a psychosocial perspective in the explanation of patients' drug-taking behavior
TRANSCRIPT
sot. sci. hied. v01. 27, NO. 3, pp. 277-285, 1988 Printed in Great Britain
0277-9536/aa 53.00 + 0.00 Pergamon Press pk
A PSYCHOSOCIAL PERSPECTIVE IN THE EXPLANATION OF PATIENTS’ DRUG-TAKING
BEHAVIOR
L. DOUGLAS RIED’ and DALE B. CHRISTENSEN’ ‘Assistant Professor of Pharmacy Practice and 2Associate Professor of Pharmacy Practice, School of
Pharmacy X-69, University of Washington, Seattle, WA 98195, U.S.A.
Abstract-The goal of this study was to examine the applicability of the Health Belief (HBM) and Theory of Reasoned Action (TRA) models in predicting drug-taking compliance behavior among female patients with uncomplicated urinary tract infections.
Thirty-eight percent of the respondents were compliant, 25% finished all of their medication, but missed one or more doses during the course of therapy, and 37% did not finish all of their medication as directed. Two HBM and three TRA variables had a statistically significant relationship with compliance: barriers and benefits (HBM) and belief strength, outcome evaluation, and behavioral intention (TRA). HBM variables explained 10% of the variance in the compliance variable. Adding the TRA variables to the model explained an additional 19% of the variance. Social influence variables (normative expectations, motivation to comply, and subjective norms) had a significant influence in the prediction of behavioral intention, but not in the prediction of compliance behavior.
Suggestions for improving compliance based on these findings include: simplification of drug therapy by customizing and simplifying the regimen, continued patient reminders of the therapy’s value, and benchmarks for patients to evaluate the success or failure of therapy. A frank and open discussion from the start of therapy about its complexity, the medication’s side effects, time and fiscal constraints, and other factors which may modify patients’ perception of the likelihood of compliance is an important key to improving patients’ compliance behavior.
Key words-Health Belief Model, Theory of Reasoned Action, compliance, health beliefs, drug-taking behavior
INTRODUCTION
For the past decade or more, attention has focused on the drug-taking compliance ‘problem’. Consid- erable effort has been directed to identify the reasons for noncompliance and to employ them in developing intervention strategies. Initial inquiries focused on patient sociodemographic and disease-related factors associated with and, presumably, predictive of non- compliance. While useful, these factors were un- satisfactory because they tended to differ across settings and patient groups. Further, they were not particularly instructive in understanding the compli- ance decision-making process, which would be useful in developing intervention strategies. It became in- creasingly clear that a theory or model of compliance behavior was needed to incorporate these disparate factors.
The earliest and the most thoroughly examined health behavior model is the Health Belief Model (HBM) (11. The HBM originally postulated that an individual’s health-related behaviors were dependent upon: (1) one’s readiness to take action, which is determined by perceptions of the seriousness of the disease and susceptibility (or re-susceptibility) to a particular disease or its sequelae; (2) one’s assessment of the benefits versus costs of alternative health behaviors, and; (3) a cue or stimulus to trigger the response. In the first case, the model suggests that persons who are more likely to be compliant perceive themselves to be susceptible to a health problem or perceive the health problem to be serious. In the
second case, benefits are evaluated in terms of the likelihood of disease prevention, cure, control, or symptom relief. Costs of alternative treatment plans (including no treatment) are defined broadly to in- clude economic costs, inconvenience, discomfort, and time costs. The assessment of costs and benefits is considered relative to the perceived threat to health. Finally, cues to action can be internal (e.g. symp- toms) or external (e.g. social pressure).
The HBM was originally proposed to explain preventive health behavior. However, it has since been expanded to explain sick role behavior and compliance with prescribed medication regimes. Becker and Maiman modified the model by adding elements of general health motivations (e.g. concerns about health, orientation toward preventive health behavior) and modifying and enabling factors (e.g. general faith in physicians and medical care, and treatment regimen complexity) [2]. Accumulated evi- dence has partially validated the model in a variety of disease circumstances [3-lo]. Individual elements of the model (e.g. perceptions of susceptibility and severity) have been shown to LX salient predictors of physician contact by persons perceiving themselves to be sick, behavior by persons diagnosed as sick, compliance behavior with short term drug therapies, and drug-taking compliance behavior for long term drug therapies by patients with chronic conditions.
While certain elements of the HBM have been supported empirically, the model in its entirety has not. The model has been criticized as incomplete and the relationships among its predictive elements have
277
278 L. DOLJGLA~ RED and DALE B. CHRISTENSEN
Salient
be1iefs \ Attitude
0utc0ms A evaluation
\ Behavloro~ ,ntention - Behavior
Normative
/ be’iefr \ Subjective
Motivotion~norms to comply
Fig. 1. Theory of reasoned action.
not been sufficiently articulated or empirically estab- lished. Second, other important determinants (e.g. intention to comply) have not routinely been tested in the HBM framework. While the original HBM has been modified to include intent to comply as a health motivation variable [2], it has been examined in only one other previous study [ll]. Third, the HBM is recognized as an adaptation of a value-expectancy model, yet the concepts of expected value of a particular outcome and the likelihood of its occur- rence have not been theoretically developed or tested. Fourth, while the role of social influences on an individual’s behavior is acknowledged in some adap- tations, investigations of social influences are not generally included in empirical tests. Langlie, who has made one of the few formal attempts to incorpo- rate the social network variables in the health belief framework, found them to have considerable joint and independent influence on preventive health be- haviors [12]. Finally, lack of standardization among measurements of psychosocial variables is among the reasons for the inconsistency found in the HBM study’s findings [2, 131.
