Food insecurity, dietary quality, and the role of social network support and barriers among immigrant Latino adults in New York City

Abstract

The goal of this study is to describe the social networks of Latino immigrants (n=80) in New York City, and how various network features are linked with dietary quality and food insecurity. Participants had higher Healthy Eating Index (HEI) scores if their social networks were more transitive (β = 6.11, p<0.001) and had a higher proportion of social supporters (β = 4.91, p<0.001). Participants who were food insecure had networks with a higher proportion of social barriers (OR = 2.6, p<0.05). Supportive and cohesive networks may benefit dietary quality, while networks with more social barriers may exacerbate food insecurity.

Poor diets are associated with a myriad of diseases, such as cardiovascular disease, obesity, and type II diabetes (1,2). The United States Department of Agriculture (USDA) defines a healthy diet as one comprised of nutrient-dense foods including vegetables, fruits, and whole grains, with limited added sugar, sodium, and saturated fat (3). However, most individuals in the United States (U.S.) do not consume diets in line with these recommendations, and diet-related diseases cause 44,000 deaths per month in the U.S. (4).

Latinos in the U.S. disproportionately bear the burden of these diet-related diseases (5,6), due to a variety of socio-ecological barriers and risk factors. Chief among these risk factors is economic instability, as those who fall short of healthy diet recommendations are more likely to be of low-income (7), and Latinos more frequently live in poverty than non-Latinos (8). Another major risk factor for Latinos is the poverty-related condition of food insecurity (9). Food insecurity, defined as experiencing disruptions in food access and regular eating because of limited money or other resources, is higher among Latino-headed households relative to the national average (10). Food insecurity in Latinos has been associated with inadequate nutrition (11,12), and greater risk for diet-related diseases such as obesity and type 2 diabetes (13).

Further, a third of Latinos in the U.S. are foreign-born (8). Research regarding diet and food insecurity in this population is warranted, given their unique experience of immigration and subsequent challenges that increase risk for both food insecurity and poor diet. For example, immigrant Latinos experience challenges such as less access to federally-funded programs like the Supplemental Nutrition Assistance Program (SNAP), and social pressure to financially support family still living in their origin country (14); both of which may restrict money that is available for food. Indeed, food insecurity among U.S. Latino adults is more prevalent among immigrant-headed households (24.4%) than their U.S-born counterparts (18.9%) (15). Despite this higher risk for food insecurity, immigrant Latinos exhibit better diets shortly following their arrival to the U.S., compared to their U.S-born counterparts (16). However, this dietary advantage seems to be lost in subsequent generations or with longer stay in the U.S. (17).

Some have proposed that Latino immigrants are somewhat protected against food insecurity and poor diet by robust social networks—i.e., a tight-knit web of close family and friends to whom they are connected (18). In line with the social buffering hypothesis (19), these robust social networks may provide protection against the negative impacts of personal barriers and socioeconomic constraints on risk for food insecurity and poor diet. These networks serve as buffers by providing social support, including emotional support, tangible support, or informational support, that could benefit access to food. Robust networks may also have a high degree of social cohesion, meaning they are characterized by dense and interconnected ties (20). Social support may improve the diet of Latinos (21,22), and both social support and social cohesion have been linked with improved diet and food access in other groups (23,24). Community-based programs and interventions that harness social networks (e.g., food assistance programs that train and employ peer leaders) could therefore be a useful strategy for supporting healthful diets amongst Latino immigrants (25,26).

However, research that examines these relationships among diet, food insecurity, and social networks among Latino immigrants, who have unique strengths and challenges, is lacking. Though Latino immigrants may have cohesive social networks with increased social support, the ways in which they interact with these social resources may differ from other populations. For example, social support has consistently been linked with better health outcomes in other groups, but there is limited evidence of this protective relationship in immigrants (27). This could be due to the social complexities and stressors that arise with the immigration process, which strains and disrupts networks (28,29). For example, early after migration, Latino immigrants have limited resources, but their networks may consist of many transnational ties, who may not be easily accessible sources of support (28,29), leading them to rely on a few key local network members (29). However, their local ties are often other immigrants who are also facing constrained economic and social resources, whose support may be easily exhausted (29). As such, Latino immigrants may be less likely to seek support when facing food insecurity, especially if they feel they have already strained their network’s resources (14,29). For those who do seek support through mechanisms such as “food sharing”, support can actually contribute to poor diet because it is financially unrealistic to refuse offers of food, even if the food is unhealthy (30). Overall, these processes likely create network functions that are unique from that of non-migrants.

