The relationship between adherence behaviors and glycemic control in childhood diabetes

MISSING IMAGE

Material Information

Title:
The relationship between adherence behaviors and glycemic control in childhood diabetes
Physical Description:
viii, 116 leaves : ill. ; 29 cm.
Language:
English
Creator:
Spevack, Marika, 1944-
Publication Date:

Subjects

Subjects / Keywords:
Health Behavior   ( mesh )
Diabetes Mellitus, Type I   ( mesh )
Blood Glucose Self-Monitoring   ( mesh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1987.
Bibliography:
Includes bibliographical references (leaves 110-114).
Statement of Responsibility:
by Marika Spevack.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 20379913
ocm20379913
System ID:
AA00012894:00001


This item is only available as the following downloads:


Full Text











THE RELATIONSHIP BETWEEN ADHERENCE BEHAVIORS
AND GLYCEMIC CONTROL IN CHILDHOOD DIABETES














BY

MARIKA SPEVACK


DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR
THE DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA

1987









ACKNOWLEDGEMENTS


I wish to take this opportunity to express special and

heart-felt thanks to Dr. Suzanne Johnson who has provided me

with more than the expected support of a Chair throughout

this investigation. Not only has her interest and input

energized this project and enhanced the quality of the

outcome, but she has been an inspiration, a role model and a

friend throughout this endeavor.

My appreciation and thanks go out to my committee,

Nathan Perry, Sheila Eyberg, William Riley, and Jon Shuster

who have been supportive throughout this endeavor.

Most of all, I would like to take this opportunity to

thank my best friend, confidant, and husband Avram who

tolerated bad moods, angst, ambition, and enthusiasm. I

would like to express deep respect, overwhelming love, and

gratitude for his unwavering encouragement and faith in my

ability to accomplish this project.















TABLE OF CONTENTS
Page

ACKNOWLEDGEMENTS......................................... ii

LIST OF TABLES ........ ................................. iv

ABSTRACT................................................ vi

INTRODUCTION............................................. 1

METHOD................................................. 31

Sample Characteristics............................ 31
Measures.......................................... 31
Adherence Measures.............................. 31
Glycemic Control Measures....................... 39
Procedure......................................... 40

RESULTS................................................ 42

Reliability of Child Report....................... 42
Effect of Diabetes Camp on Adherence............. 60
Effects of Diabetes Camp on Glycemic Control..... 66
Relationship Between Adherence and Diabetic
Control......................................... 72
Adherence/GSP Relationship........................ 73
Hierarchical regression analyses................ 73
Categorical analyses............................ 78
Adherence/HbAlc Relationships..................... 81
Hierarchical regression analyses................ 81
Categorical analyses............................ 90

DISCUSSION.............................................. 95

Reliability of Child Report....................... 95
Effect of Diabetes Camp on Adherence.............. 98
Effect of Diabetes Camp on Glycemic Control...... 102
Relationship Between Adherence and Diabetic
Control......................................... 104
Limitations and Future Directions................ 107

REFERENCES ...... ...................................... 110

APPENDIX .............................................. 115

BIOGRAPHICAL SKETCH ................................... 116


iii














LIST OF TABLES

Table Page

1 Parent-Child Correlations for Total Sample
and by Age Group (Johnson, et al.)........... 23

2 Sample Characteristics........................ 32

3 Parent-Child Correlations for Total
Sample and by Age Group: Pre-camp............ 43

4 Parent-Child Correlations for Total
Sample and by Age Group: Immediately
Post-Camp ...................................... 45

5 Parent-Child Correlations for Total
Sample and by Age Group: Six
Weeks Post-Camp....................... ........ 47

6 Parent-Child Correlations for Total
Sample and by Age Group: Three
Months Post-Camp.............................. 49

7 Parent-Child Correlations for Total Sample
and by Age Group Collapsed Across Time....... 53

8 Parent-Child Correlations for Total Sample
Over Four Time Periods: Pre and Post-Camp... 56

9 Duncan's Multiple Range Tests on the
Adherence Measures Demonstrating Within
Subject Effects............................... 61

10 Means and Standard Deviations for
Adherence Measures at the 5 Time Periods..... 64

11 Means and Standard Deviations for Adherence
Measures with Age x Time Interaction at the
5 Time Periods ............................... 65

12 Means and Standard Deviations for Adherence
Measures that Remained Stable Over the
5 Time Periods............................... 67

13 Correlations Between Glycosylated
Hemoglobins and Serum Proteins................ 69








14 Correlations Between Adherence Measures
Within and Between Factors.................... 75

15 Average Correlations between Adherence
Factors....................................... 77

16 Predicting Post-camp HbAlc by Pre-camp HbAlc,
Age, Post-camp Injection Factor, and Pre-camp
Hbalc x Post-camp Injection Interaction...... 84

17 Predicting Post-camp HbAlc: Nonstandardized
Injection Beta Weights at Varying Levels of
Pre-camp HbAlc................................ 85

18 Pre-camp HbAlc x Post-camp Injection Factor:
Descriptive Characteristics................... 86

19 Predicting Post-camp HbAlc by Pre-camp HbAlc,
Age, Insulin Dose Change, Post-camp Injection
Factor, and Insulin Dose Change x Post-camp
Injection Interaction......................... 88

20 Predicting Post-camp Hbalc: Nonstandardized
Beta Weights at Varying Levels of Change in
Insulin Dose from Pre-camp to 3 Months
Post-camp..................................... 89

21 Pre- Post-camp Change in Insulin Dose x
Post-camp Injection Factor: Descriptive
Characteristics.............................. 91














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


THE RELATIONSHIP BETWEEN ADHERENCE BEHAVIORS
AND GLYCEMIC CONTROL IN CHILDHOOD DIABETES

BY

MARIKA SPEVACK

December 1987

Chairman: Suzanne Bennett Johnson, Ph.D.
Major Department: Clinical and Health Psychology


This study was designed to investigate the following

questions: 1) do children's diabetes management behaviors

change at summer camp; 2) does attendance at summer camp

affect diabetic children's glycemic control; 3) if there are

changes at camp, are they maintained at home; 4) is there a

relationship between adherence behavior and glycemic

control. Sixty-four 7- to 12.6-year old youngsters with

diabetes were followed before, during, and after attending a

2-week summer camp. Both their diabetes management behav-

iors and their glycemic control were monitored.

Children's diabetes management behaviors (i.e., injec-

tion regularity, injection interval, calories consumed,

eating frequency, exercise duration, exercise type, exercise

frequency, and glucose testing frequency) changed signifi-

cantly while they attended summer camp. There was also a








change in children's level of glycemic control as measured

by glycosylated serum protein levels. However, children

were in poorer glycemic control at the end of camp than at

the beginning. None of these camp-related changes were

maintained once the youngsters returned home. No clear

relationship between adherence behaviors during camp and

control could be established. However, post-camp metabolic

control was predicted by an interaction between change in

insulin dose (pre- to post-camp) and injection adherence

behaviors. Youngsters who had large increases in insulin

dose exhibited a deterioration in metabolic control when

they were more adherent. Youngsters who had large decreases

in insulin dose exhibited the opposite relationship: better

adherence was associated with better control. Further

inspection of this statistically significant interaction

indicated that pre- post-camp change in insulin dose

appeared to be initiated during camp. Camp related insulin

dose changes were maintained or enhanced once the children

went home. The group of children whose insulin was

increased during camp, were given further increases in

insulin dose once they returned home and may have been

experiencing the Somogyi phenomenon (rebounding) after camp

as a consequence of excessive increments in their insulin

dose. Home physicians, appeared to have used insulin

adjustments made during camp as a model of treatment and

continued the trend initiated by the camp physicians. It


vii







appears that high levels of adherence with inappropriate

insulin dose prescriptions can lead to poorer control.


viii














INTRODUCTION


Insulin dependent diabetes mellitus (IDDM) is a chronic

illness, the result of insufficient insulin production by

the pancreas. Onset occurs before age 30--most frequently

between the ages of 8 and 12 years--although it can occur at

any time from birth to adulthood (Blevins, 1979).

