Correlates of Antiretroviral and Antidepressant Adherence Among Depressed HIV-Infected Patients

Abstract

Although crucial for efficacy of pharmacotherapy, adherence to prescribed medication regimens for both antiretrovirals and antidepressants is often suboptimal. As many depressed HIV-infected individuals are prescribed both antiretrovirals and antidepressants, it is important to know whether correlates of nonadherence are similar or different across type of regimen. The HIV Translating Initiatives for Depression into Effective Solutions (HI-TIDES) study was a single-blinded, longitudinal, randomized controlled effectiveness trial comparing collaborative care to usual depression care at three Veterans Affairs HIV clinics. The current investigation utilized self-report baseline interview and chart-abstracted data. Participants were 225 depressed HIV-infected patients who were prescribed an antidepressant (n=146), an antiretroviral (n=192), or both (n=113). Treatment adherence over the last 4 days was dichotomized as “less than 90% adherence” or “90% or greater adherence.” After identifying potential correlates of nonadherence, we used a seemingly unrelated regression (SUR) bivariate probit model, in which the probability of adherence to HIV medications and the probability of adherence to antidepressant medications are modeled jointly. Results indicated that 75.5% (n=146) of those prescribed antiretrovirals reported 90%-plus adherence to their antiretroviral prescription and 76.7% (n=112) of those prescribed antidepressants reported 90%-plus adherence to their antidepressant prescription, while 67% of those prescribed both (n=113) reported more than 90% adherence to both regimens. SUR results indicated that education, age, and HIV symptom severity were significant correlates of antiretroviral medication adherence while gender and generalized anxiety disorder diagnosis were significant correlates of adherence to antidepressant medications. In addition, antiretroviral adherence did not predict antidepressant adherence (β=1.62, p=0.17), however, antidepressant adherence did predict antiretroviral adherence (β=2.30, p<0.05).

Introduction

D espite ongoing advances in pharmacotherapy in the treatments of depression and HIV, many depressed HIV-infected individuals experience only partial response due to nonadherence to prescribed regimens. Indeed, depressive symptoms have been repeatedly identified as a consistent, yet mutable, barrier to medical treatment adherence.^1,2 Moreover, depression has been identified as a common barrier to HIV treatment adherence.³ Yet, despite advances in antidepressant treatment, there is evidence that depressed patients (both HIV-infected and non-HIV–infected) have trouble adhering to prescribed antidepressant regimens.^4,5 Although past research has explored demographic, behavioral, and health-related barriers to adherence to depression and HIV treatment separately, limited research has examined whether common or distinct factors are barriers for individuals being prescribed both types of medications. Examination of such variables will likely add to providers' knowledge of treating the “whole” patient, i.e., examination of psychosocial and medical factors that exist for HIV-infected individuals.⁶ The present investigation focuses on barriers to adherence for both antidepressant and antiretroviral medications among depressed HIV-infected individuals prior to their enrollment in an intervention trial aimed at improving depression treatment and outcomes among treatment-seeking HIV-infected individuals.

To date, the literature has examined barriers of adherence to antiretrovirals and antidepressants separately. Extensive research has been conducted on identifying barriers to HIV treatment adherence given the virologic consequences of non-adherence (i.e., virologic failure and development of resistance). The most common factors found to impact HIV treatment adherence include complexity of treatment regimens, side effect profiles, excessive substance use, psychopathology (e.g., depression and anxiety), beliefs about medication, self-efficacy, social support, coping style, and memory lapse.⁷ In contrast to HIV treatment adherence, research on antidepressant adherence is relatively sparse. Identified barriers to antidepressant adherence have included younger age, female gender, first episode versus recurrent depression, low self-efficacy, lower levels of formal education, severity of symptoms, side effect profiles, beliefs about medication-taking behavior, regimen characteristics, and forgetting.^8

–12 Finally, given evidence that many individuals require maintenance treatment beyond remission of depressive symptoms,¹³ it is of concern that several investigators have identified premature discontinuation of treatment as a more severe form of nonadherence. One retrospective chart review indicated that only 44% of their sample completed 6 months of treatment, suggesting that the majority of patients do not receive the full benefit of the antidepressants.¹⁴ Given the illness profiles of both depression and HIV, research is needed to identify barriers to adherence for both antidepressant and antiretroviral medications.

More recently, research has begun to emerge examining adherence patterns among depressed HIV-infected patients. Several investigations have demonstrated improved HIV treatment adherence outcomes with use of antidepressant treatment. First, there was preliminary data from a retrospective chart review suggesting that adherence to antiretrovirals was positively correlated with antidepressants adherence (r=0.31; p<0.02) and of improved virologic responses given evidence of adherence to antidepressants (>80%).¹⁵ Second, several investigations have provided evidence of improved adherence to antiretrovirals among depressed individuals taking antidepressants.^5,16,17 Walkup and colleagues¹⁷ reported that prescription of an antidepressant in a prior month increased the odds of adherence to antiretroviral in the subsequent month. Furthermore, Horzberg and colleagues⁵ attempted to unpack the relationship between depression, adherence to antiretrovirals, and HIV treatment adherence. They reported that depressed HIV-infected individuals who were highly adherent (>90%) to their antidepressants had similar antiretroviral adherence profiles to nondepressed individuals and that both of these groups had significantly better antiretroviral adherence than those who were both depressed and nonadherent to their antidepressants. Finally, there is evidence that antidepressant treatment also helps improve adherence to complex antiretroviral regimens.¹⁸ However, given evidence that antidepressant and antiretroviral adherence are related, it would be of interest to identify factors that predict adherence to both types of medication.

