Measuring Adherence to Antiretroviral Therapy

in a Diverse Population Using a

Visual Analogue Scale
Thomas P. Giordano, MD,1 David Guzman, MS,2 Richard Clark, MPH,4
Edwin D. Charlebois, MPH, PhD,2 and David R. Bangsberg, MD, MPH2,3
1
Sections of Infectious Diseases and Health Services Research, Department of Medicine,
Baylor College of Medicine, Houston, Texas; 2Epidemiology and Prevention Interventions Center,
Division of Infectious Diseases and the 3Positive Health Program, San Francisco General Hospital,
San Francisco, California; 4Department of Epidemiology and Biostatistics,
University of California at San Francisco, San Francisco, California, USA

Purpose: To examine the performance of an instrument to assess adherence based

on a visual analogue scale, compared to an instrument based on 3-day recall, using
unannounced pill counts in the place of residence as the gold standard. Method:
We prospectively assessed adherence to antiretroviral therapy in 84 marginally
housed indigent HIV-infected patients who were receiving stable antiretroviral
therapy in San Francisco, California, with three adherence assessments over no
more than 4 months. Results: Mean adherence using the visual analogue scale, 3-
day recall, and unannounced pill count methods were 82.5%, 84.2%, and 75.9%,
respectively. The correlation between visual analogue scale and unannounced pill
count was high (r = 0.76) and was not statistically different from that between 3-day
recall and unannounced pill count (r = 0.71; p = .52). Both methods were also
similarly inversely correlated with HIV viral load (r = -0.49 and -0.34, respectively; p =
.22 for the difference in the correlations). The visual analogue scale correlation with
unannounced pill count was stable over time and remained high in all
subpopulations examined. Conclusion: A visual analogue scale to assess
adherence was performed as well as a more complicated 3-day recall instrument in
this diverse population. Given its simplicity, the visual analogue scale adherence
instrument will be useful in research and may be useful in routine patient care. Key
words: adherence, antiretroviral therapy, visual analogue scale

n persons infected with HIV, the importance of

I
adherence to antiretroviral medications has been
well demonstrated. Adherence to HIV
antiretroviral therapy is closely associated with
viral suppression,1–4 antiretroviral resistance,5–8
progression to AIDS,5 and death.9,10 Nonetheless,
assessing adherence to medication regimens is not
straightforward, and the inability to quickly and
inexpensively obtain accurate measures of
adherence has hindered research and patient
care.11 Even though computerized medication caps
that record bottle openings are useful in research,
practical considerations and cost limit their use in
clinical practice.12,13 Pill counts are also often used
in research, but they are time consuming and too
complex for routine care.13 Patient self-assessed
adherence therefore is often used in practice and
research. 14 To improve precision, and to allay
concerns of recall bias, most self-assessment
adherence tools ask patients to recall the specific
number of missed doses in a brief time period,
typically the last 2 to 7 days.14,15 This medication-
by-medication and day-by-day recall of adherence
requires a complicated interview. For example, a 3-
day recall series of questions for a patient on three
antiretroviral medications requires detailed
instructions and at least nine questions.
In an effort to simplify the assessment of adher-
ence, we tested a visual analogue scale based on the
work of Walsh16,17 that simply asks the patient to
For correspondence or reprints contact: Dr. D.R. Bangsberg,
Epidemiology and Prevention Interventions Center, San Francisco
General Hospital, UCSF, PO Box 1372, San Francisco, California
94143-1372 USA. Email:db@epi-center.ucsf.edu
HIV Clin Trials 2004;5(2):74–79
© 2004 Thomas Land Publishers, Inc.
www.thomasland.com

