Patient, Caregiver and Regimen Characteristics Associated With Adherence to Highly Active Antiretroviral Therapy Among HIV-Infected Children and Adolescents

Abstract

Background: 

This study assesses the relationship between child and caregiver perceptions of medication responsibility, disease knowledge, regimen complexity and adherence to highly active antiretroviral therapy among HIV-positive children. We also examine the association of adherence to child and caregiver demographic characteristics and surrogate markers of HIV disease.

Methods: 

For this 6-month longitudinal study, 24 HIV-positive children (mean age = 14.0 years) being treated with highly active antiretroviral therapy and their caregivers completed measures of medication responsibility and disease knowledge. Medication Event Monitoring System caps calculated adherence across months 1 through 3 (time 1) and 4 through 6 (time 2).

Results: 

Medication Event Monitoring System data revealed adherence rates of 81% at time 1 and 79% at time 2. Only 8% (n = 2) of child–caregiver pairs reported complete agreement regarding who held responsibility for medication-related tasks. Patients' responsibility for medication was correlated with age based on child (r = .51) and caregiver (r = .57; Ps < 0.05) perceptions, although their regimen knowledge was not. Greater regimen knowledge among caregivers and fewer child–caregiver discrepancies about medication responsibility predicted better adherence (adjusted R² = .45). Finally, adherence was correlated with CD4 percentages at time 1 (r = .50) and viral load at time 1 (r = −.56) and time 2 (r= −.68; Ps < 0.05).

Conclusions: 

Medication adherence among HIV-infected children is lower than required for optimal viral suppression. Adherence is related to surrogate markers of HIV disease but not to child or caregiver demographic variables. Responsibilities for medication-related tasks should be clarified among family members, regimen knowledge should be emphasized and caregivers should avoid assigning treatment responsibility to a child prematurely.

Treatment advances during the past decade, particularly the widespread availability of highly active antiretroviral therapy (HAART), have led to a shift in the conceptualization of HIV from a fatal disease to a chronic illness. With effective medical care and antiretroviral (ARV) treatment, children with HIV currently experience fewer opportunistic infections and lower risk of disease progression.^1,2 The degree to which HAART regimens are effective is influenced by patients' adherence to their medications.³ Nonadherence substantially increases the risk of disease progression and death^4,5 as well as the likelihood of developing resistance to a particular drug or class of drugs.^3,6,7 For these reasons, adherence has become a primary focus of research in HIV disease. Although an abundance of studies have investigated adherence-related factors in HIV-positive adults, less is known about the variables associated with medication adherence among children and adolescents with HIV infection. Three factors hypothesized to be relevant to treatment adherence in HIV-infected patients include treatment responsibility (ie, who holds primary responsibility for engaging in medication-related tasks),⁸ disease knowledge^9,10 and the complexity of the medication regimen.^11–14

The first objective of this 6-month longitudinal study was to examine the relationship of baseline responsibility for medication, various aspects of disease knowledge, and regimen complexity to subsequent adherence to HAART among children and adolescents with HIV disease and their caregivers. The second objective was to assess the relationship of adherence to child and caregiver demographic characteristics as well as to the child's medical functioning over time.

METHODS

Participants

HIV-infected children from around the United States were referred to the National Cancer Institute for participation in ARV treatment protocols developed through the HIV and AIDS Malignancy Branch. These patients were referred from a variety of sources, including general pediatric practices and university hospitals. Families of these children were invited to participate in the current adherence study during a routine outpatient clinic visit at any point during their ARV treatment protocol, provided they met all eligibility requirements. To be eligible, children must have been between 8 and 18 years of age, have vertically acquired HIV disease and have been on a HAART regimen that included a protease inhibitor (PI) in solid form for at least 3 months. In addition, children must have been living with their current primary caregiver, defined as the person most involved with administration of the ARV regimen, for at least 6 months at the time of their baseline assessment on this study. Patients were ineligible if they were receiving their PI through a gastronomy tube, were unaware of their HIV diagnosis or had clinically significant medical or cognitive impairment that would interfere with their ability to complete all study requirements. The protocol for the current study was approved by the National Cancer Institute Institutional Review Board. Informed consent was obtained from the patient's parent or legal guardian before enrollment and minor assent was obtained from all patients as well.

