Racial/ethnic minority status is inconsistently associated with antiretroviral therapy adherence by HIV-infected patients.1-6 This relationship sometimes disappears after controlling for key predictors of antiretroviral therapy adherence such as depression, active alcohol and drug use, or low health literacy7,8 whereas in other instances it does not.2 Racial/ethnic minority status is likely acting as a proxy for other unmeasured factors when the association with poor antiretroviral therapy adherence persists.
One pathway by which racial/ethnic minority status may affect antiretroviral therapy adherence is through patients' negative experiences and expectations about interpersonal treatment by their health care providers. Concerns by members of racial/ethnic minority groups about inequitable provision of care have their roots in historical and present-day experiences with the health care system.9-12 Patients who are members of racial/ethnic minority groups may be more likely than white patients to perceive unfair treatment by health care providers (discriminatory health care experiences)13-15 and have lower expectations that providers will act in their best interests (health care provider distrust).16-20 Discriminatory health care experiences and health care provider distrust may, in turn, influence racial/ethnic minority patients' health-related attitudes, beliefs, and behaviors regarding treatment adherence.21,22
The quality of interpersonal HIV/AIDS care may affect achievement of desired health care outcomes.23,24 Specifically, HIV-infected patients who report better interpersonal treatment from health care providers are more likely to report better adherence to antiretroviral therapy.4,25-30 Little is known, however, about the effects of discriminatory health care experiences and health care provider distrust on antiretroviral therapy adherence. In the only previous study of its kind, HIV-infected patients who reported more discriminatory health care experiences attributed to socioeconomic status (though not race) had poorer adherence to antiretroviral therapy than those who reported fewer experiences in unadjusted analyses.31 Trust in one's physician was independently associated with better antiretroviral therapy adherence in a large, cross-sectional study of HIV-infected patients30 and in a prison-based directly observed therapy program32 but not among participants of a prospective cohort study.2 No previous study, however, has examined the contribution of these factors to racial/ethnic disparities in antiretroviral therapy adherence.
Furthermore, existing research has only examined the direct effects of discriminatory health care experiences and health care provider distrust on antiretroviral therapy adherence. Long-standing research suggests that HIV-infected patients' beliefs about and attitudes toward antiretroviral therapy are some of the most consistent and proximate determinants of adherence.33-36 Some studies suggest that racial/ethnic minority HIV-infected patients report different psychosocial barriers toward adherence and hold more negative attitudes toward antiretroviral therapy than white patients.37,38 Negative feelings toward health care providers that manifest as, or result from, discriminatory health care experiences and health care provider distrust may lead HIV-infected patients to doubt the efficacy and worth of antiretroviral therapy and thus indirectly make them less likely to adhere to their prescribed medication. Therefore, the present study examined the contribution of discriminatory health care experiences and health care provider distrust to the observed racial/ethnic disparity in antiretroviral therapy adherence among a nationally representative sample of HIV-infected patients.5
The HIV Cost and Services Utilization Study (HCSUS) collected baseline data between January and February 1996 with follow-up interviews conducted 6 and 12 months afterward; full details have been described elsewhere.39,40 The eligibility criteria for the original study were residence in the 48 contiguous states; aged older than 18 years; known HIV infection; and receipt of medical care in facilities other than emergency departments, the military, or prisons. HCSUS used multistage population-proportion-to-size sampling for urban areas and purposive sampling for rural areas.39,40 Researchers surveyed 4042 HIV-infected persons receiving care in 28 urban areas and 24 clusters of rural counties in the contiguous United States.39-41 Of those surveyed, field interviewers completed full in-person structured interviews using computer-assisted personal interview instruments at baseline with 2864 (68%) individuals.
The present study used data from 1911 participants in the public use HCSUS data set who completed all 3 interviews and reported combination antiretroviral therapy use at the second follow-up interview. The sample represented 67% of participants at baseline and 84% at the first follow-up interview. The secondary analysis conducted for the present study was approved through expedited review by the Institutional Review Board of the University of North Carolina, School of Public Health.
Exploratory factor analysis determined which items, initially selected based on face validity, defined the factors (ie, unobserved or latent variables) that represented the study constructs.42 Composite measures were created from items for each factor, and Cronbach alpha was used to assess internal consistency (proportion of variance attributable to the true score of a latent variable).
