This paper is in the following e-collection/theme issue:

Medicine 2.0: Social Media, Open, Participatory, Collaborative Medicine (1573) GIS (Geographic Information Systems) Applications in Public Health and Spatial Epidemiology (95) Depression and Mood Disorders; Suicide Prevention (1290) e-Mental Health and Cyberpsychology (1411) Infodemiology and Infoveillance (1278) Infoveillance, Infodemiology, Digital Disease Surveillance, Infodemic Management (1015)

Published on in Vol 25 (2023)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/43623, first published .
The Association of a Geographically Wide Social Media Network on Depression: County-Level Ecological Analysis

The Association of a Geographically Wide Social Media Network on Depression: County-Level Ecological Analysis

The Association of a Geographically Wide Social Media Network on Depression: County-Level Ecological Analysis

Authors of this article:

Alaina M Beauchamp1, 2 Author Orcid Image ;   Christoph U Lehmann2, 3, 4, 5 Author Orcid Image ;   Richard J Medford5, 6 Author Orcid Image ;   Amy E Hughes2, 5 Author Orcid Image
Article Authors Cited by Tweetations (13) Metrics

    Original Paper

    1Department of Epidemiology, Human Genetics, and Environmental Sciences, School of Public Health, University of Texas Health Science Center at Houston, Dallas, TX, United States

    2Peter O'Donnell Jr School of Public Health, University of Texas Southwestern Medical Center, Dallas, TX, United States

    3Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, TX, United States

    4Lyda Hill Department of Bioinformatics, University of Texas Southwestern Medical Center, Dallas, TX, United States

    5Clinical Informatics Center, University of Texas Southwestern Medical Center, Dallas, TX, United States

    6Department of Internal Medicine, University of Texas Southwestern Medical Center, Dallas, TX, United States

    Corresponding Author:

    Alaina M Beauchamp, MPH, PhD

    Peter O'Donnell Jr School of Public Health

    University of Texas Southwestern Medical Center

    5323 Harry Hines Blvd

    Dallas, TX, 75390

    United States

    Phone: 1 2146483232

    Email: alaina.beauchamp@utsouthwestern.edu


    Related ArticleThis is a corrected version. See correction statement in: https://www.jmir.org/2023/1/e47896

    Background: Social connectedness decreases human mortality, improves cancer survival, cardiovascular health, and body mass, results in better-controlled glucose levels, and strengthens mental health. However, few public health studies have leveraged large social media data sets to classify user network structure and geographic reach rather than the sole use of social media platforms.

    Objective: The objective of this study was to determine the association between population-level digital social connectedness and reach and depression in the population across geographies of the United States.

    Methods: Our study used an ecological assessment of aggregated, cross-sectional population measures of social connectedness, and self-reported depression across all counties in the United States. This study included all 3142 counties in the contiguous United States. We used measures obtained between 2018 and 2020 for adult residents in the study area. The study’s main exposure of interest is the Social Connectedness Index (SCI), a pair-wise composite index describing the “strength of connectedness between 2 geographic areas as represented by Facebook friendship ties.” This measure describes the density and geographical reach of average county residents’ social network using Facebook friendships and can differentiate between local and long-distance Facebook connections. The study’s outcome of interest is self-reported depressive disorder as published by the Centers for Disease Control and Prevention.

    Results: On average, 21% (21/100) of all adult residents in the United States reported a depressive disorder. Depression frequency was the lowest for counties in the Northeast (18.6%) and was highest for southern counties (22.4%). Social networks in northeastern counties involved moderately local connections (SCI 5-10 the 20th percentile for n=70, 36% of counties), whereas social networks in Midwest, southern, and western counties contained mostly local connections (SCI 1-2 the 20th percentile for n=598, 56.7%, n=401, 28.2%, and n=159, 38.4%, respectively). As the quantity and distance that social connections span (ie, SCI) increased, the prevalence of depressive disorders decreased by 0.3% (SE 0.1%) per rank.

    Conclusions: Social connectedness and depression showed, after adjusting for confounding factors such as income, education, cohabitation, natural resources, employment categories, accessibility, and urbanicity, that a greater social connectedness score is associated with a decreased prevalence of depression.

    J Med Internet Res 2023;25:e43623

    doi:10.2196/43623

    Keywords

    Facebook; social connectedness; depression; county-level analysis; social media; mental health; research; ecological; geography; GIS 


    Humans fundamentally require connections to other humans. Humans have lived in social groups since the dawn of civilization [1]. The type, quality, and amount of these social connections affect psychological well-being and physical health [2]. Social connectedness decreases human mortality, improves cancer survival, cardiovascular health, and body mass, results in better-controlled glucose levels, and strengthens mental health [3].

    Engaging in social media is a popular pastime for billions of people worldwide and is considered a normative means of staying connected to other people for the technological generations. Even though it purportedly offers connectedness, social media provides the least amount of happiness among the 27 leisure activities surveyed [4,5]. The social media platform with the largest number of users worldwide is Facebook (Meta) [6], where some 1 million users log in, upload 200,000+ pictures, comment on 500,000+ posts, and react with emojis to some 4M+ posts each minute [7,8]. Since 2017, researchers employed by and collaborating with Facebook have calculated and released the Social Connectedness Index (SCI) to measure the density and geographic distribution of an individual’s connections on this social media platform [9,10].

    The composition of a social network, or the individuals, and groups that make up an individual’s social connections, can play a significant role in mental health. A strong and supportive social network can have a protective effect on mental health, while a lack of social connections can increase the risk of mental health problems such as depression and anxiety [11]. Interventions that focus on building and strengthening social connections, such as community-based programs or online support groups, can be particularly beneficial for individuals who have a limited social network [12]. Additionally, interventions that target specific groups, such as older adults or individuals with disabilities, who may be at a higher risk of social isolation, can also be effective in promoting mental health [13-15]. Another important consideration is the diversity of an individual’s social network. A diverse social network can have a protective effect on mental health by providing individuals with a wider range of perspectives and experiences. Understanding the composition of an individual’s social network can inform interventions and policies aimed at promoting mental health by building and strengthening social connections, targeting specific groups, and promoting diversity and inclusivity.

