Article Text

Download PDFPDF
Sense of community-belonging and health-behaviour change in Canada
  1. Perry Hystad1,
  2. Richard M Carpiano2
  1. 1School of Population and Public Health, University of British Columbia, Vancouver, Canada
  2. 2Department of Sociology, University of British Columbia, Vancouver, Canada
  1. Correspondence to Perry Hystad, School of Population and Public Health, University of British Columbia, James Mather Building, 5804 Fairview Avenue, Vancouver, BC, V6T 1Z3 Canada; phystad{at}gmail.com

Abstract

Background Research indicates that primary prevention targeting individual behaviours should incorporate contextual factors. The objectives of this study are to examine the role of community-belonging and contextual factors on health-behaviour change in Canada, and whether the influence of community-belonging on behaviour change varies by specific types of behaviours and contextual factors.

Methods Data on individual-level community-belonging, socio-demographics and self-rated health were obtained for 119 693 respondents from the 2007/2008 Canadian Community Health Survey located within 100 health regions across Canada. Contextual factors were based on health-region groupings of socio-economic determinants of health. Multilevel models were used to estimate the influence of community-belonging and health-region contextual factors on general, and specific, health-behaviour changes in the past year.

Results After controlling for individual and contextual factors, community-belonging showed a positive dose–response relationship with health-behaviour change. Health-region contextual factors were only slightly associated with behaviour change; however, the influence of community-belonging on behaviour change showed significant variability based on health-region contextual factors. The influence of community-belonging also varied by specific health-behaviour changes, but for most prominent health behaviours (exercise, weight loss and improved diet) the effect was consistent.

Conclusions Community-belonging was strongly related to health-behaviour change in Canada and may be an important component of population health prevention strategies. Efforts to increase community-belonging, however, need to be considered along with contextual factors.

  • Health-behaviour change
  • community-belonging
  • contextual factors
  • social capital
  • primary prevention
  • health behaviour
  • multilevel modelling

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Prevention efforts to improve health behaviours have traditionally focused on individual decisions and choices. This approach has changed with the recognition of the need to move behavioural prevention beyond simply the individual and consider contextual factors, such as community social and physical environments, that may promote or constrain the practice of healthy lifestyles.1 To date, a number of studies have identified disparities in the prevalence of health behaviours (eg, physical activity, smoking, diet, preventive medical screening) across rural, suburban and urban contexts.2–5 In an effort to understand how such environments may be health-promoting or health-inhibiting, further research has identified a broad range of contextual factors, such as availability of amenities and services,4 6 7 community social and economic conditions,8 resident social networks9 and the built environment,10 11 that are associated with health behaviours. The majority of studies, however, tend to focus on specific behaviours in isolation (eg, smoking,12 physical activity,13 14 overweight15 or diet16 17) with little research focusing specifically on health-behaviour change in general. While focusing on specific behaviours is a necessary strategy, identifying how contextual factors may improve health-behaviour change in general has important implications for a variety of prevention initiatives that employ a population strategy to promote healthy lifestyles,18 19 and as a foundation for more specifically targeted behaviour-prevention initiatives.

Community-belonging—the degree to which an individual is, or perceives to be, connected to their community—has been a central element of a variety of theoretical conceptualisations regarding how community contextual factors might be linked to health-related behaviours.20 21 Community-belonging may influence the likelihood of undertaking behavioural changes through: (1) the exposure to health-related behaviour norms and attitudes in the community; (2) psychosocial mechanisms such as self-esteem, social status, control and social stress; and (3) access to material and other types of community resources.22–24 Analyses of Canadian national data found that increased community-belonging was associated with higher self-rated general and mental health.22 25 This research, however, only examined the direct relationship between community-belonging and health. Consistent with theories of how community-belonging and related constructs link individuals to contextual conditions and their inherent resources or deficits, the magnitude and direction of the influence of community-belonging on health outcomes may vary with respect to not only community context but also the health outcome, or behaviour, of interest. Some US findings indicate that, depending on the social context of the community in which one lives, increased community-belonging may be positively or negatively associated with health outcomes when it is measured either in terms of one's degree of actual community ties21 24 or as one's psychological attachment to community.26 Thus, it is important to understand how health-behaviour change may vary, based on the interaction between community-belonging and different contextual factors.

