Journal of Pediatric Psychology, Vol. 28, No. 1, 2003, pp. 67-79
© 2003 Society of Pediatric Psychology
A Follow-Up Study of Adherence and Glycemic Control Among Hong Kong Youths With Diabetes
1 University of Texas Southwestern Medical Center at Dallas, 2 The University of Hong Kong, 3 Princess Margaret Hospital, 4 Tseung Kwan O Hospital
All correspondence should be sent to Sunita Mahtani Stewart, Psychiatry, UT Southwestern Medical Center at Dallas, 5323 Harry Hines Boulevard, Dallas, Texas 75390-8589. E-mail: Sunita.Stewart{at}UTSouthwestern.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Objective To extend longitudinally an earlier study of the pathway from symptoms of emotional distress (ED) through self-efficacy (SE) and adherence to glycemic control (GC) in youths with diabetes, and to examine the contribution of different specific adherence behaviors to changes in GC. Methods Fifty-six Hong Kong youths with diabetes received a follow-up evaluation 12-24 months after initial participation. ED, SE, self-reported adherence to medical regimen (SRA), and GC were assessed at both evaluations. Results The pathway from ED to SE to SRA to GC was replicated. Participants' SRA to regular checks on blood glucose levels, and taking steps to maintain levels in the recommended range, explained significant variance in changes in GC. Conclusions The model offers strategies to enhance health care in youths with diabetes. Findings support the importance of adherence to the medical regimen but emphasize the complexity of the relationship between adherence behaviors and GC. Self-regulatory behaviors, rather than compliance with fixed instructions, appear to have the most impact on GC.
Key words: type 1 diabetes; glycemic control; Hong Kong; adherence; compliance; adolescents; self-efficacy.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This longitudinal study addresses two issues of interest in diabetes management in the pediatric and adolescent population. The first pertains to the association between symptoms of emotional distress and glycemic control, a measure of disease management in young people with diabetes. The second addresses the relationship between adherence to the medical regimen and glycemic control.
The literature on broad measures of emotional adaptation and glycemic
control in youths with diabetes is somewhat equivocal, though for the most
part, good emotional function has been linked with treatment adherence
(La Greca & Schuman,
1995
). When patients who have significant difficulty in managing
their diabetes are examined, they have been found to have emotional
difficulties. Liss et al.
(1998) found that among
children and adolescents hospitalized for diabetic ketoacidosis, 88% met
criteria for at least one psychiatric disorder. However, when a cross-section
of the population is examined, the findings are mixed. Depressive symptoms
predicted glycemic control in some studies (e.g.,
La Greca, Swales, Klemp, & Madigan,
1995) but not in others (e.g.,
Worrall-Davies, Holland, Berg, &
Goodyer, 1999).
In a review of the literature, Johnson
(1995) has suggested that more
complex models be considered to examine the predictors of effective disease
management. Johnson hypothesized that the influence of emotional status on
glycemic control is through patient adherence to the medical regimen.
Consistent with this hypothesis, children with better emotional adjustment are
more adherent (Anderson, Miller, Auslander,
& Santiago, 1981; Jacobson
et al., 1990; Littlefield et
al., 1992). If emotional adjustment acts through indirect paths,
the conflicting literature may well be explained; the association between
emotional status and glycemic control may not consistently be apparent unless
the underlying variables that carry the effect of emotional function to health
indicators are also examined.
Additional variables may be implicated in the path from symptoms of
emotional distress to glycemic control. Cognitive variables, given importance
in health belief models (Becker,
1974), have been proposed to underlie the relationship between
emotional adjustment and adherence (Ott, Greening, Palardy, Holderby, &
Bell, 2000; Senecal, Nouwen, & White,
2000). Self-efficacy, or a person's level of confidence about his
or her ability to manage the demands of challenges (frequently specified as
the medical regimen), has emerged as a highly salient variable in health care
(Anderson, Funnell, Fitzgerald, &
Marrero, 2000) and in the design of interventions to improve
adherence to medical regimens (e.g.,
Howorka et al., 2000). The
association between self-efficacy and emotional well-being has been
consistently demonstrated (Bandura,
Pastorelli, Barbaranelli, & Caprara, 1999). Self-efficacy also
predicts self-regulation and persistence in goal-oriented behaviors
(Bandura, 1997), offering a
link between emotional distress, on the one hand, and behavioral efforts to
maintain compliance with medical directions, on the other.
