Introduction

The trends and prevalence of type 2 Diabetes Mellitus (T2DM) in Latinos residing in the United States are a major public health concern. The age- and sex-adjusted prevalence of diabetes among Latinos is 19.1% compared to 12.1% among non-Latino whites [1–3], and they are at a higher risk of inferior T2DM outcomes, such as amputations, end-stage kidney disease, and death [4]. Among Latino adults, the highest prevalence is found in Mexican Americans (21.7%) [1, 3] and they have a greater risk of complications associated with T2DM, particularly retinopathy and kidney disease [4].

Effective T2DM management requires controlling blood sugar levels, which can be measured by glycated hemoglobin A1C [Hba1c] and tests for the percentage of red blood cells with sugar-coated hemoglobin in the past three months [5]. However, Latinos have a lower prevalence of regular glucose monitoring than their white counterparts [6–10]. Poor T2DM management is associated with individual risk factors such as lack of health insurance. This situation is more common among Mexican Americans than among other Latinos and is linked to poorer health outcomes [11]. Healthcare expansions for low-income populations have been associated with an increased probability of diabetes diagnosis and uptake of medication regimens, however, long-term diabetes-related outcomes have been difficult to achieve [12]. This may be because of non-adherence related to issues beyond cost, such as faith in diabetes medications [13]. To be sure, cost plays a role, as research shows that income predicts adherence among Mexican Americans with T2DM, but there is also evidence that larger neighborhood-level factors may play a role in nonadherence due to beliefs about medications [11].

Another crucial aspect of T2DM management is maintaining a healthy diet, which can be difficult due to various factors such as material conditions (e.g., cost of food, access, and availability) and complex social and cultural factors (e.g., traditional dietary habits and obesogenic environments). Adapting to a healthy eating pattern requires individuals to overcome a range of social, economic, and cultural obstacles that often serve as a barrier to behavior change. Studies have shown the efficacy of nutrition education programs and interventions in increasing awareness and understanding of the necessary dietary changes for successful T2DM management, when culturally tailored approaches are employed [14]. However, attaining sustained individual-level dietary behavior change that leads to improved T2DM long-term outcomes has proven challenging, even with programs tailored to Mexican Americans [15]. Dietary changes such as limiting the use of traditional food items are difficult for any group; however, for Mexican Americans, a typical T2DM diet restricts many important food staples, including tortillas and rice [16]. This makes adopting the diet a difficult lifestyle adjustment not only for individuals with T2DM but also for their complex social system, which includes family members and acquaintances both inside and outside the home [14].

Despite the complexity and multi-level nature of factors that could help engage individuals with T2DM in self-care behaviors [14], research has largely focused on individual-level factors based on theories of individual change [14, 17]. Specifically, among Mexican Americans with T2DM, one area of research has relied heavily on the theory of locus of control, which posits that “external” control orientations may lead to poorer T2DM outcomes, as patients may believe that the disease is beyond their control [18]. Intervention studies using this theory postulate that this psychological orientation is associated with patients taking less action, resulting in poor outcomes [19–22]. However, there is a lack of evidence that this is the case [23]. Research on Mexican Americans with T2DM has also focused on acculturation, which refers to the process by which immigrants adopt the norms, values, beliefs, and behaviors of the dominant group [24–26]. Despite a consistent relationship showing that lower levels of acculturation are associated with lower T2DM prevalence among non-Mexicans, including other Latinos [15, 27–29], those in transition or the “middle” phase of the acculturation process complicate this relationship in samples of Mexican immigrants [30, 31]. Several factors drive these potential differences, including levels of human and social capital, reasons for migration and subsequent opportunities for successful integration, and transnational social network ties to Mexico [32].

