Abstract
To examine unannounced telephone pill counts as a measure of adherence to antiretroviral therapy among adolescents and young adults living with perinatal HIV infection.
Participants were recruited from an ongoing longitudinal study to complete four monthly, unannounced telephone pill counts. Detailed notes concerning participants’ medication habits surrounding adherence were recorded
Two-thirds of 102 eligible participants aged 18–27 years participated; 57% were female, 69% were Black. Blacks and participants with viral loads >40 and >1,000 copies/ml were less likely to participate. Average adherence across calls was 77%. Those who completed all calls averaged significantly higher adherence scores than those who did not. Calls revealed adherence barriers at individual (e.g., medication disorganization), social (e.g., limited support), and system (e.g., pharmacy problems) levels
Despite challenges, this procedure can be implemented with this population and can help identify adherence barriers important for interventions that address medication-taking behaviors.
Adolescence and young adulthood are developmental stages in which young people are at highest risk for poor adherence to medications across chronic health conditions, including HIV (Kim, Gerver, Fidler, and Ward, 2014). Suboptimal adherence places adolescents and young adults (AYAs) living with HIV at risk for poor health outcomes, including treatment failure and disease progression (Bangsberg, Perry, etal., 2001). AYAs have the lowest rates of viral suppression of all age-groups (CDC, 2016a). Furthermore, it is estimated that fewer than half of the AYAs living with perinatal HIV infection (AYAs with PHIV) in the United States are virally suppressed (Kahana etal., 2015), and they are at higher risk for nonadherence to antiretroviral therapy (ART) than AYAs who acquired HIV through sexual behavior (MacDonell, Naar-King, Huszti, and Belzer, 2013).
Although rates in the United States have declined over the past several decades, vertical transmission continues to be an issue in low- and middle-income countries (LMICs) where treatment access has been delayed and inconsistent. It is estimated that nearly 2 million children worldwide are living with perinatal HIV infection (UNAIDS, 2016), and rates of mother-to-child-transmission in some LMICs are as high as 30% (UNICEF, 2016). Thus, with approximately 400 children becoming infected every day and hundreds of thousands of children reaching adolescence and young adulthood, nonadherence is a significant global public health issue (UNAIDS, 2016).
AYAs with PHIV may miss or stop taking medication owing to treatment fatigue, desire to be “normal,” concerns about stigma and inadvertent disclosure, and cognitive problems associated with early infection (Claborn, Meier, Miller, and Leffingwell, 2015). AYAs with PHIV also have been shown to be at high risk for mental health and substance use problems, both of which negatively affect adherence (Bucek etal., 2016; Mellins etal., 2011). Unfortunately, both clinical care and research have been hampered by lack of a “gold standard” measure of pill-taking behaviors in determining factors associated with nonadherence and evaluating impact of interventions (Lehmann etal., 2014).
Existing methods of measuring adherence have both advantages and considerable limitations. Self-report, the most commonly used measure, is easy to implement and inexpensive. However, social desirability and poor recall compromise its accuracy (Usitalo etal., 2014). Electronic monitoring devices (EMDs) record date, time, and frequency of container openings but not whether, or how many, pills are ingested. Moreover, EMDs can be costly and burdensome (Adefolalu & Nkosi, 2013). Pharmacy refill records of time between refills have been used to estimate adherence. However, as patients may throw out pills or share medications, it is again difficult to assess pill-taking behavior (Lehmann etal., 2014). In pill counts conducted at clinic appointments, health-care providers measure adherence by comparing the number of pills a patient brings to the appointment with the number of pills s/he should have based on prescriptions. Although this method is objective and potentially more sensitive than other methods, patients can forget to bring in medications or discard unused doses (“pill dumping”) to appear more adherent (Lehmann etal., 2014).
To address some limitations of appointment-based pill counts, Bangsberg and colleagues (2001) evaluated the utility of unannounced pill counts conducted at participants’ homes. A research assistant (RA) visited the participant’s home at unannounced times to count all pills in a participant’s possession every 2–4 weeks over 3 months. Home-based pill counts were significantly associated with simultaneously collected EMD data. However, sending staff to a participant’s home unannounced requires considerable staff time and can be costly.
To reduce the cost and staff burden of unannounced home-based pill counts, Kalichman etal. (2007) evaluated the use of unannounced telephone-based pill counts. In a primarily Black, middle-aged HIV+ cohort (N = 77, Mage = 44 years), participants completed 13 monthly unannounced telephone pill counts. An unscheduled home pill count occurred immediately after a telephone pill count, and Kalichman etal. (2007) found no differences between the two methods, supporting the use of telephone-based pill counts.
Although unannounced telephone pill counts have been successfully implemented with adults living with HIV, there are no reports of its use with AYAs with PHIV. AYAs differ developmentally and may interact differently with phones than older populations, often preferring texting (Forgays, Hyman, and Schreiber, 2014). They may be less likely to participate in a protocol in which they need to spend time talking on the phone. AYAs with PHIV are one of the highest risk groups for ART nonadherence, and demonstrating preliminary feasibility of this procedure with this population may provide researchers as well as clinicians with an important new tool for measuring adherence.
