Neighborhood Social Cohesion and Physical Activity and Obesity Outcomes among Native Hawaiian and Pacific Islanders

Abstract

Objective:

Native Hawaiian and Pacific Islanders (NHPIs) report high obesity rates and low physical activity (PA) levels. We examined associations between neighborhood social cohesion, obesity, and PA among NHPI adults.

Methods:

Cross-sectional data from the 2014 NHPI National Health Interview Survey were analyzed. Social cohesion was assessed using a summative scale and categorized as low, medium, and high. PA was measured via self-report and categorized as insufficiently active (0–149 minutes/week) or sufficiently active (150+ minutes/week). Obesity status was based on self-reported height and weight measures converted into standard body-mass-index categories. Multivariable logistic regression models estimated odds of obesity and sufficient PA associated with social cohesion level.

Results:

The study sample included N=2,590 NHPI adults. Compared to those in low social cohesion neighborhoods, participants in high social cohesion neighborhoods had increased sufficient PA odds (OR=1.59; CI:1.19–2.12, p=0.003) and decreased obesity odds (OR=0.57; CI:0.40–0.83, p=0.005). No associations were observed between social cohesion and obesity and PA outcomes comparing individuals in medium vs. low social cohesion neighborhoods.

Conclusions:

High social cohesion was associated with reduced obesity odds and increased sufficient PA odds. Findings highlight the importance of enhancing social connectivity as a potential strategy to promote PA and healthy weight among NHPIs.

INTRODUCTION

The United States (U.S.) population consists of 1.4 million Native Hawaiians and Pacific Islanders (NHPIs).¹ NHPIs have their origins in one of three main regions located in the Pacific, which include Polynesia (e.g., Samoa), Micronesia (e.g., Guam), and Melanesia (e.g., Fiji).² Historically, it has been extraordinarily difficult to assess health and health behavior data specific to the NHPI population due to institutionalized discriminatory policies that did not recognize specific racial and ethnic groups or allow for self-identification if an individual’s race or ethnicity was not represented as a category. Prior to 1990, the U.S. Census Bureau did not include NHPI as a racial category. Beginning with the 1990 Census, the category Other Asian or Pacific Islander was added and allowed NHPI individuals to write in their racial group. For the 2000 Census, more changes came with the addition of the following racial groups: Native Hawaiian, Guamanian, or Chamorro. Further, Other Asian or Pacific Islander was changed to Other Pacific Islander and individuals could write in their racial group.³ These changes to the U.S. Census have facilitated the collection and analysis of data of NHPIs as a distinct group from other Asians, a critical advancement towards identifying health inequities. Aggregating racial groups (e.g., combining NHPIs with other Asians) may mask important heterogeneity in health outcomes that exist between subgroups.

Collecting and analyzing health data on NHPIs as a distinct group is important because recent evidence suggests that they have significantly higher rates of overweight, obesity, coronary artery disease, pre-diabetes, diabetes, and hypertension compared to other Asian populations in the U.S.^4–6 More than half of the 12 studies reviewed in a systematic review indicated that NHPIs did not meet recommended aerobic activity guidelines for moderate-intensity physical activity (PA)⁷ or the combination of moderate and vigorous-intensity PA,⁸ though the meta-analysis concluded that NHPIs were no different from other Asian subgroups with regards to PA levels.⁸ NHPIs are at a much higher risk of obesity and associated chronic diseases compared to Whites, though data on the determinants driving obesity among this population is limited. Factors shown to be associated with obesity among NHPIs include lifestyle behaviors such as poor diet, low PA, and sleep,^9–11 acculaturation to Western lifestyles,¹² lower socioeconomic status, and stressors related to acculatration,¹³ discrimination,¹⁴ and environmental conditions.¹⁵ There are also minimal data on lifestyle factors associated with PA in NHPIs. For example, fruit and vegetable consumption was the only health variable found to be related to PA in a convenience sample of 100 NHPIs residing in San Diego.¹⁶ Further investigation of factors associated with obesity and obesity risk behaviors (such as PA) is particularly important to inform prevention and intervention programs tailored for this group.