Self-reported “combination antiretroviral therapy adherence” is the extent to which HIV-infected patients took their medication as prescribed. This measure included 5 items obtained at the second follow-up interview (Cronbach alpha = 0.82). For each prescribed antiretroviral medication, participants were asked about 4 aspects of their medication taking: the number of days in the last 7 days that they (1) forgot to take a dose, (2) deliberately skipped a dose, (3) took less medication than prescribed, and (4) took all medications as prescribed. The first 3 items were reverse coded so that higher scores would mean higher levels of adherence (eg, number of days in the last 7 days that they did “not forget” to take a dose). The final value of each indicator was the sum of all items across medications divided by the number of medications reported, with a range of 0-7. A fifth global indicator assessed the extent to which participants believed they took their antiretroviral medications exactly as prescribed in the last 30 days (1 = none of the time to 6 = all the time) and rescaled to match the other adherence items (1 = 0, 2 = 1.25, 3 = 3, 4 = 4.25, 5 = 5.5, and 6 = 7). One-month recall items and recall items that are global ratings may be less prone to overreporting than 3- or 7-day frequency recall items,43 and research suggests that a 30-day global adherence measure may be no less accurate than a 7-day recall measure.44 All 5 items were summed (unweighted) and averaged to create a composite score, with higher scores meaning better adherence. For bivariate analyses that used a dichotomous measure of adherence, “perfect adherence” was a score of 7, defined as no reports of forgetting a dose, skipping a dose, taking less medication than prescribed, taking all medications as prescribed, and a global belief that the participant took all medications as prescribed in the last 30 days.
“Racial/ethnic minority status” was dummy coded as 0 if participants were self-classified as white race, non-Hispanic ethnicity and 1 for any other racial/ethnic group.
“Discriminatory health care experiences” had 6 dichotomous items, 3 obtained at baseline and 3 obtained at the first follow-up interview (Cronbach alpha = 0.84). Participants reported at baseline whether, since they had become HIV infected, any health care provider had “ever” exhibited hostility or disrespect to them, paid less attention to them than other patients, or refused them service. At the first follow-up interview, participants reported whether any health care provider had ever been uncomfortable around them, treated them as inferior, or preferred to avoid them since they had become HIV infected. Although the items were collected at both baseline and first follow-up interviews, they were considered as a group because they represent a range of experiences that patients might consider discriminatory.
“Health care provider distrust” had 7 items obtained at the second follow-up interview assessing the extent to which participants trusted their current primary provider of HIV/AIDS care (Cronbach alpha = 0.92). On a 5-point scale (1 = completely to 5 = not at all), participants reported the degree that they trusted their health care providers to offer quality care, know the best treatments, provide enough information, keep personal information confidential, treat them in a nonjudgmental manner, offer high-quality care regardless of insurance status, and put participant needs ahead of research goals.
Four factors captured participants' attitudes and beliefs about antiretroviral therapy, with all items obtained at the second follow-up interview. Several items for each factor (the number specified below) were summed and averaged to create composite scores. “Psychological burden of medication” had 8 items that assessed the degree to which participants had a range of worries and concerns about taking antiretroviral therapy: had too many pills to take, wanted to avoid side effects, felt the drug was too toxic, reminded self of HIV status, health was not improving, worried about becoming immune, took a “drug holiday,” and felt depressed or overwhelmed. Each item was measured on a 4-point scale (1 = never to 4 = often), and the 8-item factor had a Cronbach alpha of 0.85. “Difficulty accessing medication” had 4 items assessing the degree to which participants felt that obtaining antiretroviral therapy was difficult: hard to get HIV medications, not easy to get HIV prescriptions, takes a lot of time and effort to get HIV medication, and HIV medication would be hard to get if it runs out. Each indicator was measured on a 4-point scale (1 = strongly disagree to 4 = strongly agree), and the factor had a Cronbach alpha of 0.71. “Difficulty scheduling medication” had 3 items assessing the extent to which changes in participants' routines impeded their adherence to antiretroviral therapy: away from home, too busy or forgot, and change in daily routine. Each indicator was measured on a 4-point scale (1 = never to 4 = often), and the factor had a Cronbach alpha of 0.76. “Weak medication efficacy beliefs” had 2 items assessing the extent to which participants believed that HIV medications make people live longer and that HIV medications improve the quality of people's lives. Both items were measured on a 4-point scale (1 = strongly disagree to 4 = strongly agree), and the factor had a Cronbach alpha of 0.72.
The control variables used in these analyses were obtained from derived variables in the public use data set.41 All variables were assessed at baseline except as noted. The sociodemographic covariates were age (recoded to category midpoint to approximate a continuous scale), sex, sexual orientation (heterosexual, bisexual, and homosexual), highest degree obtained (less than high school, high school/general education diploma (GED), associate's, and bachelor's or higher), annual income (less than $5000-$40,000 or more, recoded to category midpoint to approximate a continuous scale), insurance status (none, Medicaid, Medicare, and private), and existence of a usual HIV provider. The HIV-specific covariates were HIV risk exposure category (male-to-male sexual contact, male-to-female sexual contact, injecting drug use, and others), year first diagnosed with HIV, whether the participant was diagnosed with AIDS, lowest reported CD4 count, and number of antiretroviral medications prescribed at second follow-up interview. Other health covariates were standardized physical and mental health composite scores of the health-related quality-of-life subscales of the SF-36 (second follow-up interview), the number of depression and dysthymia (a chronic, less severe form of depression) symptoms, and 4 categories of illicit drug use in the past year (marijuana or analgesics; inhalants, hallucinogens, sedatives, and amphetamines; heroin or cocaine; and no use). Social support was a standardized composite variable on a 100-point scale that measured the degree to which participants felt they had someone who could give them money, help with daily chores, and love them. Adherence self-efficacy assessed the extent to which participants believed they would be able to take their antiretroviral medication as prescribed over the next month (1 = strongly agree to 4 = strongly disagree).