74
 VISUAL ANALOGUE SCALE FOR MEASURING ADHERENCE
indicate by marking on a linear scale their adher-
ence to each medication in the last 3 to 4 weeks. A
visual analogue scale has a number of potential
advantages over 3-day self-reported adherence: it
is more quickly assessed, it is able to obtain data
about a longer time frame, and it places a lower
response burden on the patient. Although a visual
analogue scale has been used in cross-sectional
studies and its results have correlated with HIV
viral load,16,17 in the former study there was no
objective assessment of adherence and in the latter
study the gold standard assessment was with a
computerized medication cap on a single medica-
tion read at a single scheduled visit. Such objective
assessments of adherence at scheduled visits are
subject to the “toothbrush effect” in which adher-
ence is improved or pills are dumped just prior to a
visit. Further, the study population had high ad-
herence (93%) by computerized medication cap,
86% were gay men, and only 5% had a history of
injection drug use. Thus, this population is not rep-
resentative of the present HIV epidemic, and the
performance of a visual analogue scale in a broad
population of patients with a range of adherence
remains unknown. We therefore prospectively
compared the performances of standard 3-day self-
report and visual analogue reports with unan-
nounced pill counts conducted at patients’ usual
place of residence over a 3-month period for all
antiretroviral medications in a population of low-
income HIV-positive homeless and marginally
housed individuals. In previous studies, we have
found that unannounced pill counts perform well
compared to computerized medication caps, and
they are highly associated with HIV viral load,1,18
development of resistance,6 and progression to
AIDS.5 Here, we report on the performance of the
visual analogue scale and the 3-day recall adher-
ence tools compared to unannounced pill count in
a low-income population of homeless or margin-
ally housed indigent adults.
METHOD
Participants were recruited from the REACH co-
hort, a prospective cohort of HIV-positive home-
less or marginally housed adults recruited in sys-
tematic sampling from San Francisco homeless
shelters, free-lunch programs, and a probability
sample of residence hotels charging less than $400
• GIORDANO ET AL. 75
per month between 1996 and 1998.19 All REACH
participants who were receiving highly active
antiretroviral therapy were eligible for the prospec-
tive adherence study. Once a month, a research
assistant made an unannounced visit at the
participant’s usual place of residence on a weekday
and during daylight hours. Participants were not
told in advance when to expect an assessment. Ad-
herence assessments were conducted using inter-
viewer-administered structured questionnaires
that assessed adherence for all components of the
antiretroviral regimen using 3-day recall and a vi-
sual analogue scale, and then the unannounced pill
count was performed. Based on the work of
Chesney et al.,14 the 3-day recall tool asks partici-
pants how many doses of each medication were
taken yesterday, the day before yesterday, and 3
days ago. It also asks how many pills were taken
each day. The visual analogue scale tool asks par-
ticipants to, “Put a cross on the line below at the
point showing your best guess about how much of
each drug you have taken in the last 3 or 4 weeks.
For example, 0% means you have taken no drug,
50% means you have taken half your drug, and
100% means you have taken every single dose of
your drug.”17 Participants underwent scheduled
phlebotomy for HIV viral load assessment
monthly and for CD4 count assessment quarterly.
The present study includes patients with three
successful visits to assess adherence over no more
than a 4-month period and with no changes in
antiretroviral regimen over the time period. Patients
also must have had at least two HIV viral loads
obtained during that period, the latter obtained
within 14 days of the last adherence visit. For each
method of assessing adherence, the monthly adher-
ence was calculated for each patient as the mean
adherence for each medication. Overall adherence
for each method was the mean adherence for each
patient for all medications over the three visits. To
examine the validity of the assessments, Pearson
correlation coefficients comparing 3-day and visual
analogue scale adherence to unannounced pill count
and average HIV viral load were computed. To mea-
sure the stability of the correlations, we compared
their correlation to unannounced pill count over
time. Correlation coefficients (r) were calculated and
compared using the Fisher z transformation, which
normalizes the underlying distribution of each cor-
relation coefficient.20
76 HIV CLINICAL TRIALS • 5/2 • M AR-APR 2004

Table 1. Comparison of the correlation of the results of the visual analogue scale and 3-day recall
adherence instruments to unannounced pill count and viral load in patients receiving combination
antiretroviral therapy

Sample Correlation
size coefficient 95% CI p

Correlation to unannounced pill count

Visual analogue scale 84 −0.76 −0.65, 0.84
3-day recall 84 −0.71 −0.59, 0.80 .52

Correlation to log10 viral load

Visual analogue scale 84 −0.49 −0.64, −0.31
3-day recall 84 −0.34 −0.51, −0.13 .22

Correlation to unannounced pill count, by visit

Visual analogue scale, visit 1 84 −0.57 −0.40, 0.70
Visual analogue scale, visit 2 84 −0.66 −0.52, 0.77
Visual analogue scale, visit 3 84 −0.73 −0.61, 0.81 .08*
3-day recall, visit 1 84 −0.68 −0.55, 0.78
3-day recall, visit 2 84 −0.50 −0.32, 0.64
3-day recall, visit 3 84 −0.67 −0.53, 0.77 .95*

Note: p values test whether the difference between the correlation of visual analogue scale adherence to unannounced pill count
adherence and three-day recall adherence to unannounced pill count adherence is significant, unless otherwise noted. CI =
confidence interval.
*p value for the comparison of the correlation at visit 1 to visit 3 using the same adherence tool.