In light of this prior research, we explore the role of sense of community-belonging on health-behaviour change by conducting multilevel analyses of health-region-level contextual factors and individual-level data from a national Canadian population health survey. Health regions are responsible for administering health-promotion and disease-prevention programmes to residents within a particular area of Canada. This study tests three hypotheses. First, similar to previous Canadian studies examining other health outcomes, we hypothesise that an individual's level of community-belonging will be positively associated with the likelihood of undertaking a health-behaviour change. Second, we expect to observe health-region-specific variations in the likelihood of individual health-behaviour change such that health regions characterised by greater rurality, residential instability and socio-economic disadvantage will be associated with a lower likelihood of residents reporting health-behaviour change. Third, in light of prior studies examining the interaction between context and community attachment, we hypothesise that the influence of community-belonging on health-behaviour change will be moderated by these health-region-level contextual characteristics. In addition to testing these specific hypotheses, we also evaluate to what degree the association between community-belonging and health-behaviour change reflects specific health behaviours.

This study contributes to the existing literature by: (1) determining how community-belonging and regional contextual conditions may independently influence the likelihood of individuals undertaking behavioural change in general, an important step for improving their overall health and well-being, and as such a potential platform for population-based health interventions; (2) exploring how the health influence of community-belonging may change based on regional contextual factors; and (3) examining variation in the influence of community-belonging on different health behaviours.

Methods

Data sources

Two levels of data are used in this study: individual-level data were obtained from the 2007/2008 Canadian Community Health Survey (CCHS) Cycle 4.1, and health-region data were obtained from 2007 Statistics Canada peer groupings. Ethical approval for the use of these data was covered by the publicly available data clause of the authors' academic institution. The CCHS is a national cross-sectional survey conducted by Statistics Canada to provide estimates of health determinants, health status and health-system utilisation. The CCHS uses a probability sample at the health-region level. Therefore, for this research, weighted percentages are reported, and probability weights are included in all analyses. The survey includes household residents aged 12 and older, and is representative of the Canadian population by age and gender at the health-region level. The CCHS Cycle 4.1 was administered throughout all regions in Canada during 2007/2008. After removing non-responses for all variables, there were 119 693 survey responses located within 100 health regions available for analyses.

Individual-level measures

Health-behaviour change is the primary outcome variable of this analysis and was assessed using a single, dichotomous scale item asking respondents whether or not they had done something to improve their health over the past 12 months. Respondents who reported improving their health were also asked what the single most important change was: more exercise, lost weight, changed diet, changed smoking, reduced alcohol, reduced stress, medical treatment, took vitamins and other. Each of these nine behaviour changes was included as a dependent variable in subsequent models examining specific types of behaviour change (each coded 1=yes, 0=no).

Our primary individual-level independent variable of interest is community-belonging, which consisted of a single item asking respondents ‘How would you describe your sense of belonging to your local community?’ and was measured using a four-point Likert scale (1=very strong to 4=very weak) (see also Ross22). Kim and Kaplan27 conceptualise sense of community as: (1) attachment to place, (2) community identity and (3) social interaction. While the item used here, sense of community-belonging, cannot be examined further in terms of these three dimensions using other measures included in the CCHS, the overall item has been shown elsewhere as an important determinant of self-reported health.22

We also examined several potential confounding factors that can be theorised to influence one's ability, capacity and/or need to undertake behavioural changes: gender, age, household median income, education, marital status, visible minority status (coded Caucasian and non-Caucasian (aboriginal population not included)), immigrant status and self-rated health. The non-response for individual household income was 13%, and individuals who did not report income were 12% less likely (OR: 0.88, 95% CI 0.85 to 0.91) to have undertaken health-behaviour changes compared with those that reported income. We therefore included a ‘missing income’ category within the household income variable. Overall, the weighted percentage of the sample removed due to missing data (ie, refusal or not stated in either dependent of independent variables) was 8.2%.