The Theory of Reasoned Action (TRA) is a general model of behavior which is based upon the same precepts as other value-expectancy models [14, 151. The TRA (Fig. 1) addresses some of the afore- mentioned shortcomings of the HBM. First, it estab- lishes a link between behavioral intention and behav- ior, and specifies the inter-relationships and causal processes among the variables. Second, it addresses the value-expectancy concepts through the salient belief strength and outcome evaluation variables. According to the TRA, patients evaluate the benefits and barriers to compliance in terms of their like- lihood of occurrence. Third, it recognizes the im- portance of social influences on the individual, namely perceptions of referent others’ normative expectations, subjective norms, and the individual’s degree of motivation to comply with referent others. Finally, measurement of the concepts are specified in detail [14] and can be adapted to TRA’s specific behaviors, including health-related behaviors [ 161. This would result in more consistent measurement of the TRA’s concepts, facilitating comparisons across studies.
Ried et al. examined the applicability of this model in explaining compliance intentions among 300 male veterans taking antihypertensive medications [ 171. These patients were given questionnaires inquiring about their beliefs concerning susceptibility to hyper-
tension and its sequelae, the benefits of drug-taking, the seriousness of the disease, perceptions of their physicians’ expectations about their behavior, subjec- tive norms, motivation to comply, attitudes, and drug-taking compliance intention. As predicted by the model, all independent variables were positively correlated with intention to take medicines. Overall, the variables predicted 35% of the variance Patients who expressed an intention to take their high blood pressure medicines perceived themselves to be more susceptible to the disease or its sequelae, expected to gain positive benefits, and had more favorable attitudes about taking their medication. Further, salient health beliefs and physicians’ expectations significantly affected development of favorable atti- tudes towards taking prescribed high blood pressure medicines. Results of path analysis showed that attitudes and subjective norms had the greatest direct impact upon the development of intention to take the medication as prescribed. Two other factors, sus- ceptibility and physicians’ normative expectations, also had significant direct effects on drug-taking intentions.
While the study of VA patients was instructive, it also raised several other questions. Only male hyper- tensives were studied. Is the model equally effective in explaining drug-taking behavioral intentions and compliance behavior among other populations? Among patients with other types of diseases (e.g. acute conditions)? Finally, is patient’s subsequent drug-taking behavior consistent with their intentions and can compliance be predicted as well as behavioral intentions?
Study objectives
The goal of this study was to examine these questions by exploring the applicability of the HBM and TRA models in predicting compliance behavior among female patients with uncomplicated urinary tract infections (UTI). The specific study objectives were to:
1. E,xamine the inter-relationships among attitudes, health beliefs, social influences, and behavioral in- tentions;
2. Examine the relationship between attitudes, health beliefs, social influences, and behavioral in- tentions and patients’ drug-taking compliance behav- ior, and;
3. Evaluate and compare the ability of the HBM and the TRA to predict drug-taking compliance.
Patients’ drug-taking behavior 279
METHODOLOGY Independent variables
Subjects
Two groups of subjects were identified for this study. One was female students seeking medical care at the Oregon State University Student Health Cen- ter. The other group consisted of female enrollees at Kaiser Permanente, a large prepaid group practice HMO in the Portland, Oregon area. Eligible subjects were those presenting to the student health center and HMO pharmacies a prescription to treat an uncom- plicated UTI. UT1 was chosen as the study disease because symptoms of UTIs are easily recognized and associated with the disease, it is usually of short duration and nonrecurrent, and consistent drug- taking for short periods is an established and effective treatment 118201.
The independent variables included in this study are drawn from the HBM and TBA (Appendix). Measures of patients’ beliefs about the benefits and barriers to compliance, seriousness of non- compliance, and susceptibility to the disease’s adverse consequences are included from the HBM. Measures of the patient’s outcome evaluation, salient belief strength, normative expectations, motivation to com- ply, attitude, subjective norms, and behavioral in- tention are included from the TR4. The outcome evaluation and salient belief strength variables repre- sent patients’ expected value of a particular outcome and likelihood of its occurrence. Social influences are represented by the normative expectations, subjective norms, and motivation to comply variables.
Procedures
During a 30-day period in March, 1984 at the student health center, and a 90-day period from April to June 1985 at the HMO, all patients presenting a prescription for 20 tablets of trimethoprim 160 mg/sulfamethoxazole 800 mg for treatment of an uncomplicated UT1 were asked to participate in the study. If they consented, they were asked to complete a self-administered questionnaire while waiting to have their prescription filled. When completed, pa- tients placed their questionnaire in a locked box located at the pharmacy or mailed it directly to the primary author if they could not wait. Questionnaires were distributed in this fashion to 40 patients seen at the student health center and 148 patients seen at the HMO.
Ten of the 11 independent variables were initially measured with multiple items. Only subjective norms was operationalized by a single item, patterned after the wording suggested by Ajzen and Fishbein [14]. Patients’ responses to the individual items were summed as a composite index. The internal consis- tency of nine of the indices was calculated with item analysis*. The alpha coefficients are shown in Table 1.
Dependent variable
All patients selected were prescribed the same drug regimen in order to control for length and complexity of drug therapy. Upon receipt of the filled pre- scription, each patient received specific verbal and written instructions from the pharmacist to take 1 tablet every 12 hr for 10 days.
