Effects of individual and neighborhood socioeconomic status on antiretroviral therapy adherence: The role of adherence self-efficacy

Abstract

This study aimed to examine the potential mediation effect of adherence self-efficacy on the associations between individual and neighborhood socioeconomic status and antiretroviral therapy adherence in a sample of 337 people living with HIV in South Carolina, United States. Results showed that there were no direct effects of individual or neighborhood socioeconomic status on antiretroviral therapy adherence, whereas both individual socioeconomic status and neighborhood socioeconomic status were associated with adherence self-efficacy, which in turn were related to antiretroviral therapy adherence. These findings suggest that interventions targeting adherence self-efficacy may improve antiretroviral therapy adherence among people living with HIV with low socioeconomic status or those living in socioeconomically disadvantaged neighborhoods.

Introduction

The availability of antiretroviral therapy (ART) has dramatically decreased the progression rate from HIV infection to AIDS and death (Palella et al., 2006; Sterne et al., 2005). HIV, therefore, has been reframed from an acute fatal disease to a manageable chronic disease (McGrath et al., 2014). However, people living with HIV (PLWH) may not equally benefit from the availability of ART (Cunningham et al., 2005; Rubin et al., 2010; Singh et al., 2013). Socioeconomic status (SES) disparities in HIV treatment outcomes have been observed, where PLWH with low SES are at greater risk of experiencing poor virologic (e.g. viral load) and immunological (e.g. CD4 cells) responses to ART treatment than their counterparts with high SES (for a review, see Burch et al., 2016b). The SES gap in HIV treatment outcomes remains in settings with universal free HIV care (Burch et al., 2016a).

Not only does individual SES matter, but SES at the ecological level, such as the neighborhood, may have an impact on HIV (Latkin et al., 2013). Individual SES refers to an individual’s social position relative to others, and it has been defined by income, education, occupation, or a combination of these dimensions (Adler and Ostrove, 1999; Wu et al., 2013). In parallel to individual SES, neighborhood SES is defined by neighborhood levels of income, education, occupation, or a composite of these indicators (Ross and Mirowsky, 2008), which are commonly operationalized as census tract–level or zip code–level measures (e.g. percentage of residents of the same zip code with a high school diploma) derived from administrative data (e.g. Joy et al., 2008; Wiewel et al., 2016). In the United States, HIV disproportionately affects people residing in disadvantaged neighborhoods (Denning and DiNenno, 2010). Among PLWH, socioeconomically disadvantaged neighborhoods also significantly contribute to unsuppressed viral load and low CD4 cell counts (Gueler et al., 2015; Joy et al., 2008; Shacham et al., 2013; Wiewel et al., 2017), and such effects on HIV treatment outcomes go above and beyond individual SES (Shacham et al., 2013). Therefore, to better elucidate the SES gap in HIV treatment outcomes, it is crucial to conceptualize SES at both individual and neighborhood levels (Pavlova-McCalla et al., 2012).

Many factors contribute to SES disparities in HIV treatment outcomes, including ART adherence (Burch et al., 2016b). It is well established that ART adherence plays a primary role in affecting virologic and immunological responses to ART treatment (Braithwaite et al., 2007; Mannheimer et al., 2002). ART adherence can be influenced by both environmental and individual factors, including SES (Golin et al., 2002). Studies have indicated that PLWH with low individual SES have poorer ART adherence than those with high individual SES (Catz et al., 2001; Cioe et al., 2017; Dale et al., 2016; Kleeberger et al., 2001), with exceptions that observed no effects of individual SES on ART adherence (Brown et al., 2013). For example, a cross-sectional study indicated that among men who have sex with men living with HIV, those with less than a bachelor’s degree were 2.5 times more likely to report suboptimal adherence (i.e. < 95%) than their counterparts with a bachelor’s degree or higher (Cioe et al., 2017).

Neighborhood SES may also have additional effects on ART adherence. Indeed, a few studies have begun to examine the potential effect of neighborhood SES on ART adherence (Silverberg et al., 2009; Surratt et al., 2015). In the study of Surratt et al. (2015), the authors found that zip code–level poverty was related to ART nonadherence among a sample of socioeconomically disadvantaged PLWH, whereas Silverberg et al. (2009) found census tract–level education and income had no impacts on ART adherence over 24 months after ART initiation. Although inconsistent and null findings were documented, perspectives from the ecological model have emphasized the important roles of multiple ecological systems including individual and neighborhood contexts in influencing medication adherence (Berben et al., 2012). Neighborhood factors, such as socioecological disadvantage, may act as distal risk factors to shape health-related behaviors through the proximal interpersonal and personal mechanisms (Pampel et al., 2010).