We adapted and implemented Kalichman’s pill-count protocol as part of a longitudinal study of AYAs with PHIV conducted in New York City (NYC), home to 25% of the US population of children born with HIV (CDC, 2016b). We explored preliminary feasibility of this strategy by examining three study questions: (1) would AYAs with PHIV agree to enroll in the pill-count protocol, (2) what are the challenges to enrollment and implementation, and (3) are there demographic and psychosocial differences between participants who enrolled and were retained in the protocol compared with those who did not enroll or who were not retained? Furthermore, we examined notes collected by phone assessors at each call to identify potential barriers to adherence.
Methods
Participants and Procedures
Data come from Child and Adolescent Self-Awareness and Health (CASAH; Mellins etal., 2009), a longitudinal cohort study following AYAs with PHIV and uninfected youth who were perinatally exposed to HIV and their caregivers who were recruited from four medical institutions in NYC. Inclusion criteria at baseline were (1) youth aged 9–16 years with perinatal exposure to HIV; (2) cognitive capacity to complete interviews; (3) English- or Spanish-speaking (caregivers only, all youth were English speaking); and (4) residing with a caregiver who could give permission for youth’s participation. Data for this article are from the sixth CASAH interview (Follow-up 5; FU5), when unannounced telephone pill counts were added to the protocol. To be eligible for pill counts, participants had to be enrolled in CASAH, be living with PHIV, be prescribed ART, have access to a phone, and not be enrolled in directly observed therapy.
Measures
Unannounced Telephone Pill Counts
AYAs with PHIV were invited to participate in the pill-count protocol over the 4 months following their CASAH FU5 interview. Each participant completed a pill-count training session at the end of the interview. For eligible participants who consented to audio recording of the interview and then enrolled in the pill count (n = 77), the average training session was 19.53 min (SD = 5.81, range = 9.25–34.27). Training included (1) how to read pharmacy-label information; (2) how to count pills; and (3) a medication-taking habits survey that asked about ART regimen, medication storing, pharmacy information, and other behaviors such as sharing pills. RAs also collected alternative contact information to be used if we lost contact with participants.
The CASAH Project Director and Coordinator were trained by Kalichman’s team and, in turn, trained CASAH Bachelor/Master-level RAs to conduct telephone pill counts. A trained RA completed a baseline call with the participant within a week of the psychosocial interview. The assessor confirmed the participant’s ART regimen and recorded the number of containers (bottles, blister packs, or rolls) in the home for each medication. For each antiretroviral medication or medication combination, the assessor collected pharmacy-label information (i.e., dosing instructions, prescription number, refill date, number of refills, pharmacy name, quantity prescribed, and milligrams per pill). Finally, the assessor had the participant count aloud the number of pills in each container twice to confirm the number reported.
The assessor called the participant again approximately 1 month later (28–35 days) to repeat the baseline-call procedure and ask additional questions exploring pill-taking behavior between calls. The participant was asked if s/he had any “losses” or “gains” in medication supply. A “loss” can occur if a participant throws out or loses medication, gives medication to someone else, or leaves medication at a location other than the home. “Gains” can occur if a participant is dispensed medication from the pharmacy or hospital, borrows medication from someone, or finds pills that were not counted at prior calls. The second call resulted in the first adherence score. The participant was telephoned twice more over the next 2 months, resulting in a total of three adherence scores for each medication.
Call Notes
During each call, the assessor recorded the participant’s answers to open-ended questions about losses or gains in pill supply with detailed notes, gathering more qualitative information through probes that provided insight into medication behavior. For example, if a participant had a gain in pills, the phone assessor would record it and then ask questions about the source of those pills. In addition, a participant might spontaneously offer information—for instance, that s/he had another bottle at a different location (i.e., caregiver’s house) or received extra medication while hospitalized. Assessors kept detailed notes on such issues but were instructed not to intervene.
Pill-Count Adherence Scores
An adherence score was calculated as the difference between pills counted at two consecutive calls divided by the number of pills that should have been taken in the intervening period (based on dosing instructions and any reported losses or gains in medication). For the analysis, adherence scores were averaged across medications to assess regimen-level adherence, which represents the best index of reaching drug levels sufficient for viral suppression. Furthermore, there was a high correlation of adherence scores between medications within the same time point (.94–.98).
Pill-Count Process Measures
For each participant, we documented contact attempts (number of call attempts it took for a phone assessor to complete each pill-count call) and number of days between completed pill counts.