The compliance variable was operationalized by asking the patient, “Have you finished all of your medicine? If not, how many tablets do you have remaining?’ The number of tablets which should have been remaining (20 tablets - [no. of days x 2 tablets/day]) was subtracted from the number of tablets actually remaining, as determined from the pill count. If the result was within + 1 tablet, the patient was considered to be compliant. If patients were directed to discontinue taking the medication by their physician because of an adverse drug reaction, allergy, or some other reason, they were considered to be compliant.
Patients were interviewed by telephone approx. 10 days later and asked about their actual drug-taking behavior and current health-related beliefs and atti- tudes. A questionnaire was mailed to patients if they could not be contacted by telephone within 4 days. Telephone interviews or mailed questionnaires were completed and returned by 35 student health center patients (87.5% response rate) and 78 HMO patients (52.7% response rate). The combined response rate was 60%.
Inaccurate reporting of compliance rates is a recog- nized problem. In this study, we attempted to en- hance response validity by first assuring the subjects of the confidentiality of their responses. Further, questions were phrased in the manner suggested by Sackett and others to maximize accurate recording. These approaches appeared moderately successful. The reported level of noncompliance in this study was found to be similar to others’ estimates of non- compliance [21, 221, and patients interviewed by tele- phone could be heard counting the tablets.
Patients were also asked if they had “ever forgotten a dose” and if they had taken their medication “exactly as directed for all 10 days?” Patients were considered noncompliant if they answered ‘yes’ to the first question and ‘no’ to the second. Whether or not patients had completed the prescribed medication was of paramount importance in assigning comph- ante scores to patients. Not taking the medication exactly as directed in all instances (e.g. forgetting a dose, but taking it later) was considered to be of lesser importance. Compliance scores were assigned to patients based on the hierarchy shown in Table 2.
The SPSS and SPSS/PC+ computer packages were used to perform the statistical analyses [23-251. Statistical significance is reported at P Q 0.05.
*One index, perceptions of barriers, consisted of only two items, therefore calculation of the alpha coefficient was not possible.
Student health center patients averaged 23 years (range 18-35 years) and HMO patients averaged 30 years (range 16-79). A majority of the HMO patients had at least some college education. The distribution of compliance scores showed that 38% of the re- spondents were compliant (Table 2). An additional 25% of the respondents finished all of their medica-
RESULTS
Tab
le
I.
Inte
rcor
rela
tion,
m
ean,
st
anda
rd
devi
atio
n an
d re
liabi
lity
estim
ates
of
va
riab
les’
(X,)
(X
,)
(X,)
(X
,)
(X,)
(X
,)
(X,)
(X
s)
(X,)
(X
I,)
(X,1
) (X
,2)
Bar
rier
s (X
,)
Seri
ous
(X,)
Su
scep
tibili
ty
(X,)
B
enef
its
(X,)
B
elie
f st
reng
th
(X,)
O
utco
me
eval
uatio
n (X
,)
Nor
mat
ive
expe
ctat
ions
(X
,)
Mot
ivat
ion
(Xs)
A
ttitu
de
(X,)
Su
bjec
tive
norm
(X
,,)
Inte
ntio
n (X
, ,)
Com
plia
nce
(X,,)
M
ean
sn
1.00
0.
16*
0.16
’ I .
oo
0.16
. Po
.lo
0.38
” 0.
23**
0.
08
-0.1
0 0.
01
0.28
.’
0.13
0.
17’
0.11
0.
18.
0.20
’ 0.
08
0.16
’ 0.
20.
0.20
’ 0.
14
0.31
” 0.
09
9.09
2.
85
2.68
0.
73
0.17
. -0
.11 1.00
0.
04
0.07
-0
.04
- 0.
03
-0.0
2 0.
01
0.00
0.
10
0.03
2.
85
0.49
0.43
.’
0.26
**
- 0.
05
I .I0
0.
18.
0.09
0.
17
0.09
0.
55..
0.16
. 0.
08
0.17
’ 5.
51
0.65
0.09
-0
.11
0.08
0.
22.
I .oo
0.
08
0.04
0.
02
0.07
0.
10
0.12
0.
32.’
3.
68
0.44
0.01
0.
15
0.13
0.
22.
0.16
’ 0.
31.’
0.
20’
0.21
. 0.
09
0.21
’ 0.
04
0.04
0.
03
0.01
0.
00
0.1
I 0.
22
0.12
0.
69.’
0.
18**
0.
09
0.05
0.
03
0.09
0.
01
1.00
0.
16.
0.32
.’
0.03
0.
03
0.13
1.
00
0.50
’.
0.32
” 0.
49”
0.25
.’
0.37
.’
I SK
I 0.
08
0.25
. 0.
03
0.24
” 0.
06
I .JO
0.
29**
-0
.03
0.42
” 0.
21.
0.26
.’
1.00
-0
.01
-0.0
1 - -
0.04
0.
05
0.09
0.
20’
-0.0
6 0.
01
0.1
I 0.
08
5.18
3.
83
3.16
5.
80
I .40
0.
62
n V
i n
75
n 43
n
54
-.--
--
C
ronb
ach’
s al
pha
0.93
0.
83
0.80
0.
84
0.84
0:
;; 0:
;; 0.
80
0.41
**
0.26
” 0.
12
0.27
’.
0.23
’ 0.
01
0.21
l
0.11
0.
22’
0.27
’ I .