Measures

Responsibility for Medication-Related Tasks.

The Responsibility for Medication Scale (RMS) is a 9-item questionnaire¹⁵ administered during the baseline assessment based on a similar measure that has been used with the pediatric diabetes population.¹⁶ For 9 medication-related tasks, children and caregivers were asked to indicate whether primary responsibility was held by the child, the caregiver(s) or both. Item examples include “getting pills out of the containers and ready to take” and “remembering to take the medicine when away from home.” Higher scores on this measure indicate more responsibility given to the child. A discrepancy score was obtained by noting the number of discrepant responses between the child and caregiver; this number can range from 0 (no discrepancies; complete agreement) to 9 (discrepant responses on all 9 items).

Disease Knowledge

Knowledge of HIV and Adherence.

During the baseline assessment, children and caregivers each completed the HIV Knowledge Questionnaire as a measure of general knowledge of HIV disease and knowledge about adherence-related concepts. This measure, developed by our group, consists of 2 subscales with true–false and multiple choice questions. The general knowledge of HIV subscale (K-HIV) is comprised of 24 questions pertaining to various aspects of HIV disease, including routes of infection, types of medication and the immune system. A 6-item subscale assessing knowledge about adherence (K-ADH) includes items about immunologic and virologic benefits of adherence and possible consequences of nonadherence to ARV regimens (eg, drug resistance). Scores represent the percentage of correct responses. Separate, parallel versions of this measure were created for children (8–11 years), adolescents (12–18 years) and their caregivers (of children 8–18 years). These 3 versions were pilot tested to assess the level of difficulty in terms of content and readability for each age group, and modifications were made based on information obtained during this process.

Regimen Knowledge.

The Pills Identification Test (PIT), adapted from a similar measure,¹⁷ was administered to assess knowledge of the patient's HAART regimen. Patients and caregivers separately were shown a chart containing a picture of all ARV medications (in pill form only) approved by the U.S. Food and Drug Association and were asked to (1) identify all the medications in the child's regimen, (2) name each of the identified medications, (3) specify the number of prescribed doses per day and (4) indicate the number of prescribed pills per dose. Because responses to the 2 dosing questions (3 and 4) were very similar, these 2 items were pooled to form a composite variable representing the child's dosing schedule. The PIT is scored by calculating an overall percentage of correct responses to these questions (as determined by medical chart review) summed across each medicine in the patient's ARV regimen yielding a total score of regimen knowledge and scores on 3 subscales: Visual Identification (Visual ID), Naming, and Dosing Schedule.

Regimen Complexity

A score representing regimen complexity was computed that accounted for the total number of ARV medications in a regimen; the type of administration (solid, liquid or injection); the number of doses per day; the number of pills per day for each medication in solid form; and the number and type of specific dosing instructions associated with one or more medications (take with food or on an empty stomach). This methodology was based on a weighting system used by George et al¹⁸ with modifications applied to make the system more relevant to pediatric HIV disease; the final weights were determined based on input from 4 of the coauthors (SM, PW, GR and LW). As an example, each regimen was weighted according to the number of dosing instructions included. For instance, a weight of 1.0 was given if the regimen included one or more medications that must be taken with food, whereas medications taken on an empty stomach were given a weight of 1.5. Medications that contain multiple antiretroviral drugs (eg, Kaletra, Combivir, Trizivir) were counted as one medication.

Adherence

Adherence to medication was assessed through use of an electronic monitoring system. This Medication Event Monitoring System (MEMS) relies on “Track caps” containing a microelectronic circuit that registers the exact dates and times when the pill bottle is opened. MEMS data were analyzed with respect to the percentage of doses taken within 2 hours of a target time. An adherence percentage was computed for each child as follows: [(scheduled doses − missed doses)/scheduled doses] × 100. MEMS data were analyzed as composite measures of adherence estimates averaged across months 1 through 3 (time 1) and 4 through 6 (time 2).