However, despite the evidence that underscores the importance of sociocultural factors in the study of T2DM, far fewer studies have taken a social-network perspective [17]. Interventions among Latinos with T2DM, approached from a social network perspective, suggest that leveraging the knowledge and experience of network members improves T2DM in the short term (<6 months) [33–38]. However, current studies have mainly focused on either dyads or peers, and there has been limited research on how social networks can be leveraged for long-term outcomes. To fully harness the potential of social networks for T2DM management, it is crucial to understand the structural features and attributes of personal networks among Mexican Americans. By gaining a deeper understanding of how social networks operate within this population, interventions can be tailored to optimize diabetes management and improve long-term outcomes. A large community of researchers across disciplines has developed useful theoretical frameworks to understand how ties can be viewed as conduits through which resources (e.g., information and support) can flow to and from. Public health research from a network perspective has largely focused on infectious diseases and tobacco use [39], with a focus on how knowledge and behavior can be powerfully spread across interpersonal ties [40, 41] as well as how networks comprising such ties can amplify the spread [41].

The effect of a person’s behavior on their social network has significant implications for public health and has been explored through various theoretical perspectives in multiple disciplines [39, 42]. One widely recognized theory in public health is the theory of diffusion, which postulates that individuals can influence each other through their interactions and develop similar beliefs or become aware of similar information [42, 43]. Social capital studies represent another rich tradition, in which the flow of resources or network benefits depends on varying degrees of access and permission. Within each of these traditions, most studies take a whole network approach and focus on the pure effects of the network structure (e.g., equivalence, centrality) on an outcome [42, 43]. Another set of studies explained networks from an egocentric perspective. Sometimes also referred to as personal-network studies, in such research, the focal point of the researcher is to understand the “ego” and her set of ties to others in their network called the “alters” [43]. These studies have also utilized diffusion and social capital frameworks, but it is important to note that they typically explore issues beyond structure, such as attributes of the ego, as well as attributes of the alters [42].

Focusing on the attributes of alters is an important step in characterizing the personal networks of Mexican Americans to elucidate several issues within the theoretical frameworks of social capital and diffusion. First, there is an unexplored aspect of social homogeneity as it relates to the process of acculturation for Mexican Americans. That is, the acculturation process is associated with T2DM and important self-care behaviors, but scholars have conceptualized this as an individual-level process. Almost no work has examined the acculturation attributes of a personal network in which an immigrant person is embedded. Second, there is a lack of nuanced understanding of the diffusion process by locus of agency. That is, it is unclear whether Mexican American egos adopt from their alters in a largely active or passive way and whether the alters employ active processes (e.g., through social control) or mimetic processes (e.g., through habitual practices). Third, contextual details of the lived experiences of Mexican American immigrants could shed light on how migration can disrupt social networks and relationships, which can be highly destructive to social capital, especially among those in transition or in the “middle” of the acculturation process. This mixed-method study aims to shed light on these complex issues within the context of T2DM and self-care behaviors among Mexican Americans in New York City, using qualitative and egocentric network data. Despite the importance of social networks in shaping health behaviors, few studies have explored their role in promoting long-term behavior change among this population. To address this gap, this mixed-method study asked the following research questions (1) What social network features are associated with HbA1c levels among a community sample of Mexican immigrants in New York City? (2) How do these quantitative features converge with qualitative narratives of their lived experiences? Given that over 61% of all Latinos are of Mexican origin, with diverse cultural backgrounds and deep roots in Latin America [44], understanding the social determinants of health within this population has significant public health implications.

Methods

We used a mixed-methods approach, employing a convergent design [45], with the quantitative phase preceding the qualitative phase. The study involved the same cohort of participants in both the stages, ensuring consistency and coherence across data collection and analysis, as described below.