OO
0.
29”
3.86
0.
42
0.31
” 0.
09
0.04
0.
19’
0.35
**
0.22
’ 0.
07
0.01
0.
12
0.08
0.
33**
I.
00
2.89
I.
05
l P
c 0.
05.
**P
Q
0.0
1.
‘Cor
rela
tion
mat
rix
belo
w
the
diag
onal
re
pres
ents
in
terc
orre
latio
n of
var
iabl
es
usin
g th
e si
ngle
ite
m
inte
ntio
n va
riab
le,
corr
elat
ion
mat
rix
abov
e th
e di
agon
al
repr
esen
ts
inte
rcor
rela
tion
of
vari
able
s co
rrec
ted
for
atte
nuat
ion.
~~
al
pha
coef
fici
ent
for
the
beha
vior
al
inte
ntio
n in
dex
was
lo
w
and
elim
inat
ion
of
any
one
item
re
duce
d th
e in
dex’
s re
liabi
lity
even
fu
rthe
r.
Tw
o al
tern
ativ
es
to
this
pr
oble
m
wer
e
eval
uate
d.
Firs
t, a
sing
le
item
, “H
ow
likel
y is
it
that
yo
u w
ill
take
th
e m
edic
ine
for
your
in
fect
ion
exac
tly
as d
irec
ted
for
the
next
IO
day
s ”
was
sel
ecte
d to
be
use
d as
the
m
easu
re
of
beha
vior
al
inte
ntio
n.
Thi
s pa
rtic
ular
ite
m
was
ch
osen
fr
om
the
thre
e al
tern
ativ
es
beca
use
it ha
d th
e hi
ghes
t ze
ro
orde
r co
rrel
atio
n w
ith
the
com
plia
nce
depe
nden
t va
riab
le
and
this
w
ordi
ng
is r
ecom
men
ded
by
Ajz
en
and
Fish
bein
[l
4)
and
was
us
ed
by
Bec
ker
e! a
l. [I
I).
Sec
ond,
th
e co
rrel
atio
ns
wer
e “c
orre
cted
fo
r at
tenu
atio
n”.
In
this
ca
se,
sam
ple
corr
elat
ions
ob
tain
ed
from
th
e da
ta
are
adju
sted
to
co
mpe
nsat
e fo
r th
e er
ror
or
unre
liabi
lity
of
the
mea
sure
men
t of
th
e va
riab
les.
T
his
was
do
ne
usin
g th
e fo
rmul
a:
Whe
re
I is
the
cor
rect
ed
corr
elat
ion;
ri
y is
the
ob
tain
ed
corr
elat
ion;
r_
an
d r,
, ar
e th
e re
liabi
lity
estim
ates
of
th
e tw
o va
riab
les
(26.
271.
M
anip
ulat
ions
of
the
co
rrec
ted
corr
elat
ions
sh
ould
iiv
e a
clos
er
appr
oxim
atio
n to
the
the
oret
ical
ly
unde
rlyi
ng
rela
tions
hips
th
an
wou
ld
man
ipul
atio
ns
of t
he
unco
rrec
ted
data
[Z
S].
Reg
ress
ions
w
ere
run
usin
g bo
th
corr
elat
ion
mat
rice
s an
d co
mpa
red.
T
here
w
ere
no
subs
tant
ive
diff
eren
ces
in
the
resu
lts
whe
re
the
com
plia
nce
vari
able
w
as
the
depe
nden
t va
riab
le.
How
ever
, us
e of
th
e co
rrec
ted
corr
elat
ion
mat
rix
impr
oved
th
e pr
edic
tion
of
beha
vior
al
inte
ntio
n si
gnif
ican
tly
(R’
= 0.
55
vers
us
0.18
).
The
re
sults
re
port
ed
here
in
are
base
d on
th
is
seco
nd
alte
rnat
ive.
_”
- .-
_
- -
-..-
-
-._
. .^
-.
.~
- ._
.
-.
_ --
_ _
_ -
- -
- -
- -.
-
Patients’ drug-taking behavior
Table 2. Patient corn&awe
281
Number %
Finished AND took exactly 43 38. I Finished BUT not taken exactly 28 24.8
NOT finished, took exactly 28 24.8
NOT finished, not taken exactly 14 12.3
Total 113 100
Table 3. Comparison of HBM and TRA in predicting compliance
HBM alone’ HBM plus TRA*
Susceptibility .O.Ol -0.04
Serious 0.03 -0.01
Benefits 0.06 0.04
Barriers 0.28t 0.24t
Intention - 0.20t
Normative expectations - -0.17t
Belief strength - 0.26f
Outcome evaluation - 0.227
Attitude - 0.03
Motivation to comply - -0.03
Subjective norms - 0.11
Total R’ 0.10 0.29
F ,4,,,2, = 2.80; P < 0.03 F ,,,, p51 = 3.48; P < 0.001
Sig. of RZ change = FC7,9Jl = 3.59; P -z 0.002.
*Standardized partial regression co&icient.
tP <0.05. tP < 0.01.
tion, but missed one or more doses during the course of therapy. The remaining 37% did not complete the course of therapy as directed.
The possibility of nonresponse bias was explored because of the observed differences in the response rate between the two patient groups. Respondents unable to be followed up, either by mail or telephone, may have been different in their pre-drug regimen beliefs, attitudes and perceptions of norms. Re- sponder’s and nonresponder’s mean scores for each of the independent variables were compared using Student’s t-test. The two group’s mean scores did not differ on any of the 11 independent variables.