Medical Markers

Levels of HIV-1 RNA polymerase chain reaction (PCR), or viral load, as well as percentages and absolute counts of CD4 T-lymphocyte cells were collected as required by the patient's ARV treatment protocol. These values were obtained at the time of the child's baseline adherence assessment and at the 3-month (time 1) and 6-month (time 2) evaluations.

Procedures

During the patient's baseline evaluation for the adherence protocol, each child and his or her primary caregiver were administered the measures assessing responsibility for medication (RMS), knowledge of HIV (K-HIV) and adherence (K-ADH) and knowledge of the child's ARV treatment regimen (PIT). If the primary caregiver was not present when the child was enrolled in the study (ie, another caregiver accompanied the child to the clinic), the PIT could not be administered because it must be completed in person, but the RMS, K-HIV and K-ADH scales were sent home for the primary caregiver to complete and return by mail (n = 3).

To monitor adherence, families were provided with a MEMS cap that was placed on a new bottle of the child's PI. The PI was chosen as the drug to monitor for the sake of consistency across patients. In addition to the MEMS cap, families were given a log to track any MEMS bottle openings that did not correspond to scheduled ARV doses. Based on these log entries, a research assistant made adjustments to the MEMS data to maximize the accuracy of reported adherence rates. For example, if a caregiver indicated on the log that the bottle was inadvertently opened twice on a particular morning, the extra opening was manually excluded from the data. During patients' 3- and 6-month clinic visits, MEMS caps and log entries were collected.

At the time of enrollment, families were asked whether they used a pill tray for arranging and dispensing their child's ARV medications. To avoid disrupting a routine that was working well, these families were given the option of foregoing the MEMS portion of the study and participating in all other study procedures.

Statistical Analyses

Correlated-samples t tests and mixed design repeated measures analyses of variance (ANOVA), with Greenhouse-Geisser corrections when appropriate, were used to compare child and caregiver responses on the knowledge measures (K-ADH, K-HIV and PIT) and the RMS. A correlated-samples t test was used to examine changes in MEMS adherence from time 1 to time 2. Pearson correlation coefficients were calculated to assess the relationship between variables of interest, including knowledge, treatment responsibility, adherence and medical markers. Although viral load levels and CD4 absolute counts often are analyzed categorically to increase the clinical relevance, dividing the 17 children with MEMS data into categories would have resulted in small numbers of children in each group; thus, medical markers were analyzed as continuous variables. Linear regression analyses were used to examine the influence of baseline variables on adherence. Levels of plasma viral load were subjected to a log₁₀ transformation to normalize the distribution of data.

RESULTS

Patient Sample

Of 38 eligible families approached about this adherence protocol, 30 agreed to participate. Of these, 6 were taken off study before completion for the following reasons: 2 families withdrew because the caregivers and/or patients reported discomfort with study procedures (eg, “felt we were being checked up on”); 2 patients' ARV medications were discontinued temporarily for medical reasons; one caregiver with a history of psychiatric problems became depressed during her participation and requested that she and her child be withdrawn from the study; and one child's medical care was transitioned to another clinic in the patient's hometown. Thus, 24 families provided data for this study. Seven of these families chose to enroll in the non-MEMS arm of the study, so adherence data were obtained from 17 families.

Demographic and Medical Information

The total sample of 24 patients (13 boys and 11 girls) ranged in age from 8 to 18 years (mean = 13.9 years, standard deviation [SD] = 2.6). As shown in Table 1, the majority of the children were black (46%) or white (42%). Less than half (42%) of the children were living with a biologic parent and 54% percent were living in 2-parent homes. Nearly all (92%) of the primary caregivers had completed high school and the mean education level was 13.4 years (SD = 3.1). One-third (33%) of the primary caregivers were HIV-positive. The mean number of years since the child's first HAART regimen was 6.9 years (SD = 0.7 years; range, 5.0–7.8 years).