Our earlier study (Stewart et al.,
2000) proposed a model (see
Figure 1) seeking to clarify
the pathway of effects that lie between emotional distress and disease control
among youths with diabetes. The data supported a model that linked the
emotional status of young people with diabetes to their self-efficacy, which
then related to self-reported adherence to the medical regimen. Self-reported
adherence related closely to glycemic control. However, a significant
shortcoming of the previous study was the cross-sectional design, which did
not allow investigation of the stability of the central variables and limited
the predictive inferences that could be drawn from the model. The first goal
of this study was to assess the stability of the central measures and of the
model.
|
Our second set of goals was designed to investigate the relationship
between adherence and glycemic control. Adolescents are notoriously less
adherent to diabetic regimens than are children (e.g.,
Jacobson et al., 1990;
Johnson, Freund, Silverstein, &
Hansen, 1990) or adults
(Morris et al., 1997), and
improving adherence has been identified as an important strategy to improve
glycemic control (Morris et al.,
1997). However, studies of the relationship between adherence and
glycemic control have been quite inconsistent in their results (e.g.,
Johnson, 1994). Recent
methodological concerns in the study of adherence in pediatric and adolescent
diabetes note that (a) measuring glycemic control in adolescence is
problematic because of hormonal fluctuation and changes in insulin resistance,
(b) the regimen can be complex and adherence is operationalized differently in
different studies, and (c) although adherence is frequently assessed as a
monolithic construct, in fact, patients may adhere to some aspects of the
diabetes regimen but not to others
(Johnson, 1995;
La Greca & Schuman, 1995).
Prior to analyses using the measure of glycemic control, we report on
consistency of this measure in our sample. The measurement of adherence takes
seven different behaviors into account, and the association among these
behaviors was of interest.
Many of the studies regarding nonadherence have been conducted with insulin
administration defined as the adherence measure (e.g.,
Morris et al., 1997). The
roles of components of the diabetes regimen other than insulin have been
little investigated in study of glycemic control
(Johnson, 1995). Whether
changes in specific aspects of the diabetes regimen correspond to changes in
glycemic control in a cross-section of young people with diabetes is not
presently known. These findings would have significant relevance to health
care practice.
We present longitudinal data obtained 12-24 months (Time 2) following the original study (Time 1) for a subset of our original sample. In summary, the goals of this study were to assess the following:
- The stability of measures that appear to be important predictors of
glycemic control: emotional distress, self-efficacy, and adherence to the
medical regimen;
- Whether the pathway proposed in an earlier study from emotional distress to
self-efficacy to adherence to glycemic control can be replicated in patients 1
to 2 years following the initial study;
- Whether predictors at the first evaluation relate to outcomes along the
path at the second evaluation;
- The interrelationship of different adherence behaviors;
- The relative prediction offered by specific adherence behaviors to variance
in glycemic control; and
- Whether changes in glycemic control parallel changes in specific adherence
behaviors over time.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Participants
This study includes 56 patients with type 1 diabetes mellitus. The sample consisted of 28 boys and 28 girls, ranging in age from 10 to 23 years. These patients were a sub-group of 70 patients (representing 97% of the eligible recruitment pool approached) that participated in an earlier study (Stewart et al., 2000
). The
number is smaller than the original group because 1 of the 4 clinics of
recruitment withdrew from the study. In addition, 3 patients had moved out of
the catchment area of the clinics and could not be contacted. The 14 patients
who were not included at the second survey were compared to the 56 patients of
the study on age, duration of illness, number of daily glucose checks and
injections, number of diabetes-related hospitalization, hypoglycemic episodes
reported in the previous month, glycosylated hemoglobin levels, and the seven
adherence behaviors that form the adherence scale. The overall MANOVA, which
included 13 measures, was significant, F(df) = 1.96, (14,
54), p < .05. Post-hoc tests indicated that the two groups were
different only for duration of illness, F(df) = 8.49, (1),
p < .01; the 14 patients who were not included in this study had
been diagnosed for a shorter period of time (M = 4.04 years,
SD = 3.05 years) than the 56 patients of this study (M =
7.33 years, SD = 4.00 years). However, they were not different on any
of the other variables in the MANOVA. Approval was obtained from the University of Hong Kong Medical Faculty ethics committee. Signed informed consent was obtained from patients and parents. All participants were assured of the confidentiality of the data, and forms were coded for anonymity.
Procedures
Beginning 12 months after the completion of the first data gathering, all
youths who were patients in the participating clinics at the time of the first
study were contacted and verbal consent obtained to continue with the study.
The data gathering took place during a routinely scheduled clinic visit, and a
mutually convenient upcoming visit was identified for this purpose. All youths
completed the forms independently. Hong Kong school children have been trained
from early school years to manage written material, which gave us some
confidence that the forms would be less taxing for the younger participants
than for a typical Western child this age. Typically, the testing was
administered to one or two patients at a time, and the research assistant was
available to answer questions.