Next, the possibility of differences between the student and HMO patient’s pre-drug regimen beliefs, attitudes and perceptions of norms influencing the results was investigated. Respondent’s mean scores did not differ on any of the variables, including the compliance variable. This suggets that the beliefs, attitudes and perceptions of social expectations of these two groups of female patients were similar. In the absence of any significant differences, the two patient groups were pooled together for the remain- ing analyses.
The inter-relationships among the variables under study were next explored. Means, standard devi- ations, and zero order correlations are presented in Table 1. As expected, most independent variables
were positively correlated with each other and with behavioral intention and compliance. Perceptions of barriers, benefits, seriousness, belief strength, nor- mative expectations, attitude and subjective norms were associated with behavioral intention. Compli- ance was significantly associated with two of the HBM variables, barriers and benefits, and three of the TRA variables, belief strength, outcome evalu- ation and behavioral intention. Belief strength has the highest correlation with compliance, followed by behavioral intention, perceptions of barriers, out- come evaluation, and perceptions of benefits of the drug therapy.
Stepwise multiple regression analysis was used to examine the utility of HBM variables alone in pre- dicting compliance (Table 3). Combined, the HBM variables accounted for 10% of the variance. The one statistically significant predictor was perceptions of barriers to compliance. Despite a statistically significant zero order correlation with compliance, perceived benefits did not enter into the equation once barriers had entered. This was due in part to its moderately high correlation with barriers (r = 0.38).
It was hypothesized that addition of the TRA value-expectancy, social influence and intention vari- ables would significantly improve prediction of com- pliance when compared to the HBM variables alone. This hypothesis was investigated using hierarchical
Table 4. Summary of stepwise multiple regression analysis of TRA predictor variables alone on compliance intention
and compliance behavior* (N = 107)
Compliance intention Compliance behavior
Step Variable &tat Cum R’ Variable &tat Cum R’
I Subjective norms 0.36 0.07 Belief strength 0.27 0.12
2 Normative expectations -0.61 0.23 Compliance intention 0.27 0.19
3 Belief strength 0.24 0.30 Outcome evaluation 0.19 0.23
4 Attitude 0.22 0.34
‘Stepwise inclusion and removal procedure terminated when the probability of F to enter is greater than 0.05 and
probability of F to remove is greater than 0.10
tstandardized regression coefficients.
282 L. DOUGLAS RED and DALE B. CHRISTENSEN
Table 5. Summary of hierarchical regression analysis of social influences on compliance intention and compliance behavior* (N = 107)
Variable
Compliance intention Compliance behavior
Belief/value Social Belief/value Social expectancy influences expectancy mfluences
alone added alone added
Benefits Barriers Susceptibihty Seriousness Attitude Outcome evaluation Belief strength Compliance intention Motivation to comply Normative expectations Subjective norms Total R2
- 0.30
-0.15 0.3% 0.10 0.324 0.26t
-0.05 0.285
-0.03 -0.X+
0.3% 0.55
0.04
0.20 -0.04 -0.04 -0.01
0.2 I 0.26: 0.20t
-
- 0.26
0.03 0.25t
-0.02 0.00 0.03 0.247 0.30t 0.06:
-0.04 -0.22
0.14 0.29
Sig. of R2 change F ,,,PJ) = 17.33; P c 0.001 F (3.94, = 1.15; P bO.05
*Standardized partial regression coefficient. tP < 0.05. $P <O.Ol. §P < 0.001.
regression procedures. The HBM variables were first entered into the equation as a group, followed by the grouped TFL4 variables (Table 3). The TFL4 variables made a significant contribution to the prediction of compliance [29]. The value-expectancy, social influence and intention variables explained an addi- tional 19% of the variance in the compliance vari- able, raising the total explained variance to 29%. The following four variables were found to be approxi- mately equal in importance as statistically significant predictors of compliance behavior in this combined model: barriers, intention, belief strength, and out- come evaluation.
The utility of TRA variables alone in predicting behavioral intention and compliance was next exam- ined using stepwise regression analysis (Table 4). Four variables were significant predictors of behav- ioral intention. First to enter was subjective norms, which explained 7% of the variance. The remaining three variables, in order, were: normative ex- pectations, salient belief strength, and attitude. Com- bined, these four variables explained 34% of the variance in behavioral intention.
A similar stepwise multiple regression procedure was used to examine the ability of the TRA variables, including behavioral intention, to predict compliance behavior (Table 4). Three variables were statistically significant, explaining 23% of the variance. However, these were not the same variables as those most strongly predicting behavioral intention. First to enter was belief strength, followed by behavioral intention and outcome evaluation. Behavioral in- tention and belief strength were nearly equal in importance as predictors of compliance.
Table 7. Parsimonious stepwise regression model of most significant predictor variables on compliance behavior* (N = 107)
Step Variable Betat Cum R’
I Belief strength 0.3 I 0.12 2 Barriers 0.28 0.20 3 Outcome evaluation 0.18 0.23 4 Compliance intention 0.19 0.26
*Stepwise inclusion and removal procedure terminated when the probability of F to enter is greater than 0.05 and probability of F to remove is greater than 0.10.
tstandardized regression coefficient.