Social cognitive theory suggests that psychological factors play critical roles in accounting for the effects of sociostructural factors (e.g. SES) on health behaviors (Bandura, 1998). Sociostructural factors may not impact health behaviors (e.g. ART adherence) directly, but through some psychological mechanisms, particularly self-efficacy (Bandura, 1997). Self-efficacy is defined as an individual’s belief about his or her ability to perform a specific task or achieve a certain outcome (Bandura, 1997). Individuals with a higher level of perceived self-efficacy are more likely to exert effort and persist in the face of challenging tasks (Van der Bijl and Shortridge-Baggett, 2001). It has been suggested that self-efficacy is a particularly crucial determinant of optimal behavioral health, including treatment adherence (Dunbar-Jacob and Mortimer-Stephens, 2001; Munro et al., 2007). In the HIV context, self-efficacy in HIV treatment adherence (hereafter referred to adherence self-efficacy), an individual’s perceived ability to adhere to the HIV treatment plan including ART medication (Johnson et al., 2007), has been found to be associated with better ART adherence (Brittain et al., 2018; Houston et al., 2016; Parsons et al., 2007; Zhang et al., 2016). For example, Houston et al. (2016) found that compared to PLWH with low adherence self-efficacy, those with high adherence self-efficacy were nearly three times more likely to report optimal ART adherence (i.e. ⩾ 95%).

SES has been suggested as a determinant of self-efficacy (Han et al., 2014). SES can affect task-specific self-efficacy through its influence on mastery experience, vicarious experience, verbal persuasion, and physiological and emotional states (Boardman and Robert, 2000; Clark, 1996), the four resources of the development of self-efficacy (Bandura, 1997). It suggests that SES limits an individual’s access to material, social, and cultural resources and affects the sense of control, which in turn impacts the four resources and, therefore, the task-specific self-efficacy (Clark, 1996), such as adherence self-efficacy. Indeed, studies of PLWH have documented an association between individual SES and adherence self-efficacy (Reynolds et al., 2004; Zhang et al., 2016). For example, a multicenter study in the United States and Italy found that ART-naïve PLWH with low education reported much lower adherence self-efficacy than PLWH with high education (Reynolds et al., 2004). Neighborhood SES has also been found to be associated with self-efficacy above and beyond individual SES (Boardman and Robert, 2000). It is suggested that low neighborhood SES may decrease social resources that PLWH can access (Latkin et al., 2013), which in turn contributes to low adherence self-efficacy (Turan et al., 2016).

To date, few studies have examined whether adherence self-efficacy may act as a potential psychosocial mediator of the effect of SES on ART adherence. Elucidating the mechanisms linking SES to ART adherence is critical for developing interventions to improve ART adherence for PLWH, particularly given that HIV disproportionally affects people of low SES or those living in disadvantaged neighborhoods (Denning and DiNenno, 2010). In addition, only a few studies have tested the effects of SES at both individual and neighborhood levels on ART adherence (Surratt et al., 2015). Therefore, the aim of this study was twofold: (1) to assess the direct effects of individual and neighborhood SES on ART adherence and (2) to examine the potential mediation effects of adherence self-efficacy on the associations between individual and neighborhood SES and ART adherence among PLWH who were currently on ART in South Carolina, United States. We hypothesized that there would be direct effects of individual and neighborhood SES on ART adherence. We further hypothesized that there would be significant indirect effects of individual and neighborhood SES on ART adherence through adherence self-efficacy.

Methods

Participants and procedure

Data were collected from a large immunology clinic in South Carolina during May and September 2018. Inclusion criteria were (1) living with HIV and (2) over 18 years of age. A total of 402 participants agreed to participate and completed a cross-sectional survey. Of the 402 participants, 337 reported they were currently receiving ART at the time of the survey. Given the focus of this study, only those (n = 337) reported on ART were included in the current analysis.

The survey was administered at the clinic with the support from the clinic staff, and it took about 35–40 minutes to complete. Each participant received a monetary incentive (i.e. a $20 gift card) for his or her participation after the survey. Prior to the survey, written consent was obtained from all participants. The study procedure was approved by the Institution Review Board at the University of South Carolina.