Additionally, the tumultuous nature of the immigration process and the resulting social strain emphasizes the need for studies focusing on Latino immigrants’ health to consider both positive and negative dimensions of social networks. While social networks can provide resources and social support, they can also be a source of conflict, undermining, and social barriers to health (24,31). As such, the theoretical approach adopted in this study views social ties as one of the many complex “pathways” that may both benefit, or be a barrier to, food access and healthy diets among Latino immigrants (32). This theory recognizes daily food choice as a process influenced by the interaction between structural norms (e.g., social support, social conflict, economic resources, availability of food) and cultural norms (e.g., food sharing).

Taken together, the goal of this cross-sectional study is to elucidate the social networks of Latino immigrants, and how features of these networks—social support, social cohesion, and social barriers—are associated with participants’ dietary quality and household food insecurity. Based on previous research and theory (19,24,32), our first hypothesis is that having a social network that is more cohesive (i.e., more interconnected), that has more social support (i.e. more network members that provide health support), and that has less social barriers (i.e., less network members who are health barriers) will be associated with (1a) improved dietary quality (i.e., better adherence to the USDA dietary guidelines), and (1b) reduced risk for food insecurity. And building on this, our second hypothesis is that the relationship between one’s social network and dietary quality (hypothesis 1a) will be moderated by food insecurity status; such that these positive social network features will be more strongly associated with better dietary quality among people who have experienced food insecurity, vs. people who are food secure. However, we also recognize that, though this research is theoretically driven, much of it is also exploratory, as the unique strengths and challenges of Latino immigrants have previously been understudied.

Methods

Participants (n=80) were recruited from one large Catholic Church in Bronx, New York City, at which an ongoing partnership had been established. The church liaison provided a list of potential participants who had participated in church activities to the research team. Trained bilingual research assistants contacted potential participants via phone or in person at the church/community center to explain details of study. All documents, including consent forms and surveys, were translated into Spanish by a certified translator and reviewed by the bilingual research team for content equivalence, which refers to the extent to which a construct holds similar meanings and relevance in two different cultures or languages (33).

If interested in the study, research assistants confirmed whether the person met the following eligibility criteria: at least 18 years old, and self-identifying as Chicano, Mexican, or Mexican American. Potential participants were excluded if they were not proficient in either Spanish or English. A total of 81 participants were enrolled. Standardized consent procedures were followed, and participants received a monetary incentive of $25 for each component of the study (i.e., social network and health survey; first and second dietary recall; biometrics). The study protocol underwent a full review and was approved by the City University of New York (IRB 2018–1081).

Data collection

Data were collected in person between January 2019 and June 2019 in English or Spanish, depending on the participants’ preference. Research assistants were trained in collecting social network data using EgoWeb 2.0, a social network data collection software, (34) and dietary intake data using Automated Self-Administered 24-hour recall (35).

Of the 81 participants enrolled, one was born in the U.S. and was excluded from these analyses. A final sample size of 80 was deemed appropriate for this study for several reasons. For studies utilizing social network analysis to explore ego-networks (i.e., the social ties surrounding each individual participant), smaller sample sizes are not uncommon because such data is time intensive to collect (36). Additionally, it is less biased to have a smaller sample size, yet collect data on one’s broader network, than to collect less network information (36). Finally, power analyses using G*Power for a two-tailed, fixed linear model suggested that a sample size of only 68 was needed to detect a medium effect size (f²=0.15) with 80% power and α=0.05.

Measures

Dependent variables

Food insecurity was measured at the household level using the six-item USDA Food Security Survey (37). Food insecurity was categorized as a binary outcome (food insecure, food secure). Participants with scores indicating very low, low, or marginal household food security were categorized as food insecure. Those with scores indicating high household food security were categorized as food secure.

Dietary quality was assessed using the 2015 Healthy Eating Index (38), which captures how well one’s diet aligns with USDA dietary recommendations. Diet intake was assessed via two Automated Self-Administered 24- hour recalls (ASA-24). To estimate usual dietary intake distribution (39), recalls were administered three to four days apart. Participants were strongly encouraged, but not required, to complete one recall on a weekday and one recall on a weekend day. Data was averaged across these two measures. The responses were analyzed using the Nutrition Data System for Research Software (40) to construct a HEI score. HEI scores range from 0 to 100, with higher scores representing closer adherence to the USDA Dietary Guidelines.