The role of social influences on behavioral in- tention and compliance was further investigated us- ing hierarchical regression procedures (Table 5). First examined was the role of social influence variables versus belief/value-expectancy variables alone in predicting behavioral intention. The belief/value- expectancy variables were entered into the equation first as a group. Combined, they explained 30% of the variance in behavioral intention, with three statis- tically significant predictors. Next, the following social influence variables were entered into the equa- tion as a group: normative expectations, subjective norms, and motivation to comply. Together these variables explained an additional 25% of the variance.
It was interesting to note that the addition of the social influence variables to the equation predicting behavioral intention had almost no effect on the standardized regression coefficients of the three statis- tically significant belief/value-expectancy variables. They were, however, correlated with two other vari- ables: perception of benefits and outcome evaluation,
Table 6. Impact of including social influence variables on compliance intentions in predicting compliance*
Variable I
Compliance intention 0.20 Normative expectations - Subiective norms -
step
2 3 4
0.13 0.06 0.06 -0.15 -0.23 -0.22
- 0.13 0.14 Moiivation to comply - - - -0.04
*All of the belief/value-expectancy variables are also included in each equation.
Patients’ drug-taking behavior 283
thereby weakening their contribution to the predic- tive model.
A similar hierarchical regression procedure was used to examine the role of social influence versus belief/value-expectancy variables alone in predicting compliance (Table 5). Different results were observed than for predicting behavioral intention. Belief strength, outcome evaluation, and compliance in- tention emerged as stastistically significant belief/value-expectancy predictor variables. These variables accounted for 26% of the variance. Seriousness and attitude were statistically significant predictors of behavioral intention, but did not emerge as important predictors of behavior, possibly because of the intercorrelation with the compliance intention variable.
The social influence variables were added to the equation in the second step. These variables ex- plained only an additional 3% of the variance. As was the case with compliance intention, there were no significant changes in the magnitude of the belief/value-expectancy standardized regression coefficients when the social influence variables were added, indicating again that these were independent (e.g. uncorrelated) predictors of compliance. Of the three social influence variables added, only normative expectations approached statistical significance.
Of particular interest was the finding that a significant reduction occurred in the standardized partial regression coefficient for compliance intention after the social influence variables were added to the equation (before beta = 0.20; after beta = 0.06). This again suggests substantial intercorrelation among the variables. However, the correlations with compliance intention were not large (highest r = 0.27). To further elucidate this finding, we investigated the impact of adding the social influence variables in a step- wise manner on compliance intention in this model (Table 6). The greatest reduction in the magnitude of the compliance intention standardized regression coefficient occurred when normative expectations was added to the equation and was reduced even further when subjective norms was included.
Finally, all of the variables from each model were combined in a stepwise regression procedure in order to obtain and evaluate a more parsimonious model (Table 7). The first variable to enter into the equation was salient belief strength, which accounted for 10% of the variance. This was followed by perceptions of barriers, outcome evaluation and behavioral in- tention, respectively. Together, these four variables explained 26% of the variance in compliance. Addi- tion of the seven remaining variables to the equation explained only an additional 3% of the variance. The following variables did not significantly improve the prediction of drug-taking compliance: motivation to comply, subjective norms, normative expectations, attitude, benefits of the treatment regimen, patients’ beliefs about their susceptibility or the seriousness of the disease.
DI!XUSSION
The results of this study are consistent with those of several other studies regarding the prevalence of the drug-taking noncompliance problem. Only 38%
of the patients being treated for an uncomplicated UT1 were completely compliant with their medica- tion regimen. An additional 25% completed their drug therapy, but missed one dose or more during the 10 day period. The remaining patients (37%) did not complete their medication regimen as directed.
Anecdotally, a number of patients who completed their therapy, but forgot a dose or two, stated during the telephone interview that they took their medica- tion faithfully until the symptoms subsided. Only then did they forget to take their medication. At least among these noncompliant patients, this would argue for the dynamic decision-making model similar to the one suggested by Christensen [30,31]. He suggests that patients’ perceptions of the importance of com- pliance change over time and that they consciously or unconsciously reassess their decision to comply be- cause of their exposure to new information (e.g. the pain went away). After reassessment, if prospects for recovery from a disorder appear good, further drug- taking may be considered less necessary, unnecessary, expensive, or inconvenient.
Overall, the “readiness to take action” belief di- mensions from the HBM explained only a small portion of the variance in compliance, and only perceptions of barriers had a significant impact on patient behavior. These results can be interpreted to say that patients who find taking medication to be easy and convenient are more likely to be compliant. However, this study did not investigate why the same regimen is ‘easy’ and ‘convenient’ for some patients and not for others.
The findings from this study lend support to critics of the HBM who state that the model is insufficiently articulated, and that important determinants are not included. Becker and Maiman [2] included “intent to comply” in their modification of the HBM, however, this variable is seldom added to empirical in- vestigations. We found that “intention to comply” significantly improved the prediction of compliance behavior, however it was not the sole predictor of compliance behavior as the TRA model specifies. In fact, it was not even the strongest predictor. This is consistent with the conceptual concerns and findings of others. For example, Becker et al. [I l] found that a mother’s estimate of the likelihood that she would be able to keep her child on the prescribed diet was the most influential “general motivation” item, how- ever, it was not a significant predictor of long-run clinic appointment keeping behavior. Others have reported similar findings which indicate that behav- ioral intention is probably not the best predictor of actual behavior [32,33]. Prior behavior has been shown to be among the best predictors of future behavior along with the situational variables which modify patients’ intentions towards further drug- taking over time.