Under normal circumstances, insulin is secreted

directly into the blood-stream by the Islets of Langerhans

in the pancreas. This hormone plays a vital role in the

metabolism of food since it facilitates the movement of

glucose from the bloodstream into muscle and fat tissue for

energy. Glucose that is not used for energy is then

directed into storage (primarily into the liver) in the form

of glycogen and/or fat. Fat is an additional source of

energy which does not require insulin to be converted.

However, the energy expended in its production is

considerable, leaving a negligible amount for functional

use. Normally, the pancreas secretes insulin at rates that

maintain glucose levels in the bloodstream within a narrow

range. In youngsters with IDDM, the pancreas cannot perform

this function and glucose levels build up.

The absence of insulin prevents glucose usage, fat is

broken down and glucose levels remain high. The liver






2
converts some of the fat into ketones, which increase in the

blood and, at high levels, spill into the urine. In

addition, when the blood passes through the kidney, glucose

is filtered out and, at high levels, the extra glucose

spills over into the urine. The kidney attempts to filter

out the excess glucose and ketones from the blood in order

to maintain acid alkaline (Ph) levels within normal limits.

This results in increased urination and increased thirst.

However, the kidney can not keep up this level of activity.

Unchecked, ketones and acids build up in the blood and the

child develops diabetic ketoacidosis--a serious condition

requiring hospitalization. Furthermore, the loss of urine

water and the use of body fat for energy produce weight loss

(Travis, 1969). If this condition is untreated, coma and

death can result.

Presenting symptoms of IDDM include polyuria, poly-

dipsia, polyphagia, and weight loss. There is often a high

prevalence of infections. More subtle signs of the disorder

include pruritus (severe and protracted itching of the

skin), enuresis bedwettingg), vulvitis, irascible disposi-

tion, lassitude, and abrupt onset of visual difficulties

(Sussman, 1971). As yet, there is no cure for this disease.



The treatment of IDDM consists of the following:

1) Daily insulin injections--A child with IDDM

injects insulin at least once and as many as

four times daily for the rest of his/her life







because the pancreas does not resume insulin

production.

2) Controlled diet--Food intake must be adjusted

for content and synchronized with the

injections and exercise for optimal benefit.

The youngster must eat three meals a day as

well as one to three snacks per day,

preferably at regular intervals. Low fat

nutrition is encouraged and concentrated

sweets are to be avoided.

3) Regular exercise--An exercise regimen is

important for improving cardiovascular

functioning which facilitates insulin

circulation.

4) Glucose testing--Close monitoring of

glycemic control (approximately four

times per day preferably before meals)

is recommended so that meals, snacks,

and exercise can be appropriately ad-

justed. In addition, this information

is necessary for the physician to make

appropriate changes in regimen recom-

mendations.

The various components of the regimen are tailored to

the needs of each individual youngster. The goal of

treatment is to maintain blood glucose levels within a

reasonable range in order to promote normal growth and







development and to prevent complications. With growing

youngsters, this task is particularly difficult since their

nutritional and insulin requirements are changing as they

grow and injections administered once or twice a day cannot

imitate the synchronous pancreatic action of nondiabetic

individuals (Guthrie and Guthrie, 1977).

Complications associated with IDDM can be divided into

three categories: short-term, intermediate, and long-term.

1) Short-term (acute) consequences include

diabetic ketoacidosis (DKA), hypoglycemia and

hyperglycemia. Diabetic ketoacidosis can be

precipitated by an infection although stress

or insufficient insulin can also be a

precipitant. Symptoms include rapid respira-

tion, polydipsia, polyuria, vomiting and

nausea. Hypoglycemia (insulin reaction) can

be caused by overdoses of insulin, its

improper distribution, inadequate intake (or

delayed absorption) of food, or too much

exercise without a concurrent adjustment of

the food intake (Guthrie and Guthrie, 1977).

Increased perspiration, confusion, irri-

tability and inappropriate behavior are

common symptoms (Sussman, 1971). Hyper-

glycemia may result from too little insulin,

eating improperly, decreased exercise

(without concomitant decreased food intake),








emotional stress, and infection. The most

frequent symptoms are weakness, increased

thirst, frequent urination, dry mouth,

decreased appetite or polyphagia, nausea and

vomiting, abdominal pain, coma, and acetone

breath which occurs when ketone levels rise

in advanced stages. These acute consequences

are characteristic of youngsters in poor

control (Guthrie and Guthrie, 1977).

2) Intermediate consequences include delayed

growth and development, frequent infections,

and psychosocial problems. Delayed growth

and development in IDDM youngsters is not

uncommon and has been found to correlate most

clearly with those children in poor control

(Guthrie and Guthrie, 1977). Various kinds of

infections are also common in IDDM children.

Clinical observations and research suggest

that these are also more common in patients

in poor diabetes control (Guthrie and

Guthrie, 1977). Psychological problems are

associated with poor control although the

nature of the relationship between psycho-

logical problems and health status in

youngsters with diabetes is not entirely

clear (Johnson, 1985).







3) Long-term (chronic) consequences of the

disease include retinopathy and nephropathy.

Retinopathy is characterized by the formation

of new capillaries and duplication of small

veins, retinal detachment, preretinal and

vitreous hemorrhage, and fibrosis. Blindness

may be the result (Sussman, 1971). Nephropa-

thy (renal disease) is the leading cause of

death among people with IDDM (Blevins, 1979).

Complications occur 15-20 years after

diabetes onset. Scientific data suggest that

these diseases may be preventable if the

patient's diabetes is maintained in good

control (Guthrie and Guthrie, 1977).



Clearly, metabolic control, or lack thereof, is

implicated in all three categories of complications both

directly and indirectly. Although evidence that complica-

tions occur more frequently among patients in poor control

is mounting (Rifkin, 1978; Maxwell, Luft, Clark, and

Vinicor, 1982; Unger,1982; Johnson, 1984), it is difficult

to properly evaluate this literature since the definition of

"control" is so loosely posited and varies from investigator

to investigator. Moreover, a definitive link between

adherence to prescribed regimen (i.e. behavior) and control

has yet to be established.








Studies in this area are few and present numerous

methodological problems. For example, control and com-

pliance are often confused as when hemoglobin Alc (HbAlc), a

metabolic indicator of glycemic control, is taken to

represent compliance. In some cases, although both con-

structs are treated separately, the same person judges both,

introducing the possibility of a spurious association

between the two (i.e., the rater may believe compliance and

control are related and judge them accordingly).

The few studies in the area can be divided into two

categories. The first category is composed of correlational

studies which are concerned with establishing associations

between compliance behaviors and control. One such study by

Rainwater, Jackson and Burns (1982) tried to assess the

relationship between psychological variables, metabolic

control, and behavioral measures using 115 children between

8 and 17 years and their mothers. Their most noteworthy

finding was that Mother's Health Locus of Control (MHLC) was

significantly correlated with behavioral and metabolic

measures suggesting that mothers who scored external tended

to have better controlled children. That is, MHLC corre-

lated significantly r=-.58, p< .01, with total amount of

sugar spilled in a 24-hour urine sample collected from the

patient, the number of patient hospitalizations in the past

year, -.42, (p<.01), the number of insulin reactions

experienced by the patient in the past month r=.36 (p<.01),

and the number of hours the patient exercised per week






8
during the past 3 months r=.35 (p<.01). However, HbAlc did

not correlate significantly with MHLC or Mother's General

Locus of Control. Further, there appeared to be no sig-

nificant relationship between the author's diabetes control

indices (such as 24-hour urine, number of hospitalizations,

number of reactions, hours of exercise and HbAlc) and the

child's self-esteem (measured by the Piers-Harris Inventory)

or with the child's life change scores (measured with the

Coddington Scale for Life Events).

This study claimed to have used both behavioral

(defined by the authors as the number of hours exercised per

week, number of diabetes related hospitalizations in the

past year, and number of insulin reactions in the past

month, and number of times acetone was noted in the urine in

the past month) and metabolic (hemoglobin Alc, triglyce-

rides, cholesterol, fasting blood sugar, and total amount of

sugar spilled in the urine in a 24-hour period) measures

which the authors presumed to reflect diabetes control.