Social network measures

Social network data were collected using a personal social network survey (also commonly referred to as an ‘ego-network’ survey), where participants first listed the names of 20 adults who were important people in their life over the past six months (41). They were asked to list people that they have regular contact with, and this contact could be face-to-face or mediated through technology (e.g., phone, online). They subsequently reported on (i) characteristics of each network member, including their social role (family, friend, coworker, etc.), demographics (gender, age), where they currently live in relation to the participant (same household/same neighborhood/same state/another state/another country), what country they were born in (United States, Mexico, other), and whether they were a source of social support or barriers (defined below), and (ii) if each pair of network members knew each other, which was used to compute the structural features of their network, such as density. Bottom of Form

Social network predictors

Social cohesion.

Social cohesion was operationalized as network transitivity (42), which represents the phenomena of ‘a friend of a friend is my friend’. Transitivity is computed by examining the presence or absence of ties between triads of network members, within the respondent’s personal network. In computing this measure, the participant is excluded from the network structure. A triad is considered transitive when all three network members are connected to each other (e.g., A knows B, and B knows C, and A knows C). Total network transitivity is the proportion of all triads that are transitive, expressed as a percentage ranging from 0% to 100%, with a higher score representing greater transitivity.

Proportion of network members who provided social support.

Social support from network members was assessed using two questions: “In the past 6 months, who helped you or encouraged you to have a healthy lifestyle, to eat healthy foods or to be active?” and “In the past 6 months, who did you go to for information, advice, or suggestions about your health, your family’s health, or other health concerns?”. Network members that provided either type of health-related support were coded as a “social supporter”. The proportion of participant’s network members who were social supporters was computed as the number of supporters divided by the total number of network members (N=20).

Proportion of network members who were social barriers.

This was assessed using the question “In the past 6 months, who made it difficult for you to have a healthy lifestyle, eat healthy foods, or be active?”. If a network member was identified by this question, they were coded as a “social barrier”. The proportion of participant’s network members who were social barriers was computed as the number of social barriers divided by the total number of network members (N=20). As a follow-up to this question, participants were asked “Why?”, and the open-ended responses were analyzed for this study.

Social network covariates

As food insecurity is measured at the household level, the proportion of household members was considered as a potential covariate. Network density was also considered. Density is computed as the number of observed connections among all network members, divided by the total number of potential connections among those members (i.e., if all pairs of network members were connected to one another). The formula for potential connections in a network that is undirected (where a tie from A to B is equivalent to a tie from B to A) is n*(n-1)/2, where n = the number of network members. For example, if a social network has 20 members, then the number of potential connections is 20*19/2 = 190. If there are only 40 observed connections in that network (i.e., 40 pairs of alters who know each other), then the density of the network is 40/190 = 0.21. Thus, density is a continuous measure that ranges from 0 (no density, where no network members know each other) to 1 (complete density, where all network members know each other). Density is a key attribute of networks that can influence other structural network features, and should be included as a covariate in social network analyses (42). For example, when density is very low or very high, there can be little meaning in each connection between alters, because there are so few or so many connections that the deletion of connections do not substantively alter other network structures like transitivity or average path length (42). For some types of networks, there is a “sweet spot” for density in the low-to moderate range (i.e., a density from 0.15 to 0.25), where each connection between alters can have profound impacts on network structures (42). As such, network density has been associated with a range of health outcomes, including dietary patterns and related diseases (43–47).

Individual-level covariates and control variables

Given the link between acculturation to the U.S. and dietary quality (17), acculturation was included as a covariate. Acculturation was measured using the 12-item Short Marin scale (48). An averaged score was computed by summing the scores on all scale items, and dividing this by the number of items on the scale (or the number that the participant responded to). This computed score ranged from 1 to 5, where higher scores indicate strong acculturation to the U.S.