First, descriptive statistics were produced using Stata 8.2.45 The complex sampling design of HCSUS required use of analytic weights to correct for the inflated variance caused by clustering.39-41 Unless otherwise noted, all analyses were conducted with weighted data. No transformation normalized the distribution of the negatively skewed key study variables (antiretroviral therapy adherence, discriminatory health care experiences, health care provider distrust, and antiretroviral therapy attitudes and beliefs), and nonparametric analyses were conducted as applicable. Bivariate analyses (t test, χ2 test, or Mann-Whitney z as appropriate) were conducted to examine differences by minority status. Bivariate analyses were also used to determine which factors were associated with antiretroviral therapy adherence at P ≤ 0.20, a liberal cut-off point used, because we did not want to prematurely drop conceptually important variables (eg, discriminatory health care experiences) in the analyses. Variables that were not significantly associated with adherence in bivariate analyses were not significant in later multivariate modeling and thus dropped from the final model.
Next, hypotheses regarding mediators of the relationship between racial/ethnic minority status and antiretroviral therapy adherence were tested using structural equation modeling (SEM) with MPlus 3.11.46 SEM proposes a set of relations between variables and evaluates the degree of model fit with observed data.47 This method controls for measurement error and allows estimation of models with multiple mediators or mediation/moderation combinations.48 This technique does not allow researchers to identify a “true” model. Rather, multiple models may fit the data, and researchers must use their own judgment to determine whether a given model corresponds with reality.47 The measurement model for SEM, created by confirmatory factor analysis, is the hypothesized relationships between observed items and factors (data not shown). The structural model for SEM produces standardized estimates (β) of the strength and direction of relationships among factors. Model parameters were produced using weighted least square mean- and variance-adjusted estimation, a robust method that is used when at least one categorical variable is incorporated in a model and also accounts for nonindependent, nonnormally distributed observations46; P values were obtained using online calculators.49 The magnitudes of standardized parameters were assessed using Cohen's (1988) typology of small (0.10 or less), medium (around 0.30), and large (0.50 or more) effects.50
Models were considered to have a good fit if: (1) the ratio of χ2 to degrees of freedom (CMINDF) was less than 3.0, (2) the root mean square error of approximation was 0.06 or less, and (3) the Tucker-Lewis and Comparative Fit indices were 0.95 or greater,51 and the weighted root mean residual (WRMR), used when categorical items are included in a model,52 was 1.00 or less. The use of multiple fit indices to evaluate a model is the recommended practice for SEM research.47,51 Model fit was improved by adding or deleting paths based on content knowledge and modification indices produced by MPlus. In all cases, however, theoretical plausibility trumped modification indices when deciding what paths to add to an existing model. When the model fit was satisfactory, nonsignificant paths were trimmed.
Finally, an effects decomposition of the total, direct, and indirect effects of racial/ethnic minority status on antiretroviral therapy adherence was produced using the MODEL INDIRECT command of MPlus.46 The direct effect is the parameter estimate of the relationship between an independent variable and the outcome variable. Indirect effects are parameter estimates of mediating pathways between an independent variable and outcome variable. The total effect is the sum of the direct and indirect effects.
Attrition and Selection Analyses
In multivariate analyses, participants who completed all 3 interviews (n = 2267) were less likely than those who did not complete all 3 interviews (n = 597) to: be a member of a racial/ethnic minority group [odds ratio (OR) = 0.75, P = 0.012], have an AIDS diagnosis (OR = 0.68, P = 0.006), or use heroin or cocaine in the past year (OR = 0.87, P = 0.008). Participants who completed all 3 interviews were more likely than those who did not complete all interviews to be exposed to HIV through male-to-female sexual contact (OR = 1.56, P = 0.047) and to have greater social support (OR = 1.00, P = 0.009).
Participants who completed all 3 interviews and also were on antiretroviral therapy at the second follow-up interview (n = 1911) distrusted their health care providers less (OR = 0.58, P ≤ 0.001) than those who were not on antiretroviral therapy (n = 356) in multivariate analyses. Participants who were on antiretroviral therapy were also more likely to have an AIDS diagnosis (OR = 2.44, P = 0.001) and a higher income (OR = 1.18, P = 0.028) than those who were not on antiretroviral therapy. After 25 participants were dropped for having missing values on multiple study variables, the final sample size was 1886.
Table 1 presents descriptive statistics stratified by racial/ethnic minority status. Compared with nonminorities, minority participants were more likely to be females, have less than a high school degree, earn less than $25,000, have no insurance or Medicaid, ever have a CD4 count <200, and have used heroin or cocaine in the past year. Minorities were less likely than nonminorities to report homosexual orientation, exposure to HIV through male-to-male sexual contact, and to be diagnosed with AIDS. Minority participants had been diagnosed with HIV for a shorter amount of time and reported less social support.