Glycosylated hemoglobin (HbAlc) is now widely accepted as

the single most reliable and adequate measure of glycemic

control. The authors' use of 24-hour urine as an index of

control is problematic for two reasons. Valid 24-hour

urines are difficult to collect even under controlled

conditions such as hospitals. In our experience, only

20-25% of all specimens solicited and collected were

complete collections. Glycosylated hemoglobin (HbAlc)

values are a more widely accepted measure of control than








24-hour urines and provide a more stable measure of control

over the preceding 3 months rather than just 24 hours.

Rainwater et al. did not report whether the 24-hour speci-

mens in their study were valid. Even if the 24-hour

collection was valid, it indirectly represents the young-

ster's metabolic control for a relatively brief period of

time. Since no relationship was shown between MHLC and

HbAlc, the most reliable indicator of metabolic control over

the past 3 months, the relationship between MHLC and the

24-hour collection data is difficult to interpret (i.e., why

would such a relationship exist for one measure of diabetes

control but not another?). Furthermore, the study did not

directly assess adherence behaviors. Rather, behavioral

measures, as defined by the authors, consisted of report by

family of the number of diabetes related hospitalizations in

the past year, average number of hours of exercise per week

over the past 3 months, number of insulin reactions in the

past month, and number of times acetone was noted in the

urine in the past month, thereby confusing adherence

behaviors with metabolic indicators and other sequelae of

poor control.

In another correlational study, LaGreca et al. (1982)

examined the relationship between metabolic control as

measured by hemoglobin Al and different aspects of diabetes

care such as degree of responsibility taken by the child,

level of knowledge, and degree of compliance as rated by

physicians following routine appointments with 40 IDDM







youngsters. Results of this study indicate that for

preadolescents there was no relationship between physicians

compliance ratings and metabolic control. For adolescents

the best predictor of control was the child's overall level

of compliance with the treatment regimen, especially with

eating regular snacks and carrying sugar for emergencies.

However, their findings suggest that regardless of com-

pliance ratings, the more responsibility assumed by the

youngster, the poorer their control.

It should be noted that the physician's four-point

rating of compliance was based on patient report (testing,

charting, administering insulin, eating proper foods, eating

regular snacks, carrying quick-acting sugar, and taking

sugar to treat reactions) as well as on whether the child

kept appointments. The reliability of these ratings was not

tested and it is unclear whether the physician's rating may

have been influenced by the patient's current level of

metabolic control. Furthermore, no information was provided

about the relative importance of any one aspect of the

regimen.

McCulloch et al. (1983) correlated dietary behaviors

such as accuracy of measuring carbohydrates and maintaining

consistent and regular eating patterns with HbA1 in adult

diabetics who kept 7-day diaries. Results indicated better

glycemic control in those persons who were more precise in

their measurements and compliant in the regularity of their

meals. There was, however, no attempt made to corroborate








the patients' self-reports, and nondietary aspects of the

treatment regimen were not studied.

One study that has attempted to address the more global

versus specific nature of the relationship between adherence

and metabolic control was conducted by Schafer, Glasgow,

McCaul, and Dreher (1983). Their 34 subjects were 12 to

14-year-olds with IDDM who completed questionnaires dealing

with regimen adherence for the past week as well as psycho-

social measures dealing with diabetes-specific family

behaviors, barriers to adherence, and family interactions.

They found that compliance was a highly specific construct;

the degree of adherence to one aspect of the IDDM regimen

was not related to adherence to the other aspects of the

regimen. Using multiple correlations, they also found that

glycosylated hemoglobin levels could be predicted from a

combination of three adherence measures (i.e., the extent to

which one's diet was followed, reported care measuring

insulin doses, and number of daily glucose tests). Unfor-

tunately, the reliability of the self-report compliance

measures was not assessed, nor was there any attempt to

verify their reports by actual measurement of home behaviors

or by corroboration by significant others. In addition, the

use of multiple regression with such a small sample and

such a large number of variables is likely to severely

overestimate the strength of the resultant R2.

In a more thorough study by some of the same inves-

tigators (Glasgow, McCaul, and Schafer, 1987), these same







issues were addressed with 93 adult IDDM outpatients.

Similar to the Schafer et al.(1983) study, this investiga-

tion explored adherence to various aspects of the diabetes

care regimen, the congruity of adherence of the different

behavioral components, and the relationship between adher-

ence and glycemic control as measured by glycosylated

hemoglobin. Subjects participated in three home interviews

during a 1 week period, completed psychosocial measures,

collected self-care measures on injections, glucose testing,

diet and physical activity. The entire procedure was

repeated for a subsample at two months and for the entire

sample at six months. Self-report information was organized

into five categories of adherence: insulin injections,

glucose testing, diet level, dietary adherence, and physical

activity. Glycemic control was measured with glycosylated

hemoglobin and percent negative glucose tests (both urine

and blood). Each adherence measure was compared to patient

report of regimen prescriptions. Their results indicated

that compliance was better for taking medication and for

glucose testing than for behaviors which required major

adjustments in routine (e.g., diet and exercise); adherence

with one component of the regimen did not necessarily

correlate with adherence to other regimen tasks. Compliance

behaviors were not stable over time although glycemic

control was relatively stable; no clear relationship between

adherence and glycemic control was established through

either bivariate or multivariate analyses. Although this







study is an improvement over most correlational studies,

there are some basic weaknesses. For example, no attempt

was made to verify actual diabetologists' prescriptions--

rather patient report of physician prescriptions were used

and comparisons were made to the self-reports or in some

instances to absolute levels of behaviors (although the

authors do not define "absolute levels," these are presum-

ably meant to be ideal levels). Similarly, no attempt was

made to corroborate any of the self-reports of adherence.

Moreover, in calculating dietary compliance, only evening

meals were used.

Intervention studies comprise the second category of

investigations in the area of compliance and control. These

studies concentrate on increasing compliance, assuming that

lack of adherence to the medical regimen can result in

serious diabetic complications. However, a number of these

studies seek to modify adherence but do not include measures

of control. In one such study Gross (1982) conducted a

multiple baseline intervention with four IDDM boys aged 10-

12 years to determine the utility of self-management

behaviors. The youngsters were selected based on parent

report that they failed to reliably perform their urine

tests the recommended four times per day. The boys received

six 1-hour long sessions (one per week) which consisted of

written lessons, discussions, modeling and role-playing. In

addition, the children were assigned a self-management

project which included collecting baseline data on the







frequency of their urine testing which was arbitrarily

designated by the investigator as the target behavior. The

youngsters chose rewards and self-delivered them contingent

on performing the target behavior and graphed their perfor-

mance. During the final phase, the subjects were taught

negotiating and contracting skills and in a meeting with the

experimenter, the child, and his parents a contract was

arbitrated. All contracts involved parental reinforcement

for continuation of self-management of urine testing.

Reliability data were collected by parents counting the

number of test tablets used on 1 day each week when the

child was not at home. The children were aware that the

parents would be checking but did not know when. No attempt

was made to monitor the accuracy of the children's urine

testing. Reliability was based on dividing the number of

agreements of parent and child report by agreement plus

disagreement and multiplying the quotient by 100. Reliabi-

lity averaged 80%. No data were reported on the rate of

compliance with the self-reward regimen. Pre- and post-

training tests on the principles and procedures of behavior

modification were conducted and 2- and 8-week follow-up data

were also collected. Children improved their rate of urine

testing from 9% to 74% of the time during the self-manage-

ment condition and at the 2-week follow-up. However, at the

8-week follow-up, two of the four families had discontinued

with the contract and those children's behavior had returned

to baseline levels.








Only one of the many necessary diabetes management

behaviors was targeted in this study and no attempt was made

to determine whether increasing the frequency of urine

testing had an effect on children's level of control.

An intervention study by Epstein et al. (1981) employed

parent training procedures. Twenty families were assigned

to one of three groups in a multiple baseline design.