Neighborhood social cohesion, defined as an individual’s appraisal of how much members of a neighborhood trust one another and perceive belonging to a neighborhood,¹⁷ is a community-level factor shown to be associated with obesity. Studies have demonstrated that higher perceived levels of neighborhood social cohesion are associated with health promoting behaviors such as obtaining a mammogram, getting a flu vaccine, obtaining a Pap test,¹⁷ and meeting aerobic PA guidelines set forth by the U.S. Department of Health and Human Services.¹⁸ Several studies indicate that perceived high levels of neighborhood social coheision is associated with greater PA among adults^18–24 and youth,²⁵ while studies of the association between social cohesion and obesity have yielded mixed results (inverse, positive, or null associations).²⁶ No studies to our knowledge have examined social cohesion, obesity, and PA among NHPIs.

Associations between social cohesion and obesity may be mediated by PA. A cross-sectional study of adults from a low-income and ethnically-mixed neighborhood sample indicated that PA mediated the relationship between perceived crime safety and obesity.²⁷ Thus, individuals who perceive high levels of social cohesion may feel safer to spend active time outdoors, exercise together in groups more frequently, and thus engage in overall higher levels of PA. Given that PA may be an important behavioral mechanism between social cohesion and obesity, examining the association(s) between neighborhood social cohesion, PA, and obesity among NHPIs may elucidate important areas for intervention among a group that experiences disproportionately higher obesity rates.⁵

Study Objectives and Hypotheses

Using cross-sectional data from NHPI adults living in the U.S., this study aimed to: (1) examine the association between neighborhood social cohesion and PA; (2) examine the association between neighborhood social cohesion and obesity; and (3) examine the role of PA as a mediator in the association between neighborhood social cohesion and obesity. We hypothesized that neighborhood social cohesion would be positively associated with PA, that neighborhood social cohesion would be inversely associated with having obesity, and that PA mediates the association between neighborhood social cohesion and obesity.

METHODS

Data Source

Each year, the National Center for Health Statistics (NCHS) at the Centers for Disease Control and Prevention (CDC) fields the National Health Interview Survey (NHIS) to gather information about the health of non-institutionalized civilian adults in the U.S. In 2014, the NCHS fielded the first and only NHIS focused exclusively on NHPIs⁶ through which a nationally representative sample of 2,590 NHPI adults were surveyed. This study was based on the 2014 cross-sectional data from the NHPI-NHIS, the largest national health survey of NHPIs to date. Respondents were randomly selected from family members within households. The survey data are sectioned into multiple files. For this study, the main file is the “Adult” file, which was merged with the “Person,” the “Household,” and the “Family” files to create the main analytical dataset. To ensure representation, the NCHS used stratified multistage area probability sampling. A detailed description of NHIS and NHPI-NHIS surveys can be found elsewhere.^28–30 This study received an exemption from the Institutional Review Board (#206591) because data were public and de-identified.

Study Population

The study population is the NHPI non-institutionalized civilian adults aged 18 years and older. Because adults aged 75 years and older constitute a very small proportion of the sample, inclusion of these individuals may pose confidentiality risks; thus, the NCHS caps the eligibility age of survey participants in the NHIS at 75. Out of the 2,590 NHPI adult (age ≥ 18 years) respondents, 50.8% self-identified NHPI as their sole race/ethnicity, and 49.2% self-identified as NHPI in combination with one or more other races (NHPI-mixed).

Measures

Outcomes

PA was assessed by asking participants to report their “physical activities (exercise, sports, physically active hobbies...) that [they] may do in [their] leisure time.” Participants were asked to summarize their usual PA in frequency and duration in the time unit(s) the participant self-provided (e.g., times per day, per week, per month, or per year). From these responses, the NCHS coded responses and created a three-category variable for PA levels/week: (1) inactive (no activity); (2) insufficiently active (1–149 minutes/week); and (3) sufficiently active (150 + minutes/week). For our study, this variable was dichotomized as insufficiently active (0–149 minutes/week) and sufficiently active (150+ minutes/week).

Obesity status was based on self-reported measures of height and weight that were used to calculate body mass index (BMI) (kg/m²). The NCHS used the classification of BMI-based weight status categories established by the World Health Organization:³¹ underweight (<18.5 kg/m²), healthy weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), or obese (≥30.0 kg/m²). Weight status was analyzed dichotomously for this study (obese vs. not obese).