Quantitative sample: 1,620 personal network ties were derived from 81 self-identified Mexican American adults recruited from a large Catholic Church in the Bronx, New York. Specifically, the church liaison identified potential prayer and activity groups with large numbers of Mexican American congregants, and we also used telephone rosters to identify potential participants. Three trained bilingual research assistants contacted participants in person or via telephone and provided them with an explanation of the details of the study. After screening for eligibility, interested individuals consented to participate using a protocol approved by the City University of New York Human Subjects Institutional Review Board (IRB# #2018–1081). The same research assistants scheduled all data collection at the participants’ preferred time and location for convenience. None of the research assistants had previous relationships with any of the participants. However, before the interview process began and to put the participants at ease, they occasionally shared details about themselves such as their educational background (one held an MPH and was pursuing additional training in a doctoral program in public health; the other two held BS degrees and one was obtaining additional training in a Master’s program in public health nutrition), as well as their nationality (e.g., Mexico; Spain; Poland). Data were collected in person between January 2019 and June 2019 in either English or Spanish, depending on the participant’s preference.

Quantitative data: Three different approaches were used to collect the various forms of quantitative data used for this analysis. First, EgoWeb was used to collect interviewer-assisted personal network interviews that consisted of three sections: (i) generating names of people in the respondents’ personal social networks (alters), (ii) collecting information about each alter (network composition), and (iii) eliciting information about the relationship between each unique pair of network alters (network structure) [46]. We asked respondents to name (by first name only) 20 individuals whom they knew, who knew them, and with whom they had contact sometime in the past 6 months, using standard free listing techniques employed in personal network data collection among vulnerable populations [47]. Eliciting 20 alters has been found to be enough data to capture variation of the network structure while still maintaining respondent burden at manageable levels [48]. Contact included face-to-face, phone/text, mail, e-mail, social media (e.g., Facebook), and research assistants probed participants to think about additional social groups (e.g., church members) to reach the 20 alters, which all participants were able to do. Secondly, Qualtrics [49] was used to collect interviewer-assisted survey data on the demographics and other individual-level characteristics of the participants, using standard measures when possible. Finally, whole dried blood spots from a finger prick were used to collect HbA1c levels, following the 2018 ADA Standards of Medical Care in Diabetes [50]. This minimally invasive approach for blood sampling can be performed in non-clinical settings by non-medically trained personnel [51]. Research assistants were trained to use universal health precautions and a common protocol to collect the blood spots from participants (six steps) and send the specimens to the lab (three steps) [52], ensuring data quality and participant safety. The software R (version 4.2.2) was used for statistical analysis [53].

Power considerations. Our primary aim was to determine whether it is feasible to conduct such a study with this population and to obtain a better understanding of the variability in the outcomes of interest and social network characteristics to design a larger-scale study. Our sample size of 70 was sufficient to detect effect sizes of 0.25 standard deviations, or larger, with a significance level of α = 0.05, and power of 0.8 (β = 0.8).

Results

Quantitative results

Table 1 shows the descriptive statistics for the sample according to HbA1c categories. Those with normal HbA1c levels were more likely to have high food security and health insurance and were more educated than those with abnormal HbA1c levels.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Summary statistics for individual-level and network-level variables.

https://doi.org/10.1371/journal.pone.0295499.t001

Table 2 present the estimated coefficients and significance levels for each indicator in the multiple regression model. At the significance level of 0.05, the individual-level indicators associated with HbA1c levels were body mass index (BMI; β = -0.07) and healthy eating index (HEI) total score (β = -0.03).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Coefficients for individual and network-level variables with HbA1c.

https://doi.org/10.1371/journal.pone.0295499.t002

Table 2 and Fig 1 show the estimated coefficients and significance levels for each network-level indicator in the multiple regression model. For a 25% increase in each network-level indicator (i.e., 5 more people with this attribute), we found at the 0.05 significance level that the percentage of alters who were diagnosed with diabetes was associated with an increase of 2.0.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Close-up coefficient plot of network-level covariates.

https://doi.org/10.1371/journal.pone.0295499.g001

Integration with qualitative results

Given the convergent mixed-method design of our study, we analyzed the qualitative data in light of the relationships established in the quantitative analysis. As such, we were particularly interested in narratives discussing alters who were diabetic in the wider network and not just those in the “consumption” network (e.g., alters with whom the person eats with). We describe the level of social network connectivity and key demographic characteristics of the qualitative subsample by normal, prediabetic, and diabetic status in Table 3. Most of the subsamples had high levels of connectivity, and those with diabetes were older and had a higher proportion of health insurance and males relative to those categorized as normal or prediabetic. We included participants with normal and prediabetes hba1c levels because they still had alters in their personal social networks that were diabetic.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Descriptive characteristics for qualitative sample.

