Participants were adults ≥50 years old who attended mobile health screening clinics at various locations across Iowa. Participants were invited to complete an optional PA questionnaire while waiting for their health screening. The survey probed PA participation and primary theoretical constructs underlying PA behavior (i.e., socio-ecological model, social cognitive theory, theory of planned behavior, and motives for PA). This study was approved by the Institutional Review Board and participants provided informed consent prior to data collection. Questionnaires were provided in the same order to each participant, and the questionnaires took approximately 10–15 min to administer. Participants were encouraged to contact the research team with any questions or concerns.

The U.S. Health Resources and Services Administration guidelines (U.S. Health Resources and Services Administration, 2022), which combine US Census, Office of Management and Budget, and Rural–Urban Commuting Area classifications, were used to identify rural areas as described in previous studies (Chrisman et al., 2014; Abildso et al., 2021). Participants living in zip codes identified to be rural were assigned a “rural” label while those outside of these zip codes were assigned an “urban” label.

We assessed numerous psychological variables that have previously been shown to be associated with PA. We included psychological questionnaires, described in more detail below, that assessed motivation, intention, self-efficacy, barriers, enablers, social support, perceptions, and attitudes toward PA. Given the theoretical similarities between the psychological scales, a principal components analysis (PCA) was used to identify distinct psychological correlates of PA and the resulting components used as the primary predictors are described below.

Participants self-reported sociodemographic and health characteristics including age, gender, education, marital status, health rating, height, and weight.

Given the theoretical similarities between the psychological scales, a principal components analysis (PCA) was used to identify distinct psychological correlates of PA. Missing data were addressed at the individual measure level, with individual measure scores imputed from the average of completed rural and urban scales. The number of imputations varied by scale: PASE: 1 (1.1%), barriers self-efficacy: 3 (3.2%), PA self-efficacy: 4 (4.2%), instrumental attitude: 9 (9.5%), affective attitude: 10 (10.5%), perceived behavioral control: 2 (2.1%), intention: 2 (2.1%), pros: 4 (4.2%), social support- family: 7 (7.4%), social support- friends: 8 (8.4%), motives- interest: 4 (4.2%), motives- competence: 3 (3.2%), motives- appearance: 4 (4.2%), motives- fitness: 4 (4.2%), and motives-social: 4 (4.2%). Imputation was performed prior to PCA analysis. As our sample size precludes inclusion of numerous covariates, demographic and health-related variables were examined for inclusion as covariates in the linear regression models based on relationships with PA using bivariate correlations. Following PCA analysis, individual linear regression models assessing the associations of the PCA components with PA level (PASE) were performed for the combined group, as well as separate rural and urban subgroups.

A PCA was performed with the summary scores of 16 measures of hypothesized correlates of PA in 95 participants. The cons of PA subscale and subjective norm item were dropped from the analysis as neither had a correlation coefficient >0.3 with any other scale, leaving 14 measures in the final PCA. The data was assessed for suitability of PCA prior to analysis. The correlation matrix demonstrated that all remaining measures had at least one correlation coefficient greater than 0.3 and the overall Kaiser–Meyer–Olkin (KMO) value was 0.84. Bartlett’s test of sphericity was statistically significant (p < 0.001), indicating data factorizability. The PCA reported four components with eigenvalues greater than one which explained 45.91, 11.32, 10.00, and 9.33% of the total variances, respectively. Scree plot visualization suggested the retention of four components (Cattell, 2023). Therefore, four components were retained.

The four-component model explained 76.56% of the total variance. A Varimax with orthogonal rotation assisted with interpretation. The interpretation of the data was consistent with identified correlates of PA with strong loadings of motivation for PA on Component 1, PA self-efficacy on Component 2, social support for PA on Component 3, and attitudes toward PA on Component 4. PCA results can be found in Table 1.

Bivariate correlations between PA and age, gender, education, marital status, overall health, and body mass index (by calculating weight (kg)/height (m²)) were each assessed, but only age was significantly correlated with PA [r(95) = −0.44, p < 0.001]. Therefore, only age was included as a covariate in the primary regression models.

Table 2 lists baseline characteristics of the rural and urban participants. In general, urban participants were more likely to be male with greater educational attainment than rural participants. Rural participants were more likely to be married, while urban participants were more likely to be widowed. There was no significant difference in PA levels between urban and rural adults [urban M(SD) = 175.12 (82.65); rural M(SD) = 168.70 (105.69); t(93) = −0.33; p = 0.742]. Middle aged adults (<65 years old) reported higher levels of PA compared to older individuals (≥65 years old) [t(94) = 4.09; p < 0.001]. Rural adults reported more positive attitudes toward PA than urban adults, as assessed by Component 4 of the PCA [t(92) = 3.90; p < 0.001] and affective attitude [t(93) = 3.14; p = 0.002] and instrumental attitude [t(93) = 2.17; p = 0.032] subscales. Urban adults reported significantly greater PA self-efficacy [t(93) = −2.22; p = 0.029]. All other individual scale comparisons were not significant between rural and urban groups.

Results of the combined final model indicated that age [β = −0.39, t(93) = −4.24, p < 0.001] and PA self-efficacy [β = 0.29, t(93) = 3.36, p = 0.001] significantly contributed to PA level (Table 3). For the rural sample, age [β = −0.31, t(46) = −2.54, p = 0.015], self-efficacy [β = 0.28 t(46) = 2.41, p = 0.020], social support [β = 0.34, t(46) = 2.78, p = 0.008], and attitudes toward PA [β = 0.29, t(46) = 2.50, p = 0.017] were each significantly associated with PA level. In the urban sample, only age [β = −0.51, t(47) = −3.63, p < 0.001] was significantly associated with PA level. Detailed regression results can be found in Table 3.