Measures
Medical Variables. The following information was obtained
from patients' charts: duration of disease, number of daily glucose checks,
number of daily injections, total number of diabetes-related hospitalizations,
hypoglycemic episodes reported in the last month, and glycosylated hemoglobin
(HbA1c) levels for past 12 months. These variables are presented in
Table I.
|
HbA1c levels are routinely obtained in each clinic. All levels from up to 12 months prior to the participation were obtained for each patient. The number of levels for each patient varied from 2 to 5, with the average number being 3.4. For the 52 patients who had at least three levels at both Time 1 and Time 2, Cronbach's alpha for the six levels was .91. A repeated measures ANOVA for the three scores in the year prior to Time 2 assessments, with time as a within-subjects factor, was nonsignificant, indicating that there were no consistent changes over time in these six levels. These analyses together suggest that despite hormonal fluctuations in the peripubertal age group, there is fair stability of glycemic control over time in this sample. Because the participants were patients in different clinics, to reduce the error resulting from variation in assays from different laboratories, we adjusted levels to percentage above the normal range for each laboratory for group analyses reported. Values were converted for pooling as follows. All HbA1c values obtained in the previous year were averaged for each participant. This mean HbA1c value was then subtracted from the number that indicated the upper end of the normal range for the laboratory (this upper boundary score ranged from 5.9% to 6.5%). The result was divided by the boundary value and multiplied by 100. Positive scores reported thus reflect percentage above the normal range adjusted for each laboratory; negative scores are values within the normal range. Higher levels indicated poorer control of the disease. Scores thus converted were used in all analyses (except in describing the patient population in Table I) and are described as "adjusted HbA1c."
Adolescent-Completed Measures. All measures were
administered in Chinese in counterbalanced order. Two bilingual individuals
translated those forms that did not exist in Chinese translation, using a
forward-backward translation procedure. The scales reported were used in the
previous study (Stewart et al.,
2000) and were reliable in this sample and (where applicable) in
healthy controls. Cronbach alphas are for Time 2 administration of
measures.
Emotional Distress. Emotional distress was assessed by the
General Health Questionnaire (GHQ). The GHQ is a general measure of
psychological symptoms, chosen for our study because a brief form is
available, validated in Hong Kong (e.g.,
Lee, Lam, Ong, Wang, & Kleevens,
1985), and has been used extensively in this population, including
with nonreferred adolescents (e.g., Lam,
Stewart & Ho, in press). The 12 items are simply phrased and
rated on a 4-point Likert scale; examples are "Have you recently been
able to concentrate on what you are doing?" and "Have you recently
been feeling unhappy or depressed?" The internal reliability of the
scale was .83 at this administration, suggesting that the items were being
answered coherently. High scores indicate more distress.
Adherence. Consistent with current concepts of adherence,
this construct was assessed in a multidimensional fashion. Participants rated
their adherence to seven different aspects of the diabetes regimen by giving
themselves grades on each care behavior from 20 to 100, with 20 being failure
and 100 an A+. The scale we used was virtually identical to the one used by
Littlefield et al. (1992) with
adolescents in Canada, with the change from letter grades to numbers up to
100, which are more familiar to Hong Kong students. The behaviors were (a)
taking insulin on schedule, (b) following food plan, (c) testing blood for
glucose regularly, (d) fitting exercise into treatment plan, (e) treating a
reaction, (f) taking steps to keep blood glucose at the right level, and (g)
remembering to do everything every day. Item (f) refers to the course of
action discussed with the patients, should their glucose levels fall outside a
specified range. For example, patients are instructed to increase carbohydrate
intake if blood glucose is low, and increase exercise and fluid intake if it
is high. Patients are also instructed to change their insulin dosage by
amounts detailed in individualized plans to keep their blood sugar within a
specified range.
Cronbach's alpha for these items at this administration was .88. The average was computed across the seven areas, termed the "adherence composite," and used as the measure of adherence in most analyses. However, when the interest was in examining the influence of the components of the treatment regimen on glycemic control, we considered each behavior separately. One of the measures, insulin administration, was significantly skewed (i.e., skew/standard error > 1.96). The log linear transformation of this variable was used in analyses. High scores indicated better adherence.
Self-Efficacy. Patients with diabetes indicated their
confidence about their own ability to follow each of the seven specific
adherence prescriptions (also used on the adherence scale above) using the
same scale from 20 to 100 (Littlefield et
al., 1992). The wording of the questions was "grade yourself
on how well you could do each of these tasks if you could get yourself as
organized as you could be." Cronbach's alpha for these seven items at
this administration was .86. Higher scores indicate more self-efficacy.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
All results that reached the significance level of p < .05 are reported. Table II presents descriptive data on all central measures and individual adherence scale items at both assessments. Table III presents the zero-order correlation matrix for the central measures at Time 1 and Time 2.
|
|
Sex and Age Differences on Central Measures
Prior to addressing the goals of this study, we report two sets of
analyses. These analyses describe the influence of two important demographic
variables, sex and age, on the central measures (emotional distress,
diabetes-specific self-efficacy, self-reported adherence composite and
adjusted HbA1c) of the study. In the previous study, female
participants reported more emotional distress and were less adherent to the
protocol. Although all participants were receiving their medical care at
pediatric endocrinology clinics, the age-span of this study is quite broad,
and information regarding how the variables were influenced by development
would be of interest.