https://doi.org/10.1371/journal.pone.0295499.t003

To further differentiate the qualitative narratives of participants based on the quantitative findings of percentage diabetic, we illustrate the proportion of diabetic alters along with their own Hba1c score (if available) for the qualitative subsample of 25 participants (see Fig 2: HbA1c levels on the y-axis and percentage of alters in the network that are diabetic on the x-axis). This shows N = 18 participants who were convergent with the quantitative findings and (i.e., those with normal HbA1c levels and a low percentage of people in their network with diabetes, as well as those with high HbA1c levels and a high percentage of people in their network with diabetes) from those who were divergent (N = 2) (i.e., those with normal HbA1c levels and a high percentage of people in their network with diabetes, as well as those with high HbA1c levels and a low percentage of people in their network with diabetes). The analyses of the narratives from this convergent group highlighted the following themes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. HbA1c levels on the y-axis and percentage of alters in the network that are diabetic on the x-axis.

https://doi.org/10.1371/journal.pone.0295499.g002

Convergent group: Elevated HbA1c levels and a higher percentage of alters with diabetes

I am not suffering, so I am doing OK

Despite their elevated hba1c levels, these participants emphasized their ability to manage the condition well due to a lack of acute physical complications. Specifically, some concentrated on their ability to still work, get up early in the morning, and drive at night, as exemplified by the quotes below. Being able to engage in these physical actions, as well as their positive outlook while doing so, typified the reasons why their uncontrolled diabetes was not seen as a health issue that warranted medical attention.

Download:

• PNG
  larger image
• TIFF
  original image

https://doi.org/10.1371/journal.pone.0295499.t004

Influence of personal network on diet

Participants narratives highlighted how network members shaped their diet, and not just those that they ate with on a regular basis. Most of the shaping came in the form of advice of what foods to avoid, which required changes such as reducing sugar and red meat intake, as well as portions and eating out. Sometimes the advice came from alters who struggled with their own elevated hba1c levels, though most participants referred to this as “high sugar.” For the most part, the diabetic alters highlighted in the exemplary quotes were not described as successful in their own efforts yet provided participants with a foreboding advice in the need to change their dietary habits. Of the few positive examples of successful management, was a participant 127 who described a diabetic alter that kept her weight under control by only eating homemade food (e.g., making her own salsa and tortillas) which in turn would save her money. But for the most part, participants described diabetic alters seeking help only after a major health scare, which reinforces this earlier theme of “I am not suffering so I am doing OK” as potentially normative behavior.

Download:

• PNG
  larger image
• TIFF
  original image

https://doi.org/10.1371/journal.pone.0295499.t005

Role of medical doctors and other factors in diabetes management

Although participants acknowledged the importance of social networks members for support with anything from housing issues to making doctors’ appointments, they also emphasized the authoritative role of medical doctors in regards to their health and diabetes management specifically. Some of the illustrative quotes specifically used the words like “authority” and suggested that only medical doctors had the power to make decisions about their health even if they had a supportive network. In some cases, participants questioned the validity of the health information provided by network members, and in others they struggled with other diabetic alters on best approaches to diabetes management. For example, participant (ID 17) emphasized the importance of stress for her and how it affected her mood affected adherence to diabetes medication, but found it hard to convince others in her network of this connection. On the other hand, participant (ID 134) had a supportive alter, her husband, who actually accompanied her to her doctor’s appointment and helped alleviate her guilt about taking time off work to do this. Irrespective of level of support from the network, however, none of the participants actually spoke about using Hba1c levels as a metric to assess their health and well-being in their lives, or the extent to which they spoke to their doctor about this.