Parents were instructed to use a point system and praise for

their youngsters' compliance with urine testing. Both

self-report (daily urine test results and adherence to urine

testing regimen) and biochemical measures (insulin dosage,

HbA1, plasma glucose, serum lipids) were used to assess

effects of the intervention. Epstein et al. went to

considerable lengths to determine the reliability of the

children's recordings (although only negative urine data

were used for analyses) by asking parents to test four

urines per week on a schedule determined by the experimen-

ters and without knowledge of the child's values. Reliabi-

lity was assessed for 91% of the treatment weeks and parents

and children agreed on the glucose concentration within a

one measurement interval range on each of the four reliabi-

lity tests during 83% of the weeks. Exact parent/child

agreement on the urine tests was not reported. The

reliability/validity of reported urine testing frequency was

measured by including predetermined quantities of inert,

placebo Clinitest tablets in each bottle. Parents were

informed about the correct number of placebos. At the end






16
of each week, parents were instructed to test the remaining

tablets in each bottle, and add the number they found to the

number the child reported finding. Comparisons of this

number to the actual number provided the measure of the

child's adherence to the regimen for that week. Parents and

children agreed on the number of marked items during 76% of

the weeks. Results showed a significant increase in the

number of tests performed and in the percent of urine tests

that were negative for glucose, although other metabolic

indices of diabetic control did not show improvement.

Epstein et al. concluded that behavioral techniques appear

useful for improving behavioral adherence to diabetes

regimen. However, in this study, increasing the number of

urine tests that were negative for glucose was not associ-

ated with improvement in other measures of diabetic control.

Kaplan, Chadwick and Schimmel (1985) randomly assigned

21 IDDM teenagers between 13 and 18 from middle class

backgrounds to either a daily social learning group or a

medical facts learning group when they entered a 3 week

summer program. The social learning group identified social

situations in which peer influence might prevent adherence

to diabetes regimen and as a group suggested appropriate

responses. Rehearsal exercises enacting problem situations

and their solutions were ultimately videotaped and then

filmed in a television studio. Guided practice with

reinforcement was used to develop their social skills. The

control group spent the same amount of time discussing








diabetes relevant medical information. These discussions

were videotaped and then filmed. Hemoglobin Al was drawn

during the program to measure metabolic control before the

program and at a 4-month reunion to measure metabolic

control since the program.

The groups were equivalent at the beginning of the

program. Hemoglobin A1 values at the reunion (obtained for

82% of those who provided an original sample) showed a

slight increase in values for the control group and a

contrasting decrease for the experimental group. Further-

more, the authors found a substantial correlation (r=-.78)

between self-reported self-care and diabetes control as

measured with HbA1 values. This study is an improvement

over earlier research in that intervention was directed at

many aspects of the treatment regimen instead of urine

testing alone. Furthermore, correlations between adherence

behaviors and control were provided.

However, the study's experimental and control groups

were small and restricted to white middle class adolescents.

The reliability of the self-report measures of compliance

was not assessed nor were parents surveyed to corroborate

the youngsters' report. Before treatment, the experimental

group was in slightly better diabetes control than the

control group (although the difference was not statistically

significant). At post-treatment, it was unclear whether the

slight improvement of the experimental group and the slight

deterioration of the controls was clearly related to the








study's intervention or was a natural continuation of each

group's pretreatment metabolic condition. Pre/post- changes

in diabetic control were not significant and there was no

assessment of and therefore no evidence that one group's

compliance was better than the other group's at post-treat-

ment. Consequently, the only evidence clearly in favor of

the experimental group was the post-treatment difference in

HbAI. Certainly, a replication of the study's findings is

needed if we are to have confidence that the post-treatment

HbAl differences were, in fact, due to the intervention

employed.

Schafer, Glasgow, and McCaul (1982) designed a multiple

baseline approach to assess the effectiveness of goal

setting and behavioral contracting to increase adherence and

control in three adolescents with IDDM. A self-monitoring

program of 1 week, introduced before goal setting, con-

stituted the baseline phase. Contracting was introduced

only if 90% compliance was not achieved. Compliance was

assessed by subjects records of relevant target activity

such as time, duration, etc. Reliability checks were made

periodically by mothers' independent monitoring. Diabetic

control was assessed with urine glucose tests performed and

recorded daily by subjects. Reliability checks were

conducted by having mothers spot-check the urine tests. Two

hour postprandial blood glucose levels were determined, and

24-hour urine glucose was collected at pretest, posttest,

and followup. Two of the subjects showed improvements in








compliance and control. The third was essentially unaf-

fected by the treatment.

Unfortunately, as mentioned above, the use of 24 hour

urines as indices of diabetic control is problematic since

they are difficult to collect even in hospital settings. In

addition, postprandial blood glucose levels measure control

at brief points in time and may not relate to the patient's

general level of diabetic control.

It is generally accepted that the issue of compliance

as it relates to level of control and possible complications

is vital to the area of chronic illness in general, and

diabetes in particular. However, results are mixed and

difficult to interpret. In order to establish whether there

is actually a causal link, methods of measuring the various

components must be improved. Clearly the major reasons for

this gap in the area of IDDM are that

1) The definition of control has varied among

investigators, and diabetologists have been

unable to reach a consensus on a precise

definition so that studies have used varying

measures of control (Spevack, Johnson,

Harkavy, Silverstein, Shuster, Rosenbloom,

and Malone, 1987). There have been recent

improvements in this situation with the

general acceptance of HbAlc, although some

investigators balk at using this as the sole

measure of control since it does not








accurately reflect variability in blood

glucose.

2) The measurement of adherence behaviors is

equally problematic. Little psychometric

data are provided as to the quality of the

measures used. Compliance and control are

often confused or the same rater judges both.

Compliance is often treated as a global

construct although there is mounting evidence

that it is complex, consisting of several

independent components. Studies focus on be-

haviors that do not relate directly to

control. For example, wearing a medic alert

bracelet is definitely important in case of

accident (Schafer et al., 1982). However,

wearing it may not have a direct impact on

glycemic control.

3) Correlational studies have been inconsistent

in demonstrating a relationship between com-

pliance and control. In any case, data from

these studies cannot be used to establish

causality.

4) The few intervention studies available

suggest that adherence may be improved.

However, improved adherence is not always

associated with improved control (Epstein and

Cluss, 1982). In some intervention studies,








only a single behavior is targeted for

intervention or selected for treatment by the

parent of the diabetic child. In such cases,

the adherence behaviors improved may have no

significant impact on diabetic control.

5) It is very difficult to study IDDM youngsters

under controlled conditions, where behavior

can be reliably measured, without interfering

with or disrupting the youngster's normal

behavior.

Many investigators have dealt with the first problem by

relying predominantly on the best single estimator of

control--namely, hemoglobin Alc--which is an indicator of

integrated plasma glucose levels over approximately 3

months. Consequently, to establish a link between adherenc-

e/behavior and control as measured by hemoglobin Alc, a 3-

month behavior monitoring system should be available.

Unfortunately, youngsters balk at keeping diaries even for

short intervals and even when they do keep such records,

reliability checks are problematic. To date the most

systematic procedure for measuring compliance has been

developed by Johnson, Silverstein, Rosenbloom, Carter, and

Cunningham (1986). A 24-hour recall interview technique was

used to assess daily management behaviors in four diabetes

relevant areas (diet, injection, exercise, and glucose

testing). Each child and his/her parent were interviewed

separately on three different occasions concerning the







child's diabetes management behaviors in the last 24 hours.

Interviews were conducted by phone and took approximately 20

minutes. Data from this procedure were used to measure 13

different diabetes management behaviors: injection regulari-

ty, injection interval, injection-meal timing, regularity of

injection-meal timing, calories consumed, percent calories:

fat, percent calories: carbohydrates, concentrated sweets,

eating frequency, exercise duration, exercise type, exercise

frequency, and glucose testing frequency. Pearson product

moment correlations for all 13 adherence measures were

conducted to asses parent-child agreement. The results are

presented in Table 1. Although all of the correlations were

statistically significant (p<.0001), a number of measures

differed in agreement depending on the age of the child with

the youngest children (6-9 years) showing poorest

parent/child agreement on measures involving time and the

10- to 12- and 13- to 15-year old groups showing the most

consistent parent/child agreement across all measures. The

oldest group (16-19 years) had highly variable parent/child

correlations. The authors attribute these findings to the

reduction of parental supervision during late adolescence

when children spend more time with peers outside of the

parents' purview. The 13 adherence measures were then

subjected to a principal component factor analysis resulting

in a five-factor solution accounting for 71.3% of the

variance. The five factors were rotated to simple structure

using the varimax procedure. Factor I (Exercise) consisted







TABLE 1
Parent-Child Correlations for Total Sample and by Age Group


Group 1
(6-9 Years)
r(n)
p <


Group 2
(10-12 Years)
r(n)