Antiretroviral Therapy Adherence
Given a 0-7 scale, all indicators of antiretroviral therapy adherence had high means, suggesting ceiling effects: number of days in the last 7 days that patient did not forget to take a dose (mean 6.57, standard deviation, 0.90), number of days in the last 7 days that patient did not purposely skip dose (mean 6.75, standard deviation, 0.91), number of days in the last 7 days that patient did not take a lesser amount than prescribed (mean 6.70, standard deviation, 0.92), number of days in the last 7 days that patient took medications as prescribed (mean 6.15, standard deviation, 1.52), and how well patient took medications exactly as prescribed in the last 30 days (mean 5.18, standard deviation, 1.19). The mean of the composite adherence measure was 6.40 (standard deviation, 0.92), with almost half of the participants (45.7%, n = 888) reporting perfect adherence to antiretroviral therapy in the past week. Minority participants were less likely to report perfect adherence than nonminority participants [40.3% vs. 45.3%, χ2 (1,54) = 31.5, P < 0.001].
Prevalence of Discriminatory Health Care Experiences and Health Care Provider Distrust
More than one third of participants (41%) reported ever having at least one of 6 types of discriminatory health care experiences since they were diagnosed with HIV. Minority participants were less likely to report discriminatory health care experiences than nonminorities [Mann-Whitney z (1857) = 7.18, P < 0.001]. Racial/ethnic minority and nonminority participants did not differ in the degree of health care provider distrust they experienced in an unadjusted bivariate analysis [Mann-Whitney z (1886) = −0.17, P = 0.869].
Attitudes and Beliefs About Antiretroviral Therapy
All measures of antiretroviral therapy attitudes and beliefs had low means, indicating floor effects: psychological burden of medication (mean 1.37, standard deviation, 0.55), difficulty accessing medication (mean 2.18, standard deviation, 0.39), difficulty scheduling medication (mean 1.91, standard deviation, 0.78), and weak medication efficacy beliefs (mean 1.53, standard deviation, 0.52). Racial/ethnic minority participants were more likely than white participants to report psychological burden regarding medication use [Mann-Whitney z (1886) = −6.16, P < 0.001], difficulty accessing their medication [Mann-Whitney z (1886) = −3.95, P < 0.001], and skepticism about the efficacy of antiretroviral therapy [Mann-Whitney z (1886) = −4.35, P < 0.001]. There were no racial/ethnic differences in participants' difficulty in scheduling their medication [Mann-Whitney z (1886) = −0.68, P = 0.496].
Factors Associated With Antiretroviral Therapy Adherence
Table 2 presents ORs representing unadjusted bivariate associations with antiretroviral therapy adherence. Poorer adherence was associated with minority status, female sex, younger age, not having a high school degree, lower income, having Medicaid, heterosexual orientation, exposure to HIV through injecting drug use or male-to-female sexual contact, higher viral load, more symptoms of depression and dysthymia, and use of heroin or cocaine in the past year. These variables were used as covariates in subsequent analyses. Health care provider distrust and more negative antiretroviral therapy-related attitudes and beliefs were associated with poorer adherence but discriminatory health care experiences were not.
Direct and Indirect Effects of Racial/Ethnic Minority Status on Antiretroviral Therapy Adherence
The a priori mediation model (χ2 = 76.02, df = 12, P = 0.000), which included multiple covariates, was a poor fit for the data by almost all of the goodness of fit indices (CMINDF = 6.34; Comparative Fit index = 0.92; Tucker-Lewis index = 0.91; root mean square error of approximation = 0.05; WRMR = 1.94). Guided by modification indices (which assess the expected drop in χ2 if a parameter is freely estimated) and conceptual plausibility based on existing research, parameters were sequentially added to improve the fit of the model. Nonsignificant paths were dropped from the final model. The squared multiple correlations of individual constructs ranged from 0.15 for health care provider distrust to 0.58 for psychological burden of medication. The standardized estimates are adjusted for factors associated with antiretroviral therapy adherence. Figure 1 presents the final model (χ2 = 40.48, df = 12, P = 0.000), which explained almost half the variance of antiretroviral therapy adherence (R2 = 0.49); covariates were included in the analyses but are not included in the figure. Unbroken lines in Figure 1 represent adjusted standardized parameters significant at the P = 0.05 level; broken lines represent hypothesized relations that were found to be nonsignificant. The final model provided adequate to good fit to the data across indices (CMINDF = 3.37; Comparative Fit index = 0.97; Tucker-Lewis index = 0.96; root mean square error of approximation = 0.04; WRMR = 1.40) according to published guidelines.51
The direct effects refer to the unbroken lines between the independent variables and antiretroviral therapy adherence in Figure 1. Racial/ethnic minority status was associated with poorer antiretroviral adherence (β = −0.21, P = 0.003). Contrary to our hypothesis, discriminatory health care experiences were not directly associated with poorer adherence, and thus the direct path was not included in the final trimmed model. Also, contrary to the hypothesized direction, more health care provider distrust was associated with better antiretroviral adherence (β = 0.06, P = 0.007). Most antiretroviral therapy attitudes and beliefs had direct effects on antiretroviral therapy adherence. Weaker beliefs in the efficacy of antiretroviral therapy were positively associated with adherence, contrary to hypothesis. Racial/ethnic minority status predicted fewer discriminatory health care experiences (β = −0.43, P = 0.001), a large effect. There was no direct effect of racial/ethnic minority status on difficulty accessing medication, so this path was dropped in the final model. Racial/ethnic minority status predicted greater health care provider distrust (β = 0.15, P = 0.027) and more negative beliefs about the efficacy of antiretroviral therapy (β = 0.32, P = 0.001). We also found unhypothesized direct effects among study constructs, such as the path from discriminatory health care experiences to medication efficacy beliefs (β = 0.14, P = 0.001) and the path from health care provider distrust to medication efficacy beliefs (β = 0.16, P = 0.004).