An initial MANOVA with sex as a factor was significant (Hotelling's
T = 3.42, df = 4,50, p = .02). Post-hoc analyses
(see Table IV) indicated that
emotional distress was greater and self-reported adherence lower in girls than
in boys. However, boys and girls did not differ in self-efficacy or metabolic
control. These findings indicate that boys with diabetes fare better than
girls on some variables that relate to disease management and are consistent
with Western (La Greca et al.,
1995) and Time 1 data.
|
Zero-order correlations for the association between age and the central
variables were not significant with one exception: the correlation between age
and the adherence composite was—.36 (p = .007), indicating that
older youths reported less adherence to the medical regimen. This finding is
consistent with the report that children are more adherent than adolescents
(Jacobson et al., 1990;
Johnson et al., 1990).
Stability of Measures of Emotional Distress, Self-Efficacy,
Self-Reported Adherence Composite, and Glycemic Control
Repeated measures MANOVA was used to assess whether the central measures
(emotional distress, diabetes-specific self-efficacy, adherence composite, and
adjusted HbA1c) obtained at Time 1 and Time 2 were significantly
different. To control for the effect of varying periods of time between the
two evaluations, we included duration between the two assessments as a
covariate. The MANOVA for time was not significant (Hotelling's T =
2.83, df = 1,53, p > .05), indicating that the scores
were reliable over time. The central measures at the two assessments are
presented in Table II.
Replication of Model of Pathway From Emotional Distress to Glycemic
Control
The core model obtained at Time 1 was tested with Time 2 measures. Munro
and Page (1993) describe the
template we have followed for path construction and analysis. The path
analysis elucidates direct and indirect effects of predictors on an outcome.
Figure 1 places the proposed
relationships in a sequential pattern of predictors and outcomes and shows the
basis for choosing the variables in each regression analysis. In the
schematic, all direct paths are shown as arrows connecting a variable earlier
in the path (predictor) with a variable that lies later in the path (outcome).
A regression analysis was done for each outcome along the path, for each
variable that has at least one arrow pointing to it in the diagram. The
outcome was regressed on all its possible predictors, that is, all variables
that precede it in the causal path. The data analysis supports a direct path
when there is a significant prediction to the outcome by a predictor
controlling for all other potential predictors to that outcome.
Table V presents the results of the analysis at each step. In the first step, self-efficacy was regressed on emotional distress. The significant regression coefficient for emotional distress's prediction to self-efficacy is consistent with the direct path between these two variables. In the second step, the adherence composite was regressed on both self-efficacy and emotional distress. Now only self-efficacy offered significant prediction to the adherence composite, supporting the single direct path to the adherence composite. In the final step, glycemic control was regressed on the three previous variables. Again, the single significant beta is consistent with the direct path between the adherence composite and glycemic control. The findings are consistent with the core model at Time 1. Emotional distress relates to self-efficacy, which influences the adherence composite, which affects glycemic control.
|
Prediction of the Path From Time 1 to Time 2
We also examined whether the associations found in the core pathway
"held" across variables over time (as presented in
Figure 2). We assessed whether
emotional function at Time 1 predicted self-efficacy at Time 2, whether
self-efficacy at Time 1 predicted the adherence composite at Time 2, and
whether the adherence composite at Time 1 predicted adjusted HbA1c
at Time 2. Time between evaluations was controlled. The partial correlation
coefficients between these measures are presented on the figure. We found no
association between emotional distress at Time 1 and diabetes-specific
self-efficacy at Time 2 (partial r = -.02), nor for Time 1
self-efficacy on the Time 2 adherence composite (partial r = .26).
The adherence composite at Time 1 significantly predicted glycemic control
12-24 months later (partial r = -.42, p < .005). Thus,
the predictive power of Time 1 variables to those next in the path at Time 2
weakened with distance from the endpoint of glycemic control.
|
Interrelationship Among Adherence Behaviors
The correlation matrix for the seven adherence behaviors at Time 2 is
presented in Table VI. All
correlations were significant. The overlap in variance
(r2) between behaviors ranged from 12% for exercise and
treating a reaction to 53% for taking steps to keep blood glucose at the
recommended level and remembering to do everything every day. Thus, although
there is consistency between adherence behaviors, the data from adolescents
with diabetes support the observations from Western samples that a
multifaceted approach to the study of adherence is necessary because these
behaviors do not entirely overlap.
|
Relative Influence of Specific Adherence Behaviors on Glycemic
Control at Time 2
This analysis was conducted to determine whether specific adherence
behaviors were more likely than others to relate concurrently to glycemic
control. Adjusted HbA1c at Time 2 was regressed on the seven
specific adherence behaviors at Time 2. We used the log linear transformation
of self-report of insulin administration. Results are presented in
Table VII. The specific
adherence behaviors accounted for 37% of the variance. Two predictors were
significant in contributing unique variance in the prediction: testing blood
glucose levels and taking steps to keep blood glucose in the recommended
range.