Download:

• PNG
  larger image
• TIFF
  original image

https://doi.org/10.1371/journal.pone.0295499.t006

Divergent group: Normal HbA1c levels with higher percentage of alters and high HbA1c levels and a low proportion of people in their network with diabetes

The participants who lacked a single member with diabetes in their network but had an elevated Hba1c level of 6.3% described close relationships with friends with whom they would often have meals or snacks. While their friends encouraged healthy eating and provided general health support, the participant’s family had a greater influence on what they ate at home. When asked about support for healthy eating, they described their mother “as always cooking for the family” and their father as the person who ultimately paid for all the food purchases in the home. The father’s recent health diagnosis had caused significant changes in how food was purchased and cooked in the household. The participant did not seem to be negotiating much about their diet at home, but they said, "Because my mom is always kind of looking out for what we eat at home, then like, when I am outside I kind of treat myself.” In other words, their autonomy over food was largely preserved outside the home. However, they did not evaluate the foods they consumed outside of the home as relevant to their own health, including their elevated Hba1c levels. Despite this, this participant had a largely positive evaluation of their overall health.

The other divergent participant with an HbA1c level of 4.7% had a support network mostly comprised of friends who were members of a Herbalife health club, a supplement company that has gained popularity among Latinos in the U.S and Latin America. The participants evaluated and interpreted their behavior through the prism of this cluster in their network and described how it helped them adopt healthier eating practices and increase their physical activity. Although the participant acknowledged that they could do more to improve their health, any health concerns were largely focused on weight rather than their Hba1c levels.

Discussion

Understanding the impact of social relationships on the etiology and management of diabetes is particularly important for immigrant Mexicans, who have the highest prevalence of the disease among all Latino subgroups [1, 3, 78]. Although past research has identified various individual-level factors that correlate with diabetes prevalence among Mexicans and Latinos, such as the locus of control orientation and acculturation [18, 27], few studies have focused on network-level factors [17]. Therefore, investigating how social relationships shape diabetes-related behaviors and outcomes among this population has significant public health implications.

Our study found an association between increased HbA1c levels and diabetic network members. Qualitative data highlighted that participants framed their glycemic control based on whether they were in an acute state of “sufrimiento” or pain. Those who experienced physical discomfort but, due to not being in such an acute state, could still fulfill their work or caretaking duties in their everyday lives, felt that glycemic control did not warrant further attention, including seeking medical care. To the extent that these values are also widely held by the personal network that people are embedded in (i.e., those alters who are diabetic), warrants further investigation if personal networks are to be harnessed as part of the intervention strategy. Other studies among Mexican Americans with T2DM have also found that energy levels and mood are salient symptoms in used to evaluate T2DM management [79]. Our findings emphasize the need for diabetes education that focuses on glucose monitoring is crucial, as it has been found to improve short-term T2DM outcomes among Mexican Americans [6], and may incorporates personal networks (e.g., diabetic alters in the network).

Other network-level factors were associated with HbA1c though not at the same level of statistical significance. Given the pilot nature of this study, we think it is still worth reflecting on alters who encouraged ego health were the most significant network-level variable associated with decreased HbA1c levels. Specifically, an increase in the proportion of alters who encouraged the ego’s health was associated with a decrease in Hba1c levels. The qualitative findings further shed light on the fact that network members who encouraged the ego’s health mostly did so by sharing knowledge and advice regarding the diet. Participants mentioned that reducing sugar and red meat intake, reducing food portions, and eating out less were the most salient diet-related information that flowed through the networks. Our findings are consistent with previous qualitative research among Mexican Americans, which suggests that a diet high in sweets is understood as a “provoking behavior” that causes diabetes (18). Sugar is highly predictive of a distinct dietary pattern among Mexican Americans across the acculturation continuum (75) and “heritage-specific” dietary patterns (76). However, more research is needed to fully understand the diffusion dynamics of dietary information among Mexican Americans. This includes investigating whether efforts by social network members influence social control (e.g., monitoring) or a combination of control and social support, as others have found (77).