The total, direct, and indirect effects of racial/ethnic minority status on antiretroviral therapy adherence, after controlling for multiple covariates, are presented in Table 3. The first 2 rows of Table 3 show the effects of racial/ethnic minority status on antiretroviral therapy adherence, total (direct and indirect paths combined) and direct alone, respectively. The indirect effects of racial/ethnic minority status on antiretroviral therapy adherence are listed by hypothesized mediator(s). There were a total of 9 significant indirect effects of racial/ethnic minority status on antiretroviral therapy adherence: 6 via discriminatory health care experiences, 1 via health care provider distrust, and 2 via medication efficacy beliefs. The strongest indirect effects came through medication efficacy beliefs (β = 0.04, P = 0.031) and the path between medication efficacy beliefs and psychological burden of medication (β = −0.02, P = 0.003). These 2 indirect effects also accounted for the largest proportion of the total effect of racial/ethnic minority status on antiretroviral therapy adherence (11.7%). Although these indirect pathways were statistically significant, the magnitudes of these effects were negligible. The direct effect (β = −0.21) is essentially the same as the total effect (β = −0.20), indicating lack of mediation from the hypothesized factors.
An observed relationship between racial/ethnic minority status and antiretroviral therapy adherence among a national sample of HIV-infected patients was not explained by discriminatory health care experiences, health care provider distrust, and other patient-level variables. Although the SEM analyses controlling for multiple covariates identified several statistically significant pathways leading from minority status to adherence involving one or more hypothesized patient-level mediators, the magnitude of each indirect effect was negligible (generally less than 0.01) and the sum of these effects was not statistically significant. Racial/ethnic minority patients generally reported more negative attitudes and beliefs toward antiretroviral therapy adherence than did white patients, which is consistent with findings of other studies.37,38 However, this association did not explain a significant proportion of the disparity in antiretroviral therapy adherence. The persistence of racial/ethnic minority status as a significant predictor in multivariate analyses of antiretroviral therapy adherence is consistent with other studies.1-4 Omitted variables such as health literacy and housing instability may have produced underestimated mediation effects. Given the many covariates, however, it is unlikely that the results would be dramatically different.
The lack of mediation may be partially explained by the fact that racial/ethnic minority patients reported less discriminatory health care experiences than white patients and had equally low levels of health care provider distrust. The unexpected finding that discriminatory health care experiences were reported more often by nonminority HIV-infected patients was also seen in a previous HCSUS study by Schuster et al53 but is in contrast with research by Bird et al31 who found no differences by racial/ethnic minority status. Possible explanations include underreporting by racial/ethnic minority participants, more opportunities to be discriminated against by white patients who were diagnosed with HIV longer, and racial/ethnic minority patients receiving care in more welcoming environments. Our findings on health care provider distrust are consistent with previous research that also did not show trust among HIV-infected patients varying by racial/ethnic minority status.2,32,54 In the face of a life-threatening illness, most racial/ethnic minority and white HIV-infected patients may more readily trust those who provide critical medical care.
Another reason for the weak indirect effects between minority status and adherence may be model misspecification. Although a strong conceptual and empirical rationale underlays the proposed mediation model, minority status may be better conceptualized as a contextual or moderating factor than as a predictor. The HIV-infected patients in this study who were also members of racial/ethnic minority groups lived in different and arguably more difficult circumstances (being younger, less educated, and poorer) than did the white patients. Such circumstances could produce different patterns of relationships among discriminatory health care experiences, health care provider distrust, antiretroviral therapy-related attitudes, and adherence. As either a predictor or a contextual factor, however, racial/ethnic minority status has meaning and consequence that may not be captured by the individual-level variables traditionally included in antiretroviral therapy adherence research.