Measures

Individual SES

Individual SES was measured using participants’ education and household income. Participants were asked to report their education level on a 7-point scale. Response options included 1 = never attended school, 2 = Grades 1 through 8, 3 = Grades 9 through 11, 4 = Grade 12 or General Educational Development (GED), 5 = some college, associate degree, or technical degree, 6 = bachelor’s degree, and 7 = any postgraduate studies. They were also asked to report their yearly household income on a 5-point scale. Response options included 1 = less than US$10,000, 2 = US$10,000–US$24,999, 3 = US$25,000–US$49,999, 4 = US$50,000–US$99,999, and 5 = US$100,000 or more. An individual SES index was calculated by averaging z-scored education and household income, with a higher score reflecting higher individual SES.

Neighborhood SES

Neighborhood SES was measured by linking participants’ home zip codes to US Census data. In this study, 57 five-digit zip codes were provided, with Census data available for all but five zip codes. Information on two indicators that were conceptually parallel with the individual SES measure (i.e. percentage of residents with high school education, median annual household income) was extracted from the US Census Bureau American Community Survey (ACS) 5-year estimates from 2012 to 2016 by ZIP Code Tabulation Area (https://www.census.gov/programs-surveys/acs/). The ACS is an annual nationwide survey that is sent to 3.5 million addresses each year. A neighborhood SES index was calculated by summing z-scored composites of these two indicators, with a higher score reflecting higher neighborhood SES.

Adherence self-efficacy

Adherence self-efficacy was measured using the HIV Treatment Adherence Self-Efficacy Scale (HIV-ASES, Johnson et al., 2007). The HIV-ASES is a 12-item scale with satisfactory reliability and validity that has been established among PLWH (Johnson et al., 2007). Participants were asked to report how confident they can do several treatment-related behaviors in the past month on a 3-point scale from 1 = cannot do at all to 3 = completely can do. Sample items are “stick to your treatment plan even when side effects begin to interfere with daily activities” and “stick to your treatment schedule even when your daily routine is disrupted.” A mean score for adherence self-efficacy was calculated by averaging participant’s scores on 12 items, with a higher score reflecting higher self-efficacy. The Cronbach’s α was .95 in this study.

ART adherence

ART adherence was measured using three items adapted from the Adult AIDS Clinical Trials Group adherence instrument (Chesney et al., 2000), with a focus on recent adherence to reduce recall bias. Participants were asked to report whether they missed any doses of ART adherence “yesterday,” “the day before yesterday,” and “last Saturday/ Sunday” on a binary response (yes/no). Participants’ responses on the three items were dichotomized into two categories: 0 = suboptimal adherence (i.e. doses missed at least one of these days) and 1 = optimal adherence (i.e. no doses missed on any of these days).

Covariates

Participants were asked to provide information on some demographic covariates, including sex, age, race, and marital status. Sex was coded with 0 = male and 1 = female. Race was coded with 0 = Black and 1 = White/others. Marital status was categorized as 0 = single, 1 = married/cohabitation, and 2 = separated/divorced/widowed. Participants were also asked to report the year of their HIV diagnosis.

Data analysis

Descriptive analyses were performed using SPSS 25.0. The t-test for continuous variables and the chi-square test for categorical variables were used to examine the group differences (i.e. ART suboptimal vs optimal adherence) in study variables. The correlation analysis was also performed to test the correlation coefficients among the variables.

To examine the direct and indirect effects of SES on ART adherence, two sets of path analyses were performed using maximum likelihood estimation with robust standard errors in Mplus 7.0 (Muthén and Muthén, 2012). The first set of path analyses was used to examine the effects of individual SES, and the second set of path analyses was used for neighborhood SES. The analysis procedure was as follows: (1) a direct effect model to test the direct effect of individual SES/neighborhood SES on ART adherence, and (2) a mediation model to test the potential mediation effect of adherence self-efficacy on the association between individual SES/neighborhood SES and ART adherence. In addition, a combined model including both individual and neighborhood SES was performed to test whether neighborhood SES had an indirect effect on ART adherence via adherence self-efficacy above and beyond individual SES. All mediation models were first performed without covariates and then with adjustment of covariates for the mediator (i.e. adherence self-efficacy) and the dependent variable (i.e. ART adherence). Covariates (i.e. sex, race, marital status, age, year of HIV diagnosed) were included in the analyses only if they were statistically significantly correlated with the mediator or the dependent variable to preserve model parsimony. Cluster-robust standard errors were obtained for all models to account for clustering by zip codes (McNeish et al., 2017). The product of coefficient approach was performed to test the indirect effect (MacKinnon et al., 2002), and Sobel test was used to examine the significance of the indirect effect (Sobel, 1982).

