assessed using R², F statistics, and P values. In addition, path analysis was used to examine the mediating contribution of PS (mediator) in the relationship between pandemic-specific stressors (social distancing behaviors and exposure to infection as predictors) and self-efficacy as the dependent variable. Path coefficients, SEs, and CIs were reported for direct, indirect, and total effects. The significance level was set at P<.05 for all tests.

ChatGPT (OpenAI) was used only as a supportive technical tool in this study. Specifically, it was used to provide suggestions for improving the accuracy of English grammar and phrasing. The authors carefully verified, edited, and approved all artificial intelligence–generated content and take full responsibility for the final paper.

This study was conducted in accordance with the Declaration of Helsinki and was approved by 7 ethics committees (Table 2). All participants provided written informed consent before participation. One copy of the consent form was kept by the research participant, and a second was stored separately from the questionnaire to maintain anonymity. Data collection and privacy protection were conducted in accordance with the General Data Protection Regulation (2016/679). All data were anonymized before analysis, and no identifying information was collected, stored, or reported. Participants did not receive any financial or other compensation for their participation in the study. No images or additional materials contain identifiable information about study participants.

This study aimed to evaluate how sociodemographic and psychological factors, along with exposure to pandemic-related stressors and PS, were associated with loneliness among older adults during the COVID-19 pandemic. Second, it explored how pandemic-induced changes in social routines, self-efficacy, and PS predicted levels of psychological distress in this population. Finally, it examined whether PS mediated the relationship between pandemic stressors and self-efficacy.

The findings of this study closely mirrored those of the earlier pilot study [19], highlighting consistent risk factors for loneliness among older adults during the COVID-19 pandemic. Similar to previous international research, predictors such as living alone [11], widowhood [48], psychological distress [2], and social isolation [49] were reaffirmed as significant contributors to loneliness. Older adults and women reported lower levels of loneliness, possibly due to caregiving roles and strong intergenerational ties that are common in Croatian society. This contrasts with previous research showing that older women are more vulnerable to loneliness [17,50]. In our study, men appeared more at risk, likely due to weaker social or family engagement. Similarly, rural residents reported lower loneliness than urban residents, opposing earlier findings [14], possibly due to rural self-reliance and reduced dependence on frequent social interaction. This study also showed lower loneliness among older adults, contrary to some previous studies [51,52]. This may reflect cultural or contextual factors that help explain differences observed in the Croatian sample, as identified in earlier research. For example, older adults in Croatia often benefit from strong family ties and intergenerational support [53]. They may also be more resilient due to past experiences with hardship, such as war or economic instability [54], or may feel more comfortable with smaller, stable social networks that are less affected by crises [55].

This study also examined the role of PS in older adults’ mental health and self-efficacy during the COVID-19 pandemic, revealing a complex relationship. Previous studies have shown that PS can support autonomy, emotional regulation, and self-reflection [56,57]. However, recent research [27,58], including this study, suggests that these benefits are context dependent. Our findings show that a higher PS is associated with increased loneliness and psychological distress during enforced isolation. This aligns with previous research suggesting that even voluntary solitude can become emotionally burdensome under restrictive conditions. This tension highlights the need to distinguish between voluntary solitude sought for restoration and involuntary solitude experienced during social constraints (eg, pandemic lockdowns [27]). Theoretically, PS may be protective only when individuals have the agency to choose when and how they are alone. When external circumstances remove that choice, the same preference may leave individuals more vulnerable, especially if it co-occurs with limited social support [32].

PS showed a small but statistically significant association with older adults’ well-being, accounting for only limited variance in loneliness and self-efficacy. In contrast, self-efficacy emerged as a stronger and more consistent predictor, reinforcing its role as an important psychological resource in later life. This finding is consistent with previous research highlighting the protective effects of self-efficacy against loneliness [59]. It may help buffer distress through various mechanisms, including enhanced coping, reduced helplessness, and a greater sense of control—factors emphasized in Bandura’s social cognitive theory [60]. While PS may offer certain benefits, its effects appear modest compared with the pronounced impact of self-efficacy. Therefore, psychosocial interventions may include elements that support PS but with realistic expectations of its overall influence. Intervention programs for older adults should prioritize the development of self-efficacy through mastery experiences, meaningful engagement, coping skills training, and supportive social environments. Such approaches are more likely to reduce loneliness and improve emotional well-being in later life.

The path analysis showed that pandemic-related experiences, including social distancing and infection exposure, had the strongest impact on self-efficacy. In contrast, PS had only a minor effect. No studies directly link PS with self-efficacy among older adults in the context of the pandemic. However, existing research [61-64] highlights the dominant role of situational factors such as lockdowns and isolation in shaping psychological outcomes, including self-efficacy. This supports our finding that dispositional factors such as PS are less central in explaining self-efficacy during the pandemic.