This investigation is part of the larger, three-site, two-arm, single-blinded, randomized controlled effectiveness trial, HIV Translating Initiatives for Depression into Effective Solutions (HI-TIDES).¹⁹ The goal of the HI-TIDES trial was to compare a collaborative care intervention²⁰ to treatment as usual (TAU) in the treatment of depression for depressed HIV clinic patients. The larger trial involved an intervention utilizing an offsite HIV depression care team (registered nurse depression care manager, pharmacist, and psychiatrist) who delivered up to 12 months of collaborative depression care supported by a Web-based decision support system. However, given that medication was the primary mode of depression treatment within the trial, adherence to prescribed regimens was essential. Therefore, the aim of the current investigation was to isolate demographic, mental health, and physical health related factors that were associated with preintervention adherence patterns. Identification of these factors will be helpful for future efforts to improve provision of physical and mental health interventions for HIV-infected veterans.

Methods

Participants

Eligible participants were HIV-seropositive males and females (aged 18 and over), who were being treated for HIV at one of three Veterans Administration Medical Centers (VAMC) HIV clinics and were identified as having clinically significant symptoms of depression during the screening session (i.e., Patient Health Questionnaire [PHQ-9]²¹ depression score≥10). Exclusion criteria were: no access to a telephone, current acute suicidal ideation, significant cognitive impairment as indicated by a score greater than 10 on the Blessed Orientation-Memory-Concentration (BOMC),²² self-report history of bipolar disorder or manic depression, and medical record diagnosis of schizophrenia. After completing the informed consent process, participants completed the baseline assessment and were randomly assigned to intervention or usual care. Two hundred seventy-six individuals were randomized at the baseline research session and 225 are included in these analyses as they had a baseline prescription for either antiretroviral treatment (n=192), antidepressant treatment (n=146), or both (n=113). All participants signed informed consents approved by their VAMC's Institutional Review Board (IRB) prior to the initiation of any research procedures.

Measures

Demographics

The baseline interview included questions about the veteran's demographics characteristics. Data utilized in these analyses included gender, age, race, marital/partner status, and level of formal education.

Patient Health Questionnaire—9 items (PHQ-9)²¹

The PHQ-9 is a 9-item self-report measure that was used to screen for the presence of symptoms of depression. A PHQ-9 score of greater than 10 has strong psychometric properties in primary care settings (e.g., 99+% sensitivity and a 91% specificity).

Symptom Checklist—20 items (SLC-20).²³

Depression symptom severity over the 2 weeks before the baseline interview was measured using the Hopkins Symptom Checklist SCL-20. The SCL-20 includes the 13-item depression scale plus 7 depression-related items from the Hopkins Symptom Checklist-90–Revised. The items are scored from 0 to 4 and averaged to provide a mean depression severity score from 0 to 4.

Mental health diagnoses.

The Mini International Neuropsychiatric Interview (MINI)²⁴ is a brief structured interview for the major Axis I psychiatric disorders, shown to be valid and reliable when compared to the Structured Clinical Interview for DSM-III-R and the CIDI.²⁵ The MINI was used to assess for the presence of clinically significant depression (i.e., major depression) and also for comorbid mental health conditions (i.e., generalized anxiety disorder, panic disorder, posttraumatic stress disorder, alcohol use disorder, any alcohol use over the last year and number of drinks in week before interview).

Physical comorbidity.

Chronic physical health conditions (other than HIV) were measured using the 21-item Physical Comorbidity scale from the Depression Outcomes Module.^26,27

HIV-related symptomatology.

HIV symptom severity was measured using the 20-item Symptoms Distress Module,²⁸ which summarizes the degree to which each symptom bothered the participant in the past 4 weeks on a scale from 0=I do not have this symptom to 4=this symptom bothers me a lot. We also created a count variable to address the number of self-reported symptoms.

Quality of life.

Health status was measured using the physical and mental health component summary scores from the Medical Outcomes Study SF-12V.²⁹ Health-related quality of life was measured using the Quality of Well-Being self-administered scale (QWB-SA).^30,31 The QWB-SA score is derived from general population preference weights and ranges from death (0.0) to perfect health (1.0).

Self-reported medication adherence.