TABLE 1: Demographic Characteristics of Child and Caregiver Participants (N = 24)

Table 2 presents the medical data of the children in our sample. At the baseline assessment, 9 children (38%) had an undetectable viral load (HIV-1 RNA PCR <50 copies/mL) and 7 (29%) had viral loads >10,000 copies/mL. Only 3 (13%) children showed evidence of severe immune compromise as indicated by CD4⁺ absolute counts of less than 200.

TABLE 2: Medical Characteristics of Participants (N = 24)

Families who did not complete the study (n = 6) did not differ systematically from the study completers (n = 24) on any demographic or medical variables based on ANOVAs and χ² tests of independence. No systematic differences were evident between the MEMS and non-MEMS participants with regard to baseline demographic or medical variables. However, compared with families who did not use a MEMS cap (n = 7), families who agreed to use a MEMS cap (n = 17) demonstrated higher scores on the children's K-HIV subscale (F[1,22] = 9.0, P < 0.01; mean percentage correct 82.4 ± 8.3 versus 72.0 ± 5.8, respectively), and on the caregivers' PIT Visual ID subscale (F[1,19] = 6.21, P < 0.05; mean percentage correct 86.2 ± 18.2 versus 64.8±16.5). There were no differences in time 1 or time 2 adherence rates according to the child's sex or race, or to caregiver education, caregiver–child relationship (biologic parent versus other), family composition (one- versus 2-parent home) or caregiver HIV status (Ps > 0.05).

Responsibility for Treatment

Table 3 presents child and caregiver scores on all the baseline measures. On the RMS, children perceived themselves as having a higher amount of responsibility than their caregivers perceived them to have (t = 8.61, P < 0.001). Ninety-two percent (n = 22) of families provided one or more discrepant responses regarding who was responsible for the 9 medication-related tasks and 33% (n = 8) of families responded inconsistently on more than half (≥5) of the tasks. Items on which at least half of the families disagreed involved remembering what time to take the medicine (n = 16 [67%]), reporting side effects to the doctor (n = 13 [54%]), remembering to take the medicine when not at home (n = 14 [58%]) and remembering to bring the medicine with them when they were going to be away from home (n = 12 [50%]).

TABLE 3: Comparison of Mean Scores on the Measures of Medication Responsibility and Knowledge

Child age was positively associated with the level of the child's medication responsibility according to both child (r = .51, P < 0.05) and caregiver (r = .57, P < 0.01) perceptions, whereas caregiver education level was unrelated to medication responsibility. Child age was unrelated to the number of caregiver–child discrepancies regarding responsibility for medication.

Disease Knowledge

HIV Knowledge Questionnaire.

Regarding the K-HIV and K-ADH subscales of the HIV Knowledge Questionnaire, repeated measures mixed design ANOVA showed a main effect of respondent (F[1,22] = 15.71, P < 0.001) such that caregivers scored significantly higher than their children (see Table 2). There also was a main effect of subscale (F[1,22] = 5.73, P < 0.05) indicating that mean scores on the K-HIV subscale were significantly higher than mean scores on the K-ADH subscale. Neither child age nor caregiver education was related to scores on the K-HIV or K-ADH subscales.