|
Change in Glycemic Control as a Function of Change in Specific
Adherence Behaviors From Time 1 to Time 2
We addressed this question through these analyses: does change in a
specific adherence behavior from Time 1 to Time 2 associate with a change in
glycemic control? We conducted stepwise regression analyses for each specific
adherence behavior, with Time 2-adjusted HbA1c as the dependent
variable and Time 1-adjusted HbA1c, duration between evaluations
and report of the same adherence behavior at Time 1 as controls. These
controls were included to account for change in measures. The seven
analyses are presented in Table
VIII. At the first step, Time 1-adjusted HbA1c was
entered. Thirty-one percent of the variance in Time 2-adjusted
HbA1c could be accounted for by information about adjusted
HbA1c obtained at Time 1. At the second step, duration between
evaluations was entered; this information did not add to the prediction of
Time 2-adjusted HbA1c. These two steps were identical for all seven
equations, but the next step was different for each of the adherence behaviors
that make up the composite. At step 3, the individual adherence behaviors at
Time 1 (to assess for change) and Time 2 were added. We used the log
linear transformation of self-report of insulin administration. Two specific
adherence behaviors changed in tandem with change in glycemic control: the
patient's reports that he or she tested blood glucose regularly and acted to
keep blood glucose levels within the specified range.
|
To assess whether parallel changes occurred in the adherence composite and in metabolic control, we performed the same analysis with the adherence composite. The prediction offered by change in the adherence composite was nonsignificant. This finding supports the value of separating the adherence behaviors.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article presents a follow-up to our earlier study proposing a model of the effects from emotional distress to glycemic control in youths with diabetes. Emotional distress, diabetes-specific self-efficacy, the adherence composite, and glycemic control were found to be relatively stable over time. At Time 2, we replicated the model relating emotional distress to self-efficacy to the adherence composite and then to glycemic control. Longitudinally, the Time 1 adherence composite predicted glycemic control significantly 12-24 months later. Different adherence behaviors were significantly intercorrelated. Nevertheless, taking account of each adherence behavior separately accounted for more variance in metabolic control than did the adherence composite, suggesting the importance of examining them separately. Of the different adherence behaviors, testing blood glucose regularly and taking action to keep blood glucose at the recommended level significantly contributed unique variance to concurrent glycemic control. Changes in glycemic control were associated with self-reported changes in these same two specific components of the diabetes self-care regimen. That changes in specific adherence behaviors predict change in control confirms the value of separating them.
The relationship among emotional distress, self-efficacy, adherence behaviors, and glycemic control remained relatively stable over time. Furthermore, with the exception of self-efficacy, the measures were correlated at the two evaluation times. Earlier steps in the model do not predict later steps longitudinally. Although we had not originally planned this analysis, the weak correlations between self-efficacy at the two testings, combined with the finding that it did not predict adherence longitudinally, prompted us to further investigate the path around self-efficacy and adherence. We examined whether change in the predictor (emotional distress) from Time 1 to Time 2 would predict change in self-efficacy (using the same framework for analyses as those presented in Table VIII). We also investigated whether change in self-efficacy would predict change in its outcome variable along the path (self-reported adherence). We found a significant prediction from change in emotional distress to change in self-efficacy (B = .45, t = 3.88, p = .001), and in change in self-efficacy to change in adherence (B= -.39, t = -2.70, p = .009). Thus, even though self-efficacy changes over time, the changes appear to be orderly and predict changes in the outcome. Self-efficacy may simply be more amenable to change than other model components and so may provide a handle for intervention to influence adherence and thereby improve metabolic control.
We propose that there are many contributors to metabolic control, which is a fairly stable measure. Some of these contributors are themselves quite stable (like adherence) and continue to predict the outcome over time. Others are less stable over 12 to 24 months and so do not continue to predict the steps in the model over time. In a parallel model of weight change, if intake assessed close to the time of weighing relates to weight, but does not 12 to 24 months later, influencing caloric intake on an ongoing basis may still be a strategy to influence weight.
Alternate explanations for the low correlation over time for self-efficacy and the absence of longitudinal prediction challenge our methodology and the model and also should be considered. The absence of correlated results for self-efficacy could indicate measurement weakness in an imported scale, even though the construct is fairly simple, tied to specific behaviors, and would not appear to be culture-specific. Measurement factors may link the variables and so account for the model's cross-sectional persistence over time. Also, the relationship between variables may be misspecified, and alternate models may do a better job of predicting longitudinally. Specifying and testing for these possibilities are a task for future research.
Although among Western youths it has been difficult to demonstrate a
relationship between adherence and disease control
(Johnson, 1994;
Johnson, 1995), given the
limitations of a small and convenience sample, our data support the
relationships between these measures. There have been reports of associations
among adherence behaviors and metabolic control in Western samples;
significant zero-order correlations were reported in a cross-sectional study
of adherence and glycemic control
(Littlefield et al.,
1992).