Antidepressant and HIV medication adherence were measured separately using the AIDS Clinical Trial Group assessment,³² which asks participants to report the number of pills per day they are supposed to take and the number of pills they skipped taking for each medication for each of the past 4 days. Percent adherence was calculated as follows. First, for each of the last 4 days, the number of pills prescribed minus number of pills taken divided by the total number prescribed for each medication was calculated. Then, add the percentage adherence for each of the last 4 days and divide by 4; this algorithm allowed us to take into account the number of pills and to better account for the “weight” of a missed pill (e.g., missing 1 pill of a 2-pill regimen likely has a bigger influence on efficacy than missing 1 pill of a 6-pill regimen). However, the distributions for both antidepressant and antiretroviral adherence were skewed as approximately 75% of the sample reported 100% adherence. The distributions were not amenable to transformation and as such, we decided to dichotomize the data in “less than 90%” and “90% or greater.” This cut-point was chosen as our recent work demonstrated that 90% adherence was the most sensitive cut-point for antidepressant adherence.³³ Although 95% adherence is a long accepted benchmark among HIV treatment adherence research,³⁴ more recent research indicates that level of adherence varies by regimen and resistance profile.^35,36 We chose 90% for both types of medication for consistency in analyses.

Patient knowledge of regimen.

As the adherence data were based on self-report, we did a chart review for medications prescribed to examine patient knowledge of regimen. Previous research has identified that poor knowledge of one's regimen can be associated with nonadherence to HIV medications.³⁷ As such, we included this in our examination of possible correlates of adherence to both types of medication; however, due to limitations in our data, we were only able to compare knowledge of names of medications. We measured knowledge in two ways. First, we measured if they knew the correct number of antidepressant and antiretroviral medications that they were prescribed. Second, we coded the names of the medications so that we could compare patient's knowledge of both the number and names of their prescribed antidepressant and antiretroviral medications to the ones reflected in their chart (chart was assumed to be gold standard). We calculated a percentage by dividing the number of correctly identified medications by the number that were prescribed to them as noted in their chart (e.g., chart review said they were prescribed Med A, Med B, and Med C; Veteran reported Med A and Med C but NOT Med B; their knowledge percentage would be 66.7%.)

Procedure

Veterans were screened for depressive symptoms by clinic staff during routine clinical care visits with their HIV provider. Veterans who fulfilled the screening criteria were referred to research to learn about the larger HI-TIDES trial. If interested, they completed the informed consent procedure at that visit and were called by research staff, on average, 7 days later, completed their baseline interview. The baseline interview contained questions about demographics, physical and mental health symptoms, treatment history, treatment preferences, and self-reported treatment adherence for both antiretrovirals and antidepressants. This interview was completed before the veteran was informed about which arm they were randomized to (intervention versus TAU). Following completion of the baseline interview, research staff conducted a chart review of the participant's electronic medical record at the VA to gather information about their current prescriptions and comorbid health conditions.

Data analytic strategy

Analyses were conducted using SAS 9.2³⁸ (SAS Inc., Cary, NC) and Stata 9.0 (StataCorp, College Station, TX).³⁹ There were three steps in the analyses to identify predictors, run the seemingly unrelated bivariate probit (SUR) model, and finally, to add adherence as an explanatory variable. Preliminary analyses to examine the relationship between participant characteristics and adherence (separate analyses for antidepressant and antiretroviral adherence) were conducted using appropriate tests based on the distribution of the item. Independent variables that were significant at the bivariate level p<0.20 with either antidepressant or antiretroviral adherence were included as variables in the final models. To examine correlates of adherence, we used an SUR model, where the probability of adherence to HIV medications and the probability of adherence to antidepressant medications are modeled jointly. The SUR model specifically accounts for the possibility that the unmeasured factors affecting adherence to HIV medications also affect adherence to antidepressant medications, which leads to correlated error terms across the two probit regression equations. These unmeasured factors include omitted variables impacting adherence to antidepressant and HIV medications, and the measurement error common to assessing adherence to antidepressant medication and adherence to HIV medications. The SUR model assumes this correlation follows a bivariate normal distribution (with covariance ρ) and calculates whether the error terms are significantly correlated and the direction of the correlation.^40,41 The significance of the correlation coefficient ρ is tested using a likelihood ratio test that compares the log likelihood of the model where ρ is restricted to 0 to the log likelihood of the model where ρ is unrestricted. If the correlation is not statistically significant (e.g., ρ=0), it implies that unmeasured factors influencing adherence to HIV medications and the unmeasured factors influencing adherence to antidepressant medications are different, and therefore that the two decisions are likely made independent of one another. In this case, the most appropriate statistical analysis involves estimating the parameters of two separate probit regressions. On the other hand, if the correlation is positive and significant, it implies that unmeasured factors affect the adherence to antiretroviral and antidepressant medications in the same way. Conversely, if the correlation is negative and significant, it suggests that the unmeasured factors affect adherence to HIV medications and antidepressant medications in opposite directions. In either case (significantly positive or negative correlated error terms), a significant correlation suggests that the decision to take HIV and antidepressant medications as prescribed are interdependent and should be modeled jointly using the seemingly unrelated bivariate probit model. Once the proper model specification (two independent probit models or seemingly unrelated bivariate probit model) was determined, significant correlates of nonadherence were examined based on the significance of the parameter estimates of the independent variables. The direction and magnitude of the correlate effects was determined by calculating standard marginal effects.⁴² Finally, two additional models were run to examine the impact of each type of adherence on the other. When adherence for one type of medication is included as an explanatory variable in the SUR equation predicting adherence for the other type of medication, it is referred to as the bivariate probit with endogenous dummy model⁴¹ or the recursive model for dichotomous choice.⁴³ For the model predicting antidepressant adherence to be fully identified, the probit equation includes one exogenous independent variable (depression severity) that is not included as independent variables in the probit equation predicting antiretroviral adherence. Likewise, for the model predicting antiretroviral adherence, to be fully identified, the probit equation includes one exogenous independent variable (HIV severity) that is not included as independent variables in the probit equation predicting antidepressant adherence (Fig. 1).

