Download Adherence to Antihypertensive Medications

Copyright 7997 by
The Cerontological Society of America
The Cerontologist
Vol. 37, No. 5, 609-619
Although treatment for hypertension is readily available, poor control of hypertension is a
major health problem frequently manifested in late life. Researchers believe that one of the
major causes of uncontrolled hypertension is failure to take medication as directed. In this
preliminary study, the medication-taking behaviors of 48 adults diagnosed with
hypertension, ranging in age from 35 to 87, were recorded for 2 months with credit
card-sized bar-code scanners. The social-cognitive model (Park, 1992) for understanding
medication adherence, which proposes that medication adherence is governed by both
beliefs and cognitive factors, was used as a basis for this research. Therefore, measures of
health behaviors, attitudes about health and medication taking, and cognitive function were
recorded, as well as blood pressure readings. The main findings were that (a) the oldest-old
and groups of middle-aged adults were the most nonadherent, whereas the young-old were
more likely to adhere than the other age groups; (b) high blood pressure readings
predicted adherence to antihypertensive medications; and (c) medication beliefs influenced
adherence in some situations.
Key Words: Adherence, Aging, Hypertension, Self-care
Adherence to Antihypertensive Medications
Across the Life Span1
Roger W. Morrell, PhD,2 Denise C. Park, PhD,3
Daniel P. Kidder, MS,4 and Mike Martin, PhD5
Uncontrolled hypertension is a major risk factor
blood pressure under effective control (Richardson,
for heart attack and stroke and an antecedent of
Simons-Morton, & Annegers, 1993; Sharkness &
cardiovascular and renal disease (National High
Blood Pressure Education Program Working Group,
1994). Hypertension is a serious health threat, as 50
to 60 million Americans are affected by the disease
(Joint Committee on Detection, Evaluation, and
Treatment of High Blood Pressure, 1993). Although
hypertension can be treated effectively by oral
medication, poor control of hypertension continues
to be a major health problem.
It has been estimated that 30-55% of patients with
hypertension do not adhere to their prescribed
medication regimen, with some patients taking less
medication than prescribed and others taking more
(Hamilton et al v 1993; Rudd, 1995; Rudd et alv 1989;
Shaw, Anderson, Maloney, Jay, & Fagan, 1995). Furthermore, it has been reported that only one half to
one fourth of all hypertensive patients have their
Snow, 1992; Skaer, Sclar, Markowski, & Won, 1993).
Because hypertension is a relatively common condition, adherence to antihypertensive therapies has
been widely studied, with an emphasis on determining the psychosocial and medical predictors of nonadherence. A recent review suggests that although
there are reliable findings on this topic, some of the
results that have been reported are conflicting and,
at times, confusing (see Dunbar-Jacob, Dwyer, &
Dunning, 1991). One reason for the discrepancy in
the findings is that the researchers have not used a
reliable, sensitive measure of adherence. An electronic monitoring device is used in this study to address this problem. Another reason for this discrepancy is that few researchers in this area have utilized
a comprehensive conceptual framework that incorporates a multidimensional view of medication taking to guide their research. Park and her colleagues,
however (Park, 1992, 1994; Park & Jones, 1996), have
developed a conceptual model for understanding
medication adherence that proposes that medication adherence is governed by both beliefs and
cognitive factors. This model integrates the Leventhal and Cameron (1987) self-regulatory view of
medication adherence with basic cognitive psychology and will be here termed the social-cognitive
model of medication adherence.
Leventhal and Cameron (1987) have proposed
that medication adherence is a self-regulatory be-
'This research was funded by a grant from the American Association of
Retired Persons, Andrus Foundation, to Denise C. Park (principal investigator) and Roger W. Morrell. The authors would like to thank Gary Lautenschlauger and Christopher Hertzog for comments on previous versions
of this article, and Joan Bennett and Kim Shifren for assistance with statistical analyses.
'Address correspondence to Dr. Roger W. Morrell, The Institute for Social Research, University of Michigan, 426 Thompson Street, Ann Arbor,
Ml 48106-1248.
'The Institute for Social Research, University of Michigan.
'The Georgia Institute of Technology, Atlanta, GA.
5
The University of Mainz, Mainz, Germany.
Vol. 37, No. 5,1997
609
havior. An individual will be adherent or nonadherent to a medication depending on what he or she
believes to be true about a disease and the perceived relationship of taking medications to disease
management. This illness representation is also dynamic and may change over the course of a disease.
In particular, individuals who believe they have no
control over their illness may be nonadherent because they may believe taking medication will have
no effect on their outcome. Thus, it is important to
understand individuals' personal beliefs about their
illness because these types of beliefs will "drive"
adherence behaviors (Park, 1994). In addition to illness beliefs, Park and Mayhorn (1996) have stressed
that it is also important to measure beliefs about
specific medications or a patient's medication representation. Persons will not take a medication if
they do not consider it useful or if a decision is
made to use the medication only when symptoms
occur (Park & Mayhorn, 1996). Such decisions might
ultimately affect adherence to antihypertensive
medications if a patient views hypertension as a
symptomatic condition (Meyer, Leventhal, & Cutmann, 1985).
The social-cognitive model suggests that, in addition to beliefs, cognitive functioning also plays an
important role in adherence, especially for older
adults (Park, 1994). Remembering to take medications involves many components of cognitive processing including comprehension, working memory, long-term memory, prospective memory, and
reasoning (Park, 1992; Park & Kidder, 1996). Age-related declines in some of these underlying cognitive mechanisms are well-documented, and may
contribute to forgetting to take medication as prescribed or to not understanding how to properly
take medication in older adults (see Salthouse,
1991). Furthermore, because older adults may be
more likely than younger adults to be prescribed
multiple medications, the process of taking medication correctly may be particularly difficult for older
patients because of the cognitive demands required
for organizing and maintaining complex medical
regimens (Park, Willis, Morrow, Diehl, & Gaines,
1994).
Other researchers have also proposed that health
status and disease state are important components
in understanding why people do or do not take
their antihypertensive medications in particular
(Cummings, Kirscht, Binder, & Codley, 1982; DeVon
& Powers, 1984; Nagy & Wolf, 1984; Stanton, 1987).