The effect of discriminatory health care experiences on antiretroviral therapy adherence in this sample was entirely indirect through participants' increased distrust of their health care providers and their weakened beliefs in the worth of antiretroviral therapy. Discriminatory health care experiences may subtly color HIV-infected patients' perspectives on their care providers and the treatment they recommend. This finding contrasts with the results of the only other study on the relationship between discriminatory health care experiences and antiretroviral therapy adherence, which was conducted with a relatively small convenience sample and examined only unadjusted bivariate correlates of adherence.31 The present study, with its large national sample and multiple covariates, offers an alternate view of how discriminatory health care experiences may affect self-reported antiretroviral therapy adherence.
The positive relationship between distrust and antiretroviral therapy adherence was unexpected given that a previous research with HIV-infected patients has found that greater trust in one's health care provider is associated with better adherence. In focus groups and interviews, HIV-infected patients have drawn connections between the trust they have in their health care providers and their willingness to comply with treatment recommendations.55,56 Our findings seem to indicate the opposite: that negative expectations of their health care providers may increase HIV-infected patients' vigilance about monitoring their care among the minority of patients who are distrustful. Ford and colleagues (in press), for example, found that African American sexually transmitted disease patients who perceived more racism were more likely to get tested for HIV.57 The HIV-infected patients in the present study may feel that their health care providers are not looking out for their best interests, and so they must be extra vigilant about the care that they receive. Previous research has not examined the extent to which health care provider distrust influences treatment-related attitudes and beliefs. The indirect effects of health care provider distrust on antiretroviral therapy adherence occurred as a result of increases in participants' psychological distress about having to take antiretroviral therapy and of weakened beliefs in the worth of antiretroviral therapy. HIV-infected patients who do not feel their health care providers has their best interests in mind may feel less convinced of the benefits of recommended treatment. Because the items of health care provider distrust and antiretroviral therapy attitudes and beliefs used in this analysis were measured at the same point in time, it is possible that the effect occurred in the opposite direction. That is, participants' negative feelings about antiretroviral therapy may have increased their distrust in their health care providers for encouraging use of the regimen in the first place.
Other findings of note relate to unhypothesized or unexpected associations (or lack thereof) among attitudes and beliefs about antiretroviral therapy and self-reported adherence. First, the difficulty participants had fitting antiretroviral therapy into their daily lives showed the strongest direct effect on self-reported antiretroviral therapy nonadherence. Being able to take one's medications despite disruptions is a key task for HIV-infected patients and is consistently associated with antiretroviral therapy adherence.33 Second, contrary to the study hypothesis, participants' perceptions that antiretroviral therapy is difficult to obtain was not significantly associated with adherence in the final model. Not being able to reliably obtain antiretroviral therapy because of financial problems, time pressures, or lack of transportation will interrupt patients' attempts to be adherent to their medication regimens. The lack of relationship between access difficulties and adherence was likely explained by the inclusion of financial resources (eg, annual income and insurance status) and social support measures, important determinants of antiretroviral therapy access36 that were included as covariates in the final model. Finally, greater skepticism about the benefits of antiretroviral therapy was associated with better adherence. This may be a result of patient vigilance, as described above, with the positive association between greater health care provider distrust and adherence.
The results of this study must be interpreted in light of its limitations. First, the data for HCSUS were collected when antiretroviral therapy first came into widespread use in the mid-1990s, heralding a transformation in HIV care and producing optimism about its possible effects.58,59 The context of participant responses in this study, however, may not reflect HIV-infected patients' current experiences with discriminatory health care experiences, health care provider distrust, or antiretroviral therapy adherence. Second, self-reported adherence measures may be subject to recall and social desirability biases.33,60 Third, only racial/ethnic minority status and discriminatory health experiences were assessed before antiretroviral therapy adherence, so causality among other factors cannot be taken for granted. In addition, other conceptualizations of the directionality of relationships among factors may fit the data equally well. Fourth, participants were asked to recall incidents of discriminatory health care experiences “since they had HIV”-which could have occurred at any time before the study. The variability in this measure may have underestimated the effect of discriminatory health care experiences on antiretroviral therapy attitudes, beliefs, and behavior. Fifth, attrition was more likely among members of racial/ethnic minority groups, and antiretroviral therapy use was less likely among participants who were more distrustful of their health care providers. These attrition and selection biases reduce the generalizability of the findings to all HIV-infected patients. Finally, members of racial/ethnic minority groups were collapsed into one category for these analyses because of the small sample sizes among some subpopulations (Asian/Pacific Islanders and Native Americans). We acknowledge, however, that each racial/ethnic minority group undoubtedly has unique life experiences that affect its health care and health outcomes12 and group members' understanding of discriminatory health care experiences and health care provider distrust.
Much is known about the correlates of antiretroviral therapy adherence but not about the factors that produce disparities by racial/ethnic minority status. Current guidelines encourage regular adherence counseling by health care providers for all HIV-infected patients,61 but antiretroviral therapy adherence may require more tailored messages to more targeted audiences. Our findings suggest that a simplistic approach focusing on patient-level factors, such as discrimination, distrust, or treatment-related attitudes and beliefs, may do little to reduce observed racial/ethnic disparities in antiretroviral therapy adherence. Following multifactor and multilevel models of adherence,36 patient-centered care for racial/ethnic minority HIV-infected individuals should balance interventions aimed at individual beliefs, needs, and choices with environmental approaches that address the settings in which care is delivered.24
The authors thank Barbara Turner, Giselle Corbie-Smith, and Robert DeVellis for their earlier comments on this work, Sharon Christ for her technical advice, and the anonymous reviewers for their feedback.
