Differences in Loneliness Across the Rural-Urban Continuum Among Adults Living in Washington State

Abstract

Purpose:

Rural residents may be at higher risk for loneliness than urban residents due to factors such as social isolation, poorer health, and socioeconomic disadvantage. To date, there have been few studies examining rural-urban differences in loneliness among adults in the United States. We examined differences in loneliness across the rural-urban continuum among adult residents living in Washington State.

Methods:

Stratified random sampling was used to select 2,575 adults from small rural, large rural, suburban, and urban areas who were invited to complete a survey on factors affecting health. Data were obtained from 616 adults (278 from small rural, 100 from large rural, 98 from suburban, and 140 from urban areas) from June 2018 through October 2019. Loneliness was measured using the UCLA Loneliness Scale (3^rd version). Multivariable linear and logistic regressions were used to examine geographic differences in loneliness (measured continuously and dichotomously).

Findings:

Mean unadjusted loneliness scores were lower in suburban compared to urban areas (35.06 vs. 38.57, P = .03). The prevalence of loneliness was 50.7%, 59.0%, 40.8%, and 54.3% in small rural, large rural, suburban, and urban areas, respectively. Suburban living was associated with lower odds for being lonely compared to urban living (unadjusted OR = 0.58; 95% CI = 0.34–0.98), but this association was not statistically significant in the adjusted model (OR = 0.63; 95% CI = 0.33–1.19).

Conclusion:

Loneliness is a prevalent health issue across the rural-urban continuum among Washington State adults.

Approximately a third to half of middle-aged and older adults in the United States (US) experience loneliness.^1,2 Loneliness has been associated with multiple chronic health conditions, lower activity levels, tobacco use, psychological distress, higher levels of depression, and markers of systemic inflammation in cross-sectional studies^1,3–7 and heart disease, functional and cognitive decline, and mortality in longitudinal studies.^8–10 Evidence from 2 meta-analyses have further shown that loneliness is associated with mortality¹¹ and depression.¹² In addition to evidence linking loneliness to multiple adverse health outcomes, loneliness has been associated with increased health care utilization.^2,13

While the prevalence of loneliness is high in the US, people living in rural areas may be particularly at risk. In addition to many rural areas being physically and socially isolated,¹⁴ rural residents are more likely than urban residents to report being in fair or poor health¹⁵ and are more likely to have certain chronic health conditions.^16,17 The relatively poorer health of rural residents may limit opportunities for social participation,¹⁸ and evidence suggests that poor perceived health, chronic illness, and functional limitations are associated with loneliness.^3,4,13 Poverty rates also tend to be higher in rural compared to urban areas,¹⁹ and living in socioeconomically disadvantaged areas has been associated with loneliness.²⁰

To date, few US-based studies have examined rural-urban differences in loneliness.^13,21 Based on a 3-item measure of loneliness, mean loneliness scores did not differ among a nationally representative sample of US older adults of urban, micropolitan rural, and non-core rural areas.²¹ Among primary care patients from Colorado and Virginia, there were no rural-urban differences in loneliness scores based on the 3-item University of California, Los Angeles (UCLA) Loneliness Scale.¹³ Given that US rural health disparities vary by degree of rurality,^22,23 the purpose of this study was to further examine differences in loneliness across a rural-urban continuum using the 20-item UCLA Loneliness Scale.

Methods

The Washington State University Office of Research Assurances determined this project was exempt from the need for Institutional Review Board review. From June 2018 through October 2019, the Washington State University Social and Economic Sciences Research Center administered a cross-sectional survey to Washington State adult residents. We defined rurality at the ZIP Code level based on a Washington State Department of Health classification scheme that uses Rural-Urban Commuting Area (RUCA) codes.²⁴ Urban areas were defined as those having RUCA codes of 1.0 and 1.1, and suburban areas were defined as those having codes of 2.0, 2.1, and 3.0 and a population density ≥100 people per square mile. Large rural areas were defined as those having RUCA codes of 4.0 through 6.1 with a population density ≥100 people per square mile, while small rural areas were defined as those with codes 7.0–10.6 or not an urban core with a population density <100 people per square mile. Stratified sampling was used in which a random sample of residential addresses were drawn from each of the 4 urban-rural areas. All Washington State ZIP Codes were included in the sample frame, but some ZIP Codes were not sampled due to having few residential addresses.

