The Science of Diabetes Self-Management and Care2023, Vol. 49(2) 91 –100© The Author(s) 2023Article reuse guidelines:sagepub.com/journals-permissionsDOI: 10.1177/26350106231158827journals.sagepub.com/home/tde
AbstractPurpose: The purpose of this study is to examine the extent to which perceived support and depressive symptoms might interfere with Hispanic patients’ ability to manage their diabetes and whether these effects vary by gender.
Methods: Data were collected from a cohort of 232 Hispanic men and women with type 2 diabetes mellitus (T2DM). Conditional process analysis was used to test a moderated mediation model of the time-lagged processes associating gender, diabetes support, and depressive symptoms with reported self-efficacy after 3 months.
Results: Increased depressive symptoms were associated with lower self-efficacy, but the conditional effects varied among men and women. The index of moderated mediation was significant, indicating that among women, the indirect effect of depressive symptoms on self-efficacy was contingent on lower levels of perceived support. Among men, increased depressive symptoms were directly associated with declines in self-efficacy and were not conditional on perceived support.
Conclusions: Results of the study have important implications for gender health equity. Mental health screening and an assessment of support needs may be important for determining appropriate complementary therapies when treating Hispanic women with chronic conditions such as diabetes. Attention to possible differences in gender-specific mental health needs could lead to improved self-management, better glycemic control, and more equitable health outcomes.
It is well established that diabetes is one of the leading causes of death in the United States.1 Among those with physician-diagnosed diabetes, persons of Hispanic origin have high glycemic levels on average compared to other racial-ethnic groups.2 Depression commonly co-occurs with diabetes, and the 2 together further increase the risk for severe health outcomes in Hispanic populations.3,4 This is especially the case among those who experience difficulty meeting basic needs,5-7 although the effects can vary by gender and acculturation.7-9
Unsatisfactory glycemic control is associated with poor diabetes self-care.10 Several studies have reported gender differences in barriers to self-care, with women reporting more barriers than men, including lack of support from family members for following a healthful diet or increasing physical activity.11-13 Hispanic women also tend to report less confidence in their ability to manage their diabetes.13,14 Based on Bandura’s15 social cognitive theory, self-efficacy can be defined as a person’s confidence in his or her ability to perform an action in a given situation. Numerous studies have established the importance of self-efficacy to health behavior change and, hence, to improved self-management of chronic conditions such as diabetes.13,16-21
Better social support can enhance self-efficacy.13,14 Inadequate social support has been found to be associated with depressive symptoms,7,22,23 which could interfere with self-efficacy. A few cross-sectional and qualitative studies have reported evidence that depressive symptoms may vary among Hispanic men and women with diabetes, with women reporting more depressive symptoms than men.7,24,25 One study reported longitudinal evidence that changes in support and depressive symptoms predicted levels of self-efficacy and diabetes self-management in Hispanic patients,6 but that study did not specifically compare men and women. Therefore, the purpose of the current study was to examine the extent to which perceived support and depressive symptoms might interfere with Hispanic patients’ ability to manage their diabetes and whether these effects vary by gender. Based on previous research,7,13,14 the authors hypothesized that compared to Hispanic men, (1) lower perceived support among Hispanic women would be associated with more depressive symptoms and (2) the combined effects of less support and depressive symptoms would decrease women’s confidence in their ability to manage their diabetes over time.
The authors completed a secondary analysis of survey data collected from a cohort of low-income Hispanic men and women with type 2 diabetes mellitus (T2DM) who had participated in a randomized control trial (RCT) of a culturally targeted, interactive education intervention for diabetes self-management from 2011 through 2013. The educational content of the intervention is described elsewhere.26,27 The original RCT study was approved by the Baylor College of Medicine Institutional Review Board.
The original RCT was designed to evaluate whether the intervention improved diabetes self-management knowledge, self-care, and glycemic control at follow-up. Although those participants who were randomly assigned to the intervention group were able to lower their blood glucose by the end of the study, their results were not significantly different from the control group. One reason may be that about 38 (16%) of the participants who had completed the study were missing a second A1C measure by the study’s conclusion, which may have decreased the power to detect a difference. For these reasons, the main RCT results from the original study were not published.
One of the criteria for including participants in the original RCT was that their A1C was high (≥8% or ≥64 mmol/mol). This means that they were already experiencing difficulties managing their diabetes before participating in the RCT. Three secondary analyses were carried out from data collected at the start of the study and the results subsequently published. The first article explored the sociocultural context of diabetes self-management in terms of preexisting barriers to behavior change,13 and the second looked at preexisting differences in selfmanagement by gender and acculturation.14 The third article discussed various psychosocial factors associated with depressive symptoms at baseline.7
The results reported in the 3 articles raised new questions regarding the processes associated with diabetes selfmanagement over time. One question that arose was whether changes in self-efficacy by the conclusion of the study might be contingent on perceived support and depressive symptoms. A second question was whether these conditional effects, if any, differed by gender. None of the 3 articles analyzed data collected after baseline. The current study attempts to explore these questions using data from baseline and follow-up.