FIG. 1.

Seeming unrelated regression (SUR) model. β, parameter estimates; ρ, correlation between the error terms; X, independent variables; ɛ, error terms.

Results

Sample characteristics

The sample consisted of 225 depressed HIV-infected veterans. Overall, the majority of the sample was male (97%), middle-aged (50±10 years), and had a high school diploma (93%). Approximately 60% of the sample was African American, which is similar to the U.S. HIV population.⁴⁴ Over three fourths of the participants met criteria for major depressive disorder based on the MINI. Moreover, 75% of the sample also met criteria for at least 1 other comorbid mental health disorder. The participants reported a range of comorbid physical health conditions. The sample self-reported high rates of adherence to both their antidepressants and their antiretrovirals. Results indicated that 75.5% of those prescribed antiretrovirals (n=192) reported 90% or greater adherence to their antiretroviral prescription while 76.7% of those prescribed antidepressants (n=146) reported 90% or greater adherence to their antidepressant prescription (Table 1).

Table 1.

Overview of Treatment Adherence and Medications Prescribed

	Antiretroviral medication (n =192)	Antidepressant medication (n =146)
	Mean±SD; range	Mean±SD; range
Treatment adherence
Antiretroviral adherence (%)	87.7±25.7; range=0–100
Antidepressant adherence (%)		84.3±32.3; range=0–100
No. of medications prescribed
Antiretrovirals—self-report	2.4±1.1; range=0–6
Antiretrovirals—chart review	2.1±1.3; range=0–6
Antidepressants—self-report		1.1±0.4; range=0–3
Antidepressants—chart review		0.9±0.6; range=0–3

SD, standard deviation.

Among the subset of participants with both prescriptions (n=113), adherence rates were as follows: both prescriptions less than 90%: n=11 (9.73%); antiretroviral adherence less than 90% and antidepressant adherence 90% or greater: n=16 (14.16%); antiretroviral adherence 90% or greater and antidepressant adherence less than 90%: n=10 (8.85%); and both prescriptions 90% or greater : n=76 (67.26%; see Table 2).

Table 2.

Adherence Rates Among Those Prescribed Both Antidepressant and Antiretroviral Medications

	Antiretroviral medication
Antidepressant medication	< 90% adherence	≥ 90% adherence
< 90% adherence	9.73%	8.85
≥ 90% adherence	14.16%	67.26%

Table 3 provides descriptive characteristics and bivariate relationships between correlates and treatment adherence.

Table 3.

Descriptive Characteristics and Bivariate Relationships Between Correlates and Treatment Adherence