RESULTS Mean visual analogue scale adherence was 82.5%

Complete adherence data were available for 84
participants. The participants were mostly male
(83%), non-white (58%), and low income (median
monthly income = $829; interquartile range [IQR] =
$762 to $1000) and had a history of injection drug
use (64%). In the 30 days prior to their baseline
interview, 19% had used injection drugs, 18% had
used crack cocaine, 43% had used alcohol, and 17%
were heavier alcohol users who drank at least 2
drinks per day at least four times. Most participants
had unstable housing. Only 30% spent more than
half the nights in the last 3 months in their own
house or apartment. Nearly a quarter of the partici-
pants did not complete high school (24%), whereas
43% completed high school but had no college-level
education. Most participants were on a twice-daily
antiretroviral regimen (89%), and 56% were receiv-
ing a protease inhibitor. The median CD4 count was
299 cells x 106/L (IQR = 176–490).
The adherence results are presented in Table 1.
(SD = 18.9), while the median was 86.4% (IQR =
70.0–98.6). The mean 3-day recall adherence was
84.2% (SD = 22.6), while the median was 94.4%
(IQR = 77.8–100). The mean unannounced pill
count adherence was 75.9% (SD = 27.5), while the
median was 87.5% (IQR = 55.6–96.7). The correla-
tion of visual analogue scale and unannounced
pill count results was high (r = 0.76; 95% CI 0.65,
0.84) and was not different from that between 3-
day recall adherence and unannounced pill count
(r = 0.71; 95% CI 0.59, 0.80; p = .52). The visual
analogue scale, 3-day recall, and unan-
nounced pill count measurements correlated
inversely with log HIV viral load (r = -0.49; 95% CI
-0.64, -0.31; r = -0.34; 95% CI -0.51, -0.13; and r
= -0.50; 95% CI -0.65, -0.32, respectively), and
the visual analogue scale and 3-day recall correla-
tions were not significantly different (p = .22). The
visual analogue scale correlation with unan-
nounced pill count was stable over time (r = 0.57,
0.66, and 0.73, at visits 1, 2, and 3, respectively), as
 VISUAL ANALOGUE SCALE FOR MEASURING ADHERENCE • GIORDANO ET AL. 77

Table 2. Comparison of the correlation of the visual analogue scale adherence instrument results to
unannounced pill count in patients receiving combination antiretroviral therapy, by selected participant
characteristics

Sample Correlation
Characteristic size coefficient 95% CI p

Age (years) < 45 36 0.76 0.57, 0.87 .96

> 45 48 0.76 0.61, 0.86

Sex Male 70 0.79 0.68, 0.86 .31

Female 14 0.63 0.15, 0.87

Race African American 34 0.77 0.58, 0.88 .64

All others 50 0.72 0.55, 0.83

Education (years) > 12 25 0.85 0.69, 0.93 .18

< 12 50 0.72 0.56, 0.83

Housinga Stable 25 0.82 0.63, 0.92 .38

Unstable 59 0.74 0.59, 0.83

Current IDU No 68 0.80 0.69, 0.87 .34

Yes 16 0.66 0.25, 0.87

Current crack use No 69 0.72 0.58, 0.81 .17

Yes 15 0.87 0.64, 0.96

Frequent alcohol useb No 70 0.74 0.62, 0.83 .15

Yes 14 0.90 0.69, 0.97

Antiretroviral regimen NNRTI 34 0.63 0.38, 0.80 .14

Protease inhibitor 40 0.80 0.65, 0.98

Note: p values compare the correlation of the visual analogue scale adherence to unannounced pill count adherence by the
characteristic noted. CI = confidence interval; IDU = injection drug use; NNRTI = nonnucleoside reverse transcriptase inhibitor.
a
Stable and unstable housing were defined as reporting that more than and less than or equal to 50% of the nights in the 3 months
before the interview were spent in an apartment or house owned or rented by the patient, respectively.
b
Defined as drinking at least 2 drinks daily for at least 4 days in the 30 days before the interview.

