Introduction
Aging is a complex process influenced by a wide range of factors, with physical activity identified as a key element for the well-being of middle-aged and older adults (Lin et al., Reference Lin, Chen, Tseng, Tsai and Tseng2020). Regular physical activity contributes to the prevention of premature death and several health conditions and is associated with improved mental health (WHO, 2018). However, more than 30% of adults remain insufficiently active (WHO, 2014) and the Global Action Plan on Physical Activity 2018–2030 recommends finding evidence-based policies to increase physical activity levels in this population (WHO, 2018). Participation in physical activity is influenced by a range of factors, encompassing individual traits as well as social, cultural, environmental, and economic aspects (WHO, 2018). To effectively increase opportunities for physical activity, a comprehensive approach is needed, demanding a deeper understanding of these dynamics. In recent years, there has been a growing focus on the association between physical inactivity and conditions such as social participation or loneliness. Studies have suggested that engaging in social activities promotes physical exercise and overall physical health (Ashida et al., Reference Ashida, Kondo and Kondo2016). This connection may be explained by the fact that social interactions often involve physical activity, such as leaving the house and participating in activities with others. Conversely, reduced social contact and isolation are linked to lower physical activity levels (De Koning et al., Reference De Koning, Richards, Wood and Stathi2021).
In the context of old age, both loneliness and physical inactivity are prevalent (Netz et al., Reference Netz, Goldsmith, Shimony, Arnon and Zeev2013) and seem to be related. Theory suggests that physical activity can facilitate social engagement, the development of social support networks, and therefore reduce loneliness (Pels and Kleinert, Reference Pels and Kleinert2016). For its part, loneliness can lead to lower self-regulation levels, which could result in health risk behaviors, including physical inactivity (Hawkley et al., Reference Hawkley, Thisted and Cacioppo2009; Peltzer and Pengpid, Reference Peltzer and Pengpid2019). Despite the apparent clarity of the relationship between loneliness and physical inactivity, the bidirectional associations between loneliness and physical inactivity remain unclear. Some studies show that physical inactivity is associated with feelings of loneliness (Beutel et al., Reference Beutel, Klein, Brähler, Reiner, Jünger, Michal, Wiltink, Wild, Münzel, Lackner and Tibubos2017; Giné-Garriga et al., Reference Giné-Garriga, Jerez-Roig, Coll-Planas, Skelton, Inzitari, Booth and Souza2021), and other studies indicate no association (Schrempft et al., Reference Schrempft, Jackowska, Hamer and Steptoe2019; Smith et al., Reference Smith, Banting, Eime, O’Sullivan and Van Uffelen2017). Therefore, additional longitudinal research is required.
Given the COVID-19 pandemic, attention has also been paid to social participation and loneliness in recent years, which has emerged as a challenge among older populations (Hwang et al., Reference Hwang, Rabheru, Peisah, Reichman and Ikeda2020; Watson-Borg et al., Reference Watson-Borg, Conn and Checkland2023). The implementation of social distancing measures has led to reduced social participation and physical activity among older adults (Salman et al., Reference Salman, Beaney, E Robb, de Jager Loots, Giannakopoulou, Udeh-Momoh, Ahmadi-Abhari, Majeed, Middleton and McGregor2021), as well as higher levels of loneliness (Freedman and Nicolle, Reference Freedman and Nicolle2020). Although there is evidence of a possible relationship between these variables, to our knowledge, there are few studies examining their reciprocal relations in adulthood. Some studies demonstrate that engagement in social participation and regular physical activity were independently associated with decreased loneliness (Gyasi et al., Reference Gyasi, Adu-Gyamfi, Obeng, Asamoah, Kisiangani, Ochieng and Appiah2021). However, longitudinal studies with bidirectional analysis are needed to test how these variables feedback on each other.