Furthermore, psychosocial factors such as satisfaction with one's physician might also determine how
medication is taken (see Wolf, Putnam, James, &
Stiles, 1978).
Findings from research on adherence in general
suggest that age is not a reliable factor for predicting adherence (see Lorenc & Branthwaite, 1993, for
a discussion). Park and Jones (1996), however, suggest that age may be conceptualized as influencing
medication adherence primarily through cognitive
factors. In support of this, Morrell, Park, and Poon
(1989) have reported that elderly adults consistently
manifested poorer recall of prescription information than young adults. Moreover, Park, Morrell,
Frieske, and Kincaid (1992) reported higher nonadherence rates in adults over the age of 75 (old-old
adults) compared to individuals between the ages
of 60 and 74 (young-old adults).
In this preliminary study, the impact of beliefs
and cognitive function on adherence to medication
regimen by hypertensive adults is examined in
order to guide future work in this area. This research differs from past efforts to study adherence
in hypertensive patients in a number of ways. First,
we collect measures of beliefs about hypertension
and taking medications guided by the social-cognitive model (Park, 1992,1994; Park & Jones, 1996). Traditional demographic measures and health indices
of the participant are also collected, as well as
measures concerning participants' perceptions of
their physician's behavior. The second unique characteristic of this research is a measure of cognitive
functioning — working memory capacity. Working
memory is an estimate of the amount of processing
resources or mental energy that an individual possesses (Baddeley, 1986). Reliable findings suggest
that working memory capacity declines with age
(see Salthouse, 1991). Third, although Park et al.
(1992) included only older adults in earlier adherence work, we also examine the adherence behaviors of middle-aged adults. Fourth, it is of special interest to determine what factors might influence
adherence to antihypertensive medication as opposed to other types of medications in a patient's
regimen. This comparison has not been made in
past research and is examined in detail in this project. Finally, it is important that the method used for
measuring adherence be unobtrusive, sensitive,
and reliable, so that a true and accurate picture of
medication-taking behavior can be obtained. Therefore, in the present work, adherence is recorded by
an electronic monitoring technique — the Videx
TimeWand, a portable bar-code reader. Leirer, Morrow, Pariante, and Sheikh (1988) have used this type
of bar-code reader to measure simulated medication adherence in older adults in past research. Furthermore, Park and her colleagues have found this
particular technique to be successful in measuring
adherence to actual prescribed regimens of medications in 61 older adults and 15 patients diagnosed
with rheumatoid arthritis over periods of 2 and 3
weeks (Park, 1992; Park et al., 1992, respectively).
The monitoring period has been increased to 8
weeks in the present study, exceeding the time during which medication-taking behaviors were monitored in previous research by Park et al. The extended monitoring period allows the impact of the
passage of time on adherence to be assessed.
610
Method
Participants
Forty-eight community-dwelling adults, ranging
in age from 36 to 87 (M = 59.54, SD = 14.4), particiThe Gerontologist
pated in this study. The sample was age-stratified
such that there were 9 or 10 individuals in each 10year band from ages 35 to 75, and above. All participants were volunteers recruited from family and internal medicine practices located in the Athens,
Georgia, area. Before participants could be admitted to the study, they were required to have their
personal physician submit written verification that
they had been diagnosed with essential hypertension. Their physician was also asked to rate the
severity of their disease on a 5-point scale. They
were also required to be taking at least one prescribed medication for hypertension on a daily
basis. The number of medications taken by the participants ranged from 1 to 9 (M = 3.64, SD = 2.04).
Additionally, the participants were required to be
(a) managing their own medication regimen, (b)
have at least 20/30 corrected binocular vision, (c) not
have been hospitalized in the past 6 months, and
(d) have no restriction in arm and hand movement.
Individuals were paid a total of $175 for their participation. They received $40 at the initial visit, and additional payments of $45 during each of weeks 3, 5,
and 7. A series of analyses of variance indicated that
the participants in each of the 10-year age bands did
not differ in the number of medications or daily
dosages taken, years they had been diagnosed with
hypertension, physician rating of disease severity, or
perceived health. The groups of participants also did
not differ in educational attainment. Older adults,
however, were more likely to live alone than were
younger adults (x2 = 13.04, p = .02). Scores on these
variables are shown in Table 1 by age group.
Equipment
All participants received a Videx TimeWand to
record their medication-taking behavior, the same
device used by Leirer et al. (1988) and Park et al.
(1992). The Videx TimeWand is a credit card-sized,
portable bar-code reader. It has an optical sensor at
one corner which, when swept across a bar code,
records the bar code number into the memory of
the TimeWand. The date and time of the scan is also
recorded. Each participant was given a TimeWand
placed inside a small wallet. Inside the wallet was a
set of bar codes with corresponding labels that indicated the name and description of each of the medications in the participants' regimens. They were instructed to keep the wallet with their medications
and to scan the appropriate bar code in the wallet
immediately after taking each medication in their
regimen, so that the drug's identity, and the date
and time the medication was taken, could be
recorded.
Individual Differences Battery
The following seven instruments were administered to each of the participants. The Medication Information Survey was developed for this study and
was administered individually for each medication
in the participants' regimens. Information collected
included the reason for taking each medication,
when (time of day or night) each of the medications
was taken, the dosage of each of the medications
taken each time, and the number of years that patients had been diagnosed with hypertension.
Table 1 . Means and Standard Deviations of Medication, Illness, and Perceived Health Variables and Educational Attainment
and Living Arrangement Breakdown by Age Group
Age Group
35^4
(n = 10)
45-54
(n = 9)
55-64
(n = 10)
65-74
(n = 10)
75+
(n = 9)
2.8
(1.81)
10.9
(8.89)
2.90
(1.10)
3.5
(1.94)
11.22
(9.68)
2.55
(0.73)
4.3
(2.75)
11.60
(7.62)
2.50
(0.53)
4.1
(1.97)
14.20
(10.012)
2.60
(0.52)
3.6
(1.51)
10.44
(14.54)
2.56
(0.73)
2.0
(0.47)
1.80
(0.63)
2.44
(0.53)
1.77
(0.67)
2.20
(0.42)
1.70
(0.48)
2.40
(0.52)
1.90
(0.57)
2.25
(0.70)
1.63
(0.74)
Education
High school degree or less
Some college or college degree
Graduate degree
50.0%
40.0%
10.0%
55.5%
22.2%
22.2%
30.0%
50.0%
20.0%
40.0%
40.0%
20.0%
22.2%
33.3%
44.4%
Live Alone
Yes
No
10.0%
90%
0.0%
100%
30%
70%
30%
70%
66.7%
33.3%
Demographic Variables
Number of medications
Years diagnosed with hypertension
Physician rating of severity of hypertension3
Perceived Health
Present health rating6
Health as a barrier to activity"
"1 = extremely mild; 2 = mild; 3 = moderate; 4 = severe; 5 = extremely severe.
b
1 = excellent; 2 = good; 3 = fair; 4 = poor.