Health-region-level variables

Health regions are a conceptually important geographic area to consider for this research, as they are responsible for the health-promotion and disease-prevention programmes in a particular area of Canada. These regions represent varying scales of geography and contain populations ranging from approximately 2.5 million (city of Toronto) to 80 000 (northern, rural areas).

Owing to the contextual variation that exists across health regions, we utilised Statistics Canada 2007 peer groups to assess health-region contextual factors.28 Peer groups were created using k-means clustering to achieve maximum statistical differentiation between health regions based on 24 variables representing social and economic determinants of health, using data collected at the health-region level mostly from the Census of Canada. For a detailed description of the statistical methods used to determine peer groups, see Statistics Canada.28 Table 1 describes the principal characteristics of the resulting health-region peer groups in Canada. Peer group F was not included in analyses owing to CCHS health-region sampling. For our analyses, we use Peer group B as the reference category, as this group generally represents average-sized urban centres in Canada.

Table 1

Principle socio-economic characteristics of health-region peer groups in Canada

Data analyses

Health-behaviour change was modelled using both individual characteristics and health-region peer groups. To account for clustering within health regions, random intercept models were created using the GLIMMIX procedure in SAS 9.1.

Our analyses proceeded in four steps. First, we explored unadjusted associations between individual variables, health-region peer groups and health-behaviour change in general. Second, we analysed how community-belonging and health-region peer group context are associated with health-behaviour change, adjusting for individual sociodemographic characteristics and self-rated health. Third, we examined whether the influence of community-belonging varies by specific types of behaviour change by estimating the final adjusted model for each specific type of behaviour change reported by respondents as the most important behaviour change they had undertaken in the past year. Fourth, we included a fixed-effect interaction term between peer group and community-belonging class variables into the final model to test for interaction and then stratified the final adjusted model by health-region peer groups to further examine how the influence of community-belonging on health-behaviour change varied based on health-region peer groups.

Results

Table 2 provides the descriptive statistics for individuals reporting health-behaviour change in general and the most important behavioural changes reported. Approximately 58% of individuals reported undertaking a health-behaviour change, but this varied by health region, from a low of 48% to a high of 64%. Among respondents reporting a health-behaviour change, over 52% reported ‘more exercise’ as the most important health-behaviour change they undertook. Overall, approximately 80% of those reporting a health-behaviour change listed more exercise, lost weight (10.4%) or changed diet (16.6%).

Table 2

Sample size and weighted percentages for health-behaviour change in general and the most important behavioural changes reported

Table 3 reports the sample size, ORs and 95% CIs for the unadjusted and adjusted models for behavioural change in general, incorporating individual sociodemographic and self-rated health variables, community-belonging and health-region peer groups.

Table 3

ORs of health-behaviour change associated with community-belonging, health-region peer groups, and individual characteristics

Influence of individual-level variables on health-behaviour change

A strong dose–response relationship between community-belonging and health-behaviour change remained after adjusting for individual sociodemographic variables, self-rated health and health-region peer groups. Significantly higher odds of having undertaken a health-behaviour change in the last 12 months were found among respondents who reported their community-belonging as being somewhat weak (OR=1.10; 95% CI 1.06 to 1.15), somewhat strong (1.26; 1.20 to 1.31) and very strong levels of community-belonging (1.38; 1.32 to 1.45), compared with individuals reporting very weak community-belonging. The trend for the effects of community-belonging was statistically significant (p<0.001).