Previous research highlights the complexity of the relationship between these conditions (De Koning et al., Reference De Koning, Richards, Wood and Stathi2021), but understanding how these variables interplay over the lifespan is highly relevant because all three are related to mental and physical health in old age, and can potentially be modified (Creese et al., Reference Creese, Khan, Henley, O’Dwyer, Corbett, Vasconcelos Da Silva, Mills, Wright, Testad, Aarsland and Ballard2021). A body of cross-sectional studies shows potential bivariate relationships between social isolation, loneliness, or healthy lifestyles such as physical activity among older adults. However, prospective longitudinal research assessing these reciprocal effects in complex models in older adults has been scarce. To address the existing gap in the literature, the aim of this research is to examine reciprocal relations of physical activity, loneliness, and social participation across six years of follow-up in a representative sample of European adults over 50 years old using a Cross-Lagged Panel Model (CLPM).
Materials and methods
Sample and procedure
Data from the Survey of Health, Ageing and Retirement in Europe (SHARE) (Börsch-Supan et al., Reference Börsch-Supan, Brandt, Hunkler, Kneip, Korbmacher, Malter, Schaan, Stuck and Zuber2013) was employed in this study. SHARE is a harmonized panel study from European and Israeli citizens aged 50 years old or older. The sampling protocol follows a probabilistic sampling strategy that can vary across countries (Bethmann et al., Reference Bethmann, Bergmann and Scherpenzeel2019). Starting in 2004, SHARE counts with 8 waves of publicly available panel data. In this study, data from waves 5 (Börsch-Supan, Reference Börsch-Supan2022a), 6 (Börsch-Supan, Reference Börsch-Supan2022b), 7 (Börsch-Supan, Reference Börsch-Supan2022c), and 8 (Börsch-Supan, Reference Börsch-Supan2022d) were used. All waves were reviewed and approved by the Ethics Council of the Max Planck Society (see: https://share-eric.eu/fileadmin/user_upload/Ethics_Documentation/SHARE_ethics_approvals.pdf).
The sample was formed by all respondents who participated in wave 5 and were aged 50 years or older at that moment. This resulted in 64,887 individuals, of which 55.4% were female and 44.6% were male. Their mean age at the beginning of the study was 66.68 years old (SD = 10.03). A total of 15 countries were represented in the study: Austria (6.5%), Germany (8.6%), Sweden (7.0%), Netherlands (6.3%), Spain (10.1%), Italy (7.2%), France (6.8%), Denmark (6.3%), Switzerland (4.6%), Belgium (8.5%), Israel (3.9%), Czech Republic (8.5%), Luxembourg (2.4%), Slovenia (4.5%), and Estonia (8.8%).
Instruments
Social participation included individuals’ participation in four different activities during the previous year. The activities considered were: doing voluntary or charity work, attending educational or training courses, going to sport/social/other clubs, and taking part in political or community-related organizations coded as 1 (yes) or 0 (no). The social participation index was computed as the sum of the activities, responses ranged between 0 (did not participate in any of the activities) and 4 (participated in all considered activities).
Loneliness was assessed using the Three-Item Loneliness Scale (Hughes et al., Reference Hughes, Waite, Hawkley and Cacioppo2004). This scale considers feelings of lack of companionship, isolation, and exclusion as indicators of loneliness. Each indicator is answered using a three-point Likert scale: 1 = Hardly ever or never, 2 = Some of the time, and 3 = Often. The total loneliness score is computed as the sum of the three items and hence ranged from 3 (least lonely) to 9 (loneliest).
Physical inactivity was measured as a binary indicator merging two variables asking about participation in moderate and vigorous physical activities. For each variable, the individual was asked to report their frequency of enrolling in that kind of activity, using a four-point Likert scale (1 = More than once a week, 2 = Once a week, 3 = One to three times a month, and 4 = Hardly ever or never). Individuals who reported hardly ever or never engaging in either moderate or vigorous activities were considered physically inactive (1), and the rest were not (0). This operationalization has been previously reported in other studies using SHARE data (Matos et al., Reference Matos, Barbosa, Cunha, Voss and Correia2021).