C
1 = not at all; 2 = a little; 3 = a great deal.
Note: Standard deviations are shown in parentheses.
Vol. 37, No. 5,1997
611
A Demographic Questionnaire was adapted from
the Older Americans Resources and Services (OARS)
Instrument (Duke University Medical Center, 1975)
and it included items about age, gender, education,
profession, work status, living arrangements, marital
status, and income.
The Health Practices Questionnaire was also a
newly developed instrument. It contained questions about exercise, diet, and use of vitamin supplements.
The Hypertension Patient's Perception of Physician Behavior Scale was adapted from the Medical
Interview Satisfaction Scale (Wolf et al., 1978), but
was redesigned for this research to include items
specifically about hypertension and the patientphysician relationship.
The Patient's Hypertension Beliefs Questionnaire
was derived from the Health Belief Questionnaire
(Jette, Cummings, Borck, Phels, & Naessens, 1981).
This instrument included questions designed to assess health beliefs about hypertension and included
items on health concerns, perceived health, perceived barriers to taking medications, perceived
control of hypertension, perceived control of overall
health, self-care, perceived prognosis, detection of
hypertension, perceived side effects, and knowledge of hypertension. Sample items from the instrument used in this study are shown in the appendix.
A Present and Past Health Status Questionnaire
was adapted from the OARS Instrument (Duke University Medical Center, 1975) to measure perceived
health status.
Finally, an assessment of the participants' comprehension of their medication regimen was conducted
in which participants referred to their own medication regimen and completed an hour-by-hour plan
of what medications they would take, the reason for
taking each medication, and when they would take
the medications over a 24-hour period. This instrument was used previously by Morrell, Park, and
Poon (1989,1990).
The second cognitive task was the 30-Point Word
Familiarity Survey (Gardner & Monge, 1977), which
was also administered to ascertain the participants'
vocabulary ability.
Procedures
All participants were tested individually in a local
physician's office. During their first visit, they completed the battery of instruments described above.
After completing the package of measures, participants were trained to use the Videx TimeWand and
practiced using it with their own medications. They
were also given a wallet that had a section in which
to keep the TimeWand when not in use and inserts
that were clearly labeled with bar-code stickers and
the name and description of each drug in their regimens. Participants were instructed to scan the bar
code that corresponded to the medication they
were taking whenever they took a particular medication. They were cautioned that a successful scan
occurred only when a short beep was emitted from
the TimeWand and the red light was also illuminated on the top of the TimeWand. Participants
were also instructed to scan the individual bar
codes the same number of times as the number of
pills they took (i.e., two pills equaled two scans on
the corresponding bar code). They were provided
with a brochure that described how the TimeWand
worked and what to do in case of problems with the
device. Participants were also provided with a
"Scanner's Diary" in which they could record scanning errors, problems using the TimeWand, or any
other information about their adherence behavior
they wanted to convey to the experimenters. The
diary data were carefully cross-checked with actual
scanning printouts, and accidents in scanning were
deleted from the adherence scores used in the data
analyses.
Because the batteries in the TimeWands remained charged for about 7 days, participants were
required to return for a brief appointment with the
experimenter the same day each week over the next
8 weeks. The primary purpose of these visits was to
Cognitive Battery
exchange the TimeWand they had been using for
Two cognitive tasks were administered to each of
a freshly charged device. The participants' blood
the participants. The first cognitive task was a mea- pressure was also recorded after resting 5 minutes
sure of working memory, the Listening Span Task, during four of the appointments (i.e., every other
which was collected to assess subjects' cognitive reweek).
sources (Daneman & Carpenter, 1980; Salthouse &
During the fifth-week visit, the measures of workBabcock, 1990). Working memory is a measure of
ing memory and vocabulary ability were administotal storage and processing capacity available to an
tered. At the end of the eighth week, participants
individual. This task required participants to answer
returned for a final visit with the experimenter. The
questions about simple sentences that were orally
last TimeWand that had been issued and all Scanpresented (the processing component) while they
ner's Diaries were returned at this time
simultaneously remembered the last word (the storage component) in each of the sentences. The number of sentences increased from one to six over
Results
trials so that the number of items they were to remember progressively increased. There were three
Adherence Data Analyses
trials at each level of memory load. The task was
stopped whenever participants made errors on
Data Scoring. — The adherence data were scored
three consecutive trials. An individual's span was
according to the following plan. Four types of addetermined by the total number of trials correct.
herence errors (omission errors, commission er612
The Gerontologist
rors, quantity errors, and total errors) were determined for each medication taken by the participants (a total of 175 medications over the 48 participants). Omission errors were recorded when
participants did not take or omitted dosages from
their regimens. Commission errors were noted
when participants took extra dosages of a medication (e.g., taking a medication 4 times during a day
instead of the prescribed 3 times a day). Quantity
errors involved participants taking additional
amounts of a medication during a dosing event
(e.g., taking 3 pills instead of the prescribed 2 pills
as reflected by the number of scans made with the
TimeWand). Total errors were obtained by summing the number of omission, commission, and
quantity errors. Proportions were then derived by
dividing the absolute number of errors made in
each of the four categories during each of the four
biweekly periods by the absolute number of
dosages participants were scheduled to take during
each of the biweekly periods. The biweekly scores
were used because few differences were observed
between weekly and biweekly scores.
made by all of the participants over the four biweekly monitoring periods, the data were graphed.
Figure 1 demonstrates a positive linear relationship
in the number of omission errors made over time,
with the number of these types of errors increasing
steadily as the monitoring period progressed.