Demographics
Demographic variables collected at the FU5 interview included participants’ age, gender, ethnicity, race, education, socioeconomic status (SES), and household composition. Given that more than two-thirds of CASAH participants are Blacks and/or Latino, as were the majority of children born with HIV in NYC (NYCDOH, 2016), we examined ethnicity and race as two dichotomized variables: Hispanic/Latino (yes vs. no) and Black (yes vs. no). Education was trichotomized as not having a high school diploma versus having a high school diploma/General Educational Development diploma versus having some college or higher education. Many participants could not report exact income for all household members, but were able to confirm receipt of supplemental income benefits available to low-income households (i.e., public assistance, food stamps). Therefore, receipt of supplemental income (yes vs. no) was used as a proxy for SES. Household composition was dichotomized as living with an adult caregiver versus not living with a caregiver.
Mental Health Status
Psychiatric functioning was assessed in the FU5 interview using the Diagnostic Interview Schedule for Children (DISC-IV), young adult version (Shaffer etal., 1996). This structured instrument asks participants about symptoms of psychiatric diagnoses as defined by the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders system (APA, 1994). Participants were screened for anxiety, mood, disruptive behavior, and substance use disorders (SUD). For this article, we created two dichotomized variables reflecting whether the participant met criteria for “any nonsubstance psychiatric disorder” (yes vs. no) or for SUD (yes vs. no).
Medication-Taking Habits Survey
During the pill-count training session, the CASAH RA administered quantitative questions about how participants obtain, store, and take their medications. We examined two questions from this survey: means of obtaining medication (pickup vs. home delivery) and medication packaging (original pharmacy bottles vs. other storage methods).
Self-Reported Adherence
Self-reported adherence was examined using a scale administered via Audio Computer-Assisted Self-Interview (SSI Web, 2016) during the FU5 interview and at every call in which an adherence score was obtained (Calls 2–4). Questions were (1) In the last 30 days, how good a job did you do at taking your HIV medicines in the way you were supposed to (very poor to excellent); (2) In the last 30 days, how often did you take your HIV medicines in the way you were supposed to (never to always); (3) How hard is it for you to take your HIV medicines in the way you are supposed to (extremely hard to not hard at all; Wilson, Lee, Michaud, Fowler, and Rogers, 2016); and (4) When was the last time you missed any of your medications (within the past week to never; Chesney etal., 2000)? The mean of the four items was calculated, with higher scores indicating better adherence. Alpha for the four-item scale was .87. We also examined responses to the last item (When was the last time you missed any of your medications?) dichotomized as ≤1 month versus >1 month or never.
HIV RNA Viral Load
HIV RNA Viral load (VL) data from the 12 months before the FU5 interview were collected from medical charts. The VL value from the date closest to the interview for each participant was used to characterize the cohort and examine differences in participation. VL was dichotomized both as ≤40 versus >40 copies/ml (i.e., the lower limit of the laboratory assays) and as ≤1,000 versus >1,000 copies/ml (Olds, Kiwanuka, Ware, Tsai, and Haberer, 2015).
Data Analysis
Descriptive data on all study variables are presented, including numbers of participants who were eligible and who agreed to participate. Chi-squared analyses compared characteristics (e.g., medication-taking behaviors, self-reported adherence) of enrolled participants who contributed at least one adherence score with those who did not contribute any adherence scores. To better understand preliminary feasibility and fidelity to the protocol, we averaged across time points the number of call attempts assessors made to complete a pill count and the number of days between pill-counts calls. Finally, t tests were used to compare average adherence scores of participants who completed the entire protocol (three adherence scores) with those who contributed only one or two scores. Phone assessors coded the call notes into narrative themes, reviewed all of the thematic groups, and categorized them into broader topics. Three participants were dropped from analyses because they provided invalid adherence data.
Results
Study Enrollment
Participant flow through the study is presented in Figure 1. Among AYAs with PHIV who completed the parent study psychosocial interview at the time of analysis (N = 114), 87% (n = 102) were eligible for the pill counts. Reasons for ineligibility included not currently prescribed ART (n = 4), no phone access (n = 2), and enrolled in directly observed therapy (n = 6). Only six participants refused to participate. Enrolled participants (n = 96) were 18–27 years old (M = 22.81, SD = 2.62) and 57% female; 69% identified as Black and 49% as Hispanic/Latino; and 62% were on ART regimens requiring more than one pill per day.
Enrollment and Retention of Participants in the Pill-Count Protocol
Contributing Pill Count Adherence Data
The majority (72%) of enrolled participants contributed adherence data, defined as completing a baseline call to establish number of pills and at least one additional call to determine an adherence score. Only 58% completed the entire four-call protocol.
Implementation of Unannounced Telephone Pill-Count Protocol
Phone assessors required, on average, four call attempts (range = 1–14, SD = 2.53) at each time point to complete a pill count. On average, there were 33 days between pill counts (range = 28–46, SD = 3.40). Although we were able to contact most participants within the monthly window (28–35 days), this required multiple calls at different hours of the day and evening. Some participants had multiple jobs with changing hours, making contact difficult. Others were not employed or in school and had no reliable routine. Sometimes a participant was reached at an inconvenient time (e.g., when with someone who did not know the participant’s HIV status). When a phone assessor had difficulty reaching a participant, designated alternative contact means, such as email and caregivers’ phone numbers, were used. Often participants could not afford regular cell phone service, or they changed phone numbers. Other participants either never answered or stopped answering our phone calls.