Other demographic measures that were included as covariates or control variables were education level, income, age, gender, job status, health insurance, and moving age. Respondent indicated whether they currently received benefits from the Supplemental Nutrition Assistance Program (SNAP) (yes/no). Job status was categorized as working (yes/no), with those who indicated they were working part-time or full-time, or were self-employed being assigned “yes”. As migrating at a younger age is strongly linked with acculturation (49), age of moving from country of birth to the United States (U.S.) was included as a covariate.

Analyses

Descriptive statistics were computed for the participants of the study, and for the demographics of social supporters and barriers. Using Excel, the open-ended data on network social barriers were organized into key categories by both the lead author of this study and a second research assistant.

Hypothesis 1a was addressed by using multiple linear regression with HEI score as the dependent variable (Model 1). Hypothesis 1b was addressed by using multiple logistic regression with food insecurity status as the dependent variable (Model 2, referent category: food secure). A stepwise approach was used for both models. Initial bivariate correlations were used to explore the relationships between all pairs of predictor and dependent variables. All potential predictors that were marginally correlated with either food insecurity or HEI score (p<0.1) were added into each model in related groups (i.e. demographic covariates/controls, social network covariates/controls, and social network predictors). Demographic and social network covariate and control variables were retained in both models if they were marginally significant in either models (p<0.1), to ensure that both final models had the same set of covariate/control variables. In the final models, income, education (ref=any college), acculturation, moving age, job status (ref=not working), health insurance (ref=no insurance), and social network density were retained as covariates or control variables. Social network predictors were retained in the models if they were statistically significant (p<.05). We report standardized regression estimates and standard errors for both models, and we report odds ratios with 95% confidence intervals for the logistic regression model (DV=food insecurity). Hypothesis 2 was addressed by adding the interaction term for “food insecurity” and “proportion of network members who provided social support” to Model 1. All analyses were performed using R (version 4.0.2).

Results

Descriptive statistics are summarized in Table 1 and network visualizations of select participants are provided in Figure 1. The average HEI score was 57.7 (SD=12.2), and 56.2% of participants were food insecure. On average, 29.9% (SD=23.9) of network members were social supporters, and 8.5% (SD=12.5) of network members were social barriers. Characteristics of the social supporters and barriers are summarized in Table 2. Of the 478 social supporters, 5.6% (n=27) were spouses, 37.7% (n=180) were friends, 36.2% (n=173) were close family, and 18.6% (n=89) were extended family. Additionally, 12.8% (n=61) of supporters were born in the U.S., 83.1% (n=397) currently lived in the U.S., and 16.3% (n=78) lived in the same household as the participant. Of the 136 social barriers, 8.8% (n=12) were spouses, 42.6% (n=58) were friends, 39.7% (n=54) were close family, and 11.0% (n=15) were extended family. Finally, 14.8% (n=20) were born in the U.S., 96.3% (n=131) currently lived in the U.S., and 23.5% lived in the same household as the participant. Compared to all network members, a higher proportion of social supporters were spouses and close friends, and a higher proportion of social barriers were spouses and close family, and lived in the same household.

Table 1.

Individual and social network characteristics of sample (N=80)

Individual Characteristic	M (SD) / N (%)
Demographics
Age (years)	43.1 (11.4)
Female	56 (70.0%)
Married/Living with partner	48 (60.0%)
Education level
Less than 12th grade	47 (58.8%)
High school graduate or GED	18 (22.5%)
Some college/college degree	15 (18.8%)
Part- or full-time worker, yes	52 (65.0%)
Household income
<$10,000	20 (25.0%)
$10,000–$29,999	38 (47.5%)
$30,000–$49,999	12 (15.0%)
$50,000+	10 (12.5%)
Acculturation
Age at migration to U.S. (years)	23.4 (9.4)
Marin acculturation scale	1.6 (0.5)
Score of 1.0–1.5	41 (51%)
Score of 1.51–2.0	23 (29%)
Score of 2.01–2.5	10 (13%)
Score of 2.51–3.0	4 (5%)
Score of 3.01–3.6	2 (2%)
Health Indicators
Has health insurance, yes	46 (57.5%)
Dietary Quality
Healthy eating index (HEI)a	57.7 (12.2)
Food Security
Receives SNAP benefits, yes	26 (32.5%)
USDA food security scale
Food insecure (very low, low, or marginal food security)	45 (56.2%)
Food secure (high food security)	35 (43.8%)
Social Network Characteristics	M (SD)
Network structure
Network Density (%)	51.8 (24.6)
Network Transitivity (%)	75.9 (18.0)
Social support and barriers
Proportion of supporters	29.9 (23.9)
Proportion of barriers	8.5 (12.5)

Notes:

^aScores can range from 0–100; lower scores indicate poorer diet intake quality, higher scores indicate better diet intake quality

Figure 1. Network transitivity, and social support and barriers within personal networks.