Certain caveats apply, even within Hong Kong culture. Because of sample size restrictions, demographic measures such as sex and age were not examined as moderators of the relationships proposed. Our sample size was not large enough to determine whether they interacted differently with the variables as well. Future studies should examine whether this model varies according to sex and age of the participants.
Furthermore, this study tested a model that assumes a fixed direction of effects. The analyses did not test models where some of the influences may take paths in a different direction. Poor glycemic control may influence emotional distress, with resulting low self-efficacy and poor adherence to the medical regimen. Indeed, more complicated nonlinear or bidirectional models where variables provide feedback to each other may be in effect. Some of our findings are also consistent with a model whereby past experience of good control influences adherence, and good control may reinforce the behaviors that influence continued control in the future. Future research should address whether data fit prospectively specified alternative models that compete with the schematic provided.
Our data extend the literature by examining changes in glycemic control
that can be attributed to different aspects of the disease management regimen.
Overall, the seven specific items on the adherence self-report predict 39% of
the variance in glycemic control; this effect size is large
(Cohen, 1988). The seven
specific items on the adherence scale are significantly correlated with each
other. This finding is quite similar to that reported by Littlefield et al.
(1992), who also showed high
internal reliability for the same scale with Western adolescents. Thus, in
both cultures, patients who tend to be adherent to one aspect of the regimen
also tend to adhere to other aspects.
We were also interested in examining the relative contributions of the
different aspects of diabetes care to glycemic control. Taking insulin as
directed has been found to be an important indicator of glycemic control in a
study that depended on a more objective measure than self-report because
percentage of insulin prescriptions supplied by pharmacies to the participants
were examined (Morris et al.,
1997), explaining 39% of the variance in adjusted
HbA1c. In this study, taking insulin as prescribed on its own
examined in a post-hoc analysis explained 6% of the variance. Regular testing
of blood glucose and taking steps to keep blood glucose levels within the
recommended range, rather than insulin administration as directed, contribute
significant variance to the prediction of glycemic control. Furthermore,
controlling for previous levels of glycemic control, changes in blood glucose
testing and taking action to keep blood glucose in the recommended range
accounted for a significant amount of variance in changes in glycemic
control. This finding reinforces the issues Johnson
(1995) recently raised with
regard to behavioral management of illness in adolescents. Fluctuations in
control may relate to aspects of the medical regimen other than insulin
administration and are more complex than simple adherence to a set of medical
prescriptions; furthermore, the variance accounted for by such changes is
still quite small.
There may be several explanations for differences between our findings and
those of Morris et al. (1997).
Self-report was obtained at one point in time in this study versus accumulated
information regarding dispensed insulin over at least a year by Morris and his
group. Consistent with previously reported findings
(Glasgow, McCaul, & Shafer,
1987; Johnson,
1991), the patients of this study seem to adhere to insulin
prescription (see the high mean for insulin administration in
Table I). In addition, the
measures of adherence to the medical regimen obtained in this study were
patient-based rather than "objective." However, the relatively
crude measure of the adherence composite accounts for as much variability in
glycemic control at two different points in time as the highly specific and
"objective" measures taken over a longer period of time.
Nevertheless, these findings may be specific to the context. First, the assurance of confidentiality of a research protocol may have allowed patients to be more forthcoming about their adherence behaviors than if this information was obtained in the clinical situation. Second, in a culture that emphasizes compliance with authority, patients may show little variation in their adherence to concrete aspects of the medical regimen such as insulin administration. As a result, this absence of variability would be statistically manifested in weaker associations between behaviors and outcomes. Finally, there may actually be overreporting of adherence, particularly because in this culture, obedience to authority is highly valued. Our study emphasizes the importance of more cross-cultural research to clarify these issues.
The model tested has clinical implications and may be useful as a basis for
designing interventions. Although it has been shown
(Geffken et al., 1997) that
inpatient multimodal treatment in patients with a history of recurrent
hospitalizations resulted in reduced need for hospitalization in the year
following treatment, the specific changes that resulted from this costly
treatment, and translated into better self-care, were not documented. Such
information may be helpful for the design of programs to promote diabetes care
and has been lacking in the past, even when investigators have highlighted the
central importance of behavioral variables in disease management in young
people (e.g., Morris et al.,
1997). If the association between self-efficacy and adherence and,
indirectly, glycemic control can be shown to be free of methodological
confounds, the link presents an important potential handle for promoting
health-enhancing behaviors. The literature on cognitive interventions to
decrease dysphoric mood (an important component of the "emotional
distress" step in our model) generally and self-efficacy specifically is
well developed, and many of the interventions are readily adaptable into
routine medical care of adolescents. Furthermore, given the fact that type 1
diabetes is frequently diagnosed when the child is still too young to take on
self-care, it is in the child's medical interests for health care
professionals to encourage increasing supervised involvement in self-care,
thereby enhancing the likelihood of eventual self-efficacy in medical
management.