Call length and difficulty varied. Some calls were brief and easy (approximately 10 min), with participants able to readily locate label information and account for all pills. Other calls were more challenging (up to 35 min). According to notes collected by phone assessors during a call, many participants accumulated numerous bottles of medication from periods of nonadherence, sometimes having multiple open bottles from which doses were taken at random. In the call notes, this was indicated by information such as the number of refills left and pharmacy refill dates. Collecting this information made the calls longer and more difficult, particularly given the double counting requirement.
Medication Adherence as Determined by Pill Count
The mean adherence score across participants and calls was 77%, ranging widely from 15% to 100% (SD = 20.86). Some participants reported taking more pills than prescribed, and therefore, were “overadherent” (>100%). For analysis, their scores were truncated to 100%.
Self-Reported Adherence
Average self-reported adherence across all calls was 3.9 of 6 (a higher score is indicative of better adherence). The correlation between average self-reported adherence and average pill-count score at each call was statistically significant, .38–.58, p < .01.
Differences Between Participants Who Did and Did Not Contribute Adherence Data
Chi-squared analyses were used to differentiate between eligible participants who contributed adherence data and those who did not. Group characteristics and differences are presented in Table I. Participants with VL > 40 copies/ml at the time closest to the FU5 interview were significantly less likely to contribute an adherence score (p < .02), as were those with VL > 1,000 copies/ml (p < .04). Blacks were significantly less likely than non-Blacks to contribute an adherence score, p < .02. No other variables examined were statistically significant (e.g., age, gender, SES, education, psychiatric disorder). By an additional analysis, using a chi-squared test, participants identifying as Black were not any more likely than other participants to have a VL > 40 or >1,000 copies/ml, p > .05 (data not shown).
Comparison of Eligible Participants With and Without Adherence Scores
| With (n = 69) | Without (n = 33) | t (df) | p | |
|---|---|---|---|---|
| M (SD); range | M (SD); range | |||
| Age | 22.73 (2.55); 18–26 | 22.98 (2.78); 18–27 | 0.46 (100) | .65 |
| Self-reported adherence at interview | 4.04 (1.32); 1–6 | 3.67 (1.30); 1–6 | −1.28 (96) | .20 |
| 1n (%) | 1n (%) | χ2 (df) | p | |
| Female | 42 (61) | 16 (48) | 1.40 (1) | .24 |
| Male | 27 (39) | 17 (52) | ||
| Hispanic/Latino | 38 (55) | 12 (36) | 3.13 (1) | .08 |
| Non-Hispanic/Latino | 31 (45) | 21 (64) | ||
| Black | 42 (61) | 28 (85) | 5.96 (1) | .02* |
| Non-Black | 27 (39) | 5 (15) | ||
| Receiving supplemental income | 42 (62) | 22 (67) | 0.23 (1) | .63 |
| Not receiving supplemental income | 26 (38) | 11 (33) | ||
| Education: | ||||
| Less than High School Diploma | 17 (25) | 11 (33) | 0.99 (2) | .61 |
| High School/General Educational Development Diploma | 33 (48) | 15 (46) | ||
| Some college and above | 19 (28) | 7 (21) | ||
| Living with caregiver | 32 (46) | 17 (53) | 0.40 (1) | .53 |
| Not living with caregiver | 37 (54) | 15 (47) | ||
| Uses original bottles | 40 (59) | 19 (58) | 0.14 (1) | .91 |
| Does not use original bottles | 28 (41) | 14 (42) | ||
| Medication pickup required | 10 (15) | 7 (23) | 0.93 (1) | .34 |
| Medication pickup not required | 58 (85) | 24 (77) | ||
| Viral load > 40 | 28 (45) | 17 (74) | 5.57 (1) | .02* |
| Viral load ≤ 40 | 34 (55) | 6 (26) | ||
| Viral load > 1,000 | 15 (24) | 11 (48) | 4.41 (1) | .04* |
| Viral load ≤ 1,000 | 47 (76) | 12 (52) | ||
| Missed meds in the past month | 39 (57) | 21 (72) | 1.95 (1) | .16 |
| Did not miss meds in the past month | 29 (43) | 8 (28) | ||
| Any psychiatric disorder | 20 (29) | 10 (33) | 0.19 (1) | .67 |