Each network represents the personal network of a different participant, where nodes represent the participant (black node) and their named alters (gray nodes), and ties represent social connections. For network transitivity, the first three graphs are participants with the lowest transitivity, and the last three graphs are those with the highest transitivity. For social support, alters that provided the participant with support are colored yellow. The first three graphs are participants with the lowest proportion of social supporters in their network, and the last three graphs are those with the highest proportion of social supporters. For social barriers, alters that were barriers are colored red. The first three graphs are participants with the lowest proportion of social barriers in their network, and the last three graphs are those with the highest proportion of social barriers.

Table 2.

Characteristics of participant’s network members that are social supporters and barriers

	All Network members (N=1,600)	Network members who provided social support (n=478)	Network members who were social barriers (n=136)
	Mean (SD)	Mean (SD)	Mean (SD)
Age	39.1 (14.7)	39.9 (13.3)	37.7 (12.9)
	% (N)	% (N)	% (N)
Female	57.8 (924)	65.1 (311)	54.4 (74)
Spouse	2.3 (37)	5.6 (27)	8.8 (12)
Friend	45.0 (720)	37.7 (180)	42.6 (58)
Close family	23.9 (383)	36.2 (173)	39.7 (54)
Extended family	24.0 (384)	18.6 (89)	11.0 (15)
Lives in U.S.	82.4 (1,319)	83.1 (397)	96.3 (131)
Lives in same HH	9.8 (156)	16.3 (78)	23.5 (32)
Born in U.S.	13.9 (221)	12.8 (61)	14.8 (20)

Bivariate correlational analyses are summarized in Table 3. There was no difference in the HEI scores between participants who were food secure, and those who were food insecure (r = 0.00, p=0.97). Better dietary quality was significantly correlated with having a higher proportion of social supporters in one’s network (r = 0.24, p<0.05) and greater network transitivity (r = 0.19, p<0.1). Having a higher proportion of social barriers in one’s network was correlated with food insecurity (r = 0.25, p<0.05).

Table 3:

Bivariate Associations^a Between Social Networks, HEI, and Food Insecurity

	HEI	Food insecurity	Network density	Network transitivity	Prop. of supporters
HEI
Food insecurity	0.0
Network density	0.02	0.0
Network transitivity	0.19^	−0.1	0.71***
Prop. of supporters	0.24 *	0.17	0.03	−0.15
Prop. of barriers	0.14	0.25*	0.14	0.06	0.15

^***p < 0.001;

^**p < 0.01;

^*p < 0.05;

^{^}p < 0.1

HEI=Healthy eating index; SN=Social network

^aPearson correlation coefficients (r)

Results for the regression analyses are summarized in Table 4. In testing hypothesis 1a (Model 1), HEI scores were positively predicted by higher network transitivity (β = 6.11, p<0.001) and a higher proportion of social supporters in one’s network (β = 4.91, p<0.001). HEI scores were also positively predicted by acculturation (β = 3.64, p<0.05), and were negatively predicted by having less than an 11^th grade education (vs. some college, β = 7.72, p<0.05), age at moving to the U.S. (β = −3.08, p<0.05), and network density (β = −4.76, p<0.05). In testing hypothesis 1b (Model 2), food insecurity was positively predicted by having a higher proportion of social barriers in one’s network (OR = 2.6, p<0.05). Greater risk for food insecurity was also marginally predicted by being a high school graduate/having a GED (vs. some college, OR = 6.95, p<0.1), and decreased risk for food insecurity was predicted by working (vs. not working, OR = 0.33, p<0.1), and having health insurance (OR = 0.35, p<0.1). To test hypothesis 2, an interaction term for “food insecurity” and “proportion of network members who provided social support” was added to the linear regression model for HEI score (Model 1), but was not significant (β = 0.76, p = 0.4), and was dropped from the final model.

Table 4.