Gender and education were explored as covariates in conjunction with age in the regression models given their established association with physical activity (Plotnikoff et al., 2004). With these added covariates, the attitudes toward PA component was also significantly associated with PA (p = 0.021) in the combined model. There were no significantly different conclusions in the rural or urban stratified models.

In an additional sensitivity analysis, we assessed associations of available neighborhood environment-related subscales with PA. Neighborhood Environment Walkability Scale (Saelens et al., 2003; Cerin et al., 2006) subscales were available for a subset of participants in our sample (residential density n = 78; land use mix-access n = 60; neighborhood satisfaction n = 89). Neither residential density, described as the perceived density of residences in an area, or land use mix-access, defined as the perceived ability to access places nearby through PA, differed between rural and urban groups [t(76) = 0.06, p = 0.949; t(58) = −0.60, p = 0.552]. However, rural adults reported greater neighborhood satisfaction than urban adults [t(87) = 2.45, p = 0.016]. Therefore, neighborhood satisfaction was included in a regression model with age and the PCA components. In the combined model, social support was significantly associated with PA in addition to age and self-efficacy. Social support (p = 0.012) and attitudes toward PA (p = 0.028) remained significant in the rural model. There were no significant predictors in the urban subsample (ps > 0.05).

This study aimed to identify psychological correlates of PA in rural and urban middle aged and older adults. Middle aged adults reported higher levels of PA compared older adults, fitting with prior literature (Spartano et al., 2019). We found similar levels of PA in rural and urban participants, which contrasts with other studies reporting that rural residents engage in less PA than their urban counterparts (Cohen et al., 2018; Pelletier et al., 2021; Whitfield et al., 2023). Possible reasons for similar PA levels in our rural and urban participants include similar perceived residential density and land use mix-access, as well as greater neighborhood satisfaction among rural participants although this was not associated with PA in any final regression models. Lack of perceived differences in location accessibility and residential density, as well as greater neighborhood satisfaction may have caused our rural subset to be more active than rural groups in previous literature (Stronegger et al., 2010; Wang et al., 2019). Age was also the strongest predictor of PA levels, and age was similar between our urban and rural samples, which has not always been the case in previous studies (Parks et al., 2003; Martin et al., 2005). Finally, while we followed HRSA rural–urban definitions, the range of populations for the rural and urban locales may not have been as extreme as in other studies. For example, the ‘urban’ zip codes in this study had an average population of 211,754 people while the rural areas averaged 25,934 people (US Census Bureau, 2010; U.S. Health Resources and Services Administration, 2022). Previous work reporting on rural–urban differences in PA has included much larger metro areas of 1–5+ million residents (US Census Bureau, 2010; Cohen et al., 2018). Thus, it is quite possible that Iowa is relatively more homogenous in its population density, which could contribute to less separation in rural and urban characteristics.

Despite finding no differences in self-reported PA levels, different psychological correlates of PA were identified in stratified analyses performed with rural and urban participants. Among rural participants, greater social support, and self-efficacy, as well as more positive attitudes toward PA were associated with greater amounts of PA. This finding aligns with reports of social support barriers associated with lower PA levels among rural adults, while not urban adults (Pelletier et al., 2022). Among urban participants, however, none of the psychological correlates were significantly associated with PA. These results suggest that the association between psychological correlates and PA may differ based on rural or urban residence. Resources targeting psychological correlates of PA may be particularly impactful among rural adults, while PA may be more affected by other factors in urban adults such as the individual’s physical health and environment (Weiss et al., 2007; Sallis et al., 2012).

Interestingly, motivation for exercise, while the leading component identified by the PCA analysis accounting for nearly half of the total variance, was not significantly associated with PA in the combined, rural, or urban models conflicting with findings in prior rural and urban samples (Pelletier et al., 2022). This finding suggests that motivation, which has been viewed as a vital predictor of PA performance and public health promotion target (Teixeira et al., 2012; Molanorouzi et al., 2015), may not be the primary driver of PA behavior in middle-aged and older adults. However, our results do fit with other literature demonstrating that motivation is a weaker predictor of PA than other psychological correlates, perhaps because motivation can be transient (Ryan and Deci, 2000; Braver et al., 2014). Social cohesion has been noted among many rural communities (Avery et al., 2021), and considering group attitudes and social support may be particularly important among individuals living in rural areas. On the other hand, urban areas may have more built-in support for PA with extensive infrastructure for PA, making social support and group attitudes less important in predicting PA level among urban adults. Our results indicate that considering the environment and characteristics of the targeted population and targeting self-efficacy, social support, and attitudes toward PA may produce more effective health promotion campaigns. While many PA promotion programs have targeted the built environment and motivation for PA (Brownson et al., 2009; Ferdinand et al., 2012; Sallis et al., 2020), our findings indicate that interventions targeting self-efficacy, social support, and attitudes toward PA may be more impactful, particularly among rural adults. Further research is needed to develop these intervention programs.

There were limitations to the present study. The study was cross-sectional, and it is possible that higher levels of PA may exert reciprocal influences on positive psychosocial correlates of PA, distinct from the impact of rural or urban setting. The sample size was limited, and our findings cannot be extrapolated to all rural and urban middle-aged and older adults. PA was self-reported, which is prone to reporting bias.