Postal invitations were mailed to 2,575 randomly selected household addresses obtained from the US Postal Service Delivery Sequence File. Invitation letters provided study details and instructions for completing the survey online and specified that the adult with the most recent birthday should complete the survey. Invitation letters also provided information for completing a Spanish version of the survey. A $1 bill was included with the initial mailing. Three additional follow-up letters were mailed over 6 weeks. Two of these letters included a paper version of the survey and a postage-paid return envelope. Sampled residents were assigned a unique survey access code to prohibit the completion of multiple surveys. Double data entry was used to ensure the accurate recording of mailed survey responses.

For this study, 616 Washington State adult residents (278 from small rural, 100 from large rural, 98 from suburban, and 140 from urban areas) from an estimated population of 3,024,196 residential households completed or partially completed the survey, resulting in a 25.5% response rate and a sampling error of ± 4%.

Measures

Loneliness

Loneliness was assessed using the third version of the 20-item UCLA Loneliness Scale,²⁵ and evidence supports the validity of the scale among rural populations.⁴ Participants indicated how often they experienced feelings pertaining to loneliness on a Likert-type scale with response options ranging from 1 (never) to 4 (always). Summative scores range from 20–80 with higher scores indicating greater loneliness.²⁵ The Cronbach alpha in our sample was 0.936.

Control Variables

We adjusted for multiple sociodemographic variables that could affect the association between rurality and loneliness including age, sex, ethnicity, race, marital status, employment status, and education.

Statistical Analysis

We characterized the sample using simple descriptive statistics. Bivariate comparisons were made between respondent characteristics and rurality using Pearson chi-squared tests. Unadjusted and adjusted linear regression models examined the association between rurality and loneliness. Rurality was treated as a 4-category variable with urban areas as the reference category. To examine the association between rurality and loneliness that may be clinically significant, we ran logistic regression models with loneliness as a binary variable (lonely vs. not lonely) using a UCLA Loneliness score greater than 40 to define lonely.⁴ Adjusted (multivariable) linear and logistic regression models included the covariates noted above. Post-stratification weights were applied to survey responses to account for differential responses across levels of rurality in all analyses, which were conducted using Stata/MP version 15.1 (StataCorp, College Station, Texas).

Results

Characteristics of Household Respondents

There were no differences in sex, race, ethnicity, employment status, or education level among the groups (Table 1). The small and large rural groups were on average 5 years of age older than the urban group. The suburban group had a higher proportion of married or cohabitating respondents compared to other groups.

Table 1.

Demographic Characteristics of Household Respondents by Geographic Characterization, Washington State, 2018–2019

Characteristic	Small Rural	Large Rural	Suburban	Urban
	N= 206,181	N= 194,859	N= 266,805	N= 2,356,351	P
Sex
Male	37.4	39.5	46.0	31.5	.06
Female	62.7	60.5	54.0	68.6
Age
Mean (95% CI)	60.5 (58.5–62.5)	61.0 (57.3–64.6)	58.9 (55.4–62.3)	55.5 (52.3–58.7)	.05
Range (min-max)	18–96	20–93	24–91	20–88
Ethnicity
Hispanic	3.3	3.5	4.7	4.9	.86
Non-Hispanic	96.7	96.5	95.3	95.1
Race
White	82.7	73.0	74.5	72.9	.20
Non-White	5.8	9.0	11.2	15.0
Missing	11.5	18.0	14.3	12.1
Marital status
Married or cohabitating	64.3	52.2	71.6	53.1
Not marrieda	35.7	47.8	28.4	46.9	.01
Employment status
Employed	43.6	39.8	52.3	47.3	.43
Not employed/retired	56.4	60.2	47.7	52.7
Education level
High school or less	16.7	12.2	4.6	8.6	.09
At least some college	83.3	87.8	95.5	91.4

Abbreviation: CI, confidence interval

Note: Unless otherwise indicated, values represent percentages of total population size. P values indicate the difference across groups based on design-based Pearson chi-squared test (adjusted Wald test used for age variable).

^aNot married individuals include those reported being separated, divorced, single, widowed, or other.

Differences in Loneliness Across the Rural-Urban Continuum

Respondents from suburban areas had lower mean unadjusted loneliness scores compared to respondents from urban areas (35.06; 95% CI = 32.77–37.35 vs. 38.57; 95% CI = 36.49–40.65; Wald test P = .03). There were no other statistically significant differences in mean unadjusted loneliness scores. Mean unadjusted loneliness scores were 36.65 (95% CI = 35.24–38.06) among respondents from small rural areas and 38.56 (95% CI = 35.61–41.51) among those from large rural areas.