Results

Descriptive results

Of 337 participants, 261 (77.4%) were categorized into ART optimal adherence group, and 74 (22.0%) were ART suboptimal adherence group; two participants were excluded due to missing responses on adherence measures. Compared to PLWH with ART optimal adherence, participants with ART suboptimal adherence were more likely to report lower adherence self-efficacy (t = 4.20, p < .001). However, there were no differences in individual or neighborhood SES between these two groups (t = 0.85, p = .40; t = 0.09, p = .93, respectively). PLWH in the ART optimal adherence group and their counterparts in the ART suboptimal adherence group also did not differ in the distribution of sex, race, marital status, year of HIV diagnosed (ps > .10, see Supplementary Table 1), with an exception that there was a marginally significant difference in age between these two groups (t = 1.96, p = .051). The correlation analysis showed that both individual and neighborhood SES were correlated with adherence self-efficacy (r = .27, p < .001; r = 0.18, p = .003, respectively, see Supplementary Table 2). Individual and neighborhood SES were also significantly correlated (r = 0.17, p = .005).

Direct and indirect effects of SES on ART adherence

Individual SES

The direct effect model showed that there were no direct effects of individual SES on ART adherence (standardized β = 0.10, p = .19). Mediation analyses without covariates showed that individual SES was related to adherence self-efficacy (β = 0.27, p < .001). Adherence self-efficacy, in turn, was related to ART adherence (β = 0.32, p < .001). Individual SES was not associated with ART adherence (β = −0.01, p = .90). There was a significant indirect effect of individual SES on ART adherence via adherence self-efficacy (z = 3.52, p < .001). Given that there was a significant correlation between race and adherence self-efficacy and there were marginally significant age differences in ART adherence, race and age were included as the covariates in the model. A similar pattern of results was obtained in the adjusted mediation model (see Figure 1(a)). The indirect effect of individual SES on ART adherence via adherence self-efficacy remained significant (z = 3.07, p = .002). The model accounted for nine percent of the variances in adherence self-efficacy and 11 percent of the variances in ART adherence.

Figure 1.

The model of socioeconomic status (SES), adherence self-efficacy, and antiretroviral therapy (ART) adherence: (a) the model for individual SES and (b) the combined model for neighborhood SES. Standardized path coefficients were presented. Paths with solid lines represented statistically significant, and dashed lines represented statistically nonsignificant. Demographic variables (i.e. age, race) were controlled for the model but not displayed for simplicity, with the path from race to adherence self-efficacy being statistically significant (β = 0.25, p = .002 in the model for individual SES; β = 0.22, p = .01 in the combined model for neighborhood SES). ART adherence was coded as 0 = suboptimal and 1 = optimal; race was coded as 0 = Black and 1 = White/others.

Neighborhood SES

There were no direct effects of neighborhood SES on ART adherence (β = 0.01, p = .91). In the mediation model, however, neighborhood SES was associated with adherence self-efficacy (β = 0.18, p < .001). Adherence self-efficacy, in turn, was associated with ART adherence (β = 0.35, p < .001). The association between neighborhood SES and ART adherence was not statistically significant (β = –0.07, p = .23). There was a significant indirect effect of neighborhood SES on ART adherence through adherence self-efficacy (z = 2.93, p = .003). In the combined model, neighborhood SES remained significantly associated with adherence self-efficacy (β = 0.12, p = .025), which in turn was related to ART adherence (β = 0.33, p < .001), controlling for individual SES. Neighborhood SES remained not associated with ART adherence (β = –0.07, p = .27). The indirect effect of neighborhood SES on ART adherence through adherence self-efficacy remained significant (z = 2.03, p = .042). The results were similar after controlling for both individual SES and covariates (see Figure 1(b)), with an exception that the indirect effect of neighborhood SES on ART adherence became nonsignificant (z = 1.79, p = .074). The model accounted for 10 percent of the variances in adherence self-efficacy and 12 percent of the variances in ART adherence.

Discussion

SES disparities in HIV treatment outcomes have been well established, and such disparities might be partially driven by the differences in ART adherence with regard to SES (Burch et al., 2016b). This study extends existing studies by examining the potential mediation role of adherence self-efficacy in the associations between individual and neighborhood SES and ART adherence among PLWH receiving ART. Our results showed that there were no direct effects of individual and neighborhood SES on ART adherence. However, both individual and neighborhood SES were associated with adherence self-efficacy, and adherence self-efficacy, in turn, was associated with ART adherence.