Exposure

The exposure of interest for this study was neighborhood social cohesion. This variable was measured by asking participants to rate the extent to which they agreed or disagreed (4-point Likert Scale) with the following statements about their neighborhood: (1) People in this neighborhood help each other out; (2) People in this neighborhood can be trusted; (3) This is a close-knit neighborhood; (4) There are people I can count on in this neighborhood. The response categories were reversed coded as: (1) definitely disagree; (2) somewhat disagree; (3) somewhat agree; (4) and definitely agree. The sum of the responses to the four statements were calculated to create a summative score for neighborhood social cohesion (scores ranged from 4–16; higher scores indicated higher perceived levels of social cohesion). We obtained a Cronbach’s alpha=0.84, indicating a good internal consistency.³² Using criteria established by prior research analyzing this summative variable (Cronbach’s alpha=0.89) among NHPIs participating in the NHIS, scores of neighborhood social cohesion were categorized into low (<12), medium (12–15), and high (>15).³³

Mediator

PA was first examined as an outcome with neighborhood social cohesion being the exposure of interest and, second, as a mediator through which neighborhood social cohesion is indirectly associated with obesity.

Covariates

Measured variables hypothesized to be associated with the exposure and outcomes based on the literature were included as covariates in our analyses. These covariates included sex; multiracial status (NHPI-singular vs. NHPI-mixed); age (continuous and squared); income level ($0–34999; $35000–74999; ≥$75000); household education (<college education vs. having a college degree or higher); employment status (had a job last week, no job last week but had job in the last 12 months coded as employed; no job last week and no job past 12 months; never worked coded as unemployed); and marital status (single; divorced; separated or widowed coded as not married or in partnership; and married or living in cohabitation).^20,27

Statistical Analysis

All descriptive and regression analyses were conducted with STATA 16.1/SE in November 2020 and performed on complete cases, yielding an analytic file of n=2,301. Statistical significance was determined at an a priori two-tailed alpha (α=0.05). For all analyses, we considered the complex survey design of the data by applying the sampling weights, stratification, and primary sampling units already computed by NCHS and available in the data sets for the entire n=2,590 study sample. STATA svy procedure was used to estimate parameters and adjust for NHPI-NHIS complex stratified-multistage-area-probability sampling.

For descriptive statistics, we computed means and standard errors (se) for continuous variables and percentages (95% confidence intervals) for categorical variables. To test our study hypotheses, we conducted bivariate and multivariable logistic regression models to estimate unadjusted and adjusted odds of our outcomes of interest (sufficient PA; obesity) associated with our exposure of interest (neighborhood social cohesion). For the exposure, low neighborhood social cohesion served as the referent group. Multivariable models adjusted for sex, age, multiracial status (NHPI singular/NHPI mixed), highest level of education in the family, household income, employment status, and marital status. Estimates from regression models for the study sample (n=2,590) were derived from those with complete data (n=2,301); results are presented for the entire study sample (n=2,590) due to the application of sampling weights that were created by NCHS a priori based on all n=2,590 participants. Multicollinearity among the independent predictors in our models was examined by assessing variance inflator factors. Once the regression models were estimated, we computed the Hosmer–Lemeshow global goodness of fit test statistic—which takes the survey sampling design into account—to examine the adequacy of the logistic models by testing the null hypothesis that the fitted model is correct.

The extent to which PA partially mediated the effect of neighborhood social cohesion on obesity was determined by partitioning the effects of the exposure directly associated with the likelihood of obesity from its indirect effects mediated by PA. The mediation analysis reflects the proportion of total effects mediated expressed as the sum of changes in the probability (∑Δp) of outcome and is interpreted as the ∑Δp of having obesity.^34,35

RESULTS

The sample consisted of N=2,590 NHPI adults (50.6% females; mean age was 40.4 years (standard error (se): 0.37). The majority of participants (58.4%) reported a household income of less than $75,000/year, and over two-thirds (69.6%) had less than a college degree. Additional sociodemographics are presented in Table 1. Close to half (40.9%) of our study sample of NHPI adults were insufficiently active, and 42.6% were classified as having obesity. Approximately a third of participants (34.9%) reported having low neighborhood social cohesion, 40.5% reported having medium neighborhood social cohesion, and approximately a quarter (24.6%) reported having high neighborhood social cohesion.

Table 1.