Group 3 Group 4
(13-15.6 Years) (16-19 Years)
r(n) r(n)
p < p <


Adherence Measure


Injection
Regularity

Injection
interval

Injection-
meal timing

Regularity of
injection-
meal timing

Calories
consumed

Percentage
calories-fat


.61(152)
(.0001)

.77(154)
(.0001)

.67(163)
(.0001)

.42(148)
(.0001)


.77(139)
(.0001)

.64(167)
(.0001)


.46(27)
(.02)

.36(29)
(.05)

.53(31)
(.002)

-.23(26)
(.26)


.79(30)
(.0001)

.50(31)
(.005)


.68(65)
(.0001)

.71(67)
(.0001)

.547(68)
(.0001)

.50(63)
(.0001)


.80(66)
(.0001)

.63(70)
(.0001)


.83(42)
(.0001)

.54(40)
(.003)

.60(44)
(.001)

.58(41)
(.0001)


.61(33)
(.0001)

.76(44)
(.0002)


-.04(18)
(.86)

.91(18)
(.0001)

.78(20)
(.0001)

.46(18)
(.05)


.92(10)
(.0002)

.60(22)
(.003)


Total
Sample
r(n)







TABLE 1
(continued)


Total Group 1 Group 2 Group 3 Group 4
Sample (6-9 Years) (10-12 Years) (13-15.6 Years) (16-19 Years)
r(n) r(n) r(n) r(n) r(n)
p < p < p < p < p <
Adherence Measure
Percentage .64(167) .44(31) .63(70) .81(44) .62(22)
calories- (.0001) (.01) (.0001) (.0001) (.002)
carbohydrate

Concentrated .62(167) .47(31) .71(70) .76(44) .12(22)
Sweets (.0001) (.007) (.0001) (.0001) .58

Eating .45(167) .67(31) .47(70) .54(44) .13.(22)
frequency (.0001) (.0001) (.0001) (.0002) (.57)

Exercise .59(167) .03(31) .96(70) .79(44) .47(22)
duration (.0001) (.87) (.0001) (.0001) (.03)

Exercise .54(167) .74(31) .99(70) .66(44) .32(22)
type (.0001) (.0001) (.0001) (.0001) (.15)

Exercise .62(167) .62(31) .72(70) .72(44) .32(22)
frequency (.0001) (.0002) (.0001) (.0001) (.13)

Glucose .78(164) .81(31) .73(68) .77(44) .37(21)
testing (.0001) (.0001) (.0001) (.0001) (.10)
frequency


Table from Johnson, et al.,


1986








of the three exercise measures, Factor II (Injection)

consisted of all four injection measures, Factor III (Diet

Type) was made up of measures of diet type, Factor IV

(Eating/Glucose Testing Frequency) included measures of

eating frequency and glucose testing frequency, and Factor V

(Diet Amount) included measures of total calories consumed

and the amount of concentrated sweets ingested. The Johnson

et al. (1986) results indicate that adherence to a diabetic

regimen consists of five independent constructs and should

not be viewed as a global construct.

Diabetes summer camps offer a natural but relatively

controlled environment. Here youngsters' behavior and

metabolic control could be easily monitored and both would

be expected to change. Such an environment would be

conducive to investigating relationships of compliance/

control. Diabetes summer camps have as their main goal to

create an environment in which IDDM youngsters can meet

other children with diabetes and learn to live more normally

with their disease. The focus of the program is usually to

provide a model for healthy living along with didactic

components designed to reinforce adherence skills and

increase understanding of the disease. It is a generally

accepted philosophy that research conducted at summer camps

must be nonintrusive and must not interfere with camper ac-

tivities and enjoyment. The major criticism leveled at camp

studies is that they constitute an unnatural environment

which itself affects the youngsters' behavior and level of








control. Whether this is a valid criticism and if so,

whether the effects are positive or negative, has never been

comprehensively evaluated. There is evidence that diabetes

camp may improve diabetic control although only one study

specifically addressed this issue (Strickland et al., 1984).

The Strickland et al. study examined changes in glycemic

control during a 2-week summer camp program by performing

pre-camp and post-camp values of fasting plasma glucose,

glycosylated hemoglobin, and glycosylated serum protein in a

group of 36 children. Their results suggested that there

was measurable improvement in diabetic control in some

children.

Similarly, only a few studies have addressed behavioral

changes associated with camp. For example, Stunkard and

Pestka (1962) examined physical activity at a 2-week Girl

Scout camp and found that there was significantly more

activity during camp compared with at-home behavior. The

youngsters studied did not have diabetes.

Most camp studies deal with psychological constructs.

Moffatt and Pless (1983) investigated changes in locus of

control in juvenile diabetic campers during a 3 week camp

experience and found that there were significant changes

toward internal locus of control.

Self-concept was scrutinized using the Tennessee Self-

Concept Scale with 26 IDDM adolescents (aged 13-17) attend-

ing an 8-day summer camp (Hoffman, Guthrie, Speelman, and

Childs, 1982). The authors report that self-concept may be








more resistent to change above the age of 14 and that

initial level of control was not a significant predictor of

self-concept change. The greatest change occurred in the

13- to 14-year old females who improved on self-concept

significantly. However, this study did not address the

issue of whether self-concept change was associated with

change in glycemic control.

In a more recent study, Scharf, Adams, and Leach (1987)

compared 45 IDDM adolescents aged 12 to 17 who attended a

residential diabetes summer camp, to IDDM youngsters who did

not attend camp on psychological functioning immediately

following the camp experience and at a 5 month follow- up.

They reported that campers' adjustment to diabetes, their

self-worth, and their parents' perceptions of their

behavioral competencies and general personality adjustment

did not change due to the camp experience. Moreover,

metabolic control as measured with HbAlc before camp did not

differentially influence the experimental or control

subjects in how they reacted to the camp experience or how

they adjusted to a chronic illness. No attempt was made to

assess whether the camp experience affected glycemic control

after camp.

The effects of stress on glycemic control were examined

at a 1982 Tennessee camp for diabetic children using 39

adolescents between the ages of 12 and 15 years (Hanson and

Pichert, 1986). The children were followed for 3 days while

insulin administration, dietary intake, exercise, self-








reported stress, and blood glucose levels (measured via

Chemstrip bG Reagent strips four times during the 24 hours

preceding each afternoon rest period during which stress

questionnaires were administered) were monitored. Results

suggested that negative cumulative stress, as perceived by

the youngsters, correlated significantly with blood glucose

levels. Interestingly, positive cumulative stress was

significantly and negatively correlated with girls' blood

glucose levels.

Educational experiences at camp were found to contri-

bute to the knowledge and performance of self-care techni-

ques in a study by Lebovitz, Ellis, and Skyler (1978).

Harkavy et al. (1983) found that knowledge increased in 12-

to 15-year old campers but not in 10- to 11-year olds.

Dorchy, Loeb, Mozin, Lemiere, and Ernould (1982) adminis-

tered a 27 question pre- and post-test to 63 IDDM youngsters

and found that regardless of age or duration of diabetes all

children ages 9 to 15 years benefit from the teaching of

theory and practical issues. The youngest children (9-10

years) showed progress in all areas while the oldest (13-15

years) made greater gains in the areas of nutrition and

insulin therapy.

There is, therefore, some evidence that the camp

experience may affect psychological factors such as locus of

control (Moffat and Pless, 1983) and self-esteem (Hoffman,

et al., 1982), knowledge about the disease (particularly in

12- to 15-year old youngsters, Harkavy et al., 1983), self-








care (Lebovitz et al., 1978), and physical activity levels

(Stunkard and Pestka, 1962). Moreover, there is some

evidence that metabolic control may improve during the camp

session (Strickland et al., 1984).

Improvements in adherence behaviors are assumed to

occur since camp provides exercise opportunities and the

youngsters' regimen is regulated regarding injections and

meals. There are, however, no empirical data documenting

improved adherence at camp nor attesting to the relationship

between such behavior change and improvements in metabolic

control. Further, it is unclear whether any behavioral

changes induced by the camp setting are maintained once the

child returns home.