The prevalence of loneliness using a cut-point of 40 was 50.7%, 59.0%, 40.8%, and 54.3% among respondents from small rural, large rural, suburban, and urban groups, respectively. These differences were not statistically significant (design-based Pearsons chi-squared test, P = .09).

The results of the linear regression models are presented in Table 2. In the unadjusted linear regression model, suburban location was associated with lower loneliness scores (β = −3.51; 95% CI = −6.60 to −0.41). However, in the adjusted model, geographic location was not associated with loneliness. Factors associated with higher loneliness scores in the adjusted model included being male, being non-White, and not being employed. Younger age and being married or cohabitating was associated with lower loneliness scores.

Table 2.

Geographic Characterization and Demographic Characteristics and Their Association with UCLA Loneliness Score (continuous) Among Washington State Adults, 2018–2019

	Unadjusted		Adjusted
Variable	β Coefficient (95% CI)	P value	β Coefficient (95% CI)	P value
Rurality
Small rural	−1.92 (−4.43, 0.59)	.134	−0.51 (−2.96, 1.94)	.68
Large rural	−0.01 (−3.62, 3.60)	.995	0.16 (−3.25, 3.57)	.93
Suburban	−3.51 (−6.60, −0.41)	.026	−1.02 (−4.05, 2.00)	.51
Urban	Ref	-	Ref	-
Age (continuous)	−0.08 (−0.16, 0.01)	.077	−0.18 (−0.28, −0.09)	<.01
Sex
Male	2.85 (−1.02, 6.72)	.148	4.00 (0.48, 7.53)	.03
Female	Ref	-	Ref	-
Ethnicity
Hispanic	−0.08 (−5.56, 5.40)	.978	−3.74 (−10.41, 2.93)	.27
Non-Hispanic	Ref	-	Ref	-
Race
White	Ref	-	Ref	-
Non-White	8.93 (3.61, 14.24)	.001	7.97 (3.29, 12.65)	<.01
Missing	0.25 (−4.86, 5.36)	.924	−3.72 (−9.39, 1.94)	.20
Marital status
Married or cohabitating	−7.02 (−10.27, −3.78)	.000	−6.81 (−9.93, −3.69)	<.01
Not married	Ref	-	Ref	-
Employment status
Employed	Ref	-	Ref	-
Not employed/Retired	2.58 (−0.67, 5.82)	.119	4.75 (1.07, 8.44)	.01
Education level
High school or less	6.38 (1.96, 10.81)	.005	4.23 (−0.48, 8.95)	.08
At least some college	Ref	-	Ref	-

Abbreviation: CI, confidence interval

Note: Loneliness measured using the UCLA 20-item loneliness scale, with scores treated continuously in linear regression models (F = 6.58 P < .001; R² = 0.2537).

In the unadjusted binary logistic regression model, living in a suburban area was associated with lower odds for being lonely compared to living in an urban area (OR = 0.58, 95% CI = 0.34–0.98). There were no statistically significant associations between living in a small or large rural area and being lonely compared to living in an urban area (Table 3). In the adjusted model, the association between living in a suburban area and being lonely was no longer statistically significant (OR = 0.63; 95% CI = 0.33–1.19). Being non-White or having high school or less education were associated with higher odds for being lonely. Being married or cohabitating was associated with 65% lower odds for being lonely compared to respondents who were not married or cohabitating. The odds of loneliness decreased significantly with increasing age (Table 3).

Table 3.

Geographic Characterization and Demographic Characteristics and Their Association with UCLA Loneliness Score (dichotomized) Among Washington State Adults, 2018–2019

	Unadjusted		Adjusted
Variable	Odds ratio (95% CI)	P value	Odds ratio (95% CI)	P value
Rurality
Small rural	0.87 (0.58, 1.30)	.492	0.89 (0.53, 1.48)	.64
Large rural	1.21 (0.72, 2.04)	.469	1.14 (0.59, 2.21)	.69
Suburban	0.58 (0.34, 0.98)	.042	0.63 (0.33, 1.19)	.15
Urban	Ref	-	Ref	-
Age (continuous)	0.99 (0.97, 1.00)	.120	0.97 (0.95, 0.99)	<.01
Sex
Male	1.39 (0.77, 2.49)	.273	1.68 (0.84, 3.36)	.14
Female	Ref	-	Ref	-
Ethnicity
Hispanic	0.61 (0.15, 2.39)	.475	0.28 (0.05, 1.73)	.17
Non-Hispanic	Ref	-	Ref	-
Race
White	Ref	-	Ref	-
Non-White	2.56 (1.10, 5.95)	.029	2.79 (1.12, 6.96)	.03
Missing	3.56 (1.39, 9.14)	.008	0.47 (0.10, 2.29)	.35
Marital status
Married or cohabitating	0.37 (0.21, 0.66)	.001	0.35 (0.18, 0.66)	<.01
Not married	Ref	-	Ref	-
Employment status
Employed	Ref	-	Ref	-
Not employed/Retired	1.40 (0.81, 2.42)	.222	1.98 (0.91, 4.28)	.08
Education level
High school or less	3.20 (1.22, 8.42)	.018	4.15 (1.54, 11.14)	.01
At least some college	Ref	-	Ref	-