Nonadherence Over Time. — In order to ascertain if there was a systematic trend in the number of
omission, commission, quantity, or total errors
The Effects of Age and Monitoring Period on Adherence. — For the following sets of analyses, age
was partitioned into five categories (35-44: n = 10;
45-54: n = 9; 55-64: n = 10; 65-74: n = 10; 75+: n = 9).
Individual analyses of variance (ANOVAs) were conducted for omission, commission, quantity, and
total errors, with age group as a between-subjects
variable and the four biweekly monitoring intervals
as a repeated measures variable. ANOVAs were
used instead of regression analyses so that the effect of time of measurement could be determined.
Results from the analyses conducted on commission and quantity errors are not reported because
these types of errors were very rare (an average of
about 2% for commission errors and an average of
less than 1% for quantity errors across the age
groups and biweekly periods). Mean proportion
and standard deviations of omission and total errors
are shown in Table 2 as a function of age group and
biweekly monitoring interval.
0.16
0.14
0.12
0.08
0.06
0.04
0.02
Biweekly 1
Omission Errors
Biweekly 3
Biweekly 2
Commission Errors
Figure 1. Mean biweekly proportion of errors across age groups by type of error.
Vol. 37, No. 5,1997
613
Quantity Errors
Biweekly 4
Total Errors
Total Errors. — Results from the ANOVA conducted on total errors indicated only a significant
main effect of age, F(4,144) = 5.26, MSe = .111, p =
.0006. The highest rates of nonadherence were observed in the oldest-old and the group of participants between the ages of 55 and 64 (an average of
21.5% and 18.3% across the four biweekly monitoring periods, respectively). The lowest proportion of
total errors was made by participants 65 to 74 years
old (an average of 3.8% across the four biweekly
monitoring periods). The rate of nonadherence for
the other groups (ages 35-54) fell between the two
older groups (approximately 12.5%). The proportion
of total errors for participants 65 to 74 years old was
significantly different from the oldest age group
and from the group of participants between the
ages of 55 and 64, Tukey's Studentized Range =
3.907, p < .05.
Omission Errors. — The analysis of omission errors, which were the most common errors made,
yielded a significant main effect of age, F(4,144) =
5.35, MSe = 0.083, p = .0005, and biweekly monitoring interval, F(3,432) = 3.73, MSe = 0.010, p = .01.
More omission errors were made by the oldest participants (about 19%) than the young-old participants (ages 65-74), whose proportion of nonadherence was extremely low (about 3%) across the
monitoring period, Tukey's Studentized Range =
3.907, p < .05. A comparison test on the interval data
indicated that more omission errors were also made
in the last biweekly monitoring period (approxi-
Table 2. Mean Proportion, Standard Deviations', and Marginal
Means of Total and Omission Errors as a Function of Age Group
and Biweekly Monitoring Interval
Biweekly Intervals
.. . .
1
Marginal
Age Croup Biweeki Biweek2 Biweek3 Biweek4 Means
Total Error;
35-44
45-54
55-64
65 74
75+
.112
(.171)
.145
(.211)
.159
(.251)
.018
(.050)
.155
(.184)
Omission Errors
35^4
.100
(.169)
.104
45-54
(.181)
55-64
.105
(.207)
65-74
.015
(.046)
75+
.143
(.188)
.071
(.100)
.161
(.250)
.220
(.414)
.020
(.038)
.211
(.218)
.092
(.134)
.161
(.209)
.204
(.387)
.061
(.124)
.252
(.224)
.112
(.144)
.144
(.239)
.147
(.251)
.054
(.110)
.240
(.274)
mately 12%) than in the first biweekly monitoring
period (approximately 9%), F(1,144) = 6.97, MSe =
0.012, p = .009.
Overall, these results suggest that the pattern of
total errors was driven primarily by the number of
omission errors made. Because total errors appeared to be the most global representation of the
different types of medication-taking errors that
were made, only total errors were analyzed in subsequent analyses.
The Effects of Age and Monitoring Period on Adherence to Different Types of Medications. — To
determine whether adherence was affected by different types of medications, the medications that
the participants took were divided into two categories: antihypertensive medications and all other
medications taken on a prescribed basis. Two 5 X 4
ANOVAs were conducted (one on the antihypertensive medications and one on the other medications
in the regimen) on the proportion of total errors
recorded, with age (divided into the five categories
as above) as a between-subjects factor and the four
biweekly monitoring periods as a within-subjects
factor.
Results from the ANOVA conducted on the antihypertensive medications revealed a significant
main effect for age, F(4,42) = 2.57, MSe = 0.013, p =
.05, with the oldest age group and the group of participants between the ages of 45 and 54 making significantly more errors with their antihypertensive
medications (M = .16 and .15, respectively) over the
four biweekly periods than the group of participants between the ages of 55 and 64 (M = .04),
Tukey's Studentized Range = 4.03, p < .05, as shown
in Table 3. The proportion of total errors committed
by the remaining age groups were about the same
(about 4-5%) across the four monitoring periods.
No significant effects were observed from the
ANOVA conducted on the proportion of total errors for the other prescribed medication category.
.097
.153
Table 3. Mean Proportion and Standard Deviations' of Total Errors
as a Function of Age Group Across Biweekly Monitoring Intervals
for Antihypertensive Medications and Other Medications
.183
.038
Type of Medication
.215
Age Group
.059
(.084)
.160
(.250)
.082
(.206)
.011
(.034)
.186
(.206)
.081
(.129)
.124
(.191)
.141
(.258)
.047
(.087)
.202
(.176)
.110
(.137)
.121
(.229)
.124
(.234)
.030
(.077)
.214
(.216)
.088
35^4
.127
45-54
.133
55-64
.026
65-74
.186
75+
"Standard deviations are shown in parentheses.
Antihypertensive
Medications
(n = 47)
Other
Medications
(n = 37)
.053
(.078)
.148
(.181)
.041
(.050)
.048
(.063)
.163
(.154)
.035
(.047)
.058
(.118)
.211
(.275)
.015
(.063)
.038
(.140)
"Standard deviations are shown in parentheses.