Addition of the TRA variables improved the model’s prediction of compliance. In each regression analysis, ‘belief strength”, or belief in the likelihood of symptomatic relief, was a stronger predictor of compliance than any of the “readiness to take action” variables, including perceptions of the importance, worth, safety or other benefits of compliance. It is interesting to note that none of the social variables entered into the final equation predicting compliance
284 L. DOUGLAS RIED and DALE B. CHRISTENSEN
behavior, even though they were the strongest predic- tors of compliance intention. One interpretation of our findings is that expectations of significant others, namely spouse, family and physician, play a significant role in the development of intentions, however, they do not seem to carry over to actual behavior (at least not for this acute disease). While intentions are shaped by significant persons in their environment, a patient’s own beliefs/value- expectancy and situational circumstances had a greater influence on whether they actually completed their course of medication. On the other hand, for asymptomatic, chronic diseases where discomforting symptoms do not act as internal cues to compliance, social influences may be more important and re- minders from significant others or public health announcements may serve as external cues to compliance.
In development of a more parsimonious model, two of the statistically significant predictors were value-expectancy variables, belief strength and out- come evaluation. Compliant patients believe more strongly that compliance with their medication regi- men is likely to relieve their symptoms and to evalu- ate the symptoms associated with UTI’s more nega- tively. Compliant patients are also more likely to intend to take the medication exactly as directed from the start and are less likely to feel that taking their medication as directed is too difficult or too incon- venient.
The results of this study should be generalized only with caution to other populations or other disease states. Nevertheless, our findings suggest directions for developing strategies that health care profession- als may use to improve compliance with drug ther- apies for acute conditions. Not only is it necessary for patients to feel there are benefits associated with taking the medication exactly as directed, but they must also believe that these benefits are likely to accrue. Health care practitioners need to continually remind patients of the therapy’s value and to point out benchmarks for patients to evaluate the success or failure of the therapy. This requires close super- vision and on-going communication. Maintaining patients’ perceptions of the likelihood of success and the value of the treatment over time is an important key to improving patient’s compliance behavior.
Perceptions of barriers was the second strongest predictor of compliance. Health care professionals must reduce the perception that medication taking is difficult or inconvenient. This may be done by select- ing more acceptable dosage forms, including long acting drugs, and by customizing and simplifying the regimen, particularly by reducing the number of dosage intervals. A change in condition from symp- tomatic to asymptomatic or an increase in time constraints due to resuming regular family or work responsibilities as recovery takes place could also signal the beginning of noncompliance. Barriers to compliance can be significant enough in certain situ- ations to overcome patients’ intentions, however strong their intentions were in the beginning.
Finally, the patient’s evaluation of their likelihood of taking the medication as directed is a significant benchmark in judging whether a patient will be compliant or not in the future. Modifying situations
influence even the best of intentions, but patients who do not intend to take the medication from the beginning are less likely to be compliant. A frank and open discussion about the complexity of the therapy, the medication’s side effects. time and fiscal con- straints, and other factors which may modify pa- tients’ perception of the likelihood of compliance should be discussed from the start.
Acknowledgemenr-This work was supported in part by NIH Biomedical Research Grant RR07079.
REFERENCES
1. Rosenstock I. Why people use health services. Milbank Meml Fund Q. 44(3), 94127, 1966.
2. Becker M. H. and Maiman L. A. Sociobehavioral determinants of compliance with health and medical care recommendations. Med. Care 13, l&24, 1975.
3. Becker M. H. (Ed.) The Health Belief Model and Personal Health Behavior. Slack, Thorofare, N.J., 1974.
4. Becker M. H., Haefner D. P., Kirscht J. P., Maiman L. A. and Rosenstock I. M. Selected psychosocial models and correlates of individual health related be- haviors. Med. Care suppl. 15, 27-46, 1977.
5. Kirscht J. P. and Rosenstock I. M. Patient adherence to antihypertensive medical regimens. J. Commun. Hlrh. 3, 115-124, 1977.
6. Bloom-Cerkoney K. A. and Hart L. K. The relationship between the health belief model and compliance of persons with diabetes mellitis. Diabetes Care 3,596598, 1980.
7. Becker M. H., Radius S. M., Rosenstock I. M. et al., Compliance with a medical regimen for asthma: a test of the health beliefs model. Publ. Hlrh Reo. 93.268-277.
A
1978. 8. Heinzelman F. Factors in prophylaxis behavior in treat-
ing rheumatic fever, an exploratory study. J. Hlth Hum. Behac. 3, 73-78, 1962.
9. Becker M. H., Drachman R. H. and Kirscht J. P. Predicting mother’s compliance with pediatric medical regimens: J. Pediatr. 81,-843, 1972. .
10. Mikhail B. The health belief model: a review and critical evaluation of the model, research, and practice. Adv. Nurs. Sci. 4, 65-82, 198 1,
1 I. Becker M. H., Maiman L. A., Kirscht J. P. ef al. The Health Belief Model and prediction of dietary compli- ance: a field experiment. J. Hlth sot. Behav. 18,348-366, 1977.
12. Langlie J. K. Social networks, health beliefs, and pre- ventwe health behavior. J. Hlth sot. Behav. 18, 244-260, 1977.
13. Rosenstock I. Historical origins of the Health Belief Model. In The Health Belief Model and Personal Health Behavior (Edited by Becker M. H.). Slack, Thorofare, N.J., 1974.
14. Ajzen I. and Fishbein M. Undersfanding Attitudes and Predicting Social Behaoior. Prentice-Hall, Englewood Cliffs, N.J., 1980.