Characteristics of N=2,590 Native Hawaiian and Other Pacific Islanders Adults Participating in the 2014 National Health Interview Survey

VARIABLES	Unweighted Frequencies (N)	Weighted Proportions* % (95% confidence interval)
Dependent
Physical Activity
Insufficiently active<150 min.	1,068	40.9 (38.6–43.3)
Sufficiently active>=150 min.	1,464	59.1 (56.7–61.4)
Having Obesity
0=not having	1,474	57.4 (54..5–60.3)
1=having	1,045	42.6 (39.7–45.5)
Independent Variables
Neighborhood Social Cohesion (main predictor)
Low (<12)	744	34.9 (31.3–38.6)
Medium (12–15)	1,094	40.5 (37.7–43.4)
High (>15)	695	24.6 (21.5–28.0)
Sex
Male	1,213	49.4 (45.6–53.2)
Female	1,359	50.6 (46.8–54.4)
Household Income
$0–34,999	592	24.9 (21.7–28.3)
$35,000–74,999	788	33.5 (30.6–36.7
≥$75,000	969	41.6 (38.1–45.1)
Employment Status
Employed	1,661	67.5 (63.3–71.4)
Not Employed	926	32.5 (28.6–36.7)
Education Status
< Bachelor’s degree	1,722	69.6 (66.8–72.3)
≥ Bachelor’s degree	868	30.4 (27.8–33.2)
Marital Status
Not married or in partnership	921	41.4 (38.2–44.6)
Married or in partnership	1,667	58.6 (55.4–61.7)
Continuous Age (mean and standard errors)	2,590	40.4 (0.37)

^*Because the complex sample design was taken into account (sampling weights, primary sampling units, and stratification) the weighted proportions shown may not correspond to the unweighted frequencies.

Results from the bivariate logistic regression (Model 1) suggest a positive association between living in neighborhoods with high social cohesion—as compared to low social cohesion—and being sufficiently active (OR=1.48, CI:1.00–2.18; p=0.050); however, this association was marginally significant. Results from multivariable logistic regression (Model 2) show that NHPI adults living in neighborhoods with high social cohesion have higher odds of being sufficiently active (or 59% more likely) than not, relative to those living in neighborhoods with low social cohesion (OR:1.59, CI:1.19–2.12, p=0.003). For this adjusted model, the Hosmer-Lemeshow (F-adjusted test statistic F_(9,19) = 0.736; p=0.672) is non-significant, suggesting no evidence of lack of fit. See Table 2-A.

Table 2-A.

Results from the Bivariate and Multivariable Logistic Regression Models: Associations between Neighborhood Social Cohesion and Odds of Sufficient Physical Activity among N=2,590 Native Hawaiian and Pacific Islander Adults Participating in the 2014 National Health Interview Survey

	Model 1† OR (95% CI)	p-values	Model 2†† OR (95% CI)	p-values
Neighborhood Social Cohesion
Low (reference group)	-		-
Medium	1.02 (0.83–1.66)	0.246	1.15 (0.83–1.58)	0.393
High	1.48 (1.00–2.18)	0.050	1.59 (1.19–2.12)	0.003

Note:

^†Unadjusted odds-ratios(OR) and confidence intervals (CI) obtained from bivariate logistic regression

^††Adjusted odds-ratios(OR) and confidence intervals (CI) obtained from multivariable logistic regression model adjusted for sex, age, age², multiracial status (NHPI singular/NHPI mixed), highest level of education in the family, household income, employment status, marital status.

Model 3 from the bivariate logistic regression shows that NHPI adults living in neighborhoods with high levels of social cohesion have decreased odds of having obesity (OR: 0.56, CI:0.39–0.80, p=0.003) compared to adults living in neighborhoods with low social cohesion. In the fully adjusted multivariable model (Model 4), this association remained significant; NHPI adults living in high vs. low social cohesion neighborhoods have 0.57 times the odds of having obesity (CI:0.40–0.83, p=0.005). The Hosmer-Lemeshow (F-adjusted test statistic F(9,19) = 1.801; p=0.134) is non-significant, suggesting no evidence of lack of fit. See Table 2-B. Multicollinearity among the independent variables was not detected in multivariable regression models (all variance inflator factors (VIF) were <1.27, with an overall VIF mean of 1.14). No associations were observed between social cohesion and obesity and PA outcomes comparing NPHIs who lived in medium vs. low social cohesion neighborhoods.