    Studies have shown that time spent using more recently developed social media platforms (eg, Twitter and Instagram) is linked to personal insecurity and negative mental health outcomes [16]. Measures of social media network connection have shown public health utility in predicting the spread of infectious diseases (eg, COVID-19), and text analyses of social media posts have described the public’s reaction to public health messaging and interventions [17-22]. However, few public health studies have leveraged large social media data to classify user network structure and geographic reach rather than solely using social media platforms. The incorporation of composition and social connection distance allows for the exploration of how populations may maintain networks across boundaries and how these “cumulative-type” lists of connections may differ in the user-perceived social support for behavioral health given a temporal and geographic disconnect from social media friends.

    The objective of this study was to determine the association between population-level web-based social connectedness and reach and depression in the population across geographies of the United States.


    Study Design

    Our study provides an ecological assessment of aggregated, cross-sectional population measures of social connectedness, and self-reported depression across all counties in the contiguous United States. We used measures obtained between 2018 and 2020 for adult residents in this area.

    Measures

    The study’s outcome of interest is self-reported depressive disorder published at the county level by the Centers for Disease Control and Prevention’s PLACES project [23]. The PLACES project disseminates measures pertaining to 27 chronic diseases for all US census tracts and counties. The PLACES depression measure derives from answers to the Behavioral Risk Factor Surveillance System (BRFSS), an annual, nationally representative survey designed to aid program planning for states by providing prevalence estimates of health behaviors and outcomes. To report a depressive disorder diagnosis, a survey respondent must answer “yes” to the survey question: “Have you ever been told by a doctor, nurse, or other health professional that you had a depressive disorder (including depression, major depression, dysthymia, or minor depression)?” We used the most recently published data from 2018 to estimate the percentage of county residents with a depressive disorder.

    The study’s main exposure of interest is the SCI, a pair-wise composite index describing the “strength of connectedness between 2 geographic areas as represented by Facebook friendship ties” [9]. This measure describes the density and geographic reach of county residents’ social network using Facebook friendships, and it can differentiate between local and long-distance Facebook connections. Specifically, the SCI is a measure of the probability of each Facebook user having a connection to another Facebook user across the geography of their respective residential locations, from county i to county j. We downloaded the August 2020 version of this measure from Facebook’s Data for Good platform [9].

    To create a measure that works as a composite, we calculated the relative probability of Facebook friendships (Social connectednessi,j) as a county-level representation of the relative probability of Facebook users in county i being connected to given users in county j (SCIi,j). The scale of the relative probability is classified using ranks of the county’s measure to categorize SCI into groups relative to the top quintile in the United States (ie, <1, 1-2, 2-3, 3-5, 5-10, 10-25, 25-100, or ≥100 20th percentile). The higher the SCI rank, the more likely it is that county residents have more connections and that those connections span a larger geographical area. Lower ranks imply both fewer connections and relatively more local connections.

    We included several covariate measures to control for confounding due to the county’s urban or rural status, natural surroundings of the area (eg, greenspace, lakes, and mountains), geographic connectivity, and population characteristics. Because urban or rural status may affect social media use or depression prevalence [24], we included a measure of county-level urbanicity using the urban or rural classification scheme of the Centers for Disease Control and Prevention National Center for Health Statistics from 2015 (eg, large metropolitan, medium metropolitan, small metropolitan, micropolitan, and noncore) [25].

    As physical surroundings like beautiful scenery may affect happiness and depression [26-28], we included the Natural Amenities Scale, a measure created by the US Department of Agriculture Economic Research Service, to represent the natural surroundings of a county area that enhance that location as a place to live (eg, climate, topography, and water area) [29]. We operationalized this measure by ranking counties comparatively across the continental United States, with a rank of 1 indicating the fewest natural amenities and a rank of 7 indicating the most natural amenities.

    We derived a measurement of geographic accessibility for each county from the Environmental Protection Agency’s Environmental Quality Index (EQI; 2006-2010) built environment component [30]. This measure provides a composite score describing each county’s roads, highway safety, commuting behaviors, business and housing environments, walkability, and green space. We created quintiles of the EQI to summarize the traversability of the built environments for each county, with the first quintile representing the highest accessibility and the last quintile representing the lowest accessibility.

    Increasing income increases happiness, albeit with diminishing returns [31]. Educational status, age, gender, and cohabitating may further affect depression and happiness. To adjust for these factors, we collected median household income (US dollars), the average age of county residents (years), the percentage of male residents, the percentage of residents with the highest educational attainment of high school degree (including equivalency) or greater, and the percentage of cohabitating adult residents from the American Community Survey 5-year estimates from 2016 to 2020.

    As chronic illness and depression are linked [32], we added a measure of self-reported poor physical health status from the BRFSS 2018 survey as a proportion of the adult population self-reporting 14 or more days during the past 30 days during which their physical health was not good.

    Occupation is associated with depression through job-related psychosocial stressors and socioeconomic status [33-35], and occupation-related travel expands the geographic reach of occupational and social networks [36-38]. Therefore, we incorporated information from the American Community Survey occupation by industry tables into our analysis. We included 4 classes of occupations: (1) management, business, science, and the arts; (2) service; (3) sales and office; and (4) natural resources, construction, and maintenance.

    Statistical Analysis

    Descriptive statistics of US counties and the 4 US census regions (ie, Northeast, Midwest, South, and West) were generated using frequencies, percentages, means, and SD. We used a global Moran Index test of spatial autocorrelation for each variable [39]. We identified local clusters of hot and cold spots using Anselin local Moran Index test for clustering with 9999 permutations, to which we applied a false discovery rate correction for multiple comparisons [40,41]. All independent variables were z score–standardized prior to modeling. Nonspatial generalized least squares regression bivariate models were fit for each independent variable with the outcome (depression). We fit a multivariable generalized least squares model to test the spatial autocorrelation of the regression residuals and the variance inflation factors. We assessed model fit via accepted and well-known specification tests and the Akaike Information Criterion [42-45]. We determined the statistical significance of measures using an α level of .05. We used ArcGIS Pro 2.9 (Environmental Systems Research Institute) to create study maps, and we conducted all analyses in RStudio (version 1.2; R Foundation) [46,47]. For the maps, we visualized major cities with over 500,000 residents. We obtained Dominic Russel’s script to process the Facebook social connectedness data for US county units through a web-based repository [48]. We implemented spatial autocorrelation and clustering tests and econometric models using the spdep and spatialreg packages [49,50].