| No psychiatric disorder | 49 (71) | 20 (67) | ||
| Substance use disorder | 17 (25) | 6 (20) | 0.29 (1) | .59 |
| No substance use disorder | 51 (75) | 24 (80) |
| With (n = 69) | Without (n = 33) | t (df) | p | |
|---|---|---|---|---|
| M (SD); range | M (SD); range | |||
| Age | 22.73 (2.55); 18–26 | 22.98 (2.78); 18–27 | 0.46 (100) | .65 |
| Self-reported adherence at interview | 4.04 (1.32); 1–6 | 3.67 (1.30); 1–6 | −1.28 (96) | .20 |
| 1n (%) | 1n (%) | χ2 (df) | p | |
| Female | 42 (61) | 16 (48) | 1.40 (1) | .24 |
| Male | 27 (39) | 17 (52) | ||
| Hispanic/Latino | 38 (55) | 12 (36) | 3.13 (1) | .08 |
| Non-Hispanic/Latino | 31 (45) | 21 (64) | ||
| Black | 42 (61) | 28 (85) | 5.96 (1) | .02* |
| Non-Black | 27 (39) | 5 (15) | ||
| Receiving supplemental income | 42 (62) | 22 (67) | 0.23 (1) | .63 |
| Not receiving supplemental income | 26 (38) | 11 (33) | ||
| Education: | ||||
| Less than High School Diploma | 17 (25) | 11 (33) | 0.99 (2) | .61 |
| High School/General Educational Development Diploma | 33 (48) | 15 (46) | ||
| Some college and above | 19 (28) | 7 (21) | ||
| Living with caregiver | 32 (46) | 17 (53) | 0.40 (1) | .53 |
| Not living with caregiver | 37 (54) | 15 (47) | ||
| Uses original bottles | 40 (59) | 19 (58) | 0.14 (1) | .91 |
| Does not use original bottles | 28 (41) | 14 (42) | ||
| Medication pickup required | 10 (15) | 7 (23) | 0.93 (1) | .34 |
| Medication pickup not required | 58 (85) | 24 (77) | ||
| Viral load > 40 | 28 (45) | 17 (74) | 5.57 (1) | .02* |
| Viral load ≤ 40 | 34 (55) | 6 (26) | ||
| Viral load > 1,000 | 15 (24) | 11 (48) | 4.41 (1) | .04* |
| Viral load ≤ 1,000 | 47 (76) | 12 (52) | ||
| Missed meds in the past month | 39 (57) | 21 (72) | 1.95 (1) | .16 |
| Did not miss meds in the past month | 29 (43) | 8 (28) | ||
| Any psychiatric disorder | 20 (29) | 10 (33) | 0.19 (1) | .67 |
| No psychiatric disorder | 49 (71) | 20 (67) | ||
| Substance use disorder | 17 (25) | 6 (20) | 0.29 (1) | .59 |
| No substance use disorder | 51 (75) | 24 (80) |
Note.1n may not sum to total n owing to missing data.
p < .05.
Comparison of Eligible Participants With and Without Adherence Scores
| With (n = 69) | Without (n = 33) | t (df) | p | |
|---|---|---|---|---|
| M (SD); range | M (SD); range | |||
| Age | 22.73 (2.55); 18–26 | 22.98 (2.78); 18–27 | 0.46 (100) | .65 |
| Self-reported adherence at interview | 4.04 (1.32); 1–6 | 3.67 (1.30); 1–6 | −1.28 (96) | .20 |
| 1n (%) | 1n (%) | χ2 (df) | p | |
| Female | 42 (61) | 16 (48) | 1.40 (1) | .24 |
| Male | 27 (39) | 17 (52) | ||
| Hispanic/Latino | 38 (55) | 12 (36) | 3.13 (1) | .08 |
| Non-Hispanic/Latino | 31 (45) | 21 (64) | ||
| Black | 42 (61) | 28 (85) | 5.96 (1) | .02* |
| Non-Black | 27 (39) | 5 (15) | ||
| Receiving supplemental income | 42 (62) | 22 (67) | 0.23 (1) | .63 |
| Not receiving supplemental income | 26 (38) | 11 (33) | ||
| Education: | ||||
| Less than High School Diploma | 17 (25) | 11 (33) | 0.99 (2) | .61 |
| High School/General Educational Development Diploma | 33 (48) | 15 (46) | ||
| Some college and above | 19 (28) | 7 (21) | ||
| Living with caregiver | 32 (46) | 17 (53) | 0.40 (1) | .53 |
| Not living with caregiver | 37 (54) | 15 (47) | ||
| Uses original bottles | 40 (59) | 19 (58) | 0.14 (1) | .91 |
| Does not use original bottles | 28 (41) | 14 (42) | ||
| Medication pickup required | 10 (15) | 7 (23) | 0.93 (1) | .34 |
| Medication pickup not required | 58 (85) | 24 (77) | ||
| Viral load > 40 | 28 (45) | 17 (74) | 5.57 (1) | .02* |
| Viral load ≤ 40 | 34 (55) | 6 (26) | ||
| Viral load > 1,000 | 15 (24) | 11 (48) | 4.41 (1) | .04* |
| Viral load ≤ 1,000 | 47 (76) | 12 (52) | ||
| Missed meds in the past month | 39 (57) | 21 (72) | 1.95 (1) | .16 |
| Did not miss meds in the past month | 29 (43) | 8 (28) | ||
| Any psychiatric disorder | 20 (29) | 10 (33) | 0.19 (1) | .67 |
| No psychiatric disorder | 49 (71) | 20 (67) | ||