Study participants received their primary health care in 4 community health centers that were part of a large, metropolitan public health system that functioned as the safety-net health care for vulnerable populations who lacked health insurance. Inclusion criteria for the original RCT were age 18 or older when diagnosed with T2DM, high A1C (≥8% or ≥64 mmol/mol), and Hispanic ethnicity. Exclusion criteria included age ≥ 70 years, non-Hispanic ethnicity, physical or mental impairment, and diabetes complications. About 300 potentially eligible patients were identified through the electronic medical record system and contacted to inform them of the study, screen them for eligibility, and invite them to participate. After determining eligibility and obtaining written, informed consent, 127 participants were randomly allocated to standard care and 123 to receive an intervention consisting of an interactive educational program containing culturally appropriate video segments, educational modules, and games.
Data were collected at baseline, which occurred shortly after randomization, and at the postintervention followup visit, which occurred 3 months following randomization. The length of the RCT was relatively short because investigators expected substantial loss to follow-up if the timeline was extended. Many patients in the target population tended to move into and out of the safety-net system due to fluctuating socioeconomic conditions and access to medical insurance connected to employment. Out of the original 250 participants, 235 completed the RCT, a loss to follow-up of about 6%. The subsample included in the current study consists of 232 participants with no missing data for any of the key variables included in this secondary analysis.
The dependent variable, self-efficacy (SE), was measured using the Self-Efficacy for Diabetes Scale.28 This is a 10-point Likert scale that asks a number of questions about participants’ confidence they can manage their diabetes in a variety of different situations. Response options range from 1 (no confidence) to 10 (totally confident). SE was calculated by averaging the responses to the 8 scale items so that the higher the average, the higher the selfefficacy (Cronbach’s α = .845).
Perceived support for diabetes self-management was measured using the Support Received Scale from the Diabetes Care Profile.29 This is a 3-point Likert scale that asks how much support is received from family and friends for following a meal plan, taking medicine, taking care of feet, getting enough physical activity, testing sugar, and handling feelings about diabetes. Response options are coded 1 (none), 2 (a little), and 3 (a lot). The authors averaged the 6 scale items so that the higher the average, the higher the perceived support (Cronbach’s α = .808).
Depressive symptoms were measured using a short form of the Center for Epidemiologic Studies Depression Scale (CES-D).30 The developers of the short form psychometrically validated the scale in a sample of older adults.31 This is a 4-point Likert scale containing a subset of 10 items out of the original 20 that ask respondents how often they were bothered by depressive symptoms during the past week. Response options for the original scale were coded 0 (rarely), 1 (some of the time), 2 (occasionally), and 3 (all the time). For the current study, the authors reverse-coded 2 of the items to be consistent with the remaining scale items, recoded all items to range from 1 to 4, then averaged the 10 scale items so that the higher the average, the more depressed the mood (Cronbach’s α = .857).
English language proficiency was used to represent greater acculturation. Language is a commonly used measure of acculturation and correlates well with other multidimensional scales.9,32 All surveys were available in English and Spanish, and bilingual research assistants collected the survey data from the participants during the original study. To facilitate communication in participants’ language of choice, all participants were formally assessed at baseline for English language proficiency using a modified version of the Hazuda scale.33 The distribution of average scores for participants in the parent study was bimodal; therefore, the continuous scale was collapsed into a dummy variable, English-speaking, coded 1 for English and 0 for Spanish.
The independent variable, female, was created by converting gender to a dummy variable, coded 1 for female and 0 for male. Treatment group was also recoded as a dummy variable, intervention, coded 1 for intervention group and 0 for control group. Age, English-speaking, and intervention were included in the model as covariates.
Based on theory and previous research described elsewhere,7,13,14 the authors developed a conceptual model a priori of the conditional (moderated) direct and indirect pathways leading to self-efficacy for managing diabetes. With this conceptual model as a guide, regression-based moderated mediation was used to examine the conditional processes associating gender, perceived support, and depressive symptoms reported at baseline with reported selfefficacy at follow-up.
As Figure 1 illustrates, the authors theorized that perceived support interacts with gender, thus moderating the direct pathway leading from gender to self-efficacy and the indirect pathway leading from gender to self-efficacy through depressive symptoms. Although not shown in the figure, both regression pathways included the covariates age, English-speaking, and intervention to control for their potential effects.