    Ethics Approval

    Our study used aggregated and anonymized publicly available data sets. Data were not linked to identifiable information to ensure privacy and confidentiality, accuracy and reliability, transparency and openness, and adhere to any applicable data usage policies. This study was reviewed, and a waiver of informed consent was granted by the University of Texas Health Science Center Institutional Review Board (HSC-SPH-22-0181).


    Descriptive statistics for counties in the continental United States, stratified by census regions, show regional patterns in our outcome and main independent measures (Table 1). County-level self-reported depression was lowest for counties in the Northeast (mean 18.6% of adults) and highest for Southern counties (mean 22.4% of adults). Social networks in Northeastern counties contained moderately local connections (SCI 5-10 the 20th percentile for 79, 36% of counties), whereas social networks in Midwest, Southern, and Western counties included mostly local connections (SCI 1-2 the 20th percentile for n=598, 56.7%, 401, 28.2%, and 159, 38.4%, respectively).

    Many counties (42.1/100, 42.1%) across the United States are noncore (ie, rural). The West had the most counties with high natural amenities (rank=7 for 69.3%, n=285 counties), whereas the Midwest had the most counties with low natural amenities (rank=1 for 41.5%, n=438 counties). Similarly, the West had the highest frequency of counties with the most traversable environments (EQI=1 for 42.6%, n=191 counties). The South contained the most counties with the least traversable environments (EQI=5 for 31.5%, n=448).

    The Western region had the highest percentage of people working in the 4 occupation classes. In the Northeast, counties had the highest median household income (mean US $65,277, SD US $16,591). The overall average age of residents was 41.6 (SD 5.5) years, and there was an average male proportion of 50.1% (50.1/100). The Southern United States had the highest frequency of county residents reporting being in poor general health (mean 25.2%, SD 4.7%). The Midwest and the South had the highest frequency of residents with a high school education or greater (mean 34.9%, SD 6.5% and mean 34.9%, SD 7.4%, respectively). Additionally, the Midwest had the highest proportion of residents classified as cohabitating (mean 58.2%, SD 5.4%).

    All predictor measures included in our study displayed significant spatial autocorrelation (Table 2, column 1), and all predictors except for the average age of county residents were significantly associated with self-reported depressive disorders (Table 2, column 2). These associations were robust to adjustments for covariates and spatial structure, except for urbanicity (Table 2, columns 3 and 4). The best-fitting model was a multivariable spatial eigenvector filter generalized least squares model (Table 2, column 4). As the quantity and distance that social connections span (ie, SCI) increases, the prevalence of depressive disorders is 0.3% (SE 0.1) lower per rank. In other words, counties classified as the most connected (ie, SCI≥100 the 20th percentile) had a depressive disorder prevalence that was 2.1% lower than counties considered the least connected (ie, SCI<1 the 20th percentile). Greater income, age, and percent of male residents (β=.1, β=.3, and β=.3, respectively) were all associated with fewer depressive disorders. Predictors showing positive associations with depressive disorders included natural amenities (β=.2, SE .0), the built environment (β=.2, SE .0), management, business, science, and the arts (β=.3, SE .1), service occupations (β=.1, SE .0), poor general health (β=.9, SE .1), and a high school education or more (β=.2, SE .0).

    The geographic distribution of depressive disorders was visually clustered along the Southwestern banks of the Mississippi River, in Appalachia, and in the Pacific Northwest (Figure 1, Panel A). Our local Moran I analysis shows clusters of high (ie, “hot spots”) and low (ie, “cold spots”) depressive disorder prevalence across the contiguous United States (Panel B) that confirm the visual analysis of Panel A. Local clusters of low depression prevalence comprised the state of New Jersey and surrounding counties, a large part of the northern Midwest, central California, and southern California. The majority of Tennessee, Kentucky, and West Virginia counties are part of a local cluster of high depression, as are parts of surrounding states (eg, Alabama, Arkansas, and Louisiana) and the Pacific Northwest (Panel B). In addition to these concordant clusters, the local Moran I analysis identified areas with sharp changes in prevalence over spatially contiguous regions (ie, low and high outliers). Specifically, one county (ie, Philadelphia County, Pennsylvania) was a high outlier of depressive disorder prevalence, and 3 counties (ie, Washington County, Pennsylvania; Garrett County, Maryland; and Frederick County, Virginia) were low outliers.

    Table 1. Descriptive statistics of county characteristics by census region (N=3142).

    		OverallNortheastaMidwestbSouthcWestd
    Depression (%), mean (SD)21.0 (3.8)18.6 (6.4)19.9 (3.2)22.4 (3.2)20.2 (3.2)
    Social Connectedness Index Percentile, n (%)

    		<1×—Lowest social connectedness300 (9.4)0 (0)132 (12.5)124 (8.7)36 (8.7)

    		1-2×1162 (37.4)4 (1.8)598 (56.7)401 (28.2)159 (38.4)

    		2-3×503 (16.2)25 (11.5)173 (16.4)232 (16.3)73 (17.6)

    		3-5×467 (15.0)55 (25.4)93 (8.8)250 (17.6)69 (16.7)

    		5-10×376 (12.1)79 (36.4)48 (4.6)197 (13.9)52 (12.6)

    		10-25×230 (7.4)47 (21.7)11 (1.0)151 (10.6)21 (5.1)

    		25-100×70 (2.3)7 (3.2)0 (0)59 (4.2)4 (1.0)

    		≥100×—Highest social connectedness8 (0.3)0 (0)0 (0)8 (0.6)0 (0)
    Urbanicity, n (%)