	Full sample (n=225)	Antiretroviral adherence (n=192)		Antidepressant adherence (n=146)
Correlate variable	Descriptive results	Descriptive results	Bivariate relationship with correlate	Descriptive results	Bivariate relationship with correlate
Demographics
Site	Site 1: 143 (63.6%)	Site 1: 125 (65.1%)	χ²(2)=0.94, p=0.63	Site 1: 89 (61.0%)	χ²(2)=3.08, p=0.21
	Site 2: 56 (24.9%)	Site 2: 44 (22.9%)		Site 2: 42 (28.8%)
	Site 3: 26 (11.5%)	Site 3: 23 (12.0%)		Site 3: 15 (10.2%)
Gender	218 males (96.9%)	189 males (98.4%)	FET, p=0.57	140 males (95.9%)	FET, p=0.14
Age	49.9±9.6 (range, 25–76)	50.2±9.5 (range, 25–76)	t(190)=−3.01, p<0.01	50.5±9.8 (range, 26–76)	t(144)=−3.09, p<0.01
African American	136 African Americans (60.4%)	113 African Americans (58.9%)	χ²(1)=1.70, p=0.19	81 African Americans (55.5%)	χ²(1)=0.63, p=0.43
Marital	179 single/not married (79.6%)	153 single/not married (79.7%)	χ²(1)=1.05, p=0.31	113 single/not married (77.4 %)	χ²(1)=2.98, p=0.08
Education level	208 high school graduate or more: (92.4%)	177 high school graduate or more: (92.2%)	FET, p=0.195	135 high school graduate or more: (92.5%)	FET, p=1.00
Mental health diagnosis
Major depressive disorder	171 meet criteria: (76.0%)	145 meet criteria: (75.5%)	χ²(1)=0.04, p=0.84	120 meet criteria: (82.2%)	χ²(1)=0.0008, p=0.98
Generalized anxiety disorder	136 meet criteria: (60.4%)	117 meet criteria: (60.9%)	χ²(1)=0.32, p=0.57	95 meet criteria: (65.1%)	χ²(1)=6.32, p<0.01
Panic disorder	24 meet criteria: (10.7%)	20 meet criteria: (10.4%)	FET, p=0.27	21 meet criteria: (14.4%)	FET, p=0.78
Posttraumatic stress disorder	65 meet criteria: (28.9%)	49 meet criteria: (25.5%)	χ²(1)=0.15, p=0.70	53 meet criteria: (36.3%)	χ²(1)=0.91, p=0.34
Alcohol use disorder	38 meet criteria: (16.9%)	34 meet criteria: (17.7%)	χ²(1)=0.02, p=0.89	21 meet criteria: (14.4%)	FET, p=0.78
Any alcohol use in last year?	166 drinks alcohol: (73.8%)	140 drinks alcohol: (72.9%)	χ²(1)=0.01, p=0.92	105 drinks alcohol: (71.9%)	χ²(1)=2.39, p=0.12
Number of drinks in week before baseline	3.6±7.8; range, 0–53	3.6±7.6; range, 0–53	Wilcoxon, p=0.92	3.4±7.6; range, 0–50	Wilcoxon, p=0.32
Comorbid mental health condition?	168 meet criteria: (74.7%)	141 meet criteria: (73.4%)	χ²(1)=0.04, p=0.84	119 meet criteria: (81.5%)	χ²(1)=8.30, p<0.01
Depression
SCL-20	1.8±0.6; range, 0.4–3.6	1.8±0.7; range, 0.4–3.6	t(190)=1.52, p=0.13	1.9±0.6; range, 0.6–3.6	t(144)=−0.48, p=0.63
Physical health
Self-reported no. of HIV symptoms	11.1±3.7; range, 3–19	10.8±3.7; range, 3–19	Wilcoxon, p<0.001	11.4±3.6; range, 3–19	Wilcoxon, p=0.85
Severity of self-reported HIV symptoms	35.7±14.4; range, 6–76	34.8±14.2; range, 6–76	t(190)=2.74, p<0.05	37.8±14.0; range, 7–76	t(144)=0.13, p=0.90
Physical comorbidity	3.4±2.3; range, 0–11	3.4±2.4; range, 0–11	Wilcoxon, p=0.06	3.5±2.3; range, 0–11	Wilcoxon, p=0.63
Quality of life
Physical health component score of SF-12	40.9±12.0; range, 16.1–64.9	41.5±12.2; range, 16.1–64.9	t(182)=−1.00, p=0.32	40.4±12.1; range, 19.1–64.8	t(139)=−0.49, p=0.62
Mental health component score of SF-12	35.0±10.9; range, 10.8–61.2	35.1±11.2; range, 10.8–61.2	t(182)=0.03, p=0.98	32.6±9.9; range, 10.8–57.6	t(139)=0.31, p=0.75
Quality of well-being	0.5±0.1; range, 0.2–0.7	0.5±0.1; range, 0.2–0.7	t(190)=−2.17, p=0.03	0.5±0.1; range, 0.2–0.7	t(144)=−0.70, p=0.48
Medication knowledge
Knowledge variable		76.7±36.2; range, 0–100	Wilcoxon, p=0.99	70.6±41.4; range, 0–100	Wilcoxon, p=0.21

Note: FET=Fisher's Exact Test; χ² analysis was used for categorical data, with Fisher's exact test used when small cell counts rendered the χ² inappropriate. t tests were used for continuous data, with Wilcoxon used for non-normal distributions. Analyses with SF-12 are on subset due to missing data from SF-12 data collection.

Bivariate relationships between correlates and treatment adherence

Separate analyses were conducted to examine bivariate relationships between demographic correlates, mental health diagnostic status, alcohol use measures, markers of physical health including HIV symptoms (both number and severity), measures of physical comorbidity, quality of life, medical knowledge, study-related factors (i.e., site), and treatment adherence (Table 3). Age (p<0.05), race, education, depression severity, self-reported number (p<0.05) and severity of HIV symptoms (p<0.05), physical comorbidity, and quality of well-being (p<0.05) were associated with antiretroviral adherence at p<0.20 in bivariate analyses. Given the high correlation between the “number of HIV symptoms” variable and the “severity of HIV symptoms” variable, only the severity measure was used in subsequent analyses. In contrast, antidepressant adherence was associated with the following correlates at p<0.20: gender, age (p<0.05), marital status, generalized anxiety disorder diagnosis (p<0.05), comorbid mental health disorder (p<0.05), and any alcohol use in the past year. Given that the ‘comorbid mental health condition’ variable was highly correlated with “generalized anxiety disorder” variable, only the “generalized anxiety disorder” variable was used in subsequent analyses.

Final prediction model

A seemingly unrelated bivariate probit model was estimated and the positive correlation between the two errors terms was significant (ρ=0.64, Wald test of ρ: χ²(1)=9.84, p=0.002). Therefore, correlates of adherence to antiretrovirals and antidepressants were estimated jointly. Education, age, and HIV symptom severity were significant correlates of antiretroviral medication adherence while gender and generalized anxiety disorder were significant correlates of adherence to antidepressant medications (see Table 4 for full results). The results from the bivariate probit with endogenous dummy models (available from first author upon request) indicate that antiretroviral adherence does not predict antidepressant adherence (β=1.62, p=0.17) but antidepressant adherence does predict antiretroviral adherence (β=2.30, p=0.036).

Table 4.