Glycosylated serum proteins (GSP) differ from glyco-

sylated hemoglobins (GH) in that serum protein levels in the

blood change more rapidly than do hemoglobin values with

changes in blood glucose concentration (Mehl, Wenzel,

Russel, Gardner, and Merimee, 1983). Therefore, since

glycosylated serum proteins have been shown to accurately

reflect alteration of mean glycemic levels 1 to 2 weeks

prior to testing, it becomes feasible to address the issue

of whether adherence behaviors at camp are indeed related to

metabolic control at camp. This study will, therefore,

assess the following:

1) The effect of summer camp on children's

diabetes management behavior;








2) The effects of a diabetes summer camp

environment on children's diabetes control;

3) The length of time that effects of camp are

maintained--i.e., do changes or effects of

camp dissipate once the child returns home;

4) The relationship between adherence behavior

and level of diabetic control.

This proposed investigation represents a naturalistic

study in which the experience of camp is expected to

actually change both the children's behavior and their

levels of diabetic control thereby providing the opportunity

to monitor both. Therefore, the behavior and the levels of

control are being naturally manipulated by virtue of the

youngsters' attendance at camp, providing an ideal situation

to assess the extent to which changes in diabetic control

are associated with changes in behavior. Should such

changes be demonstrated, and their association be estab-

lished, it would constitute strong support for a relation-

ship between behavior and health status.














METHOD

Sample Characteristics

Participants were 64 IDDM youngsters who attended the

1985 North Florida Camp for Children and Youth with

Diabetes. The 1985 camp session lasted 2 weeks. One parent

of each youngster was also asked to participate. Informed

consent was obtained from all participants. Youngsters were

between the ages of 7-12 (with a mean age of 10.5 years).

There were 34 boys and 30 girls. Participating families

reported income in 5 categories: 1= $ 0 to $9,999, 2=

$10,000 to $19,999, 3= $20,000 to $29,999, 4= $30,000 to

$39,999, and 4= $40,000+. Of all participating families

14.5% were in category 1, 27.3% in 2, 23.6% in 3, 18.2% in

4, and 16.4% in 5. All but one (98.4%) of the youngsters

had had diabetes for at least 1 year with a range from .9

years to 9.1 years (Table 2). All but one of the youngsters

were Caucasian. Only two youngsters dropped out of the

study after camp.


Measures

Adherence Measures

Twenty-four hour recall interviews, conducted with

children and parents, were recorded on Diabetes Daily Record

forms to assess the youngsters' diabetes related







TABLE 2
Sample Characteristics


Total Group 1 Group 2 Group 3
Sample 7-9 10-11.4 11.5-12.6
Years Years Years


Sample Size 64 20 23 21

Sex: Males 34 13 13 8
Females 30 7 10 13

Family Income
in categories: 1 14.5% 0% 28.6% 11.1%
2 27.3% 31.3% 28.6% 22.2%
3 23.6% 25.0% 23.8% 22.2%
4 18.2% 18.8% 9.5% 27.8%
5 16.4% 25.0% 9.5% 16.7%

Age: Mean 10.5 8.5 10.8 12.2
Range 7.6-12.6 7.6-9.1 10.1-11.1 11.6-12.6
Stand. Dev. 1.57 .65 .36 .47


Duration: Mean 3.9 3.6 3.3 4.9
Range .9-9.1 1-5.1 .9-4.4 1.1-8.3
Stand. Dev. 2.67 2.11 2.18 3.37






33
behaviors before, during and after camp (see Appendix). The

interviews were conducted by telephone. When possible, two

of these interviews dealt with.weekdays and one with weekend

activities. Participants were told that the interviewer was

concerned with what actually transpired rather than what

they believe should have been done. Trained, undergraduate

research assistants conducted these interviews. Partici-

pants were asked to recall the previous day's activities in

a sequential manner beginning with when the child woke up

and ending with when he went to bed. All diabetes relevant

behaviors were recorded. Interviewers prompted for informa-

tion that may have been inadvertently omitted, such as,

mid-morning snacks, exercise, glucose testing, etc.

Youngsters and their parents were interviewed separately.

Using data obtained from the 24 hour recall interviews,

13 different adherence measures were quantified. For each

measure, a range of scores is possible with higher scores

indicating relative noncompliance and lower scores indicat-

ing relative compliance (Johnson et al., 1986):

1) Injection Regularity--this measure assessed

the degree to which injections were given at

the same time every day. It was calculated

by measuring the standard deviation of

injection times reported across the inter-

views.

2) Injection Interval--this measure assessed the

youngster's average deviation from an ideal








injection interval of 24 hours between a.m.

injections. For youngsters who take p.m.

injections, an ideal.shot interval was

arbitrarily defined as 10 hours between a.m.

and p.m. injections on the same day, and 14

hours between the p.m. injection and the a.m.

injection on the following day.

3) Injection-Meal Timing--this measure evaluated

the timing of injection in relationship to

meals. It was calculated by averaging the

number of minutes between taking an injection

and eating a meal. Sixty minutes were added

to this average so that youngsters taking

their injection 60 minutes before a meal

received an adherence score of zero. Young-

sters who, on the average, took their

injections 60 minutes to 1 minute before

meals received scores from 0 to .99. Young-

sters who usually took their injections at

the time of their meals or after eating

received scores of 1.0 or greater.

4) Regularity of Injection-Meal Timing--

this measure appraised the regularity of

intervals between injections and eating.

This was measured by calculating the

standard deviation of the Injection-Meal

Timing measure described above.








5) Calories Consumed--for each youngster, an

ideal number of total daily calories was

identified based on the youngster's age, sex,

and ideal weight for height. Ideal weight

for height was obtained from tables provided

in a standard textbook on childhood obesity

(Collipp, 1980). Ideal calorie consumption

per kilogram was calculated using standards

provided by the U.S. Public Health Service

(Anderson et al., 1981). Each child's ideal

total number of daily calories was then

subtracted from the youngster's reported

daily caloric consumption. Reported calorie

consumption was obtained from the diet

information obtained by interview. All diet

information was coded in exchange units from

which calorie estimates were derived (Franz,

1983). A high score on the calorie consumed

measure indicated that the youngster ate more

than his ideal total number of daily calo-

ries. A zero score indicated that the child's

actual calorie consumption equalled that of

his ideal calorie consumption. A negative

score indicated that the youngster ate less

than the ideal.

6) Percent Calories: Fat--based on the exchange

unit information collected from the 24-hour








recall interviews, the percent of total

calories consumed which consisted of fat was

calculated. Ideal fat consumption was based

on the lower limit (25%) recommended by the

American Diabetes Association (Nuttal and

Brunzall, 1979). Ideal fat consumption (25%)

was subtracted from actual fat consumption.

Scores above zero indicated that the child

consumed more than 25% of his calories in

fats. Scores below zero indicated that the

child's fat intake was less than 25% of his

total calories.

7) Percent Calories: Carbohydrates--this

measure was also based on the exchange unit

information collected from the 24-hour recall

interviews. Ideal carbohydrate consumption

was based on the upper limit (60%) recom-

mended by the American Diabetes Association

(Nuttal and Brunzall, 1979). In this case

actual carbohydrate consumption was sub-

tracted from the ideal (60%) so that scores

above zero indicated insufficient carbo-

hydrate ingestion. A score of zero indicated

the child's diet consisted of 60% carbo-

hydrates. Scores below zero indicated that

more than 60% of the calories consumed

consisted of carbohydrates. No measure of








protein consumption was developed since it

can be automatically determined by knowing

the child's fat and carbohydrate consumption.

8) Concentrated Sweets--forty calories of any

concentrated sweet was considered equivalent

to one concentrated sweet exchange unit. The

average number of these exchange units

consumed per day was calculated.

9) Eating Frequency--based on an ideal of six

meals/snacks per day, the percent of snacks

or meals not eaten across the interviews was

calculated and multiplied by 100. A high

score indicates that the child ate infre-

quently. A low score indicated frequent

eating occasions. A score of zero indicated

the child averaged six meals or snacks per

day.

10) Exercise Duration--the average amount of time

the child spent exercising during each

exercise occasion across the interviews was

calculated and a constant (1) was added to

avoid subsequent division by zero. The

reciprocal of this score was used so that low

scores indicated lengthy exercise and high

scores indicated little or no exercise.