    		Noncore1310 (42.1)41 (18.9)520 (49.3)569 (40.0)180 (43.5)

    		Micropolitan637 (20.5)46 (21.2)232 (22.0)261 (18.4)98 (23.7)

    		Small metro355 (11.4)27 (12.4)112 (10.6)167 (11.7)49 (11.8)

    		Medium metro369 (11.9)38 (17.5)78 (7.4)207 (14.6)46 (11.1)

    		Large metro437 (14.1)65 (30.0)113 (10.7)218 (15.3)41 (9.9)
    Natural Amenities Index Rank, n (%)

    		1—Fewest natural amenities440 (14.3)0 (0)438 (41.5)0 (0)2 (0.5)

    		2440 (14.3)32 (15.3)263 (24.9)141 (10.0)4 (1.0)

    		3440 (14.3)26 (12.4)104 (9.9)306 (21.8)4 (1.0)

    		4440 (14.3)50 (23.9)67 (6.4)312 (22.2)11 (2.7)

    		5440 (14.3)72 (34.5)96 (9.1)243 (17.3)29 (7.1)

    		6440 (14.3)28 (13.4)55 (5.2)281 (20.0)76 (18.5)

    		7—Most natural amenities440 (14.3)1 (0.5)32 (3.0)122 (8.7)285 (69.3)
    Environmental Quality Index Quintiles, n (%)

    		1—Highest accessibility628 (20.0)21 (9.7)319 (30.2)97 (6.8)191 (42.6)

    		2629 (20.0)37 (17.1)265 (25.1)235 (16.5)92 (20.5)

    		3628 (20.0)64 (29.5)201 (19.1)301 (21.2)62 (13.8)

    		4629 (20.0)71 (32.7)164 (15.6)341 (24.0)53 (11.8)

    		5—Lowest accessibility628 (20.0)24 (11.1)106 (10.1)448 (31.5)50 (11.2)
    Occupation (%), mean (SD)

    		Management, business, science, and arts33.1 (7.0)37.9 (7.0)33.6 (6.0)31.4 (7.0)35.1 (7.2)

    		Service17.7 (3.8)18.0 (2.7)16.8 (3.3)18.0 (3.9)19.0 (4.6)

    		Sales and office20.0 (3.1)20.4 (1.8)19.4 (2.8)20.4 (3.1)19.7 (3.8)

    		Natural resources, construction, and maintenance12.5 (4.2)9.9 (3.0)12.1 (3.5)12.6 (4.2)14.0 (5.2)
    Median household income (US $), mean (SD)54,850 (14,600)65,277 (16,591)56,558 (10,594)50,882 (14,973)58,653 (16,388)
    Age (years), mean (SD)41.6 (5.5)43.1 (4.1)42.1 (5.2)41.0 (5.2)41.7 (7.2)
    Male (%), mean (SD)50.1 (2.4)49.6 (1.9)50.2 (1.7)49.8 (2.9)50.8 (2.5)
    Poor general health (%), mean (SD)21.8 (5.5)16.8 (6.3)19.2 (3.7)25.2 (4.7)19.6 (4.1)
    ≥High school education (%), mean (SD)33.9 (7.4)33.6 (8.5)34.9 (6.5)34.9 (7.2)28.2 (7.0)
    Cohabitation (%), mean (SD)56.4 (6.7)56.6 (5.2)58.2 (5.4)54.6 (7.4)57.9 (6.9)

    aNortheast census region: Pennsylvania, New York, New Jersey, Rhode Island, Connecticut, Massachusetts, New Hampshire, Vermont, and Maine.

    bMidwest census region: North Dakota, Minnesota, Wisconsin, Michigan, South Dakota, Iowa, Illinois, Indiana, Ohio, Nebraska, Kansas, and Missouri.

    cSouthern census region: Texas, Oklahoma, Arkansas, Louisiana, Mississippi, Tennessee, Kentucky, Alabama, Georgia, Florida, West Virginia, Maryland, Delaware, Washington DC, Virginia, North Carolina, and South Carolina.

    dWestern census region: Washington, Idaho, Montana, Wyoming, Oregon, California, Nevada, Utah, Colorado, Arizona, and New Mexico.

    Table 2. Bivariate and multivariate regressions of SCI and depression.a

    		UnadjustedAdjusted

    		Moran IndexNonspatial GLM,b β (SE)Nonspatial GLM,b β (SE)Spatial filter LM,c β (SE)
    Social Connectedness Index Percentile0.7d−.7 (.1)d−.3 (.1)d−.3 (.1)d
    Urbanicity0.5d−.3 (.1)d.1 (.1).1 (.0)
    Natural Amenities Index Rank0.8d.2 (.1)d.2 (.1)d.2 (.0)e
    Environmental Quality Index Quintiles0.5d1.0 (.1)d.2 (.1)e.2 (.0)d
    Occupation (%)

    		Management, business, science, and arts0.4d−1.2 (.1)d.2 (.1)e.3 (.1)d

    		Service0.3d.3 (.1)e−.0 (.1).1 (.0)e

    		Sales and office0.1d.1 (.1).0 (.1).1 (.0)e

    		Natural resources, construction, and maintenance0.3d.0 (.1)−.9 (.1)d−.0 (.0)
    Median household income (US $)0.3d−1.1 (.1)d−.5 (.1)d−.1 (.0)d
    Age (years)0.3d.1 (.1)−.1 (.1)−.3 (.0)d
    Male (%)0.1d−.4 (.1)d−.4 (.1)d−.3 (.0)d
    Poor general health (%)0.7d2.0 (.1)d2.1 (.1)d.9 (.1)d
    ≥High school education (%)0.5d1.4 (.1)d.7 (.1)d.2 (.0)d
    Cohabitation (%)0.3d−.2 (.1)e.8 (.1)d.0 (.0)

    aResults indicating lower depression associated with higher social connectedness, less natural amenities, higher built environment accessibility, more of the workforce in management, business, science, and arts, service, or sales and office occupations, higher median household income, older age, larger male population, lower proportion of population in poor general health, and higher educational attainment.

    bGLM: generalized linear model.

    cLM: linear model.

    dP<.001.

    eP<.05.