| Substance use disorder | 17 (25) | 6 (20) | 0.29 (1) | .59 |
| No substance use disorder | 51 (75) | 24 (80) |
| With (n = 69) | Without (n = 33) | t (df) | p | |
|---|---|---|---|---|
| M (SD); range | M (SD); range | |||
| Age | 22.73 (2.55); 18–26 | 22.98 (2.78); 18–27 | 0.46 (100) | .65 |
| Self-reported adherence at interview | 4.04 (1.32); 1–6 | 3.67 (1.30); 1–6 | −1.28 (96) | .20 |
| 1n (%) | 1n (%) | χ2 (df) | p | |
| Female | 42 (61) | 16 (48) | 1.40 (1) | .24 |
| Male | 27 (39) | 17 (52) | ||
| Hispanic/Latino | 38 (55) | 12 (36) | 3.13 (1) | .08 |
| Non-Hispanic/Latino | 31 (45) | 21 (64) | ||
| Black | 42 (61) | 28 (85) | 5.96 (1) | .02* |
| Non-Black | 27 (39) | 5 (15) | ||
| Receiving supplemental income | 42 (62) | 22 (67) | 0.23 (1) | .63 |
| Not receiving supplemental income | 26 (38) | 11 (33) | ||
| Education: | ||||
| Less than High School Diploma | 17 (25) | 11 (33) | 0.99 (2) | .61 |
| High School/General Educational Development Diploma | 33 (48) | 15 (46) | ||
| Some college and above | 19 (28) | 7 (21) | ||
| Living with caregiver | 32 (46) | 17 (53) | 0.40 (1) | .53 |
| Not living with caregiver | 37 (54) | 15 (47) | ||
| Uses original bottles | 40 (59) | 19 (58) | 0.14 (1) | .91 |
| Does not use original bottles | 28 (41) | 14 (42) | ||
| Medication pickup required | 10 (15) | 7 (23) | 0.93 (1) | .34 |
| Medication pickup not required | 58 (85) | 24 (77) | ||
| Viral load > 40 | 28 (45) | 17 (74) | 5.57 (1) | .02* |
| Viral load ≤ 40 | 34 (55) | 6 (26) | ||
| Viral load > 1,000 | 15 (24) | 11 (48) | 4.41 (1) | .04* |
| Viral load ≤ 1,000 | 47 (76) | 12 (52) | ||
| Missed meds in the past month | 39 (57) | 21 (72) | 1.95 (1) | .16 |
| Did not miss meds in the past month | 29 (43) | 8 (28) | ||
| Any psychiatric disorder | 20 (29) | 10 (33) | 0.19 (1) | .67 |
| No psychiatric disorder | 49 (71) | 20 (67) | ||
| Substance use disorder | 17 (25) | 6 (20) | 0.29 (1) | .59 |
| No substance use disorder | 51 (75) | 24 (80) |
Note.1n may not sum to total n owing to missing data.
p < .05.
One additional analysis was conducted comparing the average adherence of participants who completed the entire four-call protocol with that of those who contributed only one or two adherence scores. Those who completed the protocol had significantly higher average adherence scores (79% [n = 56; SD = 19.52]) than participants who only contributed one or two adherence scores (65% [n = 13; SD = 23.37], p < .04).
Narrative Analysis of Call Notes
During the calls, assessors gained insight into the participant’s pill-taking behaviors and circumstances. These “adherence stories,” which were expressed spontaneously or in response to probes about gains or losses in pills, illustrated barriers to adherence faced by some AYAs with PHIV. As presented in Table II, barriers were categorized as individual, social, and system level. Individual barriers included lack of understanding of medications prescribed (i.e., participants could not differentiate between ART and non-HIV medications). Other participants said that feelings of sadness or feeling overwhelmed interfered with adherence. Social issues, such as limited support and experiencing major life events (e.g., death of a caregiver or breaking up with a partner), also surfaced during the calls. Major life events often led participants to leave their places of residence, leaving medication behind. Finally, participants disclosed system barriers to maintaining adherence, such as suboptimal pharmacy services (e.g., late delivery or delivery of the wrong medication), and health insurance disruptions that left them without medication. Some barriers were more widely reported, such as medication disorganization and privacy concerns. Conversely, some were uncommon in our population (e.g., sharing medication), but were important to report as they interfered with the implementation and feasibility of the protocol.