11) Exercise Type--each exercise or activity was

given an energy expenditure rating (Katch and








McArdle, 1977). Higher ratings indicated

more strenuous exercise. The youngster's

average expenditure rating across the

interviews was calculated and a constant (1)

was added to avoid subsequent division by

zero. The reciprocal of this score served as

the measure of Exercise Type; low scores

indicated more strenuous exercise while high

scores indicated less strenuous exercise.

12) Exercise Frequency--the occurrence or

nonoccurrence of exercise on three occasions

per day--morning, afternoon, and evening--

was noted for each of the interviews. The

percent of nonoccurrence of exercise across

these occasions was used as an estimate of

exercise frequency. A score of zero indicated

that the child reported exercise on all

occasions. A score of 100 indicated no

reported exercise on any occasion.

13) Glucose Testing Frequency--the frequency of

glucose testing across the interviews was

calculated. Using an ideal testing frequency

of four times per day, the number of glucose

tests was divided by this ideal (e.g., 12 for

3 days) and multiplied by 100. The total was

subtracted from 100 so that high scores indi-

cated few glucose tests and low scores








indicated frequent glucose tests. A score of

zero indicated the youngster tested 4 times

per day. A score of 100 indicated no

reported glucose tests over the interviews.

Glycemic Control Measures

Glycosylated hemoglobin Alc levels were obtained at the

beginning of camp and at 3 months following camp. The

assays were performed at the Shands Teaching Hospital

Laboratory using a column chromatography method performed

with Glyco-Gel kits. The expected normal range for Glycosy-

lated Hemoglobin Alc (HbAlc) using this procedure was 3.5-

6.2% and represents an overall estimate of glycemic control

for a period of approximately 3 months prior to testing.

Glycosylated serum proteins (GSP) were obtained at the

beginning and end of the 2-week camp session to evaluate any

changes in glycemic control associated with the camp

experience. The assays were performed at the Shands

Teaching Hospital Laboratory using the Pierce GlycoTest in

vitro diagnostic kits which contain GLYCO-GEL Analytical

Columns. This is an affinity chromatography based method

used after serum is first dialyzed against normal saline to

prevent spurious elevations of GSP levels due to free

glucose (Kennedy, 1981). GSP represents the average

glycemic control for a period of 10 to 14 days before

testing. Normal ranges were not available. However, in this

analysis the control value for a nondiabetic subject ranged

from .47 to 1.1% with a mean of .79%.








Procedure

In the 2 weeks prior to attending diabetes summer camp,

each child and one of his parents were interviewed by phone

on three separate occasions using the 24-hour recall

technique described earlier. The interviews were recorded

on the Diabetes Daily Record (Appendix).

On the first or second day of camp, fasting blood was

obtained from each participant. Serum obtained from venous

samples was stored at -20 degrees centigrade to be assayed

for GSP at a later time. Heparinized blood samples were

submitted for HbAlc assays. During their 2-week camp

experience, each child and his counselor were interviewed on

three separate occasions (on 2 weekdays and 1 weekend day)

by trained interviewers who recorded diabetes management

behaviors for the previous day on the Diabetes Daily Record.

However, counselor interviews were discontinued as it became

clear that they had difficulty differentially remembering

the individual children's activities and food consumption.

During the last 2 days of camp, fasting blood was again

obtained for GSP analysis. Those children whose blood was

drawn on the first day of camp were drawn on the second to

last day of camp. Similarly, if their blood was first drawn

on the second day of camp, the second sample was obtained on

the last full day of camp.

Within the 2 weeks following camp, study participants

and their parents again participated in three telephone

interviews using the 24-hour recall technique. This






41

procedure was repeated at 6 and 12 weeks post-camp. At 12

weeks post-camp, a second HbAlc sample was obtained.














RESULTS


Reliability of Child Report

One possible measure of child report reliability is

parent-child agreement. Perfect agreement was not expected

since parents are not always able to observe everything the

child does. However, since counselors were unable to

provide useful data concerning the campers activities, the

youngsters were the only consistent reporters available and

the issue of their reliability in reporting their behaviors

was critical.

Estimates of agreement between parent and child were

calculated to indicate the reliability of the information

obtained from the youngsters at camp, using Pearson product

moment correlations. These analyses replicated the Johnson

et al. study (1986) and used their 13 adherence measures.

Agreement was assessed for the total sample and for three

age groups (7-9, 10-11.4, and 11.5-12.6) at four points in

time: pre-camp and post-camp on 3 separate occasions

(immediately after camp, 6 weeks after camp, and 3 months

after camp). These data are presented in Tables 3, 4, 5,

and 6.

For the pre-camp period, parent/child correlations for

the total sample were all significant at p < .02, except for

regularity of injection-meal timing. One or both of the two







TABLE 3
Parent-Child Correlations for Total
Pre-Camp


Sample and by Age Group


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
Adher< ence< p< pleasure
Adherence Measure


Injection
regularity

Injection
interval

Injection-
meal timing

Regularity of
injection-
meal timing

Calories
consumed

Percentage
calories-fat


.65(60)
(.0001)

.75(59)
(.0001)

.54(62)
(.0001)

.09(59)
(.5010)


.73(62)
(.0001)

.78(63)
(.0001)


.80(18)
(.0001)

.51(19)
(.0262)

-.03(19)
(.8885)

-.08(18)
(.7382)


.81(20)
(.0001)

.86(20)
(.0001)


.28(21)
(.2156)

.92(19)
(.0001)

.57(22)
(.0052)

-.08(20)
(.7450)


.62(22)
(.0023)

.64(22)
(.0013)


.82(21)
(.0001)

.64(21)
(.0018)

.75(21)
(.0001)

.42(21)
(.0526)


.84(20)
(.0001)

.73(21)
(.0002)






TABLE 3
(continued)


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < p < p < p <
Adherence Measure

Percentage .72(63) .80(20) .65(22) .69(21)
calories- (.0001) (.0001) (.0011) (.0005)
carbohydrate

Concentrated .63(63) .35(20) .71(22) .63(21)
sweets (.0001) (.1292) (.0002) (.0020)

Eating .72(63) .83(20) .64(22) .73(21)
frequency (.0001) (.0001) (.0013) (.0020)

Exercise .29(63) .68(20) .07(22) .96(21)
duration (.0211) (.0010) (.7603) (.0001)

Exercise .64(63) .76(20) .56(22) .62(21)
type (.0001) (.0001) (.0071) (.0029)

Exercise .68(63) .81(20) .50(22) .68(21)
frequency (.0001) (.0001) (.0190) (.0007)

Glucose .90(63) .89(20) .97(22) .68(21)
testing (.0001) (.0001) (.0001) (.0006)
frequency







TABLE 4
Parent-Child Correlations for Total Sample and by Age Group
Immediately Post-Camp


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < p < p < p <
Adherence Measure


Injection
regularity

Injection
interval

Injection-
meal timing

Regularity of
injection-
meal timing

Calories
consumed

Percentage
calories-fat


.54(55)
(.0001)

.73(56)
(.0001)

.68(57)
(.0001)

.44(54)
(.0008)


.62(59)
(.0001)

.74(60)
(.0001)


.54(15)
(.0381)

.67(16)
(.0045)

.31(18)
(.2152)

-.40(15)
(.1432)


.36(18)
(.1471)

.68(18)
(.0020)


.52(20)
(.0198)

.66(20)
(.0016)

.91(19)
(.0001)

.42(19)
(.0726)


.62(22)
(.0022)

.87(22)
(.0001)


.65(20)
(.0021)

.83(20)
(.0001)

.56(20)
(.0098)

.72(19)
(.0003)


.70(19)
(.0009)

.62(20)
(.0035)






TABLE 4
(continued)


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < p < p < p <
Adherence Measure

Percentage .75(60) .63(18) .85(22) .70(20)
calories- (.0001) (.0048) (.0001) (.0006)
carbohydrate

Concentrated .59(60) .58(18) .83(22) .34(20)
sweets (.0001) (.0108) (.0001) (.1415)

Eating .67(60) .77(18) .59(22) .63(20)
frequency (.0001) (.0002) (.0037) (.0026)

Exercise .21(60) .53(18) .21(22) .13(20)
duration (.1106) (.0234) (.3568) (.5965)