SUR with Significant Correlates (p<0.20) and Treatment Adherence (n=113)

	Antiretroviral adherence 90% cutoff		Antidepressant adherence 90% cutoff
Correlate variable	β (SE)	(95% CI)	β (SE)	(95% CI)
Age	0.05 (0.02)^a	0.01, 0.08	0.03 (0.02)	−0.01, 0.06
Gender	0.18 (0.80)	−1.39, 1.75	−5.60 (1.60)^a	−8.74, −2.48
Race	0.20 (0.33)	−0.46, 0.85	−0.19 (0.30)	−0.79, 0.40
Education	−6.16 (1.55)^a	−9.20, −3.12	−0.42 (0.63)	−1.66, 0.81
Marital	0.28 (0.36)	−0.42, 0.98	−0.03 (0.36)	−0.73, 0.67
Generalized anxiety disorder	0.27 (0.35)	−0.41, 0.94	0.67 (0.29)^a	0.09, 1.25
Drink or abstain in last year?	0.17 (0.33)	−0.47, 0.82	−0.25 (0.32)	−0.87, 0.38
SCL-20	0.13 (0.28)	−0.42, 0.68	0.33 (0.29)	−0.24, 0.89
Severity of SR HIV symptoms	−0.03 (0.02)^a	−0.07, −0.002	−0.01 (0.02)	−0.04, 0.02
Physical comorbidity	−0.09 (0.08)	−0.23, 0.06	0.01 (0.07)	−0.12, 0.15
Quality of well-being	−0.66 (1.31)	−3.23, 1.91	0.68 (1.43)	−2.12, 3.47

p<0.05; The correlation between the two errors terms of the seemingly unrelated bivariate probit model was significant (ρ=0.64, Wald test of ρ: χ²(1)=9.84, p<0.01).

SUR, seemingly unrelated regression; CI, confidence interval.

Discussion

The present study attempted to isolate demographic, mental health, and/or physical health factors that were associated with preintervention adherence patterns among depressed HIV-infected patients. While antiretroviral adherence did not predict antidepressant adherence, antidepressant adherence predicted antiretroviral adherence. Antidepressant adherence predicting antiretroviral adherence is consistent with previous research.^5,16
–18 Certain demographic factors (i.e., age and level of education), physical health (i.e., HIV severity), and mental health (i.e., generalized anxiety disorder diagnosis) were associated with adherence but there were relatively few significant correlates indicating the presence of omitted variables. We did note that older adults and those with comorbid generalized anxiety disorder were more adherent, suggesting that there may be additional attention to adherence with increasing age and also among those with a tendency to worry. The finding that a higher level of education was predictive of poorer adherence in the SUR equation predicting antiretroviral adherence was surprising. In a recent review of results from developed and developing nations, education was not identified as a predictor of antiretroviral adherence.⁷ However, in our sample there was a relatively small number of subjects who did not have a high school diploma, therefore, this finding should not be over-interpreted. The negative relationship between adherence and HIV symptom severity suggests to us that those who are feeling better may be motivated to maintain this status via adherence to their regimens. Overall, our findings were modest, suggesting the variables we measured were of limited utility in predicting nonadherence behaviors.

The clinical implications of our findings may be helpful for clinical providers. Results suggested that antidepressant adherence affected adherence to antiretrovirals but not vice versa. Given that the subjects enrolled in this investigation are active patients in HIV clinics and were suffering from depression, this suggests that clinical providers' efforts to improve antidepressant adherence may also improve antiretroviral adherence, which is especially important given the virologic impact of inconsistent antiretroviral adherence. Our findings mirrors several other investigations that have also provided evidence of improved adherence to antiretrovirals among depressed individuals taking antidepressants.^5,16
–18 However, we also simultaneously found that antiretroviral adherence did not impact antidepressant adherence. Even so, more research is needed to examine potential moderators of the relationship between antidepressant and antiretroviral adherence.

The present investigation has a number of strengths as well as noteworthy limitations. First, a major strength of the present investigation is the use of real-world patients, i.e., depressed HIV-infected veterans who were patients seeking care in three VA HIV clinics. The eligibility criteria for this investigation were minimal so as to facilitate generalizability of these finding and the overall trial to HIV-infected veterans seeking care in the VA more generally. Moreover, the racial makeup of our sample (60% African American) is similar to the U.S. HIV population,⁴⁴ which further improves generalizability. Second, we were able to examine correlates for both antiretroviral and antidepressant adherence simultaneously using the SUR modeling techniques. Limitations of the present investigation include reliance on self-reported treatment adherence (which involves considerable measurement error), limited variables available due to limitations in data collection, restriction of sample to veterans, use of cross-sectional data, few women in the sample, and modest sample size in the SUR model. Although there are benefits and drawbacks to self-reported HIV treatment adherence, it can be used to effectively estimate adherence.⁴⁵ Even so, future trials may want to build on our results using more objective methods of measurement such as MEMS caps, pill count, or pharmacy record review and utilize a window of measurement that is longer than 4 days.⁴⁶ Additionally, we believe that limitations in the variable “patient's knowledge of regimen” may be responsible for the nonassociation between the knowledge and adherence variables. Furthermore, though the Veterans Affairs network of health care (VAMCs) is the largest provider of HIV care in the United States, the veteran population does not mirror the overall HIV-infected population in the United States. Indeed, veterans in care for HIV infection are more likely to be male and older than the average U.S. HIV-infected adult. Although the percentage of female veterans receiving care at VAMCs is increasing, women are underrepresented in comparison to the percentage of HIV-infected women in the general U.S. population. We were limited in the interpretation of gender-related and possibly education-related findings given the limited variability of these factors in our sample. Finally, although we had 225 participants with either an antidepressant or antiretroviral prescription in our baseline sample, our sample size was much smaller (N=113) when we restricted the sample to those with both prescriptions.