Influence of health-region peer groups on health-behaviour change

In the adjusted model of table 3, few health-region peer group differences were associated with health-behaviour change. There was a general trend in the findings, however, that tended to reflect increased urbanicity/rurality. Compared with peer group B (medium-sized urban centres), predominantly rural areas—that is, peer groups D, H and I—had lower odds of individuals undertaking behaviour changes, but only the ORs for peer groups D (0.82; 0.72 to 0.93) and I (0.87; 0.76 to 1.00) were statistically significant. In contrast, the OR for individuals living in the largest Canadian cities (ie, Vancouver, Toronto and Montreal—Peer group G) was 1.08 but not statistically significant (0.95 to 1.23).

Influence of community-belonging and specific health-behaviour changes

In order to better understand what specific types of health behaviours might be underlying individuals' responses that they undertook a health-behaviour change, we examined the variation in community-belonging according to nine specific types of changes measured by the question: ‘What is the single most important change you have made (to improve your health in the past year)?’ Figure 1 illustrates the associations found between community-belonging and each behaviour change after controlling for all variables shown in table 3. The influence of community-belonging varied between specific health-behaviour changes. Slightly negative or non-significant associations were present for health-behaviour changes with small prevalence rates—reduced alcohol consumption (0.8%), took vitamins (1.6%) and other behaviour changes (4.4%). The influence of community-belonging on specific behaviour changes, however, remained positive with a dose–response relationship for some of the most prevalent changes: exercise, weight loss, changed diet and, although not statistically significant, changed smoking behaviour. Together, these four behaviours were reported by over 85% of the respondents who indicated they undertook a health-behaviour change.

Figure 1

Odds ratios of specific health-behaviour changes associated with community-belonging (very weak as reference), adjusting for individual sociodemographic variables, self-rated health and health-region contextual variables.

Variation in the influence of community-belonging on health-behaviour change by health-region peer groups

The influence of community-belonging on health-behaviour change varied significantly (F=2.02; p=0.0038) by health-region peer groups when incorporating a fixed effect interaction term into the final model. Community-belonging and peer groups were represented by class variables with very weak community-belonging and peer group B as reference groups. For simplicity, we present only the overall f test, instead of all 21 interaction terms. To further understand how the association between community-belonging and health-behaviour change varied by health-region contextual factors, we estimated the final model stratified by health-region peer groups. Figure 2 illustrates the associations between community-belonging and health-behaviour change for individuals located in each peer group, after adjustment for all other individual-level variables included in the previous analysis.

Figure 2

Odds ratios of health-behaviour change with community-belonging (very weak as reference) by health-region peer groups in Canada, adjusting for individual sociodemographic variables and self-rated health.

With the exception of Peer Group I, there is a general positive relationship observed across different peer groups between higher community-belonging and an increased odds of making a health-behaviour change. Large variations, however, can be seen, especially between peer groups representing large urban areas (Peer Groups G) and rural and remote areas (Peer Groups D, H, I). The largest effects of community-belonging on health-behaviour change were seen in peer group G (ie, Vancouver, Toronto and Montreal—the three largest metropolitan areas in Canada). In these health regions, higher odds of undertaking a change to improve health were found among individuals reporting their community-belonging as somewhat weak (1.30; 1.18–1.44), somewhat strong (1.47; 1.34–1.62) and very strong (1.58; 1.41–1.77), compared with individuals reporting very weak community-belonging.

We also examined whether the association between community-belonging and specific health behaviours varied by health-region peer groups (data not shown). While the general direction of the associations between community-belonging and different behaviour changes remains, the magnitude of these associations varied with different behaviours in different peer groups. For example, in peer group G, community-belonging showed the strongest association with changed diet, while in peer group B community-belonging showed the strongest association with increased exercise.

Discussion

The present study sought to determine how sense of community-belonging and health-region context may influence the likelihood of individuals undertaking health-behaviour change, how the influence of community-belonging may vary by specific types of health-behaviour changes and whether the influence of community-belonging varies by different health-region contextual factors (ie, peer groups).