Table 2-B.

Results from the Bivariate and Multivariable Logistic Regression Models: Associations between Neighborhood Social Cohesion and Odds of Obesity among N=2,590 Native Hawaiian and Pacific Islander Adults Participating in the 2014 National Health Interview Survey

	Model 3† OR (95% CI)	p-values	Model 4†† OR (95% CI)	p-values
Neighborhood Social Cohesion
Low (reference group)	-		-
Medium	0.86 (0.64–1.17)	0.323	0.94 (0.67–1.30)	0.684
High	0.56 (0.39–0.80)	0.003	0.57 (0.40–0.83)	0.005

Note:

^†Unadjusted odds-ratios(OR) and confidence intervals (CI) obtained from bivariate logistic regression

^††Adjusted odds-ratios(OR) and confidence intervals (CI) obtained from multivariable logistic regression model adjusted for sex, age, multiracial status (NHPI singular/NHPI mixed), highest level of education in the family, household income, employment status, marital status.

Source: Centers for Disease Control and Prevention-National Center for Health Statistics. 2014 Native Hawaiian and Pacific Islander (NHPI)- National Health Interview Survey (NHIS).

On average among NHPI, 2.80% (CI:1.78%- 8.53%) of the effect of neighborhood social cohesion on obesity was mediated by PA, after controlling for the other covariates.

DISCUSSION

This study is the first to investigate neighborhood social cohesion and its associations with PA and obesity among a large, nationally representative sample of NHPI adults. The study demonstrated that NHPI adults in our study who reported living in neighborhoods with high social cohesion were 1.6 times more likely to engage in adequate levels of PA and are 0.43 times less likely to be obese compared with those who reported living in neighborhoods with low social cohesion, adjusting for covariates. Although there was no significant association between neighborhood social cohesion and PA, neighborhood social cohesion was associated with obesity through the indirect effect of PA. The mediation only accounted for 2.8% of the total effect, however. Considering the lack of significant association between neighborhood social cohesion and PA in this study population, this small indirect effect was not surprising. This result suggests that PA still plays a mechanistic—though not direct—role in the association between neighborhood social cohesion and obesity in this population. Since PA as a modifiable factor may be easier to promote and implement than neighborhood social cohesion, promoting PA within neighborhoods may indirectly influence obesity. However, since mediation imply a temporal ordering where neighborhood social cohesion precedes PA, which, in turn, influence obesity, the indirect association found needs to be confirmed with longitudinal data to account for this temporality.

The associations between neighborhood social cohesion, PA, and obesity observed in this current study support findings in prior research that have assessed these associations among other racial/ethnic adult populations, with neighborhood social cohesion shown as a protective factor against obesity and positively associated with PA.^21,36,37 One cross-sectional study of a racially and ethnically diverse sample from the 2013 NHIS study of adults 65 and older found no association between social cohesion and obesity, though this divergent finding compared to our study may be partially attributed to age differences in study populations (the maximum age of participants in our study was 75 years).³⁶

One mechanism for the observed association between neighborhood social cohesion, PA, and obesity is that social cohesion may provide a sense of neighborhood connection and safety that facilitates and encourages residents to spend more active time outdoors, resulting in overall higher activity levels and subsequently reduced obesity risk.³⁷ Perception of crime and lack of safety are cited as major barriers that prevent or limit individuals from engaging in outdoor activity and spending time outside.^38–40 Thus, individuals who report high neighborhood social cohesion may perceive their communities to be more conducive and safe to spend active time outdoors. Another mechanism is that perceived high levels of neighborhood social cohesion may facilitate increased time spent together in groups and shared PA goals.^19,41,42 Findings from a mixed methods pilot study testing a PA intervention suggest that PA interventions may also promote social connectivity among other adult populations and indicate potential to integrate social cohesion and connectivity into PA interventions.⁴³

Neighborhood social cohesion may also be associated with obesity by influencing obesity-related behaviors other than PA. In a cross-sectional analysis using the same population of NHPIs as in this current study, researchers found that low social cohesion was associated with poorer sleep outcomes—a behavioral factor shown to be independently associated with obesity.^44,45 Another study of adults ages 45–84 years (40% White, 25% Black, 21% Hispanic) residing in six metropolitan areas across the U.S. found that social cohesion was positively associated with other health-promoting behaviors related to obesity and/or PA, such as consuming at least five fruit and vegetable servings per day and decreased smoking.²¹ These findings suggest that social cohesion may be associated with a variety of behaviors that influence health and healthy weight management.