In addition to the three measures employed over time, age, gender, and disability were also considered as time-invariant control variables. Disability was measured as a binary marker of whether the individual was limited in activities because of health (1) or not (0), based on the Global Activity Limitation Index (Van Oyen et al., Reference Van Oyen, Van der Heyden, Perenboom and Jagger2006).
Statistical analyses
The statistical analyses include the calculation of descriptive statistics and correlations in SPSS 28) and structural models in Mplus 8.9 (Muthén and Muthén, Reference Muthén and Muthén2011). The specific structural equation model employed was the CLPM, a model widely used to analyze the relationships between two or more variables longitudinal measured for two or more occasions. In this model, the variables are measured at each time point and the model examines the relationships between the variables across time. The “cross-lagged” aspect of the model refers to the fact that the model examines the lagged effects of one variable on another and the lagged effects of the other variable on the first. It allows examining the directionality and longitudinal relationships between the variables. The panel model refers to the fact that the same individuals are measured at multiple time points, making it possible to examine the autoregressive paths, the stability in the variables over time, and to control for individual differences. Model fit of the structural models was assessed with the most prevalent statistics and indices: a) the chi-square test; the Comparative Fit Index (CFI); the Root Mean Squared Error of Approximation (RMSEA); and the Standardized Root Mean Squared residual (SRMR) (Kline, Reference Kline2023), a CFI of .95 or higher, an RMSEA less than .06, and an SRMR less than .08 together can be considered an excellent fit (Hu and Bentler, Reference Hu and Bentler1999).
Results
Descriptive statistics and correlations among variables
Descriptive statistics of all variables in the study and for all waves are presented in Table 1. Additionally, correlations among the measures of social participation, loneliness, and physical inactivity are presented in Table 2.
Table 1. Means, percentages, standard deviations minimum, and maximum of the variables in the study
SD, Standard deviation; Min., Minimum; Max., Maximum; SP1 to SP4: Social Participation waves 5 to 8 SHARE; LO1 To LO4: Loneliness waves 5 to 8 SHARE; PI1 to PI4: Physical inactivity waves 5 to 8 SHARE.
Table 2. Correlations among social participation, loneliness, and physical inactivity across all waves of the study
**= p < .001; SP1 to SP4: Social Participation waves 5 to 8 SHARE; LO1 To LO4: Loneliness waves 5 to 8 SHARE; PI1 to PI4: Physical inactivity waves 5 to 8 SHARE.
Cross-lagged panel models
CLPMs for examining reciprocal effects of social participation, loneliness, and physical inactivity using four waves of the SHARE. We followed this modeling strategy: Firstly, we estimated the CLPM freely estimating within-wave associations and autoregressive and cross-lagged effects between adjacent waves. Secondly, we are constrained to equality of the autoregressive paths. If this model does not deteriorate model fit compared to the first model, it means that the stability (or lack of) across waves is the same. Thirdly, we further restricted cross-lagged effects to equality across waves. Again, if this model fit remains the same as the first model, it means that the effects among the variables of interest are constant across waves. All constraints were tested in the unstandardized coefficients. These three models controlled for age, gender, and disability.
Table 3 offers model fit indexes for this sequence of models. The model with autoregressive effects constrained to equality did not deteriorate model fit compared to the freely estimated model. For example, the differences in CFI’s were low (.007), and therefore, this means that stability of the autoregressive models is maintained across waves of data. In the same vein, when all cross-lagged effects were constrained to equality across waves, the fit not only not deteriorated but improved. Both the SRMR and the RMSEA improved, while the decrease in the CFI was negligible (.005). Overall, the best CLPM is the one with equal stability and cross-lagged effects across waves, which means that the effects among the three variables of interest remain constant across them.