Consistent with our hypothesis, community-belonging showed a positive, dose–response association with health-behaviour change. These findings extend prior Canadian research indicating the importance of community-belonging to self-rated mental and physical health.22 25 The strength of the association between community-belonging and health-behaviour change in general suggests that community-belonging could be an important component of primary prevention initiatives aimed at increasing health through broad population initiatives. Community-belonging may represent greater social integration within one's community, increased access to social support, as well as material, cultural and psychosocial resources for improving one's health. However, given that we were unable to decompose community-belonging into the three dimensions theorised by Kim and Kaplan27 (attachment to place, community identity, social interaction), more health research on community-belonging is needed before it can be used in prevention initiatives.

In terms of health-region context, Statistics Canada-derived peer-group classifications revealed small influences on individuals' health-behaviour change. The significant differences we did observe supported our hypothesis: the odds of personal health-behaviour change were significantly lower in rural locations characterised by negative population growth and a high percentage of residents receiving transfer income (peer groups D and I) compared with rapidly growing urban locations with low percentages of government transfer income (peer group B). These findings are consistent with prior literature documenting differences in rural–urban health status and behaviours in Canada and elsewhere.3 4 7

Furthermore, the influence of community-belonging on health-related behaviour change varied significantly by health-region peer groups. This was most pronounced between urban and rural areas of Canada. For example, in metropolitan areas, the odds of individuals undertaking health-behaviour change when reporting very strong (vs very weak) community-belonging increased by 58%, in comparison with a 5% decrease for individuals living in rural eastern Canada who reported very strong community-belonging. These findings reflect the need to consider the types of contexts community-belonging may be linking people to (in terms of area resources, amenities, services and networks), how these linkages differ between rural and urban areas, and how such interactions may serve to promote or inhibit the adoption of healthy and unhealthy behaviours. Prior studies have found that community-belonging had differential associations with adult smoking and health status, as well as child mental health, depending on the conditions of the communities in which respondents were living.9 24 26 For example, Carpiano24 found that the odds of daily smoking were higher among female primary care givers of children who reported high levels of community-belonging in locations characterised by relatively low levels of social capital. Similarly, Caughy et al26 found that, in deprived areas, children of mothers who reported knowing many (vs few) of their neighbours had worse mental health outcomes, while in wealthier areas, children of mothers who reported knowing many (vs few) of their neighbours had better mental health outcomes. Thus, these prior findings suggest that the influence of community-belonging on health should depend on context—in the case of the present study, contextual differences between large metropolitan and rural areas of Canada—and the specific behavioural outcome of interest.

Lastly, we were able to demonstrate how the influence of community-belonging varied by specific health-behaviour changes. For the most prominent health behaviours (exercise, weight loss and improved diet), the influence of community-belonging was consistent with a dose–response relationship. Community-belonging had the strongest positive association with exercise and a negative association with lowering alcohol consumption. This highlights the need to recognise that community-belonging may be both beneficial and detrimental, depending on different types of health-related behaviours, and contextual circumstances, and that detrimental health behaviours that are socially acceptable, such as alcohol consumption, may be a negative result of increasing community-belonging.

Limitations

There are several study limitations that must be addressed. First, the CCHS is a self-reported cross-sectional survey and suffers from the limitations associated with such study designs. The temporal relationship between community-belonging and health-behaviour change cannot be examined in this study and greater community-belonging may be either an antecedent or a consequence of making changes to improve one's own health as a result of poor health status.29 30 We adjusted for self-rated health, which only slightly reduced the significant association between community-belonging and health-behaviour change, suggesting that the observed relationship between these two variables is not spurious as a result of their respective association with self-rated health. Self-rated health was inversely associated with behaviour change, suggesting that individuals who report excellent health are less likely to undertake behaviour changes, as they may view such changes as unnecessary.

Second, both health-behaviour change and community-belonging were measured using self-reported items, which may present bias in our study. Most notably, self-reported general and specific health-behaviour changes may not capture true behaviour change; however, given the scope of the CCHS dataset, assessing actual behaviour change is not feasible. Furthermore, the use of self-reported measures for both independent and dependent variables may introduce the possibility for same-source bias. While we were unable to fully control for all variables that might account for such confounding, the observed positive relationship between community-belonging and health-behaviour change remained after controlling for self-rated health (a variable that captures underlying mental and physical health statuses) as well as other factors that could contribute to such reporting bias. The fact that we did not observe similarly patterned positive dose–response associations for community-belonging across all specific health-behaviour changes and across different health-region peer groups also suggests that the influence of same-source bias may be limited.