Pills Identification Test

On the PIT, results are based on data from all 24 children and only 21 caregivers because 3 primary caregivers did not accompany their children to the clinic during the baseline evaluation. As shown in Table 2, repeated measures mixed design ANOVA with Greenhouse-Geisser correction revealed no main effect of respondent (F[1,20] = .76, nonsignificant), but there was a significant main effect of subscale (F[2,40] = 13.71, P < 0.001). Post hoc comparisons showed that caregivers' mean scores were significantly higher on the Dosing Schedule subscale compared with the Visual ID (F[1,20] = 14.73, P < 0.01) and Naming (F[1,20] = 9.33, P < 0.01) subscales. Similarly, children's mean scores on the Dosing Schedule subscale were significantly higher than their mean scores on the Visual ID (F[1,23] = 8.19, P < 0.01) and Naming (F[1,23] = 18.78, P < 0.001) subscales. Additionally, children's Visual ID scores were significantly higher than their Naming scores (F[1,23] = 7.44, P < 0.05). Among the caregivers, the percentage correct on the Dosing Schedule subscale was 100% for all but 2 individuals. As a result of this lack of variability, caregiver PIT Dosing Schedule scores were not included in the correlational analyses that follow. None of the knowledge or PIT subscale mean scores were related to years of caregiver education or age of the child (Ps > 0.05).

Regimen Complexity

Although the number of antiretroviral medications in children's regimens ranged from 2 to 6, most regimens were comprised of 3 (n = 17 [71%]) or 4 (n = 4 [17%]) medications. Eighty-seven percent (n = 20) of the children were on twice daily dosing schedules and the remaining 17% (n = 4) were taking at least one of their ARV medications 3 times a day. Regimen complexity scores ranged from 5.5 to 15.25 (mean = 9.1, SD = 2.0) and higher scores represent more complex regimens. The lowest score was obtained by a child taking 2 ARV medications, both in tablet form; one taken with food, the other with no specific dosing instructions; both taken twice daily for a total of 6 pills taken per day. The highest score was obtained by a child taking 6 ARV medications, 5 in tablet or capsule form, one an injection involving special preparations (enfuvirtide); 3 taken with food; one taken once a day and 5 taken twice a day for a total of 21 pills per day. Regimen complexity scores were not significantly correlated with any demographic variables, including age, caregiver education and length of time since the child's first HAART regimen nor were they related to viral loads, CD4 percentages, or absolute CD4 counts.

Adherence

Mean adherence rates based on MEMS data were 80.9% (SD = 14.6) at time 1 and 78.5% (SD = 21.9) at time 2. The slight decrease in adherence rates across time was not statistically significant at the 0.05 level. Moreover, only 17% (n = 4) of children at time 1 and 21% (n = 5) at time 2 maintained adherence rates of ≥90%, a rate commonly referred to as the minimum level necessary for obtaining adequate viral suppression.³

Relationship Among Treatment Responsibility, Disease Knowledge and Regimen Complexity

No significant correlations emerged between child or caregiver perceptions of the child's medication responsibility (as indicated by RMS scores) and child or caregiver knowledge of HIV or adherence. In addition, child and caregiver RMS scores were unrelated to children's performance on the PIT total and subscale scores (Ps > 0.05), indicating that children's level of responsibility for managing their ART regimen was not associated with their regimen knowledge. Both caregivers' and children's perceptions of the caregiver holding more responsibility for the child's medication (lower RMS scores) were related to better caregiver regimen knowledge as measured by the total PIT scale (r = −49, P < 0.05) but not the Visual ID and Naming subscale scores. (As mentioned previously, the Dosing Schedule subscale scores were not included in these analyses as a result of the restricted range.) More discrepancies between children and caregivers on the RMS were related to worse caregiver performance on the K-HIV (r = −.46, P < 0.05) and PIT Naming (r = −.55, P < 0.01) subscales. Complexity of the patient's ARV regimen was not associated with child or caregiver scores on the RMS, K-HIV, K-ADH or PIT.

Relationship of Treatment Responsibility, Disease Knowledge and Regimen Complexity to Adherence

On the RMS, more child–caregiver discrepancies about how much responsibility the child held for medication-related tasks was significantly related to worse adherence at time 2 (r= −.54, P < 0.05) but not at time 1. Individually, neither child nor caregiver RMS scores were related to adherence at time 1 or time 2 (Ps > 0.05).