Table 3. Model fit indexes
ΔCFI, differences between CFIs always against the freely estimated model; CLPM, Cross-lagged panel model.
As already said, age, gender, and disability were control variables in the CLPM. Table 4 shows all these standardized effects. In general, age had positive effects on social participation and positive ones on loneliness and physical inactivity. Women participated less than men in social activities, were less active, and had more feelings of loneliness. Finally, old adults limited were less socially participative and less active, while having more feelings of loneliness. All these effects decreased when social participation, loneliness, and physical inactivity were also predictors. That is, in waves 6, 7, and 8.
Table 4. Standardized effects of the control variables (age, gender, and activity limitations) on social participation, loneliness and physical inactivity across waves
Note: all estimates are statistically significant (p < .05) unless ns (not significant) is marked.SP1 to SP4: Social participation waves 5 to 8 SHARE; LO1 To LO4:Loneliness waves 5 to 8 SHARE; PI1 to PI4: Physical inactivity waves 5 to 8 SHARE.
Regarding the within-waves associations, in almost all cases were statistically significant (p < .05), but of small magnitude. Within wave 5, social participation correlated −.134 with loneliness and −.34 with physical inactivity, and physical inactivity correlated .151 with loneliness. The same correlations in waves 6, 7, and 8 were: .005 (p > .05), .084, and .061 for wave 6; .074, .121, and .062 for wave 7; and .006 (p > .05), −.021 (p > .05), and .015 (p > .05).
The autoregressive and cross-lagged standardized effects are presented in Figure 1. These estimates are the main ones for the aim of this research, and it is important to bear in mind that all effects are equal across waves. The autoregressive effects for the three variables show great stability across waves. Regarding the cross-lagged effects, all of them are statistically significant. However, the magnitude of the relationships shows a pretty clear pattern. On one hand, psychical inactivity and social participation had medium to large effects on each other (Orth et al., Reference Orth, Meier, Bühler, Dapp, Krauss, Messerli and Robins2022). Indeed, the effect of social participation on physical inactivity was −.144 (95% CI: −.134, −.153), while the effect of physical inactivity on social participation was −.156 (95% CI: −.143, −.169). On the other hand, the impact of loneliness on social participation and physical as well as the impact of these two variables on loneliness were much smaller. In these cases, the effects were small to medium in size. Specifically, the effect of loneliness on social participation was −.038 (95% CI: −.032, −.045) and the effect on physical inactivity was .051 (95% CI: .041, .061), while the effect of social participation on loneliness was −.045 (95% CI: −.037, −.052), and finally, the effect of physical inactivity on loneliness was .046 (95% CI: .036, .055).
Figure 1. Standardized estimates of the autoregressive and cross-lagged effects of the best fitting cross-lagged panel model. Autoregressive effects are in gray and cross-lagged effects in black, within waves associations and effects of the control variables not shown for the shake of clarity; all estimates are statistically significant (p < .05). SP1 to SP4: Social Participation waves 5 to 8 SHARE; LO1 to LO4: Loneliness waves 5 to 8 SHARE; PI1 to PI4: Physical inactivity waves 5 to 8 SHARE. Standardized coefficients between parentheses. It must be borne in mind that equality of parameters is stablished with non-standardized parameters.
Discussion
As far as we know, this is the first study to examine the longitudinal relationship between social participation, physical activity, and loneliness. Our model confirmed previous research findings that all these variables were interrelated at each time point, but with a clear pattern across waves, social participation, and physical activity feedback each other, both are related bidirectionally. However, loneliness has less predictive capacity in the model and is less related to the other variables.