The importance of an active role in medical care is suggested by the
finding that the most passive parts of the medical regimen do not appear to
relate to changes in glycemic control, but those that require seeking feedback
about one's disease and making necessary adjustments do. The most salient item
that emerged in prediction to glycemic control was the extent to which the
patient reported that he or she took measures to keep blood glucose within the
recommended range. The patient who tests and acts to keep blood glucose at
recommended levels is also likely to follow the more prescribed aspects of
care. However, acting in response to blood glucose levels involves a
higher-level function in that it requires interactions with information and,
therefore, better education and application of principles to self-care. High
self-efficacy beliefs and a relationship that emphasizes shared decision
making with health care professionals would seem to provide optimal conditions
for such collaboration. This finding is particularly interesting for Hong
Kong, where the culture emphasizes obedience and respect for powerful figures,
and the typical doctor-patient relationship is quite authoritarian. The
finding supports the evidence that, in this culture as well, a sense of agency
and efficacy may be important in adjustment to health challenges
(Sun & Stewart, 2000).
Efficacy can be subtly enhanced in multiple ways during the health care professional's interaction with families. Making a point of addressing the child as well as the parent when providing explanations, asking the child's opinion in making regimen changes, problem solving with the child in a nonjudgmental fashion about the most difficult components of the regimen, monitoring the degree of self-care with the explicitly stated goal of decreasing parental involvement to the point of eventual self-management, all would be strategies that support a graded move toward eventual self-care, increasing self-efficacy, and, thus, enhancing disease management. Furthermore, translating good glycemic control into explicit statements supporting efficacy in diabetes care, providing internal attributions to success to following the regimen and external, unstable causes to "slips" in self-care, would increase efficacy and decrease maladaptive cognitions that accompany dysphoric mood.
This study has several limitations in addition to the ones already
described. The findings are based on a small sample in a single culture.
Although attempts were made to offer help with the questionnaires, the survey
methodology may result in miscommunications. Furthermore, assessments were
made largely by self-report. This may be specially limiting in the case of
reliance on self-reported adherence, because of tendencies to overreport on
such scales. This limitation is somewhat balanced by the findings of moderate
to large effect sizes, some fairly consistent results, and prediction of
variance in the outcome that is equivalent to "objective"
indicators of adherence, such as those in the Morris et al.
(1997) study.
However, future studies should include a broader range and larger groups of patients across more cultures and should use interview as well as survey methodologies and informant report as well as self-report. Unraveling the nonlinear and bidirectional complexities of the relationships among emotional function, efficacy beliefs, adherence, and disease status, and translating findings into well-designed, practical intervention programs, remains a challenge for the future.
|
Acknowledgments |
|---|
This research was supported in part by a grant from the Committee on Research and Conference Grants of The University of Hong Kong.
|
Notes |
|---|
This paper was reviewed and accepted during the term of the previous editor, Anne E. Kazak, PhD, ABPP.
Received August 2, 2001; revision received December 1, 2001; accepted May 13, 2002
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Anderson, R. M., Funnell, M. M., Fitzgerald, J. T., & Marrero, D. G. (2000). The Diabetes-Empowerment scale: A measure of psychosocial self-efficacy. Diabetes Care, 23, 739-743.
Anderson, B., Miller, J., Auslander, W., & Santiago, J. (1981). Family characteristics of diabetic adolescents: Relationship to metabolic control. Diabetes Care, 4, 586-594.[Abstract]
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Bandura, A., Pastorelli, C., Barbaranelli, C., & Caprara, G. V. (1999). Self-efficacy pathways to childhood depression. Journal of Personality and Social Psychology, 76, 258-269.[CrossRef][Web of Science][Medline]
Becker, M. H. (1974). The Health Belief Model and personal health behavior. Thorofare, NJ: Charles B. Slack.
Cohen, J. (1988). Statistical power for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.
Geffken, G. R., Lewis, C., Johnson, S. B., Silverstein, J. H., Rosenbloom, A. L., & Monaco, L. (1997). Residential treatment for youngsters with difficult-to-manage insulin dependent diabetes mellitus. Journal of Pediatric Endocrinology & Metabolism, 10, 517-527.[Web of Science][Medline]
Glasgow, R., McCaul, K., & Shafer, L. (1987). Self-care behaviors and glycemic control in Type I diabetes. Journal of Chronic Disease, 40, 399-412.[CrossRef][Web of Science][Medline]
Howorka, K., Pumprla, J., Wagner-Nosiska, D., Grillmayr, H., Schlusche, C., & Schabmann, A. (2000). Empowering diabetes outpatients with structured education: Short-term and long-term effects of functional insulin treatment on perceived control over diabetes. Journal of Psychosomatic Research, 48, 37-44.[CrossRef][Web of Science][Medline]
Jacobson, A., Hauser, S., Lavori, P., Wolfsdorf, J., Herskovitz,
R., Milley, J., Bliss, R., Gelfand, E., Wertlieb, D., & Stein, J.