614
The Gerontologist
Analyses of Age and Individual Difference Measures. — Individual one-way analyses of variance
were conducted on the Health Practices subscales,
the Perception of Physician Behavior subscales, the
Hypertension Beliefs Questionnaire subscales, the
two Health Status Questionnaire subscales, the
measures of working memory and vocabulary ability, and the Medication Comprehension measure,
with age as a between-subjects variable (using the
five age categories previously described). No age
differences were found in any of the psychosocial
measures, with the exception of the Past Health
scale. The oldest and youngest age groups reported
having had a heart attack, stroke, or having been
hospitalized for more than 5 days more often than
the other age groups, Tukey's Studentized Range =
3.90, p < .05. There were also no significant differences in scores from the working memory task or
the measure of vocabulary ability as a function of
age. Mean scores on the cognitive measures for
each of the age groups are outlined in Table 4.
beliefs about self-care (items shown in the appendix), and the hypertension knowledge questionnaire subscale from the hypertension beliefs questionnaire (shown in Table 5). Although the systolic
blood pressure reading in week 7 was correlated
with total errors, it was also highly correlated with
the blood pressure reading from week 3 (.72).
Therefore, the week 7 reading was not included in
the regression analyses conducted in order to avoid
multicollinearity.
Regression Analyses. — Because more errors in
general were observed toward the end of the monitoring period, regression equations were calculated
only for the last two biweekly monitoring periods.
Separate regression analyses were conducted for
the "antihypertensive medications only" category
and the "other medications" category, with total errors included as the dependent variable in each
case. Because the adherence scores for the antihypertensive medications and the other medications
were highly correlated (.81 for both biweek 3 scores
Correlations Between Individual Difference Mea- and biweek 4 scores), the same variables were included as predictors in both sets of the regression
sures and Adherence. — Because the number of
analyses according to the following plan.
participants in this study was relatively small, including all of the demographic items, questionnaire
The variables entered in the hierarchical regressubscales, and cognitive measures in a regression
sion analyses were the scores from the measure of
working memory, the scores from the beliefs about
analysis would not be meaningful. Therefore, zeroself-care and the knowledge of hypertension suborder correlations were calculated between these
scales, the measure of systolic blood pressure taken
variables and the proportion of total errors for antiin week 3, and chronological age (outlined in Tables
hypertensive medications only, and for the other
6 and 7). According to the social-cognitive model,
medications categories recorded in the last two bicognitive functioning and personal beliefs about illweekly periods, to determine which variables
ness were designated as possible predictors of nonshould be included as predictors in the regression
adherence. Furthermore, other researchers have
analyses that were conducted. Results from the corsuggested that disease state may also influence
relational analyses revealed that only the measures
medication-taking behavior. We used this as backof systolic blood pressure during weeks 3 and 7
ground
to devise the plan for this set of analyses. In
were significantly related to the proportion of total
the first step, We assessed the effect of general cogerrors recorded for antihypertensive medications
nitive functioning on adherence by entering the
only in either of the biweekly periods (outlined in
scores on the working memory task alone. The efTable 5). Five of the variables had significant relafects of illness beliefs and knowledge about illness
tionships to the proportion of total errors recorded
were
examined in the second step, after the meafor other medications in either of the last two bisure of working memory. The third step included
weekly monitoring periods: the measure of working
the proximal measure of systolic blood pressure
memory, systolic blood pressure readings taken
taken
in week 3 as a global indicator of disease
during weeks 3 and 7 of the monitoring period, the
Table 4. Means and Standard Deviations' of Cognitive Variables by Age Croup
Age Group
Cognitive Variables
Vocabulary score
Working memory
Levels
Trials
35^4
(n = 10)
45-54
(n = 9)
55-64
(n = 10)
65-74
(n = 10)
75+
(n = 9)
15.56
(5.57)
13.22
(4.71)
16.80
(6.01)
17.00
(6.53)
16.88
(6.68)
2.00
(1.15)
5.40
(2.59)
2.00
(0.87)
5.11
(2.42)
2.10
(1.20)
5.50
(3.27)
1.90
(0.99)
4.90
(2.51)
1.67
(1.22)
4.44
(3.36)
'Standard deviations are shown in parentheses.
Vol. 37, No. 5,1997
615
Table 7. Outline of Multiple Regression Analyses Conducted
on Other Prescribed Medications for Biweekly Periods Three
and Four for Total Errors
Table 5. Significant Correlations Between Cognitive Variables,
Blood Pressure Readings, Questionnaire Subscales, and the
Proportion of Total Errors for Biweekly Monitoring Periods Three
and Four for Antihypertensive Medications and Other Medications
Predictor
Time Period
Medication Category
Biweek3
Antihypertensive Medications Only
Systolic BP, week 3
Systolic BP, week 7
Other Medications
Working memory
Systolic BP, week 3
Systolic BP, week 7
Hypertension beliefs questionnaire
Self-care subscale
Hypertension knowledge subscale
Biweekly Period Three
Step One
Working memory -.215
Step Two
Working memory
078
Self-care subscale
252*
Knowledge
,274*
Step Three
Working memory ,004
Self-care subscale -.283*
Knowledge
-.189
Blood pressure
reading, week 3 .467**
Step Four
.004
Working memory
Self-care subscale .313*
.166
Knowledge
Blood pressure
reading, week 3 .432**
.119
Age
Biweek4
.48
.31
.43
n.s.
n.s.
.45
.36
-.30
.38
n.s.
-.30
-.30
-.31
n.s.
Nofe: *p < .05.
Table 6. Outline of Multiple Regression Analyses Conducted
on Antihypertensive Medications for Biweekly Periods Three
and Four for Total Errors
Predictor
Biweekly Period Three
Step One
Working memory
Step Two
Working memory
Self-care subscale
Knowledge
Step Three
Working memory
Self-care subscale
Knowledge
Blood pressure
reading, week 3
Step Four
Working memory
Self-care subscale
Knowledge
Blood pressure
reading, week 3
Age
Biweekly Period Four
Step One
Working memory
Step Two
Working memory
Self-care subscale
Knowledge
Step Three
Working memory
Self-care subscale
Knowledge
Blood pressure
reading, week 3
Step Four
Working memory
Self-care subscale
Knowledge
Blood pressure
reading, week 3
Age
Beta
R2Change Cumulative/?2
-.194
-.132
-.077
-.181
.038
.038
.076
Biweekly Period Four
Step One
Working memory
Step Two
Working memory
Self-care subscale
Knowledge
Step Three
Working memory
Self-care subscale
Knowledge
Blood pressure
reading, week 3
Step Four
Working memory
Self-care subscale
Knowledge
Blood pressure
reading, week 3
Age
F
1.67
1.13
-.051
-.121
-.058
.466**
.193**
.269
3.68*
-.043
-.150
-.031
.440*
.105
.008
-.201
-.113
-.079
-.208
.277
2.99*
.040
1.81
.049
.089
1.34
.142*
.231
3.00*
.002
.233
2.37*
.046
2.15
.133*
.179
3.05*
.203**
.382
6.34*
.011
.393
5.19*
.107
5.28*
.199
3.49*
.315
4.71'
.321
3.78*
-.328*
-.211
-.231
-.209
F
.092
-.157
-.254*
-.145
.351** .115*
-.151
-.276*
-.128
.325*
.088
.006
state, as well as the variables previously entered. Finally, chronological age was entered in the fourth
step after all the variables to determine if this variable would explain any additional variance.