Another hierarchical regression analysis in this study revealed several significant predictors of psychological distress (measured by Clinical Outcomes in Routine Evaluation scores). The findings indicate that lower self-efficacy is an important risk factor for psychological distress among older adults, reinforcing existing evidence of its protective role in mental health [61-64]. PS also emerged as a significant contributor to distress, even after accounting for self-efficacy and social behaviors. This finding indicates that solitude may function as a vulnerability factor in this context [27,32]. Moderate amounts of time spent alone were associated with lower distress, suggesting a potential restorative value of solitude. In contrast, limited or superficial social interactions—particularly with acquaintances—were linked to higher levels of psychological strain. This pattern underscores the importance of relationship quality over quantity during periods of social disruption. Our findings resonate with previous research showing that emotionally close ties offer more effective emotional support than weaker social connections, particularly in times of crisis [65,66]. Weak ties, by contrast, may fail to provide meaningful support and, in some cases, may even intensify feelings of loneliness or stress [67]. Interventions aimed at reducing psychological distress in older populations should not simply encourage more social interaction but should instead prioritize the development of deeper, emotionally meaningful connections.

This study, based on a large national sample of older adults, makes a valuable scientific contribution. It highlights important risk and protective factors for mental health and self-efficacy in later life during a global crisis. One of the main contributions lies in examining the underexplored construct of PS, which sheds light on the complex role PS plays in shaping mental health outcomes during enforced social isolation. Although often beneficial, a PS may increase loneliness and distress during crises, particularly when social autonomy is limited. The study also reinforces the importance of psychological well-being and social connectedness as central to maintaining mental health and self-efficacy in older populations during disruptive events.

In practical terms, the findings can inform public health policies and guide targeted interventions aimed at preventing mental health decline among older adults during future pandemics or similar emergencies. Recognizing that loneliness and distress can arise even in individuals who prefer solitude is crucial, as it highlights the need for nuanced strategies that respect individual preferences while maintaining strong social support systems. In addition, the role of self-efficacy as a buffer against distress is significant, highlighting the importance of empowering older adults with clear information, accessible support services, and opportunities to maintain independence during crises. Interventions tailored to rural and urban differences and designed to be sensitive to gender- and age-specific vulnerabilities could further improve outcomes.

Despite its strengths, the study has several limitations. The use of snowball sampling was efficient in reaching a large number of older adults across multiple Croatian cities. However, it may have introduced selection bias and limited the representativeness of the findings. Participants were likely drawn from similar networks, resulting in a more homogeneous sample in terms of socioeconomic status, education, health awareness, and other factors. This may restrict the generalizability of the results, particularly to more vulnerable or socially isolated older adults who may have been underrepresented. This sampling method was chosen due to the lockdown conditions at the time of data collection, as epidemiological measures severely limited older adults’ ability to leave their homes. Snowball sampling was the only feasible way to reach them under those circumstances. Furthermore, the cross-sectional design prevents causal inferences regarding the observed relationships. Future research should consider using probability-based sampling to enhance representativeness, especially among marginalized or hard-to-reach populations. In addition, future research should use longitudinal designs, as they can help test causality by establishing temporal order, clarifying whether changes in solitude precede changes in psychological distress or vice versa. Such designs allow researchers to track how variables evolve and examine whether shifts in one predict shifts in another. By controlling for baseline levels and assessing potential mediators or moderators, such as social support or health status, longitudinal studies can provide deeper insight into the mechanisms underlying these relationships and offer stronger causal inferences. Finally, qualitative approaches may offer deeper insights into how older adults experience and navigate solitude, stress, and resilience during crises, supporting the development of more targeted mental health interventions.

This study examined how sociodemographic, psychological, and pandemic-related factors affected loneliness and distress in older adults during COVID-19. It also explored whether PS mediated the relationship between pandemic stressors and self-efficacy. Most of these factors affected loneliness, accounting for about one-third of its variance. This finding supports and extends previous regional findings using a nationally representative sample. Changes in social routines, PS, and self-efficacy also predicted psychological distress, explaining roughly one-third of its variance. Spending a moderate amount of time alone was linked to lower distress, while limited or superficial social interactions increased it. Self-efficacy emerged as a strong protective factor against psychological distress. Although PS had only a small mediating role between pandemic stressors and self-efficacy, it still affected mental health. Overall, situational factors had the greatest impact on mental well-being. Future interventions should focus on strengthening self-efficacy and promoting meaningful social connections among older adults during times of crisis.