Individual and Social Predictors of HEI scores and Food Insecurity (N=80)

	Model 1 HEI scores Linear Regression	Model 2 Food Insecuritya Logistic Regression
	β (SE)	Odd’s Ratio (95% CI)
(Intercept)	47.08*** (4.43)	1.16 (0.17, 7.94)
Income Below 30k (Yes)	3.88 (2.74)	1.45 (0.45, 4.75)
High School Graduate or GED (ref: some college)	7.03 (3.99)	6.95^ (1.03, 46.70)
Less than 11th Grade (ref: Some college)	7.72* (3.67)	4.14 (0.68, 25.15)
Acculturation	3.64* (1.41)	1.29 (0.66, 2.51)
Age at Moving to U.S.	−3.08* (1.24)	0.93 (0.53, 1.63)
Working (Yes)	−1.32 (2.61)	0.33^ (0.10, 1.08)
Insurance (Yes)	4.34 (2.47)	0.35^ (0.12, 1.02)
Network Density	−4.76** (1.76)	0.81 (0.48, 1.35)
Network Transitivity	6.11*** (1.73)
Prop. of Supporters	4.91*** (1.27)
Prop. of Barriers		2.60* (1.06, 6.35)
R2	0.38
Adj. R2	0.29
AIC		111.81
BIC		135.63
Log Likelihood		−45.90

^***p < 0.001;

^**p < 0.01;

^*p < 0.05;

^{^}p<0.1

^aReference category: Food secure

In addition, of the 136 network members who were identified as social barriers by the study participants, 130 responses to the open-ended question of “Why?” were given. Four categories of responses were identified: (1) they buy, bring, or share unhealthy food with the participant (n=59), such as “Loves to cook sweets and share” and “Buys me coffee, bread, cookies”, (2) they overeat, eat unhealthy foods, or prefer unhealthy foods (n=42), such as “Does not like to eat vegetables” and “Eats a lot of fried foods and sweets”, (3) they encourage the participant to eat unhealthy foods/unhealthy amounts of food (n=15), such as “He does not like to eat healthy and tells me I should do the same” and “Promotes bad food”, and (4) other/unclear responses (n=13), such as “desserts”.

Discussion

This study examined relationships between dietary quality, food insecurity, and social network support and barriers among immigrant Latinos in the U.S. Over half the participants lived in a household with food insecurity, which is remarkedly higher than the national 2019 averages of 10.5% for all U.S. residents, and 15.6% for Latinos (10). As could be expected, there was some evidence that individuals with greater levels of education, that had health insurance, or a job, had decreased risk for food insecurity. Income was not associated with food insecurity. This may be due to low variability, as most participants had annual household incomes <$30,000. The average HEI score of participants in this study was 57.7, which indicates poor adherence to the USDA dietary guidelines, but is slightly higher than what has been found in other studies with Mexican immigrants (50). Interestingly, individuals with higher acculturation scores had better HEI scores, which contradicts the findings of other studies that document Latinos’ dietary quality decreases with greater acculturation (16). This may be explained by the distinct context for acculturation in New York City for individuals of Mexican descent. That is, despite the growing numbers Latinos of Mexican descent in NYC, ethnographic research suggests that they are often settle in neighborhoods with other Latino immigrants, which makes it more difficult to practice their distinct in cultural and social traditions (51,52). This may result in acculturation patterns that could be characterized as marginalized, and these marginalized acculturation patterns. These marginalized acculturation patterns have been associated with greater risk for diet-related diseases (53), thus making lesser acculturation protective of dietary patterns in this specific context.

Hypothesis 1a was partially supported, as having a network with more social support and network transitivity (the latter reflecting social cohesion) predicted better dietary quality, but having a network with less social barriers did not. In line with previous studies with non-immigrant Latinos (21,22), this study found that having more social support from one’s social network was associated with a 4.9 increase in HEI score. This is meaningful as some studies have suggested an HEI increase as small as 4–6 points is associated with lower risk for adverse health outcomes (54,55). Most of the participant’s social supporters lived in the United States, and compared to their network members overall, a higher proportion were female, spouses, close family members, and people who lived in the same households. This suggests that geographical proximity and kindship are important dimensions of garnering support, which has been previously discussed by Latinos in qualitative studies (28,29). Network transitivity also independently and positively predicted better dietary quality, which is in line with the social buffering hypothesis (19), and other studies showing that social cohesion improves various health outcomes (23,42). It is important to note that cohesive networks create strong social norms (42). The networks of the participants in this study primarily consisted of first generation immigrants living in the U.S., and less acculturation to the U.S. is linked with healthier diets (17). As such, these cohesive networks may have stronger norms for healthy diets than more acculturated groups. Future research should consider the interaction between network cohesion, network members’ acculturation, and dietary quality in Latino immigrants.