Child and caregiver scores on the K-HIV and K-ADH subscales were unrelated to MEMS adherence percentages at time 1 and time 2 (Ps > 0.05). On the PIT, caregivers' regimen knowledge total scores were positively related to adherence at both time points (r= .63, P < 0.05 at time 1; r= .72, P < 0.01 at time 2). Regarding the PIT subscales, caregiver Visual ID and Naming scores were positively related to adherence at time 1 (r = .69, P < 0.01 and r = .59, Ps < 0.05, respectively) and time 2 (r = .52 and r = .54, respectively; Ps < 0.05). Children's mean composite and subscale scores on the PIT were unrelated to adherence at both time points (Ps > 0.05).

Higher regimen complexity scores were associated with MEMS adherence percentages at time 1 (r = .57, P < 0.05) and time 2 (r = .51), although the latter relationship did not reach statistical significance (P = 0.051). These relationships indicate that more complex HAART regimens were related to better adherence.

Based on results of the correlation matrix, linear regression analyses were conducted to determine the ability of caregiver mean PIT composite scores and RMS discrepancy scores to predict time 2 adherence percentages. The resulting significant model accounted for 54% of the variance in adherence (F[2,11] = 5.62, P < 0.05, adjusted R² = .45). Individually, caregiver PIT scores contributed a significant amount of unique variance (24%) in time 2 MEMS percentages (F[1,12] = 6.12, P < 0.05). RMS scores accounted for 2% of unique variance, which was not statistically significant (F[1,12] = .52, P > 0.05).

Relationship Between Adherence and Medical Markers

The relationship between adherence to medication based on MEMS data and medical markers at baseline, time 1 and time 2 was investigated. Better adherence at time 1 was associated with lower viral loads (r = −.56) and higher CD4 percentages (r = .50) at time 1 (Ps < 0.05). Better adherence at time 2 was related to lower viral loads (r = −.68, P < 0.01) but not CD4 percentages (r = .43, P > 0.05) at time 2. Adherence was not correlated with CD4 absolute counts at any time point. Of the 4 children at time 1 and 5 children at time 2 who demonstrated adherence rates of ≥90%, 3 of these cases involved children with adherence ≥90% at both time points. All 3 of these patients had an undetectable viral load at the baseline, time 1 and time 2 evaluations.

DISCUSSION

The primary objective of this 6-month longitudinal study was to assess how treatment responsibility, disease knowledge and ARV regimen complexity relate to subsequent medication adherence in HIV-positive children. We also sought to examine the association among adherence, demographic characteristics and medical markers in this population.

Better regimen knowledge among caregivers, fewer child–caregiver discrepancies about responsibility for medication and more complex regimens predicted better adherence outcomes in our study. Several studies have suggested that a more thorough understanding of a patient's ARV regimen relates to better adherence,^19,20 but only one study was identified that focused on the pediatric HIV population and various aspects of disease knowledge were not examined individually in this investigation.⁹ The association between caregiver regimen knowledge and adherence suggests that clinicians should present caregivers with information about their child's regimen in a clear, organized format to ensure optimal understanding. Assessing regimen knowledge at the time of initiation or change of the child's regimen and at periodic intervals thereafter might serve to identify families at risk for nonadherence so that educational interventions can be tailored toward this group.

The relationship between fewer caregiver–child discrepancies regarding treatment responsibility and better adherence has not been described previously in the literature. This finding has important clinical implications and suggests that if care providers are able to help families establish consistent expectations about who holds responsibility for medication, adherence may be improved. Future interventional studies are needed to explore this hypothesis.

At both time points, children with more complex regimens were more adherent than those with less complex regimens. Other researchers have reported similar findings in adults with various illnesses and have speculated that patients with more complex regimens may be less healthy and therefore more motivated to be adherent.^21,22 Also, clinicians may be more hesitant to prescribe complex regimens to children with a history of poor adherence.

According to MEMS data, adherence rates for most of the children in our sample were lower than 90%, the level considered necessary to attain consistent viral suppression. Some caregivers expressed surprise to learn that their child's adherence was notably lower than they had thought. Thus, medical staff may want to emphasize the importance of caregivers directly observing their child or adolescent taking his or her medicine.