Furthermore, future research should investigate the extent to which the diffusion of dietary information across networks among Mexican Americans is associated with specific T2DM outcomes and symptoms. It is important to understand how those in the “middle” phase of the acculturation process think about and practice healthy eating, and how these may adhere to segmented patterns of assimilation rather than purely linear approaches (24). Additional research is necessary to understand these processes at the network level since the qualitative and quantitative findings suggests this may matter for glycemic control.

Finally, our study found individual-level variables associated with decreased levels of Hba1c, namely body mass index and the health eating index total score. These findings are consistent with previous research among both Latinos and non-Latinos that show diets that score highly on the HEI are associated with a significant reduction in the risk of T2DM [80, 81]. Indeed, the qualitative narratives focused on participant’s concepts of a healthy diet, though we did not directly address how these concepts deviated from the way in which HEI is derived during the interviews. Our quantitative results regarding BMI highlight the complex relationship between this anthropometric measure and HbA1c [82], though the qualitative narratives did make clear that reaching an ideal weight was difficult for most participants irrespective of their HbA1c score. Interestingly, a major qualitative theme was the extent to which medical doctors were regarded as the ultimate authority on T2DM information and advice, and not other alters in their network (e.g., family, friends). Future social network interventions should explore the possibility of including physicians as part of the web of social resources for Mexican Americans with T2DM, or at the very least incorporate effective ways in which participants could harness the clinical encounter for the purposes of increasing their diabetes awareness and control [78].

While our study provides valuable insights into the social and individual determinants of Hba1c levels among Mexican Americans with T2DM, several limitations should be considered. First, our study is its observational nature precludes us from making causal claims at the individual or network level. However, our findings are consistent with previous research, lending credibility to their potential causal associations. Second, our sampling strategy was non-probabilistic, in the sense that the sample was derived from a specific community in New York City and may not represent important segments of the entire Mexican American population living in the United States. Future research should replicate our study in other Mexican American communities to assess the extent to which our results generalize to other populations. In a related manner, our data were collected before the 2019 COVID-19 pandemic, which may further limit the generalizability of our results, depending in part on how the pandemic progresses in the future. Therefore, caution should be exercised in generalizing our findings to the current context. Third, our reliance on a single measurement of Hba1c is atypical, and in clinical practice, would require confirmatory testing. These can include either the same or a different test on a different day or a different test using the same blood sample [83]. Fourth, a criticism of personal network research is that it is limited by virtue of obtaining only the view from the ego [43], so discrepancies in perceptions between the ego and the alter are missed. However, perceptions are still critical components in understanding social connections (e.g., perceptions of social support) [84] and these, along with structural characteristics, are key elements in our conceptual framework. Fifth, the mixed method nature of our study limited our capacity to employ more in-depth qualitative methods, such as ethnographic work that necessitates years in the field. Further, integrating social network visualizations into the in-depth qualitative interviews entailed asking highly specific questions about each alter, a departure from the typical in-depth interview format within the interpretative phenomenology framework. Yet it is important to note that methodological applications of this and other qualitative approaches can sometimes suffer from disciplinary silos [85], and perhaps those outside public health might interpret these deviations in our qualitative approach as a flaw. Future research should address the tensions between the qualitative and quantitative arms of the study as they emerge, while also considering the extra resources needed to integrate different methods effectively. It would also entail cross-collaborations between scholars in different fields that others have called for [85] in order to foster an appreciation for innovative, mixed-method work.

Despite these limitations, our mixed-method study provides critical evidence on the role of social networks in T2DM outcomes in an important Latino subgroup, which is often aggregated in national surveillance efforts. Mexican Americans represent the largest and fastest growing subgroup in the United States and are disproportionately burdened by T2DM. Researchers argue that this disease may erode the health advantages seen in this group among immigrant generations in other health areas [86]. Our study addresses a significant gap in the research on Mexican Americans by focusing on network processes that may be crucial to understanding the well-documented disparities in T2DM-related outcomes experienced by this population.

Supporting information