    The highest-ranking category of social connectedness in the United States (ie, SCI≥100) was spatially concentrated in 1 area and comprised 8 counties: the District of Columbia; Alexandria, Arlington, Fairfax, and Falls Church counties in Virginia; and Prince George’s, Montgomery, and Charles counties in Maryland (Figure 1, Panel C). Although this map shows a clear pattern of high SCI in metropolitan counties with low SCI in their surrounding counties, the concentrated SCI area around Washington, DC, with the 8 highest-ranked SCI counties, extended far beyond the urban centers. Our local Moran I analysis indicated a cluster of high social connectedness along the Eastern coast; in this cluster, 2 counties were classified as low outliers of social connectedness (ie, Washington County, New York and Mifflin County, Pennsylvania; Panel B). Additionally, there was a cluster of high social connectedness in Colorado with 3 low outliers (ie, San Miguel County, Mesa County, and Jackson County; Panel D). In Florida, Putnam County stood out as having low SCI compared to surrounding areas (Panel D). Clusters of low social connectedness were present across states in the Midwest and Southern regions. We identified 13 counties across the United States where social connectedness was high compared to low neighboring counties, suggesting localized hubs of social connectedness that did not extend to surrounding populations (Midland, Lubbock, and Taylor counties in Texas; Jackson County, Oklahoma; Kiowa County, Colorado; Blaine County, Idaho; Stevens County, Minnesota; Boyle County, Kentucky; Pulaski and Greene counties, Missouri; Isabella County, Michigan; McDonough and Peoria counties, Illinois; and Allen County, Indiana).

    ‎
    Figure 1. County social connectedness, depression, and local cluster measures in the continental United States. Displayed here in depression quantiles (A), local clusters of depression (B), Social Connectedness Index percentile ranks (C), and local clusters of social connectedness (D).

    Principal Findings

    Exploring the county-level SCI and depression showed that social connections on Facebook that span larger geographic areas are associated with fewer county residents reporting depressive disorders. This association was robust to both the type of model and adjustments for confounding factors such as income, education, cohabitation, natural resources, employment categories, accessibility, and urbanicity. Further, a visual analysis of these 2 population-based measures shows clear local and regional patterns that are robust to formal spatial analysis.

    Contextualization

    We found the proportionally highest SCI value in areas where policy-making (Washington, DC) and highly localized, exclusive socialization happens (eg, Aspen, Colorado, and Jackson Hole, Wyoming). Washington, DC (policy making) and Atlanta, Georgia (public health and basic science) had higher SCI values than New York City, New York, and Los Angeles, California. To make connections across larger distances, individuals must have the means and opportunity to travel and make connections in far-flung places. The high SCI scores in counties surrounding Washington, DC, could be driven by the proportion of residents that work for, support, or lobby various government or congressional entities [51]. These individuals are likely to have many long-distance Facebook relationships with friends, family, and constituents from their home states, have frequent meetings with out-of-town connections or travel for work. Other clusters of places representing geographic hubs of high social connectedness include locations favored by US individuals with affluence. Therefore, SCI arguably measures the ability and means to travel, make novel connections, and have high levels of influence; members of these groups likely have higher social status through their income, education, and occupation. If SCI measures one’s affluence and ability to build a supportive or successful environment, then the effect on the reduction of depression may be largely influenced by access to favorable systems (eg, financial, health care, and occupational).

    The absence of personal autonomy and agency has been linked to depression; thus, counties with a concentration of individuals who are not only highly autonomous but also highly influential may have lower rates of depressive disorders [52]. Further examples of this relationship can be observed in counties identified through our tests of local spatial autocorrelation. In Texas, Midland, Lubbock, and Taylor counties all share a similar vein of occupational influence and connection. These counties contain concentrations of oil, gas, wind energy, and cotton production compared to the rest of East Texas. Counties such as Taylor, Texas, Jackson, Oklahoma, and Pulaski, Missouri, all have large military bases. Lastly, Boyle County, Kentucky, has a small liberal arts college that has hosted multiple vice presidential debates and trained numerous alumni who became politicians. In addition, certain regions across the United States displayed commonly seen patterns of health outcomes: the weather in the Pacific Northwest is well-known as a cause of the seasonal affective disorder [53], and the high prevalence of depressive disorders in Southern states and Appalachian regions reflect similar patterns of poor health outcomes in diabetes, cancer, and opioid use disorders in this region.

    As seen in the US map of depression prevalence, there is a notable pattern of low prevalence in certain states, such as Nebraska, North and South Dakota, Iowa, and Wyoming. This pattern is not an artifact of the BRFSS data and has been noted in prior years of collection [54]. These states are largely underserved by mental health professionals and have low availability of services, leading to artificially low rates of mental illness diagnoses [55]. In addition, these states contain large low-income populations, which may lead to low health care access and make them difficult to reach by the state-based BRFSS survey.

    Limitations

    This study has limitations requiring discussion. First, this study uses areal data drawn from separate cross-sectional sources; both the ecological fallacy and the study design limit our ability to detect evidence of truly causal relationships and apply our macro-level inferences to individual-level relationships. Despite these challenges, we were able to ascertain a novel association between social connectedness and depression, and including data for all areas of the contiguous United States enabled us to provide a detailed, nationally relevant analysis of this association.

    Facebook usage rates can vary depending on factors such as age, education level, and income. Additionally, some individuals may choose not to use Facebook for personal reasons, such as privacy concerns or a lack of interest in social media. Although the Facebook data were derived from all 2.74 billion users [56], providing one of the largest samples of social media networks across the United States, younger generations have left Facebook for newer platforms with different foci and sharing methods (eg, Snapchat, TikTok, and YouTube) [57]. Therefore, younger Americans are underrepresented in the measure of social connectedness.

    Finally, we employed a measure of depression based on self-report, which is subject to known biases; however, these data represent the only source of publicly available, nationally representative, county-level measures of depression [58].