was the 3-day recall adherence (r = 0.68, 0.50, 0.67,
at visits 1, 2, and 3, respectively).
The correlation of visual analogue scale to unan-
nounced pill count in various subpopulations is
presented in Table 2. The correlation remained in
participants with less than 75% adherence (r =
0.54), no more than a high school education (r =
0.72), and unstable housing (r = 0.74). There was
also a correlation in African Americans (r = 0.77); in
participants with active injection drug use (r =
0.66), heavier alcohol use (r = 0.90), and active crack
cocaine use (r = 0.87); and in participants on a
protease inhibitor–based regimen (r = 0.80) and on
a nonnucleoside reverse transcriptase inhibitor–
based regimen (r = 0.63). Too few participants were
on once-daily therapy to evaluate the visual ana-
logue scale’s performance in that setting.
To assess whether shifts in the mean between the
different measures of adherence were affecting our
results, we calculated the intraclass correlation co-
efficients (ICC).21 The ICC for adherence calculated
by visual analogue scale and unannounced pill
count was 0.71 (compared to the Pearson r of 0.76),
and the ICC for 3-day recall and unannounced pill
count was 0.70 (compared to the Pearson r of 0.71).
To assess the potential for selection bias, we com-
pared the baseline characteristics (listed in Table 2)
of the 84 participants in the study to the 26 patients
78 HIV CLINICAL TRIALS •
excluded from the study because of incomplete
data, and we found no significant differences (p
values were 0.17 or greater; data not shown).
DISCUSSION
A visual analogue scale to measure adherence
had been used in a population with a narrow range
of adherence (median adherence of 97%)16 and with
largely gay male patients on protease inhibitors,
over 40% of whom were on thrice-daily therapy.17
In the present study, a visual analogue scale dem-
onstrated good validity compared to unannounced
pill count and HIV viral load, performing as well as
a 3-day recall adherence tool in a diverse popula-
tion. The visual analogue scale’s correlation with
pill count was stable over time and the instrument
performed well in African Americans; in patients
on both protease inhibitor– and nonnucleoside re-
verse transcriptase inhibitor–based regimens; and
in patients with lower education, active substance
use, and lower adherence. Compared to the 3-day
recall tool, the visual analogue scale is simpler to
administer and answer, thereby offering a substan-
tial advantage over the 2- to 7-day recall instru-
ments currently in use.14,15
This study has several limitations. First, the
sample size is small and the data are for only three
assessments; small differences in the validity and
correlations of the instruments may have gone un-
detected as a result. However, this is the first study
to assess the performance of a visual analogue scale
that includes both more than a single assessment of
adherence and an objective measure of adherence
such as pill count.16,17 Second, study personnel ad-
ministered the visual analogue scale, and we can-
not be certain that it would perform as well if it
were self-administered. Still, the simplicity of the
instrument argues for this possibility. Third, the
visual analogue scale was administered after the
participant answered the 3-day recall questions.
These questions may have stimulated the
participant’s recall, prompting the participant to
report a more accurate result in the visual analogue
scale. Also, the presence of a concurrent pill count
may have influenced the participants’ reports of
adherence. Nonetheless, it is unlikely that concur-
rent pill count measures had a differential impact
on 3-day recall relative to the visual analog scale, so
the lack of difference in these measures’ correla-
tions to unannounced pill count is likely unbiased.
5/2 • M AR-APR 2004
Finally, the study only included participants on
stable antiretroviral regimens, so the scale’s perfor-
mance in patients starting a new regimen remains
unknown.
Adherence to antiretroviral therapy is one of the
most important predictors of outcome, but re-
searchers and clinicians alike have been hampered
by the lack of a valid, quick, simple, and inexpen-
sive instrument to accurately measure adherence.
In this diverse sample followed prospectively, the
visual analogue scale adherence instrument per-
formed as well as a more complicated 3-day recall
instrument both in comparison to unannounced
pill count and to virologic suppression. It is simple,
brief, inexpensive, and easily disseminated. It
could also be adapted for use in chronic diseases
other than HIV. Further, because it generates a
single number for each medication, it simplifies
data analysis compared to pill count, computerized
caps, and 3-day recall instruments. Adopting the
visual analogue scale instrument may save the time
of both researchers and participants. Therefore, un-
like other adherence assessment instruments, the
visual analogue scale instrument could be easily
self-administered and could become an integral
part of routine clinical care of patients with HIV in
both resource-rich and resource-constrained areas.
ACKNOWLEDGMENTS
This study was funded by National Institutes of
Health (NIH) grants MH54907, MH66654, and
MH64388. Dr. Bangsberg receives support from
The Doris Duke Charitable Foundation. Dr.
Giordano receives support from NIH grant
K23MH67505 and the resources and use of facilities
of the Houston Center for Quality of Care and
Utilization Studies, Department of Veterans Af-
fairs, Houston, Texas, USA.
REFERENCES
1. Bangsberg DR, Hecht FM, Charlebois ED, et al. Adherence
to protease inhibitors, HIV-1 viral load, and development of
drug resistance in an indigent population. AIDS.
200;14:357–366.
2 Paterson DL, Swindells S, Mohr J, et al. Adherence to
protease inhibitor therapy and outcomes in patients with
HIV infection. Ann Intern Med. 200;133;21–30.
3. Arnsten JH, Demas PA, Farzadegan H, et al. Antiretroviral
therapy adherence and viral suppression in HIV-infected
drug users: comparison of self-report and electronic moni-
 VISUAL ANALOGUE SCALE FOR MEASURING ADHERENCE • GIORDANO ET AL. 79