Abbreviation: CI, confidence interval

Note: Loneliness measured using the UCLA 20-item loneliness scale, with scores treated as a binary variable in logistic regression models (F = 3.22, P < .001, pseudo r-squared: 0.1286)

Discussion

Findings from this study suggest that rural residents of Washington State are not at an increased risk for experiencing loneliness compared to their urban counterparts. Our findings support recent research in which mean loneliness scores based on a 3-item scale were not statistically different among residents living in urban, micropolitan rural, and non-core rural areas.²¹ However, 41%−59% of adults living in small rural, large rural, suburban, and urban areas in our study were lonely based on a UCLA Loneliness Scale cut point of 40. This finding is consistent with evidence from the Health and Retirement Study in which the prevalence of loneliness among older adults was nearly 57% in 2012.² Our findings provide additional support that loneliness is a prevalent public health issue across the rural-urban continuum.

There are multiple reasons why rural residents may not experience greater loneliness than urban residents despite living in areas commonly characterized as geographically and socially isolating. Evidence suggests that residents of highly rural areas feel close to more relatives and have more living children and grandchildren compared to those in urban areas, and rural residents in general report having more friends than urban residents.²¹ Consequently, many rural residents may have both a stronger and larger social network of friends and family that protect against lower participation in other social activities such as attending cultural events, restaurant dining, and exercising with others.²⁶ Indeed, older rural residents have reported that relationships or interaction with others is the primary factor associated with successful aging.²⁷ Rural children have also reported having strong connections with family, neighbors, and their community yet live in sparsely populated areas and travel long distances for social activities.²⁸ These strong social connections and sense of community may be protective against loneliness for rural residents.

Although our findings suggest that rural and urban residents are at similar risk for experiencing loneliness, certain subpopulations of rural residents may be particularly vulnerable to being lonely. Researchers have reported that approximately 5% of micropolitan rural residents compared to 2% of urban residents report having no friends²¹ and that loneliness is a significant health concern for chronically ill rural residents.^4,29,30 In one study of chronically ill adults from rural Appalachia, 97% of the sample experienced significant loneliness.⁴

Given the high prevalence of loneliness and associated adverse health outcomes, effective interventions to address this public health concern are needed. In rural areas, key informants representing diverse sectors have recommended mandatory clinic-based screenings and resources for addressing social isolation and loneliness.³¹ Theeke and colleagues’ loneliness intervention based on story theory and cognitive restructuring may be a promising approach for addressing loneliness in rural areas given that the intervention was feasible to implement in primary care settings, highly accepted by participants, and decreased loneliness after a follow-up of 12 weeks.^32,33 Others have reported that an asynchronous online peer support group for chronically ill rural women was not effective in reducing feelings of loneliness,³⁴ and an Internet mental health support group intervention did not significantly decrease loneliness among suburban and urban adult men and women.³⁵ Additional research is needed to develop effective loneliness interventions for lonely rural residents who may lack access to health care services.

Limitations

Limitations to this study should be noted. Our sample was from rural and urban areas across Washington State, so findings may not be generalizable to adults living in other US regions. Unmeasured confounding variables may influence the relationship between rurality and loneliness, and researchers should consider modeling additional variables that emerge as being relevant. Findings may also differ depending on the definition of rural being used, as there are multiple definitions of rural that have implications when analyzing health-related data.³⁶ A notable strength of the current study is that our definition of rural incorporated measures of population density, which could play an important role in the risk for loneliness. Finally, the relatively modest sample size prohibited examining interactions between rurality and socioeconomic indicators as well as subgroup analyses by race and sex that may be important when considering geographic disparities in loneliness. These considerations should be addressed in future research.

Conclusions

We found no substantial evidence of significant geographic differences in loneliness among Washington State adult residents. The high prevalence of loneliness observed in this study is consistent with previous research and suggests that loneliness is a prevalent public health concern.