    Conclusions

    In this study, we uncovered a previously unexplored correlation between county-level social connectedness and depression. Given the known association between the use of social media and depression—where use and interaction with social media causally increase depression—our results suggest that SCI is not a measure reflecting the use of social media but instead functionally measures social network size, affluence, and social influence. To determine if increasing a location’s SCI can be used to reduce depression prevalence, the next investigation should focus on localities that have engaged in cultural exchange programs, which increase interchanges of people and cultural activities, and controls.

    Implications for Clinical Practice or Health Policy

    Social media has become a prevalent aspect of modern communication and is linked to both positive and negative effects on mental health. Therefore policy makers and community leaders need to consider the potential impact of social media use on individuals and communities and develop strategies to promote healthy use. One key aspect of social media policy making is establishing guidelines for the responsible use of social media platforms. This can include measures such as age restrictions and education on the potential risks of excessive social media use [59]. Another important consideration is the development of community-based interventions to prevent depression and other mental health issues related to social media use. These interventions can include support groups, counseling services, and educational programs that teach individuals how to use social media in a healthy and balanced way [60]. Additionally, community leaders can work with social media companies to promote positive content and provide resources for individuals who may be struggling with mental health issues. In summary, social media policy-making and community-based interventions can play an important role in promoting healthy social media use and preventing depression and other mental health issues [61]. By establishing guidelines for responsible use, addressing cyberbullying, and providing support and resources for individuals, policy makers and community leaders can help ensure that social media is used in a way that promotes well-being for all.

    Major health policy developers, like the US Preventative Task Force and the US Department for Health and Human Services, have identified goals for the current decade to address access to screening and treatment of depressive disorders for the US populace [14,15]. Equitable access to mental illness diagnosis and treatment has yet to be established in the United States for all persons and geographies. Specifically, many studies of accessibility fail to consider the geographic differences in economic accessibility of mental health services (eg, transportation availability and cost, and local proximity to various providers and facility types) [62,63]. The use of technology and social networks has been suggested as a promising approach to reaching these populations in need [64,65]. New clinical approaches like telehealth and behavioral intervention technologies allow for services to reach those often underserved or isolated, whether through financial disparities or geographic isolation [66,67]. The relationship between Facebook’s SCI and the indicated relationship to social capital in the United States points toward the potential for new approaches to mental health intervention through social networking sites, through which clinicians and policy makers can target and reach populations in need [68].

    Acknowledgments

    This research has received no external funding.

    Conflicts of Interest

    None declared.

    ‎
    Multimedia Appendix 1

    High resolution map of county social connectedness, depression, and local cluster measures in the continental United States.