IBM SPSS Statistics34 was used for all analyses. The focus of the analysis was on the variables included in the model and how they might combine to quantify differences in the effects on self-efficacy by gender. These effects include the conditional direct and indirect effects by gender, moderated by perceived support. This approach differs from structural equation modeling in that the emphasis is not on determining whether there are any latent effects but on how relationships among the observed variables affect the outcome.35
To take time order into consideration, the authors used regression analysis to examine the lagged, conditional effects of perceived support and depressive symptoms reported at baseline on reported self-efficacy after 3 months. Preliminary analyses using paired samples t tests (Table 1) suggested that it was appropriate to use a lagged model because the means and standard deviations for perceived support and depression did not change significantly over the time period while mean self-efficacy significantly increased (P < .001). Perceived support was mean-centered for the regression analyses to facilitate interpretation of its interaction with gender.
Recent methodological developments described in the literature35-39 have led to consensus that one of the most robust methods for quantifying indirect pathway effects, especially with smaller samples, is to create an index of the indirect effects using bootstrapped confidence intervals to analyze the effects.35,39 Conditional process analysis, using Hayes’s35,40 Process macro for SPSS was used to run the regression analyses for the direct and indirect pathways shown in Figure 1 and to calculate the Index of Moderated Mediation. This index estimates the conditional, indirect effects of the moderated mediation. Confidence intervals for the effects were calculated based on 5,000 bootstrap samples.
Table 2 shows the baseline characteristics of this subsample. The subsample consisted of 92 men and 140 women (40% and 60%, respectively). The percentages of men and women in each of the 2 treatment groups from the original study were similar. Participants ranged in age from 27 to 68, with a mean age of 50.4 for men and 51.7 for women. Half of the men were English-speaking and the other half Spanish-speaking. Among women, 32.9% were English-speaking and 67.1% Spanish-speaking. The educational attainment was low, with about 61% of the men and 72% of the women having completed less than a high school education.
The results of the regression analyses for both pathways are shown in Table 3. Model 1 displays the regression results for the pathway leading to depressive symptoms. The unstandardized regression coefficients indicate that women, English speakers, and women who reported lower perceived support were significantly more likely to report depressive symptoms at baseline. The R2 of .2270 indicated that the variables included in this model accounted for 22.7% of the variance in depressive symptoms at baseline.
Model 2 displays the regression results for the conditional direct pathway leading to self-efficacy at 3 months. After controlling for all the variables included in Model 1, only the mediator, depressive symptoms, was significantly correlated with SE. Because the direct effects pathway controlled for female gender, the negative correlation with self-efficacy meant that as depressive symptoms went up, self-efficacy went down, but for men, the association was not contingent on perceived support. This means that there is not enough evidence to support moderation of the direct effects of gender after adjusting for depressive symptoms. The relatively low but significant R2 of .0787 for Model 2 indicates that the baseline variables included in the direct pathway accounted for 7.87% of the variance in SE at 3 months.
A visual depiction of the conditional effect of support on depressive symptoms by gender is shown in Figure 2. The graph plots values at the centered 16th, 50th, and 84th percentiles of perceived support. The negative slope shown in the figure clearly illustrates the significant decrease in depressive symptoms among women as perceived support increases. In comparison, there is a slightly positive but negligible slope for men.
Although the conditional direct effects of gender on self-efficacy were not significant, this could be because Hispanic women started with lower levels of perceived support on average compared to men. Figure 3 illustrates the conditional direct effects. As perceived support increased, SE after 3 months also increased for both men and women.
The Process macro probed the conditional direct effects of gender on depressive symptoms and self-efficacy at the centered 16th, 50th, and 84th percentiles of perceived support to determine where the effects may be strongest. Table 4 shows the conditional direct and indirect effects for each of the moderated mediation pathways. The conditional direct effects of perceived support on depressive symptoms by gender are significant at the median and lower percentiles of perceived support, and the conditional direct effects on self-efficacy are not significant at any of the percentiles of perceived support.
The Process macro also probed the conditional indirect effects on self-efficacy through depressive symptoms at the 16th, 50th, and 84th percentiles of perceived support. The 95% confidence intervals for these effects were calculated based on 5,000 bootstrap samples. Because the confidence interval does not include 0 at the 16th and 50th percentiles, the results indicate that at lower levels of perceived support, increases in depressive symptoms significantly decrease self-efficacy in women (P < .05).
The evidence supports the moderated, indirect model. The index of moderated mediation of 0.1438 shown at the bottom of Table 4 quantifies the change in the indirect effect of gender on self-efficacy through depressive symptoms as perceived support changes by 1 unit. The confidence interval, based on 5000 bootstrap samples, does not contain 0. This provides evidence that the indirect effect of gender on self-efficacy through depression is contingent on perceived support (P < .05). In other words, among women, the indirect effect through depressive symptoms was 0.1438 units larger as perceived support changed by 1 unit.