toring. Clin Infect Dis. 200;33:1417–1423. A. Comparing objective measures of adherence to HIV
4. McNabb J, Ross JW, Abriola K, Turley C, Nightingale CH, antiretroviral therapy: electronic medication monitors and
Nicolau DP. Adherence to highly active antiretroviral unannounced pill counts. AIDS Behav. 2001;5:275–281.
therapy predicts virologic outcome at an inner-city human 19. Zolopa AR, Hahn JA, Gorter R, et al. HIV and tuberculosis
immunodeficiency virus clinic. Clin Infect Dis. 200;33:700– infection in San Francisco’s homeless adults. Prevalence
705. and risk factors in a representative sample. JAMA.
5. Bangsberg DR, Perry S, Charlebois ED, et al. Non-adher- 1994;272:455–461.
ence to highly active antiretroviral therapy predicts pro- 20. Zar JH. Biostatistical Analysis, 3rd ed. Englewood Cliffs,
gression to AIDS. AIDS. 2000;15:1181–1183. NJ: Prentice Hall; 1974.
6. Bangsberg DR, Charlebois ED, Grant RM, et al. High levels 21. Fleiss J. Statistical Methods for Rates and Proportions, 2nd
of adherence do not prevent accumulation of HIV drug ed. New York: John Wiley and Sons; 1981.
resistance mutations. AIDS. 2003;17:1925–1932.
7. Walsh JC, Pozniak AL, Nelson MR, Mandalia S, Gazzard
BG. Virologic rebound on HAART in the context of low
treatment adherence is associated with a low prevalence
of antiretroviral drug resistance. J Acquir Immune Defic
Syndr. 2002;30:278–287.
8. Howard A, Arnsten JH, Gardner L, et al. Lack of multi-drug
resistance in nonresponders to antiretroviral therapy with
poor adherence. In: Program and abstracts of the First
International AIDS Society Conference on HIV Pathogen-
esis and Treatment; Buenos Aires; 2002. Abstract 604.
9. Hogg RS. Heath K, Bangsberg D, et al. Intermittent use of
triple-combination therapy is predictive of mortality at
baseline and after 1 year of follow-up. AIDS.
2002;16:1051–1058.
10. Garcia de Olalla P, Knobel H, Carmona A, Guelar A, Lopez-
Colomes JL, Cayla JA. Impact of adherence and highly
active antiretroviral therapy on survival in HIV-infected pa-
tients. J Acquir Immune Defic Syndr. 2002;30:105–110.
11. Dunbar-Jacob J, Mortimer-Stephens MK. Treatment ad-
herence in chronic disease. J. Clin Epidemiol .
2001;54(suppl):S57–60.
12. Department of Health and Human Services and the Henry
J. Kaiser Family Foundation. Guidelines for the use of
antiretroviral agents in HIV-1-infected adults and adoles-
cents. Available at http://AIDSinfo.nih.gov. Last updated
July 14, 2003. Accessed September 18, 2003.
13. Liu H, Golin CE, Miller LG, et al. A comparison study of
multiple measures of adherence to HIV protease inhibitors.
Ann Intern Med. 2001;134:968–977.
14. Chesney MA, Ickovics JR, Chambers DB, et al. Self-re-
ported adherence to antiretroviral medications among par-
ticipants in HIV clinical trails: the AACTG adherence instru-
ments. Patient Care Committee & Adherence Working
Group of the Outcomes Committee of the Adult AIDS
Clinical Trials Group (AACTG). AIDS Care. 2000;12:255–
266.
15. Mannheimer S, Friedland G, Matts J, Child C, Chesney M.
The consistency of adherence to antiretroviral therapy pre-
dicts biologic outcomes for human immunodeficiency vi-
rus-infected persons in clinical trials. Clin Infect Dis.
2002;34:1115–1121.
16. Walsh JC, Dalton M, Gazzard BG. Adherence to combina-
tion antiretroviral therapy assessed by anonymous patient
self-report. AIDS. 1998;12:2361–2363.
17. Walsh JC, Mandalia S, Gazzard BG. Responses to a 1
month self-report on adherence to antiretroviral therapy
are consistent with electronic data and virological treat-
ment outcome. AIDS. 2002;16:269–277.
18. Bangsberg D, Hecht FM, Charlebois ED, Chesney M, Moss