Our study showed that low individual SES might indirectly contribute to poor ART adherence through low adherence self-efficacy. The result is in line with the social cognitive theory that emphasizes the critical role of self-efficacy in health behaviors (Bandura, 1997). Our result also corroborates previous empirical studies indicating the effects of adherence self-efficacy on ART adherence (Johnson et al., 2007; Parsons et al., 2008; Zhang et al., 2016). It has been indicated that PLWH with low SES are likely to perceive a low sense of control and have limited access to resources, resulting in low adherence self-efficacy (Clark, 1996; Janssen et al., 2000). PLWH with low adherence self-efficacy, in turn, may be less likely to stick to their HIV treatment and therefore have an increased risk of not achieving ART adherence. This finding suggests that adherence self-efficacy may be a potential psychological mediator linking individual SES to ART adherence, which implies that enhancing adherence self-efficacy may be a possible way to improve ART adherence for PLWH of low SES.

This study also showed that neighborhood SES was associated with adherence self-efficacy, which in turn was associated with ART adherence. A similar pattern of results was found when controlling for individual SES. This result is consistent with prior studies demonstrating that the effect of neighborhood SES on health-related behaviors and outcomes is not a simple by-product of individual SES (Cano and Wetter, 2014; Ross and Mirowsky, 2008), highlighting that neighborhood SES may be a risk contextual factor contributing to low adherence self-efficacy and therefore poor HIV management among PLWH. However, we observed a statistically nonsignificant indirect effect of neighborhood SES on ART adherence when controlling for both individual SES and covariates (i.e. age, race). One possible explanation for this null finding is the lack of statistical power. Future studies with PLWH from a larger set of neighborhoods are needed to more precisely identify the possible indirect effects. Another possible explanation is that the indirect effect of neighborhood SES on ART adherence through adherence self-efficacy may be partially explained by racial differences in adherence self-efficacy. Blacks are disproportionately segregated and concentrated in socioeconomically disadvantaged neighborhoods in the United States (Firebaugh and Acciai, 2016). Moreover, this study found a significant relationship between race and adherence self-efficacy, which is line with previous studies suggesting that black PLWH are more likely to experience low adherence self-efficacy than white PLWH (e.g. Johnson et al., 2007).

Interestingly, our study did not observe the direct effects of individual or neighborhood SES on ART adherence. These results are inconsistent with previous studies that demonstrated direct links between individual SES and ART adherence (Catz et al., 2001; Kleeberger et al., 2001). One potential explanation is that the direct effects of SES on ART adherence might be decreased in the settings where universal or nearly universal HIV care is available. In this study, all participants had access to HIV care through the Ryan White Program. These results suggest that when universal HIV care is accessible, SES may not affect ART adherence directly. Instead, it may impact ART adherence in an indirect and invisible way, possibly through some psychosocial mechanisms, including adherence self-efficacy.

The results in this study, however, should be cautiously interpreted due to several limitations. First, the nature of cross-sectional data excluded the possibilities to draw any definitive causal interpretations regarding the relationships among SES, adherence self-efficacy, and ART adherence. Second, the self-report measure for ART adherence may not be very reliable, though this study attempted to increase the reliability by measuring adherence within a very short time frame (i.e. within one day to one week). Self-report measures have been suggested to overestimate ART adherence, compared to objective measures, such as electronic drug monitoring, which in turn may affect the results of this study (Pearson et al., 2007). Third, using zip codes to define neighborhoods, though widely used in previous studies, might not be optimal compared to the use of the smaller geographic units, including tract- and block-levels (Krieger et al., 2002). Finally, our participants were recruited from one state that ranked high for HIV infection, poverty, and low educational attainment compared to other states in the United States (South Carolina Department of Health and Environmental Control, 2017). Our findings may not be generalizable to other settings that have substantially different sociodemographic characteristics.

Despite these limitations, our study provided the empirical evidence showing that both individual and neighborhood SES might indirectly impact ART adherence through adherence self-efficacy among PLWH receiving ART. The findings highlight that adherence self-efficacy may play a critical role in determining ART adherence. Interventions with a focus on enhancing adherence self-efficacy should be considered in the future to facilitate ART adherence for PLWH, particularly for those with low SES or living in socioeconomically disadvantaged neighborhoods.