Results supported the hypotheses; lower perceived support among Hispanic women was associated with more depressive symptoms, and the combined effects of less support and more depressive symptoms decreased women’s confidence in their ability to manage their diabetes. The results of the conditional process analysis suggested that the higher the depressive symptoms, the lower the self-efficacy and that the effects varied among men and women. The index of moderated mediation was significant, meaning that among women, the indirect effect of depressive symptoms on self-efficacy was contingent on lower levels of perceived support. Among men, increases in depressive symptoms were directly associated with declines in self-efficacy and were not conditional on perceived support.
These results are consistent with the literature documenting gender differences in perceived support and depressive symptoms in Hispanic populations.7,16,17,41-43 The results of this study complement and extend this literature, suggesting that the complex pathways leading to improved self-management in Hispanic populations could vary considerably, depending on gender. Thus, it is important to consider gender differences when promoting diabetes self-management in low-income, Hispanic populations.
Longitudinal research including objective measures is needed to determine how well the model actually predicts diabetes self-management in Hispanic men and women. A growth curve type analysis that can examine changes in perceived support, depressive symptoms, self-efficacy, and A1C at multiple points over time would be a good approach. A multipronged RCT that studies different combinations of interventions targeting support, depressive symptoms, and changes in SE and A1C would be especially informative. Regardless of study design, gender comparisons should be included.
This study has several limitations. First, it is a secondary analysis of data that were collected for a different purpose. Although the original RCT study included survey data on diabetes-related support and depressive symptoms, the intervention tested in that study was not designed to experimentally manipulate either. Therefore, the sample for the current study was treated as a cohort, and the effects cannot conclusively be determined to be causal. Second, the sample was limited to low-income Hispanic adults with high A1C who were already experiencing difficulties managing their diabetes. This limits the generalizability of the results, although the results do shed some light on some of the processes that could lead to poor health outcomes in the most vulnerable populations. Third, the burden of diabetes self-care that includes managing diet, exercise, and medication recommendations and diabetesrelated complications are likely to impact depression and self-efficacy among patients with diabetes. The authors could not use change in A1C as an objective measure of diabetes self-management because 16% of the sample was missing the second measure of A1C, reducing the sample size considerably, thereby limiting the power to determine a significant effect.
The study has strengths as well. First, the authors utilized time lags between predictors collected at baseline and self-efficacy collected 3 months later. The results of paired t tests indicated that measures of the predictors, perceived support and depressive symptoms, were persistent among the participants because they did not change significantly over the 3-month time period. Second, the study was designed to examine the conditional and indirect effects of gender differences. To the authors’ knowledge, this is the first study to incorporate gender into the conditional pathways leading to differences in self-efficacy for managing diabetes among Hispanics.
The results of this study have important implications for gender health equity. Mental health screening and an assessment of support needs may be an important complement to treatment for Hispanic women with chronic conditions such as diabetes. A multifaceted approach that includes mental health therapy and diabetes selfmanagement could lead to improved health outcomes in more vulnerable populations.5,6 Because gender differences in depressive symptoms may be contingent on lower levels of perceived support among women, it is crucial that women are screened for depression prior to implementation of behavior change interventions. Otherwise, long-term management may be difficult to achieve. If the support needs of Hispanic women are addressed, comorbid depressive symptoms among them could be reduced, leading to higher self-efficacy, better self-management, improved glycemic control, and more equitable health outcomes.
The parent study was generously underwritten by Ms Trinidad Mendenhall. Special thanks to Kikuko Omori, PhD, for her invaluable recommendations concerning conditional process analysis.
The authors declare that there are no conflicts of interest.
The first author was awarded a 2022 HHS Summer Faculty Fellowship by the College of Health and Human Services at California State University, Sacramento, to support the research and writing of this article.
Carol L. Mansyur https://orcid.org/0000-0002-6175-686X
Department of Public Health, California State University, Sacramento, California (Dr Mansyur); Department of Family and Community Medicine, Baylor College of Medicine, Houston, Texas (Dr Rustveld, Dr Nash); School of Health Professions, Baylor College of Medicine, Houston, Texas (Dr Jibaja-Weiss); and Dan L Duncan Comprehensive Cancer Center, Baylor College of Medicine, Houston, Texas (Dr Jibaja-Weiss).
Corresponding Author:Carol L. Mansyur, Department of Public Health, California State University, Sacramento, 6000 J Street, Sacramento, California 95819-6073, USA.Email: carol.mansyur@csus.edu