Our findings showed that effects were consistent across waves, indicating that the impact of loneliness, physical activity, and social participation tends to remain stable over time. In this sense, the effects of being physically inactive, having feelings of loneliness, or lacking social engagement at younger ages do not usually change over the years. Therefore, encouraging physical activity or social participation at an early stage is not only beneficial at that moment but can also establish positive changes with long-term benefits for healthy aging trajectories. Although motivating older people to start and maintain regular physical activity can be challenging, there are some programs as the neighborhood-based walking programs that proved to be effective (Iolascon et al., Reference Iolascon, de Sire, Calafiore, Benedetti, Cisari, Letizia Mauro, Migliaccio, Nuti, Resmini, Gonnelli and Moretti2020) and were easy to implement.
Being physically active is especially important at old age for maintaining the independence (Hirsch et al., Reference Hirsch, Diehr, Newman, Gerrior, Pratt, Lebowitz and Jackson2010), mental health, and well-being (Du et al., Reference Du, Tan, Yi, Zou, Gao, Zhao and Wang2015). Given the importance of promoting physical activity, various studies have analyzed the impact of different strategies on sedentary lifestyle, including social participation. The evidence suggests that social interaction is the most relevant interpersonal motivator, including communication with friends or others, peer support, and exercising with friends, among others (Yarmohammadi et al., Reference Yarmohammadi, Saadati, Ghaffari and Ramezankhani2019). There is also evidence that social isolation is associated with reduced levels of physical activity (De Koning et al., Reference De Koning, Richards, Wood and Stathi2021) and that, in early life, it predicts future physical inactivity (Caspi et al., Reference Caspi, Harrington, Moffitt, Milne and Poulton2006). On the other hand, it is stated that physical activity can facilitate social engagement and the development of social support networks (Pels and Kleinert, Reference Pels and Kleinert2016). It is also interesting to note that much of the physical activity in older adults is accumulated through short trips outside of the home such as visiting relatives and friends, exercise classes, attending cultural events, or walking (Schrempft et al., Reference Schrempft, Jackowska, Hamer and Steptoe2019). All in all, this research supports the hypothesis suggested by a cross-sectional study (Salman et al., Reference Salman, Beaney, E Robb, de Jager Loots, Giannakopoulou, Udeh-Momoh, Ahmadi-Abhari, Majeed, Middleton and McGregor2021), confirming the bidirectional relationships between social participation and levels of physical activity.
A body of cross-sectional studies demonstrates relationships between social isolation, loneliness, and healthy lifestyles such as physical activity among older adults (Kobayashi and Steptoe, Reference Kobayashi and Steptoe2018; Salman et al., Reference Salman, Beaney, E Robb, de Jager Loots, Giannakopoulou, Udeh-Momoh, Ahmadi-Abhari, Majeed, Middleton and McGregor2021), although prospective longitudinal research assessing these reciprocal effects in later life has been scarce. Social isolation and loneliness were associated cross-sectionally with low levels of physical activity (Shankar et al., Reference Shankar, McMunn, Banks and Steptoe2011). The interaction effects between social isolation and loneliness were attenuated by 50% after adjustment for health behaviors including physical activity (Hakulinen et al., Reference Hakulinen, Pulkki-Råback, Virtanen, Jokela, Kivimäki and Elovainio2018). Previous studies showed that social isolation, but not loneliness, was associated with sedentary behaviors (Schrempft et al., Reference Schrempft, Jackowska, Hamer and Steptoe2019; Tully et al., Reference Tully, McMullan, Blackburn, Wilson, Coll-Planas, Deidda, Caserotti and Rothenbacher2019). In addition, a longitudinal study of older English adults over 10 years showed that social isolation, but not loneliness, had an impact on health-related behaviors as being physically active (Kobayashi and Steptoe, Reference Kobayashi and Steptoe2018). These findings are in line with the notion that social isolation has stronger links with physical inactivity than loneliness does, supporting the results of the present study.