In conclusion, among our clinically depressed, treatment-seeking HIV-infected participants, demographic factors (i.e., age and level of education), physical health (i.e., HIV severity), and mental health (i.e., generalized anxiety disorder diagnosis) were correlates of self-reported adherence. We found limited support for additional mental and/or physical health factors as correlates of adherence to these regimens. Our findings also demonstrated that antidepressant adherence was associated with antiretroviral adherence in the SUR model but not vice versa. Future research is needed to identify and test additional factors and interventions that jointly impact antiretroviral and antidepressant adherence among treatment-seeking HIV-infected patients.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

DiMatteo

. Variations in patients' adherence to medical recommendations: A quantitative review of 50 years of research. Med Care, 2004; 42:200–209.

Ingersoll

, Cohen

. The impact of medication regimen factors on adherence to chronic treatment: A review of literature. J Behav Med, 2008; 31:213–224.

Kacanek

, Jacobson

, Spiegelman

, Wanke

, Isaac

, Wilson

. Incident depression symptoms are associated with poorer HAART adherence: A longitudinal analysis from the Nutrition for Healthy Living study. J Acquir Immune Defic Syndr, 2010; 53:266–72.

Bulloch

AGM

, Patten

. Non-adherence with psychotropic medications in the general population. Soc Psychiatry Psychiatr Epidemiol, 2010; 45:47–56.

Horzberg

, Silverberg

, Hurley

et al. Effects of depression and selective serotonin reuptake inhibitor use on adherence to Highly Active Antiretroviral Therapy and on clinical outcomes in HIV-infected patient. J Acquir Immune Defic Syndr, 2008; 47:384–390.

Shacham

, Nurutdinova

, Onen

, Stamm

, Overton

. The interplay of sociodemographic factors on virologic suppression among a U.S. outpatient HIV clinic population. AIDS Patient Care STDs, 2010; 24:229–235.

Mills

, Nachega

, Bangsberg

et al. Adherence to HAART: A systematic review of developed and developing nation patient reported barriers and facilitators. PLoS Med, 2006; 3:e438.

Brown

, Battista

, Bruehlman

, Sereika

, Thase

, Dunbar-Jacob

. Beliefs about antidepressant medications in primary care patients: Relationship to self-reported adherence. Med Care, 2005; 43:1203–1207.

Burra

, Chen

, Mclntyre

, Grace

, Blackmore

, Stewart

. Predictors of self-reported antidepressant adherence. Behav Med, 2007; 32:127–134.

10.

Chakraborty

, Avasthi

, Kumar

, Grover

. Attitudes and beliefs of patients of first episode depression towards antidepressants and their adherence to treatment. Soc Psychiatry Psychiatr Epidemiol, 2009; 44:482–488.

11.

Raue

, Schulberg

, Heo

, Klimstra

, Bruce

. Patients' depression treatment preferences and initiation, adherence, and outcome: A randomized primary care study. Psychiatr Serv, 2009; 60:337–343.

12.

Sheehan

, Keene

, Eaddy

, Krulewicz

, Kraus

, Carpenter

. Differences in medication adherence and healthcare resource utilization patterns: Older versus newer antidepressant agents in patients with depression and/or anxiety disorders. CNS Drugs, 2008; 22:963–973.

13.

Kaiser Permanente Care Management Institute. Depression Clinical Practice Guidelines. Oakland, CA: 0Kaiser Permanente Care Management Institute, 2006.

14.

Sawada

, Uchida

, Suzuki

et al. Persistence and compliance to antidepressant treatment in patients with depression: A chart review. BMC Psychiatry, 2009; 9:38.

15.

Furtek

, Martin

, Burgi

, Crum-Cianfl

, Pope

, Hale

. Antiretroviral adherence and virologic response rates in HIV-infected patients with mood disorders [Abstract 994] Poster presented at the 44th annual meeting of IDSA, Toronto, October 12–15, 2006.

16.

Yun

LWH

, Marvai

, Kobayashi

, Barton

, Davidson

. Antidepressant treatment improves adherence to antiretroviral therapy among depressed HIV-infected patients. J Acquir Immune Defic Syndr, 2005; 38:432–438.

17.

Walkup

, Wei

, Sambamoorthi

, Crystal

. Antidepressant treatment and adherence to combination antiretroviral therapy among patients with AIDS and diagnosed depression. Psychiatr Q, 2008; 79:43–53.

18.

Kumar

, Encinosa

. Effects of antidepressant treatment on antiretroviral regimen adherence among depressed HIV-infected patients. Psychiatr Q, 2009; 80:131–141.