1. Gifford AL, Bormann JE, Shively MJ, et al. Predictors of self-reported adherence
and plasma HIV concentrations in patients on multidrug antiretroviral regimens. J Acquir Immune Defic Syndr
2. Golin CE, Liu H, Hays RD, et al. A prospective study of predictors of adherence
to combination antiretroviral medication. J Gen Intern Med
3. Gordillo V, del Amo J, Soriano V, et al. Sociodemographic and psychological variables influencing adherence
to antiretroviral therapy
4. van Servellen G, Chang B, Garcia L, et al. Individual and system level factors associated with treatment nonadherence in human immunodeficiency virus-infected men and women. AIDS Patient Care STDS
5. Wenger N, Gifford A, Liu H, et al. Patient characteristics and attitudes associated with antiretroviral adherence
. Annual Conference on Retroviruses and Opportunistic Infections
. Chicago, IL; 1999. Abstract no. 98.
6. Fogarty L, Roter D, Larson S, et al. Patient adherence
to HIV medication regimens: a review of published and abstract reports. Patient Educ Couns
7. Paterson DL, Swindells S, Mohr J, et al. Adherence
to protease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med
8. Wutoh AK, Brown CM, Kumoji EK, et al. Antiretroviral adherence
and use of alternative therapies among older HIV-infected adults. J Natl Med Assoc
9. Corbie-Smith G, Thomas SB, Williams MV, et al. Attitudes and beliefs of African Americans toward participation in medical research. J Gen Intern Med
10. Gamble VN. Under the shadow of Tuskegee: African Americans and health care. Am J Public Health
11. Jones CP. Invited commentary: “race,” racism, and the practice of epidemiology. Am J Epidemiol
. 2001;154:299-304; discussion 305-296.
12. Williams DR. Race and health: basic questions, emerging directions. Ann Epidemiol
13. Lillie-Blanton M, Brodie M, Rowland D, et al. Race, ethnicity, and the health care system: public perceptions and experiences. Med Care Res Rev
. 2000;57(Suppl 1):218-235.
14. Piette JD, Bibbins-Domingo K, Schillinger D. Health care discrimination, processes of care, and diabetes patients' health status. Patient Educ Couns
15. Trivedi AN, Ayanian JZ. Perceived discrimination and use of preventive health services. J Gen Intern Med
16. Boulware LE, Cooper LA, Ratner LE, et al. Race and trust in the health care system. Public Health Rep
17. Corbie-Smith G, Thomas SB, St George DM. Distrust, race, and research. Arch Intern Med
18. Doescher MP, Saver BG, Franks P, et al. Racial and ethnic disparities
in perceptions of physician style and trust. Arch Fam Med
19. Schnittker J. Social distance in the clinical encounter: interactional and sociodemographic foundations for mistrust in physicians. Soc Psychol Q
20. Taira DA, Safran DG, Seto TB, et al. Do patient assessments of primary care differ by patient ethnicity? Health Serv Res
. 2001;36(pt 1):1059-1071.
21. Bogart LM, Bird ST, Walt LC, et al. Association of stereotypes about physicians to health care satisfaction, help-seeking behavior, and adherence
to treatment. Soc Sci Med
22. Smedley BD, Stith AY, Nelson AR, eds. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health
. Washington, DC: National Academy of Sciences; 2003.
23. McKinney MM, Marconi KM, Cleary PD, et al. Delivering HIV services to vulnerable populations: an evaluation and research agenda. Public Health Rep
24. Stone VE. Optimizing the care of minority patients with HIV/AIDS. Clin Infect Dis
25. Beach MC, Keruly J, Moore RD. Is the quality of the patient-provider relationship associated with better adherence
and health outcomes for patients with HIV? J Gen Intern Med
26. Bakken S, Holzemer WL, Brown MA, et al. Relationships between perception of engagement with health care provider and demographic characteristics, health status, and adherence
to therapeutic regimen in persons with HIV/AIDS. AIDS Patient Care STDS
27. Demmer C. Relationship with health care provider and adherence
to HIV medications. Psychol Rep
28. Ingersoll KS, Heckman CJ. Patient-clinician relationships and treatment system effects on HIV medication adherence
. AIDS Behav
29. Russell J, Krantz S, Neville S. The patient-provider relationship and adherence
to highly active antiretroviral therapy
. J Assoc Nurses AIDS Care
30. Schneider J, Kaplan SH, Greenfield S, et al. Better physician-patient relationships are associated with higher reported adherence
to antiretroviral therapy
in patients with HIV infection. J Gen Intern Med
31. Bird ST, Bogart LM, Delahanty DL. Health-related correlates of perceived discrimination in HIV care. AIDS Patient Care STDS
32. Altice FL, Mostashari F, Friedland GH. Trust and the acceptance of and adherence
to antiretroviral therapy
. J Acquir Immune Defic Syndr
33. Chesney M. Adherence
to HAART regimens. AIDS Patient Care STDS
34. Viswanathan H, Anderson R, Thomas J. Evaluation of an antiretroviral medication attitude scale and relationships between medication attitudes and medication nonadherence. AIDS Patient Care STDS
35. Richter DL, Sowell RL, Pluto DM. Attitudes towards antiretroviral therapy
among African American women. Am J Health Behav
36. Ickovics JR, Meade CS. Adherence
to antiretroviral therapy
among patients with HIV: a critical link between behavioral and biomedical sciences. J Acquir Immune Defic Syndr
. 2002;31(Suppl 3):S98-S102.