The associations observed between social cohesion and obesity in our study and previous studies suggest that enhancing social cohesion and connectedness may be high impact intervention targets for addressing obesity and related behaviors. Developing and evaluating interventions that integrate or address social cohesion among NHPIs is particularly relevant given the collectivist culture of NHPIs.^46,47 A review of culturally-adapted diabetes interventions indicated that integrating social networks and extended family members in intervention approaches is a promising strategy to target diabetes outcomes for this population.⁴⁶ The interplay of behavioral, community, cultural, and environmental factors warrant examination in future studies to further elucidate the mechanisms between social cohesion and obesity outcomes.

Limitations

Findings should be considered in the context of the following study limitations. The cross-sectional design of the 2014 NHPI-NHIS survey limits our ability to determine directionality and infer causality. Although survey data is drawn from a nationally representative population of NHPIs, this group is heterogeneous; hence, this study’s findings may not fully capture cultural heterogeneity. With respect to outcomes of interest, PA is a complex multidimensional behavior, and correct assessment via self-report is often over-estimated,⁴⁸ may be prone to cultural interpretation,⁴⁹ or social desirability bias.⁵⁰ The PA variable available in the dataset does not elaborate on the type of activity incurred or the metabolic equivalent of task (MET). Our finding that PA partially mediated the association between neighborhood social cohesion and obesity should be interpreted with caution, as the correct application of causal mediation relies on the validity of the assumption that neighborhood social cohesion temporarily precedes PA—an assumption that could not be held with cross-sectional data. Future studies on the indirect effect of PA based on longitudinal data are needed.

Dichotomizing obesity as an outcome as in this study limits our understanding of associations for other weight status groups (e.g., individuals with overweight status). Measures of obesity were derived from self-reported measures of height and weight and are subject to bias, with adults tending to slightly overreport height and underreport weight. However, a recent validation study of a national cohort of U.S. adults suggests that BMI derived from self-reported height and weight is a valid measure among men and women from different sociodeomographic groups.⁵¹ Finally, our study was limited by the availability of the measured variables included in the dataset. Other potential confounders (e.g., measures of the built food and PA environment, rurality) were not able to be included in the analyses.

Strengths

Despite the limitations, this article helps fill an important gap in the literature. This study is first to examine the association between neighborhood social cohesion and obesity and PA among NHPIs (an understudied population that has historically been aggregated with other racial/ethnic categories in data collection efforts and in scientific studies). This study is strengthened by use of a nationally representative and random sample of NHPIs. Notably, the NHPI NHIS is the largest national health survey of NHPI adults in the country. Considering the higher rates of obesity and associated comorbidities experienced by this underrepresented population, this study identifies a critical protective factor that can be leveraged upon to potentially promote PA and curb obesity in this population.

CONCLUSIONS

Perceived high levels of neighborhood social cohesion were associated with increased odds of sufficient PA and lower odds of obesity among NHPIs, a racial/ethnic group that experiences among the highest rates of obesity in the U.S. Our findings highlight the need for further investigation of mechanisms underlying the association between neighborhood social cohesion and obesity outcomes among NHPI and the role of social cohesion as a potential intervention target that may guide the development of future community-based interventions addressing obesity and PA among this population.

Study Importance Questions.

What is already known on this subject?

• Native Hawaiian and Pacific Islanders (NHPIs) report high rates of obesity and low levels of physical activity.

• Studies have demonstrated that higher perceived levels of neighborhood social cohesion are associated with health promoting behaviors among non-NHPI study populations, such as higher levels of physical activity.

What are the new findings in your manuscript?

• High neighborhood social cohesion is associated with reduced odds of obesity among a national sample of NHPI adults.

• High neighborhood social cohesion is associated with increased odds of engaging in adequate levels of physical activity among a national sample of NHPI adults.

How might the results change the direction of research or the focus of clinical practice?

• The findings highlight the role of social cohesion as a potential intervention target that may guide the development of future community-based interventions addressing obesity and physical activity among NHPIs.