Hypothesis 1b was also partially supported, as having a network with less social barriers predicted less risk for food insecurity. However, contrary to the social buffering hypothesis (19), social support and network transitivity were not protective of food insecurity. This could be because immigrants are not always able to access support as easily as non-immigrants (27), and Latino immigrants who experience food insecurity are sometimes reluctant to seek tangible support (i.e., financial support) due to concerns of overburdening their network (14,29). Furthermore, as supporters tended to be geographically close kin, they may have been in constrained economic situations similar to that of the participants’, preventing them from providing the tangible resources needed to alleviate food insecurity (14,29). This suggests that it may be beneficial for community-based programs to focus on not only providing resources to individuals in need, but also to broader communities and social networks. Furthermore, programs and interventions could consider fostering supportive social connections between at-risk individuals and culturally and demographically similar individuals with less risk for food insecurity (e.g., encouraging the exchange of support among established social ties; connecting recently immigrated Latinos with other Latinos who have lived in the U.S. longer and have better access to adequate finances, food, etc.). By adopting this whole-of-network approach and bolstering larger communities’ resources, individuals may increase their collective capacity for social support, leading to better outcomes (26).

Network members who were social barriers were more likely to be spouses or close family members, and one-quarter lived in the same household. Individuals who were food insecure had more network members who were social barriers. This relationship may be bidirectional, where having more social barriers inhibits food security, and the experience of food insecurity leads to more food related conflict with one’s network members. Food insecurity can constrain individuals from having autonomy over their food choices, and make it more difficult to manage the intersection between social influence and food choice, leading to conflict (30). The open-ended survey results helped to elucidate this complex dynamic. Specifically, most of network barriers were people who bought, brought, or shared unhealthy food with the participant. Participants may have had networks that responded to their food insecurity by offering food (30), but since the participants didn’t have the financial freedom to refuse the food when it was unhealthy, it led them to identify these network members as social barriers. Another commonly cited reason that a network member was a barrier was because they overate or preferred unhealthy food. Notably, the food insecurity measure captures food insecurity at the household level, and participants living in food insecure households frequently named other household members as social barriers. A household experiencing food insecurity and financial constraints may have more conflict over food purchasing decisions, as they try to balance various food preferences and priorities with a limited budget (28,30), creating difficulties for individuals who manage the food purchasing and preparing. As such, future studies may want to consider the relationships examined in this study among individuals who are the main food shoppers and preparers in their household.

This study is one of the first to examine the links between social networks, dietary quality, and food insecurity in Latino immigrants, and demonstrates the complexity of the pathways by which these social networks may influence food choice and food access. Despite this, the study has a few limitations worth noting. First, the cross-sectional data does not allow us to determine causal links, but only to identify associations. Second, like most dietary assessments, 24-hour dietary recalls are subject to bias, as they require participants to accurately remember and describe their diets over the past day. Third, though a small sample size is common for social network research (36,56,57), the findings of this study are not generalizable to all Latino immigrants in the U.S. Furthermore, these findings may not be generalizable to other Latino subgroups (e.g., Puerto Ricans, Dominicans), as there are important cultural and social difference between these various ethnic subgroups (58,59). However, they are likely generalizable to an important population - other Mexican immigrants living in Bronx. Last, there is controversy regarding the HEI measure, as it is not culturally adapted, and though it reflects adherence to the USDA dietary guidelines, some other healthy diets do not score high on the HEI (60).

Having a supportive and cohesive network was beneficial for dietary quality, but not food security; while individuals who were food insecure identified a higher number of network members who were social barriers to healthy living. Latino immigrants have unique network dynamics, that may benefit dietary quality but that may be related to food security in different ways than what has been shown with other populations. Future interventions with Latino immigrants facing food insecurity should consider helping them to reduce barriers among their close social ties and maintain cohesive and supportive social networks to benefit diet quality. Future studies should also use longitudinal methods to explore the causal mechanisms of the relationships identified in this study.