(1990). Adherence among children and adolescents with insulin
dependent diabetes mellitus over a four year longitudinal follow up: I. The
influence of patient coping and adjustment. Journal of Pediatric
Psychology, 15,
511-526.
Johnson, S. B. (1991). Compliance with complex medical regimens: Assessing daily management of childhood diabetes. In R. J. Prinz (Ed.), Advances in behavioral assessment of children and families (Vol. 5, pp. 113-137). London: Jessica Kingsley.
Johnson, S. (1994). Health behavior and health status:
Concepts, methods and applications. Journal of Pediatric
Psychology, 19,
129-141.
Johnson, S. B. (1995). Insulin-dependent diabetes mellitus in childhood. In M. R. Roberts (Ed.), Handbook of pediatric psychology (2nd ed., pp. 263-285). New York: Guildford Press.
Johnson, S. B., Freund, A., Silverstein, J., & Hansen, C. A. (1990). Adherence and health status relationships in childhood diabetes. Health Psychology, 9, 606-631.[CrossRef][Web of Science][Medline]
La Greca, A. M., & Schuman, W. (1995). Adherence to prescribed medical regimens. In M. R. Roberts (Ed.), Handbook of pediatric psychology (2nd ed., pp. 55-83). New York: Guildford Press.
La Greca, A., Swales, T., Klemp, S., & Madigan, S. (1995). Adolescents with diabetes: Gender differences in psychosocial functioning and glycemic control. Children's Health Care, 24, 61-78.
Lam, T. H., Stewart, S. M., & Ho, L. M. (in press). Prevalence and correlates of smoking and sexual activity among Hong Kong adolescents. Journal of Adolescent Health.
Lee, P. W. H., Lam, T. H., Ong, S. G., Wang, C. M., & Kleevens, J. W. L. (1985). A modified version of the General Health Questionnaire as a measure of occupational mental health in Hong Kong. Chinese Journal of Psychology, 26, 101-106.
Liss, D. S., Waller, D. A. W., Kennard, B. D., McIntire, D., Capra, P., & Stephens, J. (1998). Psychiatric illness and family support in children and adolescents with diabetic ketoacidosis: A controlled study. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 536-544.[CrossRef][Web of Science][Medline]
Littlefield, C. H., Craven, J. L., Rodin, G. M., Daneman, D., Murray, M. A., & Rydall, A. C. (1992). Relationship of self-efficacy and bingeing to adherence to diabetes regimen among adolescents. Diabetes Care, 15, 90-94.[Abstract]
Morris, A. D., Boyle, D., McMahon, A. D., Greene, S. A., MacDonald, T. M., & Newton, R. W. (1997). Adherence to insulin treatment, glycemic control, and ketoacidosis in insulin-dependent diabetes mellitus. Lancet, 350, 1505-1510.[CrossRef][Web of Science][Medline]
Munro, B. H., & Page, E. B. (1993). Statistical methods for health care research (2nd ed.). Philadelphia: J. B. Lippincott.
Ott, J., Greening, L., Palardy, N., Holderby, A., & De-Bell, W. K. (2000). Self-efficacy as a mediator variable for adolescents' adherence to treatment for insulin-dependent diabetes mellitus. Children's Health Care, 29, 47-63.
Senecal, C., Nouwen, A., & White, D. (2000). Motivation and dietary self-care in adults with diabetes: Are self-efficacy and autonomous self-regulation complementary or competing constructs? Health Psychology, 19, 452-457.[CrossRef][Web of Science][Medline]
Stewart, S. M., Lee, P. W. H., Low, L. C. K., Cheng, A., Yeung, W.,
Huen, K. F., & O'Donnell, D. (2000). Pathways from emotional
adjustment to glycemic control in youths with diabetes in Hong Kong.
Journal of Pediatric Psychology,
25, 393-402.
Sun, L. N. N., & Stewart, S. M. (2000) Psychological adjustment to cancer in a collective culture. International Journal of Psychology, 35, 177-185.[CrossRef]
Worrall-Davies, A., Holland, P., Berg, I., & Goodyer, I. (1999). The effect of adverse life events on glycemic control in children with insulin dependent diabetes mellitus. European Child and Adolescent Psychiatry, 8, 11-16.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  L. Greening, L. Stoppelbein, C. Konishi, S. S. Jordan, and G. Moll Child Routines and Youths' Adherence to Treatment for Type 1 Diabetes J. Pediatr. Psychol., May 1, 2007; 32(4): 437 - 447. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. Stewart, U. Rao, G. J. Emslie, D. Klein, and P. C. White Depressive Symptoms Predict Hospitalization for Adolescents With Type 1 Diabetes Mellitus Pediatrics, May 1, 2005; 115(5): 1315 - 1319. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||