Barriers to Medication Adherence and Implementation of Pill-Count Protocol
| Barrier level | Challenges to completing a call | Barriers to adherence |
|---|---|---|
| Individual | Busy/changing schedules | Busy/changing schedules |
| Lack of a reliable routine or unemployment | Lack of a reliable routine or unemployment | |
| Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | |
| Failure to keep a phone number in service; poor cell phone reception | Feeling down or having a depressive episode | |
| Sharing phones with others | ||
| Inability to correctly read pharmacy-label information | ||
| Sharing medication with someone else | ||
| Social | Lack of privacy/privacy concerns | Lack of privacy/privacy concerns |
| Homelessness; housing instability | Homelessness; housing instability | |
| Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | |
| Limited support systems to assume medication management | ||
| System | Pharmacy incorrectly prints medication label information | Pharmacy incorrectly delivers medication |
| Medication deliveries are not on time | ||
| No health insurance |
| Barrier level | Challenges to completing a call | Barriers to adherence |
|---|---|---|
| Individual | Busy/changing schedules | Busy/changing schedules |
| Lack of a reliable routine or unemployment | Lack of a reliable routine or unemployment | |
| Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | |
| Failure to keep a phone number in service; poor cell phone reception | Feeling down or having a depressive episode | |
| Sharing phones with others | ||
| Inability to correctly read pharmacy-label information | ||
| Sharing medication with someone else | ||
| Social | Lack of privacy/privacy concerns | Lack of privacy/privacy concerns |
| Homelessness; housing instability | Homelessness; housing instability | |
| Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | |
| Limited support systems to assume medication management | ||
| System | Pharmacy incorrectly prints medication label information | Pharmacy incorrectly delivers medication |
| Medication deliveries are not on time | ||
| No health insurance |
Barriers to Medication Adherence and Implementation of Pill-Count Protocol
| Barrier level | Challenges to completing a call | Barriers to adherence |
|---|---|---|
| Individual | Busy/changing schedules | Busy/changing schedules |
| Lack of a reliable routine or unemployment | Lack of a reliable routine or unemployment | |
| Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | |
| Failure to keep a phone number in service; poor cell phone reception | Feeling down or having a depressive episode | |
| Sharing phones with others | ||
| Inability to correctly read pharmacy-label information | ||
| Sharing medication with someone else | ||
| Social | Lack of privacy/privacy concerns | Lack of privacy/privacy concerns |
| Homelessness; housing instability | Homelessness; housing instability | |
| Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | |
| Limited support systems to assume medication management | ||
| System | Pharmacy incorrectly prints medication label information | Pharmacy incorrectly delivers medication |
| Medication deliveries are not on time | ||
| No health insurance |
| Barrier level | Challenges to completing a call | Barriers to adherence |
|---|---|---|
| Individual | Busy/changing schedules | Busy/changing schedules |
| Lack of a reliable routine or unemployment | Lack of a reliable routine or unemployment | |
| Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | Medication disorganization (i.e., multiple bottle of the same medication open at the same time, storing medication at several locations) | |
| Failure to keep a phone number in service; poor cell phone reception | Feeling down or having a depressive episode | |
| Sharing phones with others | ||
| Inability to correctly read pharmacy-label information | ||
| Sharing medication with someone else | ||
| Social | Lack of privacy/privacy concerns | Lack of privacy/privacy concerns |
| Homelessness; housing instability | Homelessness; housing instability | |
| Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | Major life events (i.e., death of a loved one, breaking up with a boy/girlfriend) | |
| Limited support systems to assume medication management | ||
| System | Pharmacy incorrectly prints medication label information | Pharmacy incorrectly delivers medication |
| Medication deliveries are not on time | ||
| No health insurance |
Discussion
To our knowledge, CASAH is the first study to use unannounced telephone pill counts to assess ART adherence among AYAs with PHIV. Preliminary feasibility was explored by addressing three study aims. Our first aim was to determine whether participants would agree to enroll in the unannounced telephone pill-count protocol. Based on our participants’ low refusal rate (6%) and high enrollment rate (94%) as well as the high proportion of enrolled participants providing at least one adherence score (72%), our data suggest that AYAs with PHIV can be successfully engaged in this protocol.
Addressing our second aim, we identified challenges to enrollment and implementation. More participants than expected were ineligible for the pill-count protocol (n = 12), and only 58% of those enrolled completed all four calls. Although the calls were completed on average within the designated window of 28–35 days, this required staff time during nontypical work hours, which may be time- and cost-prohibitive in some settings. For participants who were difficult to contact because their cell phones went out of service, collecting alternative contact information was critical for retention. Although giving participants dedicated study cell phones (Kalichman etal., 2007) might increase retention rates, this may be cost-prohibitive. In addition, other non-phone-related factors such as unpredictable schedules and lack of privacy made it difficult to complete these unscheduled calls. Further research is needed to determine in greater detail why some participants were maintained in the protocol and others were not.
Our third aim was to evaluate differences between participants who were enrolled and retained in the protocol and those who were not. We found that participants who contributed all three adherence scores had a higher average adherence than those who contributed one or two scores. Thus, nonparticipation in the pill counts may have been a proxy for suboptimal adherence if those having trouble taking medications as prescribed did not want to participate in a protocol that tracked their adherence behavior. These participants may have been more disorganized or less well functioning, and thus, found the calls more challenging to complete. This highlights the importance of establishing a nonjudgmental context for the pill-count calls. In addition, participants who have trouble adhering to a medication regimen may also experience similar barriers to sustaining adherence to a study protocol over several months.