37. Ferguson TF, Stewart KE, Funkhouser E, et al. Patient-perceived barriers to antiretroviral adherence
: associations with race. AIDS Care
38. Siegel K, Karus D, Schrimshaw E. Racial differences in attitudes toward protease inhibitors among older HIV-infected men. AIDS Care
39. Frankel MR, Shapiro MF, Duan N, et al. National probability samples in studies of low-prevalence diseases. Part II: designing and implementing the HIV cost and services utilization study sample. Health Serv Res
. 1999;34(pt 1):969-992.
40. Shapiro MF, Berk ML, Berry SH, et al. National probability samples in studies of low-prevalence diseases. Part I: perspectives and lessons from the HIV cost and services utilization study. Health Serv Res
. 1999;34(pt 1):951-968.
41. Duan N, McCaffrey DF, Frankel MR, et al. HCSUS Baseline Methods Technical Report: Weighting, Imputation and Variance Estimation (Report MR-1060-AHCPR)
, Santa Monica, CA: RAND; 1999.
42. DeVellis RF. Scale Development: Theory and Applications
. 2nd ed. Thousand Oaks, CA: Sage Publications; 2003.
43. Lu M, Safren SA, Skolnik PR, et al. Optimal recall period and response task for self-reported HIV medication adherence
. AIDS Behav
44. Walsh JC, Mandalia S, Gazzard BG. Responses to a 1 month self-report on adherence
to antiretroviral therapy
are consistent with electronic data and virological treatment outcome. AIDS
[computer program]. Version 8.0. College Station, TX; STATA Corporation, 2003.
[computer program]. Version 3.11. Los Angeles, CA; 2004.
47. Kline RB. Principles and Practice of Structural Equation Modeling
. 2nd ed. New York: The Guilford Press; 2005.
48. Hoyle R, Smith G. Formulating clinical research hypotheses as structural equation models: a conceptual overview. J Consult Clin Psychol
50. Cohen J. Statistical power analysis for the behavioral sciences
. 2nd ed. Hillsdale, NJ: Lawrence Earlbaum Associates; 1998.
51. Hu L, Bentler PM. Fit indices in covariance structure modeling: sensitivity to underparameterized model misspecification. Psychol Methods
52. Muthén LK, Muthén BO. Mplus user's guide
. 3rd ed. Los Angeles, CA: Muthén & Muthén; 2005.
53. Schuster MA, Collins R, Cunningham WE, et al. Perceived discrimination in clinical care in a nationally representative sample of HIV-infected adults receiving health care. J Gen Intern Med
54. Whetten K, Leserman J, Whetten R, et al. Exploring lack of trust in care providers and the government as a barrier to health service use. Am J Public Health
55. Roberts KJ. Physician-patient relationships, patient satisfaction, and antiretroviral medication adherence
among HIV-infected adults attending a public health clinic. AIDS Patient Care STDS
56. Stone VE, Clarke J, Lovell J, et al. HIV/AIDS patients' perspectives on adhering to regimens containing protease inhibitors. J Gen Intern Med
57. Ford C. Perceived racism, neighborhood characteristics and human immunodeficiency virus (HIV) testing behaviors among African American community residents and transmitted disease (STD) clinic patients
. Chapel Hill, NC: Health Behavior & Health Education, University of North Carolina; 2005.
58. Kelly JA, Kalichman SC. Behavioral research in HIV/AIDS primary and secondary prevention: recent advances and future directions. J Consult Clin Psychol
59. Kelly JA, Otto-Salaj LL, Sikkema KJ, et al. Implications of HIV treatment advances for behavioral research on AIDS: protease inhibitors and new challenges in HIV secondary prevention. Health Psychol
60. Turner BJ. Adherence
to antiretroviral therapy
by human immunodeficiency virus-infected patients. J Infect Dis
. 2002;185(Suppl 2):S143-S151.
61. Centers for Disease Control and Prevention. Guidelines for using antiretroviral agents among HIV-infected adults and adolescents: recommendations of the Panel on Clinical Practices for Treatment of HIV. MMWR