19.

Pyne

, Fortney

, Curran

et al. Effectiveness of collaborative care for depression in HIV clinics. Arch Intern Med, 2011; 171:23–31.

20.

Fortney

, Pyne

, Edlund

et al. A randomized trial of telemedicine-based collaborative care for depression. J Gen Intern Med, 2007; 22:1086–1093.

21.

Kroenke

, Spitzer

. The PHQ-9: A new depression diagnostic and severity measure. Psychiatr Annu, 2002; 32:509–515.

22.

Katzman

, Brown

, Fuld

, Peck

, Schechter

, Schimmel

. Validation of a short orientation-memory-concentration test of cognitive impairment. Am J Psychiatry, 1983; 140:734–739.

23.

Derogatis

. SCL-90-R. Administration, Scoring, and Procedures Manual, 3rd. Minneapolis, MN: National Computer Systems, Inc., 1994.

24.

Sheehan

, Lecrubier

, Sheehan

et al. The validity of the Mini International Neuropsychiatric Interview (MINI) according to the SCID-P and its reliability. Eur Psychiatry, 1997; 12:232–241.

25.

Lecrubier

, Sheehan

, Weiller

et al. The Mini International Neuropsychiatric Interview (MINI): A short diagnostic structured interview: Reliability and validity according to the CIDI. Eur Psychiatry, 1997; 12:224–231.

26.

Rost

, Smith

, Burnam

. Measuring the outcomes of care for mental health problems. Med Care, 1992; 30:MS266–MS273.

27.

Smith

Jr. , Burnam

, Burns

, Cleary

, Rost

. Depression Outcomes Module (DOM) Handbook of Psychiatric Measures. Washington, D.C.: American Psychiatric Association, 2000; 213–215.

28.

Justice

, Holmes

, Gifford

et al. Development and validation of a self-completed HIV symptom index. J Clin Epidemiol, 2001; 54,Suppl 1:S77–90.

29.

Jones

, Kazis

, Lee

et al. Health status assessments using the Veterans SF-12 and SF-36: Methods for evaluating outcomes in the Veterans Health Administration. J Ambul Care Manage, 2001; 24:68–86.

30.

Kaplan

, Anderson

. The general health policy model: An integrated approach. Spiker

. Quality of Life and Pharmacoeconomics in Clinical Trials. Second ed. Philadelphia: Lippincott-Raven Publishers, 1996; 309–321.

31.

Kaplan

, Bush

, Berry

. Health status: Types of validity and the index of well-being. Health Serv Res, 1976; 11:478–507.

32.

Chesney

, Ickovics

, Chambers

et al.

Self-reported adherence to antiretroviral medications among participants in HIV clinical trials: The AACTG adherence instruments. Patient Care Committee & Adherence Working Group of the Outcomes Committee of the Adult AIDS Clinical Trials Group (AACTG)

AIDS Care, 2000; 12:255–266.

33.

Fortney

, Pyne

, Edlund

, Mittal

. Relationship between antidepressant medication possession and treatment response. Gen Hosp Psychiatry, 2010; 32:377–379.

34.

Paterson

, Swindells

, Mohr

et al. Adherence to protease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med, 2000; 133:21–30.

35.

Bangsberg

. Less than 95% adherence to nonnucleoside reverse-transcriptase inhibitor therapy can lead to viral suppression. Clin Infect Dis, 2006; 43:939–941.

36.

Shuter

. Forgiveness of non-adherence to HIV-1 antiretroviral therapy. J Antimicrob Chemother, 2008; 61:769–773.

37.

Miller

, Liu

, Hays

et al. Knowledge of antiretroviral regimen dosing and adherence: A longitudinal study. Clin Infect Dis, 2003; 36:514–518.

38.

SAS Institute Inc. SAS software, Version 9.2. Cary, NC, 2008.

39.

StataCorp. Stata Statistical Software: Release 9. College Station, TX: StataCorp LP, 2005.

40.

Costa-Font

, Gil

. Obesity and the incidence of chronic diseases in Spain: A seemingly unrelated probit approach. Econ Hum Biol, 2005; 3:188–214.

41.

Fabbri

, Monfardini

, Radice

. Testing exogeneity in the bivariate probit model: Monte Carlo evidence and an application to health economics. Dipartimento di Scienze Economiche, Università di Bologna, Working Paper No.514; 2004.

42.

Park

. Regression Models for Binary Dependent Variables Using Stata, SAS, R, LIMDEP, and SPSS. Working Paper. The University Information TechnologyServices (UITS) Center for Statistical and Mathematical Computing, Indiana University, 2009.

43.

Maddala

. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press, 1983.

44.

VA Clinical Case Registry. http://vaww.hiv.va.gov/. 2012 February 14.

45.

Simoni

, Kurth

, Pearson

, Pantalone

, Merrill

, Frick

. Self-report measures of antiretroviral therapy adherence: A review with recommendations for HIV research and clinical management. AIDS & Behav, 2006; 10:227–245.

46.

, Safren

, Skolnik

et al. Optimal recall period and response task for self-reported HIV medication adherence. AIDS Behav, 2008; 12:86–94.