Our results also showed that lower levels of social participation and higher levels of physical inactivity predicted loneliness, as in previous research (Zhao and Wu, Reference Zhao and Wu2022). These results have practical implications, as physical activity and social participation are recommended as possible multidisciplinary strategies to reduce loneliness in older people (Gyasi et al., Reference Gyasi, Adu-Gyamfi, Obeng, Asamoah, Kisiangani, Ochieng and Appiah2021; Vancampfort et al., Reference Vancampfort, Lara, Smith, Rosenbaum, Firth, Stubbs, Hallgren and Koyanagi2019). However, the impact of social and physical activity on loneliness, as well as the impact of loneliness on these two variables, was small compared to the impact found between the other two variables. These findings indicated a weak relationship between loneliness and social isolation (Shankar et al., Reference Shankar, McMunn, Banks and Steptoe2011). One possible explanation for these results is that not everyone benefits from these activities to alleviate feelings of loneliness; some studies have considered individual factors and found that for those who are extraverted, high social involvement reduces loneliness, whereas less social participation is better for those who are introverted, have social anxiety, or enjoy being alone (Schutter et al., Reference Schutter, Koorevaar, Holwerda, Stek, Dekker and Comijs2020).
This study has several strengths, including the use of a large and representative sample of older adults, which provided greater statistical power. Another strength is the long six-year follow-up of participants, as well as the control for confounding factors such as age, gender, and disability. However, there are several potential limitations that need to be acknowledged. Firstly, our data is based on self-reported questionnaires, which may introduce reporting bias. Secondly, while our research attempted to control for the most relevant confounding variables, future studies should consider other variables such as participants' levels of depression, which have been associated with loneliness and low social and physical activity (Wang et al., Reference Wang, Bishwajit, Zhou, Wu, Feng, Tang, Chen, Shaw, Wu, Song, Fu, Feng and Naugle2019). In our study, social participation was measured as a sum of attendance to certain social activities, but other aspects such as the frequency of participation that have also been linked to better mental health (Tomioka et al., Reference Tomioka, Kurumatani and Hosoi2017) could be considered in future studies.
Conclusions
This study found a significant bidirectional relationship between social participation, physical inactivity, and loneliness in older adults with an especially large effect between physical inactivity and social participation. Furthermore, these effects remained stable across aging trajectories. Thus, strategies aimed at promoting social participation and physical activity in earlier stages of age should be considered mutual goals to promote quality aging. Additionally, older adults who experience loneliness, although this condition requires a person-centered intervention, could benefit from attending social participation activities or engaging in physical activity.
Conflicts of interest
None.
Source of funding
This work was supported by project PID2021-124418OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by “ERDF A way of making Europe.” Zaira Torres is a researcher beneficiary of the FPU program from the Spanish Ministry of Universities “[grant number FPU20/02482].”
Description of authors’ roles
ZT, JMT, and TSM designed the study. JMT and IF analyzed the data. JMT and IF wrote the results section. ZT, TSM, and NPS wrote the paper. All authors provided critical input on the written manuscript and data interpretation.
Acknowledgments
This paper uses data from SHARE Waves 5, 6, 7, 8 (DOIs: 10.6103/SHARE.w5.800, 10.6103/SHARE.w6.800, 10.6103/SHARE.w7.800, 10.6103/SHARE.w8.800,) see Börsch-Supan et al. (2013) for methodological details. The SHARE data collection has been funded by the European Commission, DG RTD through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812), FP7 (SHARE-PREP: GA N°211909, SHARE-LEAP: GA N°227822, SHARE M4: GA N°261982, DASISH: GA N°283646) and Horizon 2020 (SHARE-DEV3: GA N°676536, SHARE-COHESION: GA N°870628, SERISS: GA N°654221, SSHOC: GA N°823782, SHARE-COVID19: GA N°101015924) and by DG Employment, Social Affairs & Inclusion through VS 2015/0195, VS 2016/0135, VS 2018/0285, VS 2019/0332, and VS 2020/0313. Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C, RAG052527A), and from various national funding sources is gratefully acknowledged (see www.share-project.org).
Open access