    PNG File , 2887 KB
    1. Festinger L. The social organization of early human groups. In: Changing Conceptions of Crowd Mind and Behavior. Berlin: Springer; 1986:203-215.
    2. Martino J, Pegg J, Frates EP. The connection prescription: using the power of social interactions and the deep desire for connectedness to empower health and wellness. Am J Lifestyle Med 2017;11(6):466-475 [FREE Full text] [CrossRef] [Medline]
    3. Holt-Lunstad J, Smith TB, Layton JB. Social relationships and mortality risk: a meta-analytic review. PLoS Med 2010;7(7):e1000316 [FREE Full text] [CrossRef] [Medline]
    4. Bryson A, MacKerron G. Are you happy while you work? Econ J 2017;127(599):106-125 [FREE Full text] [CrossRef]
    5. Poushter J, Bishop C, Chwe H. Social media use continues to rise in developing countries but plateaus across developed ones. Pew Research Center. 2018.   URL: https:/​/www.​pewresearch.org/​global/​2018/​06/​19/​social-media-use-continues-to-rise-in-developing-countries-but-plateaus-across-developed-ones/​ [accessed 2022-03-12]
    6. Auxier B, Anderson M. Social media use in 2021. Pew Research Center. 2022.   URL: https://www.pewresearch.org/internet/2021/04/07/social-media-use-in-2021/ [accessed 2022-03-13]
    7. Chaffey D. What happens online in 60 seconds in 2021? Smart Insights. 2021.   URL: https://www.smartinsights.com/internet-marketing-statistics/happens-online-60-seconds/ [accessed 2022-03-13]
    8. Dobrilova T. Things that happen on the Internet every 60 seconds in 2022? Tech Jury. 2022.   URL: https://techjury.net/blog/internet-in-real-time/#gref [accessed 2022-03-13]
    9. Bailey M, Cao R, Kuchler T, Stroebel J, Wong A. Social connectedness: measurement, determinants, and effects. J Econ Perspect 2018;32(3):259-280. [CrossRef]
    10. Bailey M, Cao RR, Kuchler T, Stroebel J, Wong A. Measuring social connectedness. National Bureau of Economic Research. 2017.   URL: https://www.nber.org/system/files/working_papers/w23608/w23608.pdf [accessed 2022-03-13]
    11. Santini ZI, Koyanagi A, Tyrovolas S, Mason C, Haro JM. The association between social relationships and depression: a systematic review. J Affect Disord 2015;175:53-65. [CrossRef] [Medline]
    12. Zagic D, Wuthrich VM, Rapee RM, Wolters N. Interventions to improve social connections: a systematic review and meta-analysis. Soc Psychiatry Psychiatr Epidemiol 2021;57(5):885-906. [CrossRef] [Medline]
    13. Tan JH, Abdin E, Shahwan S, Zhang Y, Sambasivam R, Vaingankar JA, et al. Happiness and cognitive impairment among older adults: investigating the mediational roles of disability, depression, social contact frequency, and loneliness. Int J Environ Res Public Health 2019 Dec 06;16(24):4954 [FREE Full text] [CrossRef] [Medline]
    14. Pronk NP, Kleinman DV, Richmond TS. Healthy people 2030: moving toward equitable health and well-being in the United States. EClinicalMedicine 2021;33:100777 [FREE Full text] [CrossRef] [Medline]
    15. Siu AL, US Preventive Services Task Force (USPSTF). Screening for Depression in Adults: US Preventive Services Task Force Recommendation Statement. JAMA 2016 Jan 26;315(4):380-387. [CrossRef] [Medline]
    16. Memon AM, Sharma SG, Mohite SS, Jain S. The role of online social networking on deliberate self-harm and suicidality in adolescents: a systematized review of literature. Indian J Psychiatry 2018;60(4):384-392 [FREE Full text] [CrossRef] [Medline]
    17. Basit MA, Lehmann CU, Medford RJ. Managing pandemics with health informatics: successes and challenges. Yearb Med Inform 2021;30(1):17-25 [FREE Full text] [CrossRef] [Medline]
    18. Diaz MI, Hanna JJ, Hughes AE, Lehmann CU, Medford RJ. The politicization of ivermectin tweets during the COVID-19 pandemic. Open Forum Infect Dis 2022;9(7):ofac263 [FREE Full text] [CrossRef] [Medline]
    19. Medford RJ, Saleh SN, Sumarsono A, Perl TM, Lehmann CU. An "infodemic": leveraging high-volume Twitter data to understand early public sentiment for the coronavirus disease 2019 outbreak. Open Forum Infect Dis 2020;7(7):ofaa258 [FREE Full text] [CrossRef] [Medline]
    20. Saleh SN, Lehmann CU, McDonald SA, Basit MA, Medford RJ. Understanding public perception of coronavirus disease 2019 (COVID-19) social distancing on Twitter. Infect Control Hosp Epidemiol 2021;42(2):131-138 [FREE Full text] [CrossRef] [Medline]
    21. Tsao SF, Chen H, Tisseverasinghe T, Yang Y, Li L, Butt ZA. What social media told us in the time of COVID-19: a scoping review. Lancet Digit Health 2021;3(3):e175-e194 [FREE Full text] [CrossRef]
    22. Cinelli M, Quattrociocchi W, Galeazzi A, Valensise CM, Brugnoli E, Schmidt AL, et al. The COVID-19 social media infodemic. Sci Rep 2020;10(1):16598 [FREE Full text] [CrossRef] [Medline]
    23. PLACES: Local Data for Better Health.: Centers for Disease Control and Prevention; 2018.   URL: https://www.cdc.gov/places [accessed 2021-12-01]
    24. Glaeser EL, Gottlieb JD, Ziv O. Unhappy cities. J Labor Econ 2016;34(Suppl 2):S129-S182 [FREE Full text] [CrossRef] [Medline]
    25. Ingram DD, Franco SJ. 2013 NCHS Urban-Rural Classification Scheme for Counties. Vital Health Stat 2 2014 Apr(166):1-73 [FREE Full text] [Medline]
    26. Seresinhe CI, Preis T, MacKerron G, Moat HS. Happiness is greater in more scenic locations. Sci Rep 2019;9(1):4498 [FREE Full text] [CrossRef] [Medline]
    27. Winters JV, Li Y. Urbanisation, natural amenities and subjective well-being: evidence from US counties. Urban Studies 2017;54(8):1956-1973 [FREE Full text] [CrossRef]
    28. Abraham A, Sommerhalder K, Abel T. Landscape and well-being: a scoping study on the health-promoting impact of outdoor environments. Int J Public Health 2010;55(1):59-69 [FREE Full text] [CrossRef] [Medline]
    29. McGranahan D. Natural Amenities Scale.: U.S. Department of Agriculture; 2012.   URL: https://www.ers.usda.gov/data-products/natural-amenities-scale/ [accessed 2021-12-30]
    30. Environmental quality index - technical report (2006-2010). Science Inventory. 2020.   URL: https:/​/cfpub.​epa.gov/​si/​si_public_record_report.​cfm?dirEntryId=350545&Lab=NHEERL&subject=Air%20Research&showcriteria=0& searchall=Air%20and%20Modeling&acttype=Product&timstype=PUBLISHED+REPORT&sortby=revisionDate [accessed 2023-03-13]
    31. Killingsworth MA. Experienced well-being rises with income, even above $75,000 per year. Proc Natl Acad Sci USA 2021;118(4):e2016976118 [FREE Full text] [CrossRef] [Medline]
    32. Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B. Depression, chronic diseases, and decrements in health: results from the world health surveys. Lancet 2007;370(9590):851-858 [FREE Full text] [CrossRef]
    33. Bender A, Farvolden P. Depression and the workplace: a progress report. Curr Psychiatry Rep 2008;10(1):73-79. [CrossRef] [Medline]
    34. Bonde JPE. Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med 2008;65(7):438-445. [CrossRef] [Medline]