The results of the regression analyses revealed that
the blood pressure reading taken in week 3 was a
predictor for adherence not only to antihypertensive medications but also to other prescribed medications included in the participants' regimens. Beliefs about self-care was a significant predictor only
for other prescribed medications. Furthermore,
chronological age did not significantly increase the
observed R2 when it was entered after all of the variables in any of the sets of analyses conducted.
These findings must be taken with caution, however, because the results reported here are subject
to simultaneity bias, as blood pressure influences
-.068
-.101
-.116
.414*
-.057
R2 Change Cumulative/? 2
*p<.05;**p<.001.
-.064
-.116
-.102
.400*
Beta
*p<.05;**p<.001.
616
The Gerontologist
adherence behavior and vice versa. Unfortunately,
limitations in our data prevent us from statistically
controlling for this by conventional methods.
The analyses conducted on the antihypertensive
medications revealed similar results for both biweekly monitoring periods, as shown in Table 6. For
the third biweekly period, approximately 28% of the
variance (R2 = .277) was accounted for by the systolic
blood pressure reading taken during week 3 in the
fourth step. For the fourth biweekly period, the systolic blood pressure reading taken during week 3
accounted for 23% of the variance (R2 = .233) in the
fourth step. None of the other variables were found
to be significant predictors of total errors in either
of the biweekly periods, when all of the variables
were entered into the regression equation in the
last step.
Results from the analyses conducted on the other
prescribed medications included in the participants'
regimens also revealed similar findings for the last
two biweekly monitoring periods. For the third biweekly period, approximately 39% of the variance
{R2 = .393) was accounted for by the beliefs about
self-care subscale and the systolic blood pressure
reading taken during week 3 (as shown in Table 7)
in the fourth step. For the fourth biweekly period,
the same two variables accounted for approximately
32% of the variance (R2 = .321) in the fourth step.
The knowledge of hypertension subscale and the
measure of working memory were not found to be
significant predictors of total errors in either of the
biweekly periods by the time all of the variables had
been entered into the regression equation.
Discussion
There are a number of important findings from
this study and the results may be summarized as follows: (a) the oldest-old and middle-aged adults have
the highest risk of nonadherence to medications,
whereas the young-old are more likely to adhere
than middle-aged adults; (b) high blood pressure
readings predicted nonadherence to antihypertensive medications; and (c) beliefs about self-care
were predictors of adherence to other types of medications only. Each of these major findings is discussed below.
Age, Medication Type, and Adherence
Clear effects of age on nonadherence were observed in this sample of hypertensive patients, as
the oldest adults and the adults aged 55-64 made
more errors in medication taking than the other age
groups across the monitoring period. The types of
errors they made were primarily omissions. The
lowest number of errors was observed in the
young-old age group (approximately 4%). The rate
of nonadherence for the other participants fell in
between the oldest groups (65-74 and 75+). Furthermore, there is some evidence to indicate that the
oldest-old and those participants between the ages
of 45 and 54 years of age were more likely to make
Vol. 37, No. 5,1997
errors with their antihypertensive medications than
the other age groups. These findings are consistent
with previous work conducted by Park et al. (1992),
which documented a greater likelihood of the oldest-old to omit dosages of medications relative to
young-old adults. The present study suggests that
this is a reliable finding, and also provides important data about middle-aged adults that Park et al.
did not study. Middle-aged adults may also be at
risk for nonadherence.
It is tempting to speculate that the reason for the
oldest-old not taking their medication as prescribed
may be due to declines in cognitive functioning. We
did not, however, find age differences in the measures of working memory and vocabulary that were
administered, as is typically found in aging research.
Working memory capacity has been reliably shown
to decrease with age, and vocabulary ability generally remains stable or increases with age (see Salthouse, 1991). There is a trend in the expected direction for both of these measures in the data, but
perhaps low statistical power, as a result of the limited number of participants in each cell, restricted
the ability to detect significant differences. It should
also be noted that the usual comparison in which
age-related differences in working memory are reported is between young adults (ages 18-25) and
older adults (over the age of 60). Our comparisons
were between middle-aged and older adults, a factor that may have also contributed to our inability to
detect these differences. Furthermore, there are significant moderate correlations between the working memory measures and the level of education of
participants (correlation between working memory
level score and education = .35, and correlation between working memory span score and education =
.43). Therefore, it is possible that educational attainment moderated the decline in working memory, as
the oldest age group was the most highly educated.
Finally, the working memory span scores for the
participants are somewhat low when compared to
other samples, which have ranged from 2.91 to 2.33
for individuals between the ages of 60 and 87 (see
Salthouse & Babcock, 1990). This finding may be
due to our presenting the task verbally, which
would increase the cognitive demands and probably lower scores relative to the standard method of
presenting the task in a visual format. Thus, it is possible that our findings may not be completely generalizable to the total population because of these
low scores.
It is interesting that certain groups of middleaged adults were also more nonadherent than other
groups in the sample. No evidence, however, is available from this study to suggest that beliefs about illness or medications might have influenced the adherence rates of the middle-aged differentially, as
there were no main effects of age observed on any
of the medication beliefs subscales. Thus, it is unlikely that differences in illness or medication beliefs
between the age groups accounted for the findings
that some of the middle-aged adults were also
highly nonadherent. Perhaps the busy schedules of
617
same time. The results from the regression analyses
also provide some support for the social-cognitive
model as originally proposed by Park and her colleagues (see Park & Mayhorn, 1996). The measure of
beliefs about self-care was shown to be the primary
predictor for adherence to prescribed medications
other than antihypertensive drugs. Thus, beliefs
about taking medications may be more important
for some types of medications than others. It would
be interesting to conduct further research to ascertain which types of medications are affected by
what kinds of beliefs.