Better adherence rates were significantly related to lower viral loads at both time points and higher CD4 percentages at time 1, a finding that lends support to the validity of MEMS data. To our knowledge, only one other study has used MEMS data to examine this relationship and found that MEMS adherence percentages significantly predicted viral load,²³ but the authors did not present data on CD4 cell counts.

The children in our study reported having a significantly higher level of responsibility for their treatment than their caregivers indicated and discrepant perceptions were fairly common in this study like in others.²⁴ In the current study, 2 tasks about which discrepancies were most common involved taking medication when away from home. Families may benefit from planning ahead with regard to who will be responsible for medication-related tasks when they are not in their home environment.

The 3 areas of knowledge assessed in our study (general HIV, adherence and regimen knowledge) yielded different results in terms of mean scores and correlations with other variables. Thus, conceptualizing and measuring knowledge as one broad construct may be an oversimplification and researchers should consider multiple aspects of disease knowledge in future studies with HIV-positive children and their caregivers. The majority of children and caregivers in our study provided accurate dosing schedule information. Because caregivers' regimen knowledge was related to adherence, repeated assessments of and education about the child's ARV regimen is important to ensure comprehension, particularly for individuals with documented cognitive difficulties such as memory or executive deficits.

Children's responsibility for medication increased with age according to both children and caregivers in the current investigation, but knowledge did not. Furthermore, child age was unrelated to adherence at both time points. This suggests that caregivers are giving children more responsibility for adherence as they get older despite the fact that the children are not demonstrating an increase in regimen knowledge or adherence as they mature. Caregivers and members of the treatment team should educate patients about their disease and ARV regimen before responsibility is increased. Consultation with a psychologist regarding the child's cognitive capacity can help tailor the information to a developmentally appropriate level.

Several threats to reliability and validity of MEMS data must be acknowledged. First, of the 17 families who agreed to use a MEMS cap, there were 10 instances in which unexplained bottle openings, as many as 9 in one day, had to be manually excluded. Two MEMS caps had to be replaced because they malfunctioned or broke. In addition, we chose to define an adherent dose as one taken within a 2-hour window of the scheduled dose time, although no data exist regarding whether 2 hours is a more appropriate time frame than 1 hour or 3 hours.

With respect to demographics, our patient sample included a higher percentage of whites compared with recent Centers for Disease Control and Prevention data²⁵ and a slightly higher caregiver education level than samples described in other recent publications.^10,26 However, no demographic characteristics were related to adherence in our study, which is consistent with several other investigations.^8,27 Other study limitations include the relatively small sample size and the fact that families who declined to participate or who chose not to enroll in the MEMS arm of the study may have been less adherent than those who agreed to participate and use MEMS.

Based on findings from this study, specific aims for providers should include clarifying responsibility for medication-related tasks to prevent or resolve discrepancies; solidifying caregiver and child knowledge about the ARV treatment regimen; and guiding caregivers through the gradual process of assigning more responsibility to older children and adolescents while still maintaining a supervisory role. These goals should be sought even when families claim perfect or near perfect adherence. Research is needed to design interventions geared toward achieving these goals. Future studies may want to incorporate multiple assessments of knowledge and perceptions of treatment responsibility to determine the effectiveness of these interventions and whether changes in these variables impact changes in adherence over time. The growing body of literature focused on medication-taking behavior among HIV-infected children affirms that adherence is being recognized as an important issue. Given the numerous questions about factors relevant to adherence that have not been answered, future investigations are needed to progress toward the goal of overcoming obstacles to adherence to HAART in the pediatric HIV population.

ACKNOWLEDGMENTS

The authors express their gratitude to the children and families who agreed to participate in this research. The authors also appreciate the contributions of Sarah Calabrese, BS, Rohan Hazra, MD, Lori Weiner, PhD, and the multidisciplinary team of the Pediatric Working Group, HIV and AIDS Malignancy Branch, National Institutes of Health.

This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, and by federal contracts N01-SC-07006 and HHSN261200477004C.