    35. Madsen IEH, Nyberg ST, Magnusson Hanson LL, Ferrie JE, Ahola K, Alfredsson L, IPD-Work Consortium. Job strain as a risk factor for clinical depression: systematic review and meta-analysis with additional individual participant data. Psychol Med 2017 Jun;47(8):1342-1356 [FREE Full text] [CrossRef] [Medline]
    36. Aultman-Hall L, LaMondia JJ, Ullman H, Suender M. Surveying social network geography to model long-distance travel. Transp Res Proc 2018;32:542-552 [FREE Full text] [CrossRef]
    37. Villarreal A. The U.S. occupational structure: a social network approach. SocScience 2020;7:187-221. [CrossRef]
    38. Wall TA, Macfarlane GS, Watkins KE. Exploring the use of egocentric online social network data to characterize individual air travel behavior. Transp Res Rec 2014;2400(1):78-86 [FREE Full text] [CrossRef]
    39. Moran PA. Notes on continuous stochastic phenomena. Biometrika 1950;37(1-2):17-23. [Medline]
    40. Anselin L. Local indicators of spatial association: LISA. Geogr Anal 1995;27(2):93-115 [FREE Full text] [CrossRef]
    41. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol 1995;57(1):289-300 [FREE Full text] [CrossRef]
    42. Akaike H. A new look at the statistical model identification. IEEE Trans Automat Contr 1974;19(6):716-723. [CrossRef]
    43. Fotheringham AS, Brunsdon C, Charlton M. Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. Hoboken: John Wiley & Sons; 2003.
    44. Leung Y, Mei CL, Zhang WX. Statistical tests for spatial nonstationarity based on the geographically weighted regression model. Environ Plan 2000;32(1):9-32 [FREE Full text] [CrossRef]
    45. Pace RK, LeSage JP. A spatial Hausman test. Econ Lett 2008;101(3):282-284 [FREE Full text] [CrossRef]
    46. ArcGIS Pro. Redlands: ESRI; 2021.   URL: https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview [accessed 2022-03-13]
    47. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2021.   URL: https://www.gbif.org/tool/81287/r-a-language-and-environment-for-statistical-computing [accessed 2022-03-13]
    48. Russel D. Social Connectedness Index: Example Scripts. 2021.   URL: https://github.com/social-connectedness-index/example-scripts [accessed 2021-11-01]
    49. R Core Team. The Comprehensive R Archive Network. 2015.   URL: https://cran.r-project.org/ [accessed 2022-03-13]
    50. Bivand RS, Piras G. Comparing implementations of estimation methods for spatial econometrics. J Stat Softw 2015;63(18):1-36 [FREE Full text] [CrossRef]
    51. Geographic profile for political scientists. US Bureau of Labor Statistics. 2021.   URL: https://www.bls.gov/oes/current/oes193094.htm [accessed 2022-03-13]
    52. Bekker MH, Belt U. The role of autonomy-connectedness in depression and anxiety. Depress Anxiety 2006;23(5):274-280. [CrossRef] [Medline]
    53. Kerr DC, Zava DT, Piper WT, Saturn SR, Frei B, Gombart AF. Associations between vitamin D levels and depressive symptoms in healthy young adult women. Psychiatry Res 2015;227(1):46-51 [FREE Full text] [CrossRef] [Medline]
    54. Price JD, Barbour KE, Liu Y, Lu H, Amerson NL, Murphy LB, et al. State-specific prevalence and characteristics of frequent mental distress and history of depression diagnosis among adults with arthritis - United States, 2017. MMWR Morb Mortal Wkly Rep 2020;68(5152):1173-1178 [FREE Full text] [CrossRef] [Medline]
    55. Grimm B, Watanabe-Galloway S, Palm D, Pounds L, Deras M, Dawson S, et al. An academic and practice partnership to assess the behavioral health needs of Nebraska. Community Ment Health J 2019;55(4):561-568 [FREE Full text] [CrossRef] [Medline]
    56. Facebook Statistics and Trends. DataReportal. 2022.   URL: https:/​/datareportal.​com/​essential-facebook-stats#:~:text=Here's%20what%20the%20latest%20data,)%3A%202.​934%20billion%20(July%202022)&text=Number%20of%20people%20who%20use,)%3A%201.​968%20billion%20(July%202022)&text=Share%20of%20Facebook's%20monthly%20active,%3A%2067%25%20(July%202022) [accessed 2022-03-12]
    57. Wagner K, McLaughlin D. Facebook, alarmed by teen usage drop, left investors in the dark. Bloomberg Businessweek. 2021.   URL: https:/​/www.​bloomberg.com/​news/​articles/​2021-10-25/​facebook-files-show-growth-struggles-as-young-users-in-u-s-decline [accessed 2022-03-12]
    58. Strine TW, Mokdad AH, Balluz LS, Gonzalez O, Crider R, Berry JT, et al. Depression and anxiety in the United States: findings from the 2006 behavioral risk factor surveillance system. Psychiatr Serv 2008;59(12):1383-1390. [CrossRef] [Medline]
    59. Shah J, Das P, Muthiah N, Milanaik R. New age technology and social media: adolescent psychosocial implications and the need for protective measures. Curr Opin Pediatr 2019;31(1):148-156. [CrossRef] [Medline]
    60. Hou Y, Xiong D, Jiang T, Song L, Wang Q. Social media addiction: its impact, mediation, and intervention. Cyberpsychology 2019;13(1). [CrossRef]
    61. Cemiloglu D, Almourad MB, McAlaney J, Ali R. Combatting digital addiction: Current approaches and future directions. Technol Soc 2022;68:101832. [CrossRef]
    62. Cummings JR, Allen L, Clennon J, Ji X, Druss BG. Geographic access to specialty mental health care across high-and low-income US communities. JAMA Psychiatry 2017;74(5):476-484 [FREE Full text] [CrossRef] [Medline]
    63. Smith-East M, Neff DF. Mental health care access using geographic information systems: an integrative review. Issues Ment Health Nurs 2020;41(2):113-121. [CrossRef] [Medline]
    64. Ramos G, Chavira DA. Use of technology to provide mental health care for racial and ethnic minorities: evidence, promise, and challenges. Cogn Behav Pract 2022;29(1):15-40 [FREE Full text] [CrossRef]
    65. Sen K, Prybutok G, Prybutok V. The use of digital technology for social wellbeing reduces social isolation in older adults: a systematic review. SSM Popul Health 2021;17:101020 [FREE Full text] [CrossRef] [Medline]
    66. Brooks H, Devereux-Fitzgerald A, Richmond L, Bee P, Lovell K, Caton N, et al. Assessing the effectiveness of social network interventions for adults with a diagnosis of mental health problems: a systematic review and narrative synthesis of impact. Soc Psychiatry Psychiatr Epidemiol 2022;57(5):907-925 [FREE Full text] [CrossRef] [Medline]
    67. Butzner M, Cuffee Y. Telehealth interventions and outcomes across rural communities in the United States: narrative review. J Med Internet Res 2021;23(8):e29575 [FREE Full text] [CrossRef] [Medline]
    68. Hussenoeder FS. The bright side of social network sites: on the potential of online social capital for mental health. Digit Health 2022;8:20552076221093133 [FREE Full text] [CrossRef] [Medline]

    ‎
    BRFSS: Behavioral Risk Factor Surveillance System
    EQI: Environmental Quality Index
    SCI: Social Connectedness Index

    Edited by A Mavragani; submitted 18.10.22; peer-reviewed by M Bailey, H Kim; comments to author 12.12.22; revised version received 06.02.23; accepted 27.02.23; published 27.03.23

    Copyright

    ©Alaina M Beauchamp, Christoph U Lehmann, Richard J Medford, Amy E Hughes. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 27.03.2023.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.