Additionally, the results from the correlational
analyses suggested that working memory capacity
might play a role in one's ability to adhere to a
medical regimen, because the measure of working
memory was correlated with nonadherence for the
other prescribed medications the individuals were
taking. It would be expected that working memory
might be more influential in managing a regimen of
multiple medications relative to taking a single
medication because the organizational demands of
maintaining a large regimen would be greater than
with a small one (see Salthouse, 1991). This may be
especially true for the elderly as the regimen becomes more complex over time (Cerrella, Poon, &
Williams, 1980; Salthouse, Mitchell, Skovronek, &
Babcock, 1989). This measure of working memory,
however, was not found to be a significant predictor of total errors in the regression analyses conducted, possibly because of low variability of scores
in this sample. Perhaps the use of a more sensitive
measure of working memory or other measures of
cognition might help to "tease out" this effect in future work.
Therefore, in summary, the results from this research provide substantial information about adherence to prescribed medical regimens by individuals
diagnosed with hypertension. These findings also
clearly demonstrate the influence of the complex
intermix of a number of different types of factors on
adherence to prescribed medical regimens.
the middle-aged adults might have interfered with
their ability to take medications as prescribed. This
possible influence on nonadherence was not addressed in this study. Therefore, future research is
needed in this area to make this distinction.
These data suggest that young-old adults are the
most adherent of all of the age groups, as was
shown in previous work by Park and colleagues. To
account for this finding, it is possible that the
young-old adults may be more focused on their
medication-taking behaviors than middle-aged
adults because of less hectic schedules. Furthermore, it is unlikely that young-old adults are experiencing major cognitive deficits that would undermine their ability to adhere. Therefore, they would
be better able to handle the cognitive demands imposed on them by complex prescribed medical regimens than would the oldest-old.
Finally, the most common mistakes made in medication taking by the participants were errors of
omission. Few quantity or commission errors were
observed. This finding suggests that cognitive aids
such as telephone reminder systems (see Leirer,
Tanke, & Morrow, 1993; Tanke & Leirer, 1993) or the
use of organizers and organizational charts (see
Park, Morrell, Frieske, Blackburn, & Birchmore,
1991) might be beneficial in helping individuals remember to take medications.
Adherence to Antihypertensive Medications and
Blood Pressure. — Results from the correlational
and regression analyses conducted on only antihypertensive medications indicated that readings of
high blood pressure were the primary predictor of
nonadherence for this particular class of drugs.
Higher systolic readings in week 3 of monitoring
were shown to be related to later nonadherence.
Because a major purpose of this study was to look
at predictors of nonadherence, it is of clinical importance to demonstrate that high blood pressure
readings may not only be a sign of uncontrolled hypertension. Such elevated readings may, in fact, be
indicative of nonadherence, just as are beliefs, cognitive function, and other variables. Clinicians who
detect high blood pressure readings in patients
might be able to use this measure as a warning that
medication has not been taken as prescribed.
Adherence to Other Medications, Beliefs, and
Cognition. — Another finding was that the best predictor (a measure of systolic blood pressure) for adherence to antihypertensive medications was also a
significant predictor for other prescribed medications in the regimen. This finding suggests that individuals may form strategies for taking medications,
such as taking all of them in the morning. In this
sample, all participants were taking antihypertensive medications. Thus, it is possible that the relationship that surfaced between the blood pressure
reading and adherence to antihypertensive medications was also observed for the other medications
in their regimens, as most of the participants in this
study were taking all of their medications at the
References
Baddeley, A. D. (1986). Working memory. Oxford: Clarendon.
Cerella, J., Poon, L W., & Williams, D. (1980). Age and the complexity hypothesis. In L W . Poon (Ed.), Aging in the 1980s: Psychological issues.
Washington, DC: American Psychological Association.
Cummings, K. M., Kirscht, J. P., Binder, L R., & Codley, A. J. (1982). Determinants of drug treatment maintenance among hypertensive persons in
inner city Detroit. Public Health Reports, 97, 99-106.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in working
memory and reading. Journal of Verbal Learning and Verbal Behavior,
19,450-466.
DeVon, H. A., & Powers, H. J. (1984). Health beliefs, adjustment to illness,
and control of hypertension. Research in Nursing and Health, 7,10-16.
Duke University Center for the Study of Aging and Human Development.
(1975). OARS Multidimensional Functional Assessment Questionnaire.
Durham, NC: Author.
Dunbar-Jacob, J., Dwyer, K., & Dunning, E. J. (1991). Compliance with antihypertensive regimens: A review of the research in the 1980s. Annals
of Behavioral Medicine, 13, 31-39.
Gardner, E. F., & Monge, R. H. (1977). Adult age differences in cognitive
abilities and educational background. Experimental Aging Research, 3,
337-383.
Hamilton, C. A., Roberts, S. J., Johnson, J. M., Tropp, J. R., Anthony-Odgren,
D., & Johnson, B. F. (1993). Health Values, 17, 3-11.
Jette, A. M., Cummings, K. M., Brock, B. M., Phelps, M. C , & Naessens, J.
618
The Gerontologist
(1981). The structure and reliability of health belief indices. Health Services Research, 76,81-98.
Joint National Committee on Detection, Evaluation, and Treatment of High
Blood Pressure. (1993). The fifth report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure
(JNC V). (1993). Archives of Internal Medicine, 153,154-183.
Leirer, V. O., Tanke, E. D., & Morrow, D. C. (1993). Commercial cognitive/
memory systems: A case study. Applied Cognitive Psychology, 7,
675-689.
Leirer, V. O., Morrow, D. C , Pariante, G. M., & Sheikh, J. I. (1988). Elders'
nonadherence, assessment, and computer assisted instruction for medication recall training. Journal of American Geriatrics Society, 36,877-994.
Leventhal, H., & Cameron L. (1987). Behavioral theories and the problem of
compliance. Patient Education and Counseling, 10,117-138.