15. Fishbein M. and Ajzen I. Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research. Addison-Wesley, Reading, Mass. 1975.
16. Sejwacz D., Ajzen I. and Fishbein M. Predicting and understanding weight loss: intentions, behaviors and outcomes. In Understanding Attitudes and Predicting Social Behauior (Edited by Ajzen I. and Fishbein M.). Prentice Hall, Englewood Cliffs, N.J., 1980.
17. Ried L. D., Oleen M. A., Martinson 0. B. ef al. Explaining intention to comply with antihypertensive regimens. J. Sot. Admin. Pharm. 3, 42-52, 1985.
18. Farrar W. E. Infections of the urinary tract. Med. C/in. N. Am. 67, 187-201, 1983.
Patients’ drug-taking behavior 285
19. Massoud N. Update on the treatment of bacterial urinary tract infections. Drug Intell. Clin. Pharm. 15, 738-750, 1981.
20.
21.
22.
23.
24.
25.
26.
Stamm W. and Turck M. Urinary tract infections, pyelonephritis and related conditions. In Harrison’s Principles oflnrernal Medicine, 10th edn. McGraw-Hill, New York, 1983. Marston M. Compliance with medical regimens: a review of the literature. Nurs. Res. 19, 312, 1970. Sackett D. The magnitude of compliance and non- compliance. In Compliance with Therapeutic Regimens (Edited by Sackett D. and Haynes B.). Johns Hopkins Press, Baltimore, Md, 1976. Nie N. H., Hull C. H., Jenkins J. G. et al. SPSS: Statistical Package for the Social Sciences, 2nd edn. McGraw-Hill, New York, 1975. Hull C. H. and Nie N. H. SPSS: Update 7-9. McGraw-Hill, New York, 1981. Norusis M. J. SPSS/PC+. SPSS Inc., Chicago, Ill., 1986.
21.
28.
29.
30.
31.
32.
33.
(I) “IS
Levine R. L. and Hunter J. E. Regression methodology: correlation, meta-analysis, confid&e intervals, and un- reliabilitv. J. Leisure Res. 134). 323-343. 1983. Nunnally J. C. Psychometric Theory, 2nd edn. McGraw-Hill, New York, 1978. Duncan 0. D. Contingencies in constructing causal models. In Sociology Methodr--1969 (Edited by Bor- gatta E.). Jossey-Bass, Washington, D.C. 1971. Kerlinger F. N. and Pedhazur E. J. Multiple Regression in Behavioral Research. Holt, Rinehart &Winston, New York, 1973. Christensen D. B. Drug-taking compliance: a review and synthesis. Hlrh Serc. Res. 13, 171-187, 1978. Christensen D. B., Understanding patient drug-taking compliance. J. Sot. Admin. Pharm. 3, 70-77, 1985. Bentler P. M. and Speckart G. Attitudes “cause” behav- iors: a structural equation analysis. J. Person. sot. Psychiat. 40, 226-238, 1981. Liska A. E. A critical examination of the causal struc- ture of the Fishbein/Ajzen attitude-behavior model. Sot. Psychiat. Q. 47, 61-74, 1984.
APPENDIX
Index Measures of TRA and HBM Variables
Barriers
taking the medicine. ”
easy-hard convenient-inconvenient
(2) Perceptions of seriousness of noncompliance (serious)
If you were NOT to take the medicine for your infection exactly as directed for the next 10 days, how serious would the consequences be for you?
PAIN when urinating. DIFFICULTY when uri- nating
BLOOD in the urine.. CHILLS . FEVER. BURNING when urinating
(3) Susceptibility
“My body seems to resist illness well . ” “1 seem to get sick easier than other people do. . ” “When there is something going around I usually catch it . ” “Most people get sick easier than 1 ”
(4) Benefits
Below are pairs of opposite words,
important-unimportant safe-dangerous worthwhile-worthless healthy-unhealthy
(5) Belief strength
If you take the medicine for your infection exactly as directed for the next 10 days, how likely is it that you will have:
PAIN when urinating . DIFFICULTY when uri- nating
BLOOD in the urine.. CHILLS FEVER.. BURNING when urinating
(6) Ourcome evaluation
Think about how ‘good’ or ‘bad’ you feel each of the following things is.
PAIN when urinating. DIFFICULTY when uri- nating
. . BLOOD in the urine. CHILLS . FEVER. BURNING when urinating
(7) Normative beliefs
Generally speaking, which of the following are true about you? cc. . . My - think I should take my medicine for the next 10 days.”
doctors. family. . .
. spouse.
. . . good friends.
(8) Motivation to comply
Generally speaking, which of the following are true about you? “ . . I want to do what my - think I should do. . ”
doctors. . . . family. . . . spouse. . . . good friends. .
(9) Artirude
Below are pairs of opposite words which have been found to describe people’s attitudes about taking their medicine.
good-bad harmful-helpful foolish-wise smart-dumb
(IO) Subjective norms
“Most of the people who are important to me think I should take the medicine for my infection exactly as directed for the next 10 days...”
( I I ) Compliance inlenrion
How likely is it that you will take the medicine for your infection exactly as directed for the next 10 days? “I intend to take the medicine exactly as directed for the next 10 days” “I will probably not be able to take the medicine exactly as directed”
(12) Compliance
“Have you finished all of your medicine?” “Did you ever forget a dose?’ “Did you take your medicine exactly as directed for all IO days?’