Black participants were significantly less likely to contribute adherence data; this may be for a number of reasons, including distrust of research (George, Duran, and Norris, 2013). Given that we identified many barriers to study implementation that had a financial component (i.e., phone service), we expected that participants would differ in participation by SES; yet, we found no significant differences. However, there is limited variance in SES among our participants. Further research is needed in this area given the concentration of the US HIV epidemic among Blacks (CDC, 2016a).
Additionally, we found a higher adherence rate determined by pill count in our cohort than that measured by VL and self-report in other studies. A recent meta-analysis revealed ART adherence in adolescents was 53% in North America and 62% globally (Kim, Gerver, Fidler, and Ward, 2014), whereas overall pill-count adherence in our cohort was 77%. This could be owing to our participants underreporting the number of pills in their possession, which would artificially inflate their adherence score. It is also possible that because our participants were enrolled in CASAH they may be representative of the higher-functioning AYAs with PHIV. Similarly, participants were originally recruited from NYC HIV primary care clinics that have substantial resources for treating this population. Thus, our participants may not reflect AYAs with PHIV who live in more resource-constrained areas of the United States or the world, or who are behaviorally infected.
Unannounced telephone pill counts are a unique method of adherence measurement. Although the protocol requires effort to implement, unlike other measures (i.e., self-report, EMDs), it provides information on participants’ pill-taking behaviors and barriers that offers lessons for providers and researchers alike. The “adherence stories” recounted by participants revealed individual, social, and system barriers previously shown to have a negative impact on medication adherence (Chandwani etal., 2012; Kahana etal., 2016; MacDonell etal., 2013).
Retention in the protocol and ART adherence may be related—or at least driven by similar factors—that is, the same behaviors and situations that made calls more difficult to complete may contribute to poor adherence. For example, a participant’s inconsistent schedule was sometimes a barrier to completion of pill counts and, based on our call notes, could be a barrier to medication taking, as documented in previous studies (Chandwani etal., 2012). Many participants who were not employed or not in school did not have a reliable routine. Not only can this be problematic for completing unannounced pill-count calls, it can also negatively affect adherence, as participants do not have a structured day with potential reminders to take their medicines. Unemployment can be detrimental in other ways that adversely influence adherence (e.g., resulting in food insecurity; Kalichman etal., 2011). Moreover, several participants identified “feeling down” as a reason for not taking their medication, again similar to previous studies (Grenard etal., 2011).
Similar to these individual barriers, social barriers may contribute both to poor ART adherence and poor retention in the pill-count protocol. Phone assessors identified various social barriers to adherence, particularly unstable housing, which has been shown to be associated with poor adherence among adults (Aidala etal., 2016). Related, living situations that do not afford privacy can lead to accidental disclosure—a known barrier to adherence in AYAs (MacDonell etal., 2013). Likewise, having roommates and friends who did not know the participant’s status made it challenging to complete the calls.
Lastly, system barriers made medication adherence challenging for participants. For example, participants reported that pharmacy delivery problems, although rare, left them without medication for various periods of time. Lack of health insurance was another cited barrier to adherence. Access to prescription services and treatment is critical to keeping AYAs with PHIV adherent. Note that, although this was not an intervention study, referrals back to clinics were made if participants identified system-level barriers.
There are several limitations to interpreting our results. We did not document reasons for refusal (n = 6) or call duration, which would be helpful in future studies. We also found some adherence scores that indicated adherence >100%. Although it is possible that some participants miscounted pills or intentionally inflated their adherence scores, Kalichman etal. (2007) determined that it would be nearly impossible for participants to make the calculations needed to determine how many pills should be in their possession at the unannounced call. Thus, adherence scores >100% are most likely owing to participants accidentally double dosing or underestimating losses in medication supply.
It was beyond the scope of this study to validate the pill-count protocol as done by Kalichman etal. (2007) in conducting a home-based pill count immediately after a telephone pill count; however, future research should replicate the original validation study. Future studies also should validate this pill-count protocol with biomedical markers of adherence. We did find a significant correlation between self-reported adherence and pill-count scores, but our self-report measure did not ask participants to estimate the number of pills missed in the past month, and so the data were not directly comparable. Also, the protocol did not include a priori method of rigorously assessing adherence challenges although those challenges did emerge in our calls. Future studies should consider adding questions on barriers and facilitators to further inform evidence-based adherence interventions for HIV as well as other chronic illnesses.
Acknowledgments
The authors thank all of the individuals who participated in this study. In addition, they thank Amy Weintraub, Christina Amaral, Tamar Grebler, Ginger Hoyt, and Cynthia Merly for their continued support of the CASAH phone assessors.
Funding
This work was supported by a grant from the National Institute of Mental Health (R01-MH69133, PI: Claude Ann Mellins, PhD) and center grant from the National Institute of Mental Health to the HIV Center for Clinical and Behavioral Studies at NY State Psychiatric Institute and Columbia University (P30-MH43520; PI: Robert H. Remien, PhD).
Conflicts of interest: None declared.
References
Center for Disease Control and Prevention (CDC). (12 December
New York City Department of Health and Mental Hygiene (NYCDOH). (December