Lorenc, L, & Branthwaite, A. (1993). Are older adults less compliant with
prescribed medication than younger adults? British Journal of Clinical
Psychology, 32,485-492.
Meyer, D., Leventhal, H., & Cutmann, M. (1985). Common sense models of
illness: The example of hypertension. Health Psychology, 4,115-135.
Morrell, R. W., Park, D. C , & Poon, L. W. (1989). Quality of instructions on
prescription drug labels: Effects on memory and comprehension in
young and old adults. The Cerontologist, 29, 345-353.
Morrell, R. W., Park, D. C , & Poon, L. W. (1990). Effects of labeling techniques on memory and comprehension of prescription information in
young and older adults. Journal of Gerontology: Psychological Sciences, 45, P166-P172.
Nagy, V. T., & Wolfe, C. R. (1984). Cognitive predictors of compliance in
chronic disease patients. Medication Care, 22,912-921.
National High Blood Pressure Education Program Working Group. (1994).
National High Blood Pressure Education Program Working Group Report on Hypertension in the Elderly. Hypertension, 23, 275-285.
Park, D. C. (1992). Applied cognitive aging research. In F. I. M. Craik & T. A.
Salthouse (Eds.), The handbook of aging and cognition (pp. 449-493).
Hillsdale, NJ: Lawrence Erlbaum.
Park, D. C. (1994). Self-regulation and control of rheumatic disorders. International Review of Health Psychology, 3,189-217.
Park, D. C , & Jones, T. R. (1996). Medication adherence and aging. In A. D.
Fisk & W. A. Rogers (Eds.), Handbook of human factors and the older
adult. Hillsdale, NJ: Lawrence Erlbaum.
Park, D. C , & Kidder, D. P. (1996). Prospective memory and medication adherence. In M. Brandimonte, G. O. Einstein, & M. A. McDaniel (Eds.),
Prospective memory theory and applications (pp. 369-390). Hillsdale,
NJ: Lawrence Erlbaum.
Park, D. C , & Mayhorn, C. B. (1996). Remembering to take medications:
The importance of nonmemory variables. In D. Herrmann, M. Hohnson, C. McEnvoy, & C. Hertzog (Eds.), Research on practical aspects of
memory, Vol. 2. Hillsdale, NJ: Lawrence Erlbaum.
Park, D. C, Morrell, R. W., Frieske, D., Blackburn, A. B., & Birchmore, D.
(1991). Cognitive factors and the use of over-the-counter medication
organizers by arthritis patients. Human Factors, 33, 57-67.
Park, D. C , Morrell, R. W., Frieske, D., & Kincaid, D. (1992). Medication adherence behaviors in older adults: Effects of external cognitive supports. Psychology and Aging, 7, 252-256.
Park, D. C , Willis, S. L, Morrow, D., Diehl, M., & Gaines, C. (1994). Cognitive function and medication usage in older adults. The Journal of Applied Gerontology, 13, 39-57.
Richardson, M. A., Simons-Morton, B., & Annegers, J. F. (1993). Effect of
perceived barriers on compliance with antihypertensive medication.
Health Education Quarterly, 29, 489-503.
Rudd, P. (1995). Clinicians and patients with hypertension: Unsettled is-
Vol. 37, No. 5,1997
sues about compliance. Journal of the American Heart Association, 130,
572-579.
Rudd, P., Byyny, R. L, Zachary, V., LoVerde, M. E., Titus, C , Mitchell, W. D.,
& Marshall, G. (1989). The natural history of medication compliance in
a drug trial: Limitations of pill counts. Clinical Pharmacology Therapy,
46,169-176.
Salthouse, T, A. (1991). Theoretical perspectives in cognitive aging. Hillsdale, NJ: Lawrence Erlbaum.
Salthouse, T. A., & Babcock, R. (1990). Computation span and listening span
tasks. Technical Report. Atlanta, GA: Georgia Institute of Technology.
Salthouse, T. A., Mitchell, D. R. D., Skovronek, E., & Babcock, R. (1989). Effects of adult age and working memory on reasoning and spatial abilities. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15,507-516.
Sharkness, C. M., & Snow, D. A. (1992). The patient's view of hypertension
and compliance. American Journal of Preventive Medicine, 8,141-146.
Shaw, E., Anderson, J. G., Maloney, M., Jay, S. J., & Fagan, E. (1995). Factors
associated with noncompliance of patients taking antihypertensive
medications. Hospital Pharmacy, 3, 207.
Schulman, N., Cutter, G., Daugherty, R., Sexton, M., Pauk, G., Taylor, M. J.,
& Tyler, M. (1982). Correlates of attendance and compliance in hypertensive detection and follow-up program. Controlled Clinical Trials, 3,
13-27.
Skaer, T. L, Sclar, D. A., Markowski, D. J., & Won, J. K. H. (1993). Effect of
value-added utilities on prescription refill compliance and health care
expenditures for hypertension. Journal of Human Hypertension, 7,
515-518.
Stanton, A. L. (1987). Determinants of adherence to medical regimens by
hypertensive patients. Journal of Behavioral Medicine, 10, 377-394.
Tanke, E. D., & Leirer, V. O. (1993). Use of automated telephone reminders
to increase elderly patients' adherence to tuberculosis medication appointments. Proceedings of the Human Factors and Ergonomics Society, 193-196.
Wolf, M. H., Putnam, S. M., James, S. A., & Stiles, W. B. (1978). The medical
interview satisfaction scale: Development of a scale to measure patient
perceptions of physician behavior. Journal of Behavioral Medicine, 1,
Received August 20,1996
Accepted April1,1997
Appendix. Items From the Self-Care Subscale
of the Patient's Hypertension Beliefs Questionnaire
1. How likely is it that you will get other illnesses that are
not life-threatening on a regular basis (such as a cold)?
1 = very likely; 2 = likely, 3 = neutral; 4 = unlikely,
5 = very unlikely.
2. How likely is it that you would stop taking medicine if
it was costing a lot of money?
1 = very likely; 2 = likely; 3 = neutral; 4 = unlikely;
5 = very unlikely.
3. How well do you take care of your health?
1 = extremely well; 2 = very well; 3 = fairly well; 4 = not
very well; 5 = not at all.
619