The results of this study indicate that a significant racial disparity exists in HIV medication adherence. In our study population of 7034 patients, black race was significantly associated with decreased odds of ART adherence. Although almost 40 % of non-black patients achieved optimal adherence over a one-year period, less than 30 % of black patients were able to do so. In fact, black patients were consistently 30 % less likely to achieve optimal adherence compared to non-black patients, even after adjusting for a number of covariates. While previous smaller studies presented inconsistent results, 9–16
our study indicates that a true racial disparity in antiretroviral medication adherence is likely to be present in the HIV-infected population. In part, these racial disparities may be attributed to findings that show black patients are more likely to postpone medical care, 6
have less access to care, 30
and less trust in healthcare providers than white patients.42
Although this study found a high prevalence rate of depression among HIV-infected patients, our analysis did not find that comorbid depression moderated the association between race and adherence during the one-year study period. In other words, the presence of depression did not further enhance the already existing racial disparity between blacks and non-blacks with regards to ART adherence. However, this finding should be interpreted with caution since there is evidence that African Americans face a disparity in both diagnosis 43
and treatment of depression compared to whites.44
Thus, while African Americans are generally reported to have lower rates of depression than whites, 45,46
true prevalence may actually be higher.
We also saw that patients categorized as having depression were slightly more adherent to their HIV medication than those who were not. Initially, this may seem contradictory since the presence of comorbid depression generally worsens outcome measures. However, it seems that the improvement in adherence may be attributed to the receipt of antidepressant medication. This is consistent with two previous studies.19,27
It is understandable that this may be due to the improved emotional and mental function provided by antidepressant treatments. It may also represent a proxy for the quality of the physician–patient relationship, i.e., patients with a better relationship with their physician are more likely to have their depressive symptoms recognized and treated for.
In our analysis, depressed patients who did not receive antidepressant medication not only had worse adherence than their medicated counterparts, but also had lower adherence compared to the non-depressed patients. This implies that depression does have the expected negative impact on ART adherence if left untreated. Conversely, if depression symptoms are treated for, a patient's adherence to ART; thus, overall disease management, may in fact improve despite the initial negative connotation that comes with being diagnosed with depression. For this reason, we believe that future studies on ART adherence should include not only depression as a covariate, but also additional details on the treatment of depression, since this can significantly influence the interpretation and implication of study results. Furthermore, these findings underscore the importance of mental health evaluation in the successful management of HIV infection. Healthcare providers must increase their vigilance for symptoms of depression in HIV-infected patients and offer appropriate treatment as this may not only effectively treat those symptoms, but also improve their ART adherence.
This study was subject to several limitations. Given the inherent limits of all administrative claims data studies, causality cannot be established. Additionally, medical conditions were identified using ICD-9 and NDC codes, which if recorded inaccurately, may lead to incorrect patient classification. Data such as laboratory values and clinical outcomes are unavailable in administrative claims data and could not be included in the analysis. Since prescription claims data were used to assess this adherence, it was not possible to verify whether a patient actually consumed the filled prescription. Thus, it was assumed that any prescription filled was also consumed. Administrative claims data is also prone to omitted variable bias since many personal, social, and physician factors (such as severity of depression, socioeconomic status, education level, stress, social support, and physician trust) could not be controlled for. In an attempt to address this bias, we used proxy variables in place of some covariates when possible. However, proxy variables are not true replacements and may be prone to covariate misidentification.
To identify evidence of depression, both ICD-9 codes for depression and NDC codes for antidepressants were used. This method may improve depression detection, since it is generally under-diagnosed, 47
but could also overestimate the true number of patients with depression. Nevertheless, we used this method because a study using multi-state Medicaid claims found that compared to using either diagnostic code or prescription medication claims, the combination of both yielded much better overall performance in correctly classifying depression and other chronic illnesses.48
Furthermore, we only classified a patient as having evidence of depression if he/she did not also have a diagnosis for anxiety disorder or other disorders that would be treated with antidepressants. Finally, since we examined HIV-infected individuals from multi-state Medicaid claims, study results may not be generalizable to other HIV-infected populations.
Despite these limitations, this study's large sample size allowed for robust quantitative analysis of the association between race, depression, and antiretroviral medication adherence. Our results demonstrate that black patients are at a disadvantage compared to other races when it comes to ART adherence, and that these racial differences persist independent of whether depression is also present. Disease management programs within Medicaid could be especially beneficial to disadvantaged groups by providing customized disease education and self-management strategies with sensitivity to cultural and socioeconomic differences. With clear evidence that adherence is crucial in the successful management of HIV infection and prevention of disease progression, continued improvement in our understanding of the factors associated with any racial disparity is of utmost importance. Prospective studies using clinical and qualitative assessments may provide further insight into the causes of the racial disparities seen in ART adherence, and shed light on possible solutions to successfully target vulnerable HIV-infected populations.