Third, our use of a single indicator of individuals' sense of community-belonging is also limited. This question captures ‘sense’ of belonging, which may potentially differ from actual community-belonging, as it incorporates perceptions, feeling and relationships within a subjective community.31 Nevertheless, this item is conceptually consistent with theories linking individuals and context. Overall, this measure's association with health-behaviour changes—both independently and in interaction with the health-region peer-group context—provides direction for future research to decompose community-belonging and examine its health implications in terms of its potential constituent parts (eg, network ties, place attachment, psychological sense of community) as well as linking it to other key constructs regarding social relationships (eg, personal network- and community-based social capital).

Fourth, the CCHS sampling strategy was based on health regions, which may not represent contextual factors spatially commensurable with community-belonging. To assess health-region context, we also used health-region peer groups, which provide more interpretable information than the use of specific contextual measures, such as median household income or population density. While health regions are an important level to examine, as health regions are responsible for the health promotion and disease prevention programmes in a particular area of Canada, future research needs to examine how contextual factors at a more local scale level—both independently and in conjunction with health-region context—are associated with both health-behavioural change and community-belonging.

Conclusions

Examining community-belonging and contextual influences on health-behaviour change constitutes an important direction for prevention efforts in identifying and acting upon community factors that may facilitate or restrain individuals' efforts to adopt healthy lifestyles. Research that aims to identify such factors can help inform population-based efforts to transform communities into salutogenic environments that can improve health directly and indirectly—through fostering a more general health-promoting foundation upon which other, more outcome-specific prevention initiatives and interventions can be built.

Our findings provide further evidence for the importance of community-belonging and contextual factors for health-behaviour change, but also highlight the need to further examine how the influences of community-belonging vary with different types of behaviours and contextual factors. Overall, community-belonging showed a strong dose–response relationship with behaviours to improve health; however, associations varied slightly by type of behaviour change, with the largest positive association with changes to exercise. The role of community-belonging for health-behaviour change also showed a significant variability based on health-region contextual factors with the largest differences between urban and rural geographical regions of Canada. These results contribute to a growing literature on community-belonging and attachment by highlighting the nuanced relationship between community-belonging and health-behaviour change and the need to further examine the mechanisms through which community-belonging may be influencing health-behaviour change as well as other health outcomes.

What is already known on this subject

  • Primary prevention targeting individual behaviours should consider contextual factors; however, little is known regarding the contextual influences on health-behaviour change in general, and how contextual influences vary with different behaviours.

  • Research has shown that community-belonging is an important determinant of health, but little research has examined its role in health-behaviour change.

What this study adds

  • Community-belonging demonstrated a strong, positive dose–response relationship with health-behaviour change in general, and with several specific behaviour changes.

  • The influence of community-belonging varied by different health behaviours; however, the association remained particularly strong for exercise, weight loss and improved diet.

  • Health-region groupings, based on socio-economic determinants of health, had few associations with health-behaviour change after accounting for community-belonging, self-rated health and individual socio-demographic variables.

  • The influence of community-belonging on health-behaviour change varied by health-region contextual factors, indicating that the influence of community-belonging is sensitive to context.

Acknowledgments

The authors would like to thank M Koehoorn, University of British Columbia, for her invaluable help on a prior draft of this manuscript and for her guidance in working with the CCHS data.

References

Footnotes

  • Funding PH acknowledges funding from a Canadian Institute for Health Research Frederick Banting and Charles Best Canada Graduate Scholarship, a Michael Smith Senior Research Trainee Award and a Bridge Fellowship. RMC acknowledges funding from an Investigator Award from the Michael Smith Foundation for Health Research.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.