Exercise .64(60) .66(18) .50(22) .85(20)
type (.0001) (.0028) (.0171) (.0001)

Exercise .68(60) .67(18) .53(22) .92(20)
frequency (.0001) (.0024) (.0117) (.0001)

Glucose .78(60) .55(18) .92(22) .68(20)
testing (.0001) (.0171) (.0001) (.0009)
frequency







TABLE 5
Parent-Child Correlations for Total Sample and by Age Group
Six Weeks Post-Camp


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < P < p < p <
Adherence Measure


Injection
regularity

Injection
interval

Injection-
meal timing

Regularity of
injection-
meal timing

Calories
consumed

Percentage
calories-fat


.64(53)
(.0001)

.66(53)
(.0001)

.76(56)
(.0001)

.26(52)
(.0615)


.69(56)
(.0001)

.86(57)
(.0001)


.57(14)
(.0320)

.45(15)
(.0942)

.11(16)
(.6915)

.11(14)
(.7032)


.49(17)
(.0456)

.96(17)
(.0001)


.66(20)
(.0017)

.79(20)
(.0001)

.76(21)
(.0001)

.34(19)
(.1501)


.65(21)
(.0014)

.76(21)
(.0001)


.67(19)
(.0019)

.81(18)
(.0001)

.95(19)
(.0001)

.46(19)
(.0461)


.85(18)
(.0001)

.84(19)
(.0001)






TABLE 5
(continued)


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
P < p < p < p <
Adherence Measure
Percentage .83(57) .93(17) .70(21) .86(19)
calories- (.0001) (.0001) (.0004) (.0001)
carbohydrate

Concentrated .68(57) .92(17) .70(21) .55(19)
sweets (.0001) (.0001) (.0004) (.0141)

Eating .74(57) .62(17) .74(21) .85(19)
frequency (.0001) (.0080) (.0001) (.0001)

Exercise .30(57) .20(17) .61(21) .23(19)
duration (.0235) (.4428) (.0035) (.3383)

Exercise .56(57) .49(17) .63(21) .76(19)
type (.0001) (.0442) (.0021) (.0001)

Exercise .58(57) .40(17) .75(21) .70(19)
frequency (.0001) (.1150) (.0001) (.0009)

Glucose .91(57) .85(17) .96(21) .93(19)
testing (.0001) (.0001) (.0001) (.0001)
frequency






TABLE 6
Parent-Child Correlations for Total Sample and by Age Group
Three Months Post-Camp


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < p < p < p <
Adherence Measure


Injection
regularity

Injection
interval

Injection-
meal timing

Regularity of
injection-
meal timing

Calories
consumed

Percentage
calories-fat


.81(55)
(.0001)

.89(52)
(.0001)

.77(60)
(.0001)

.25(54)
(.0665)


.69(60)
(.0001)

.66(61)
(.0001)


.81(17)
(.0001)

.76(18)
(.0002)

.47(19)
(.0422)

.22(17)
(.4001)


.11(20)
(.6344)

.75(20)
(.0001)


.88(18)
(.0001)

.95(15)
(.0001)

.78(21)
(.0001)

-.11(17)
(.6710)


.64(21)
(.0016)

.62(21)
(.0026)


.74(20)
(.0002)

.93(19)
(.0001)

.86(20)
(.0001)

.38(20)
(.1011)


.75(19)
(.0002)

.65(20)
(.0019)






TABLE 6
(continued)


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < p < p < p <
Adherence Measure

Percentage .65(61) .68(20) .69(21) .65(20)
calories- (.0001) (.0009) (.0006) (.0017)
carbohydrate

Concentrated .43(61) .37(20) .25(21) .69(20)
sweets (.0005) (.1078) (.2666) (.0008)

Eating .80(61) .51(20) .85(21) .90(20)
frequency (.0001) (.0228) (.0001) (.0001)
Exercise .66(61) .68(20) .67(21) .77(20)
duration (.0001) (.0011) (.0009) (.0001)

Exercise .68(61) .51(20) .74(21) .77(20)
type (.0001) (.0227) (.0001) (.0001)

Exercise .71(61) .71(20) .69(21) .76(20)
frequency (.0001) (.0005) (.0006) (.0001)

Glucose .91(61) .89(20) .96(21) .85(20)
testing (.0001) (.0001) (.0001) (.0001)
frequency








younger groups exhibited poor correlations on measures

involving time (injection regularity, injection-meal timing,

regularity of injection-meal timing, and exercise duration).

However, the oldest children were significantly correlated

with their parents (p < .05) on all measures (Table 3).

Immediately after camp (Table 4), the parent/child

correlations for the total sample were all significant with

the exception of exercise duration. Injection-meal timing

and regularity of injection-meal timing were still problema-

tic for the 7- to 9-year olds. The 10- to 11.4-year olds

exhibited poor correlations on regularity of injection-meal

timing and exercise duration, while the oldest youngsters

were not significantly correlated with their parents on

measures of concentrated sweets and exercise duration (Table

4).

The 6 weeks and 3 months after camp interviews demonst-

rate a similar pattern. These results are presented in

Tables 5 and 6 respectively. Percentage calories: fat, per-

centage calories: carbohydrates, eating frequency, exercise

type, and glucose testing demonstrated significant correlat-

ions for all children. Regularity of injection-meal timing

and exercise duration demonstrated the weakest (nonsig-

nificant) correlations in each of the study periods except

for the last one in which parent/child agreement for

exercise duration improved (r=.66). However, parent/child

agreement for regularity of injection-meal timing remained

weak (r=.25).






52
In addition to the parent/child correlations calculated

for each time period, the interviews were also collapsed

across time periods to determine the overall correlations by

age group (Table 7). Parents and children's reports for the

total sample were significantly correlated for all 13

measures, ranging from r=.36 (p<.004) on regularity of

injection-meal timing to r=.92 (p<.0001) for glucose testing

frequency. Comparisons between independent correlations

using Fisher's z' transformation were conducted to assess

differences in the degree of parent/child agreement across

the three age groups (p < .05, Cohen and Cohen, 1983, pages

53-56). These tests revealed that on 5 of the 13 adherence

measures (injection interval, injection-meal timing,

calories consumed, exercise duration, and glucose testing

frequency) there were significant differences in the

parent/child correlations among the three age groups (see

Table 7). The 10- to 11.4-year old demonstrated the highest

agreement (r=.92) with their parents on the injection

interval measure while the 7- to 9-year olds' reports did

not correspond as well (r=.51). On the injection-meal timing

measure, the oldest youngsters' reports corresponded best

with their parents (r= .75), while the 7- to 9-year olds

demonstrated the worst parent/child agreement (r= -.03).

Similarly, on the calories consumed measure, the oldest

group exhibited the highest parent/child correlation

(r=.89) and the youngest group displayed the weakest

(r=.53). The 11.5- to 12.6-year olds again demonstrated the







TABLE 7
Parent-Child Correlations for Total Sample and by Age Group
Collapsed Across Time


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
pAdher< ence< p< pleasure
Adherence Measure


Injection
regularity

Injection *
interval

Injection- *
meal timing

Regularity of
injection-
meal timing

Calories *
consumed

Percentage
calories-fat


.81(63)
(.0001)

.75(59)
(.0001)

.69(63)
(.0001)

.36(63)
(.0049)


.81(62)
(.0001)

.91(63)
(.0001)


.75(20)
(.0001)

.51(19)
(.0262)

-.03(19)
(.8885)

-.08(18)
(.7141)


.53(20)
(.0163)

.96(20)
(.0001)


.88(22)
(.0001)

.92(19)
(.0001)

.57(22)
(.0052)

-.08(20)
(.7382)


.81(22)
(.0001)

.88(22)
(.0001)


.77(21)
(.0001)

.64(21)
(.0018)

.75(21)
(.0001)

.43(21)
(.0526)


.89(20)
(.0001)

.89(21)
(.0001)






TABLE 7
(continued)


Total Group 1 Group 2 Group 3
Sample (7-9 Years) (10-11.4 Years) (11.5-12.6 Years)
r(n) r(n) r(n) r(n)
p < p < p < p <
Adherence Measure

Percentage .89(63) .96(20) .88(22) .84(21)
calories- (.0001) (.0001) (.0001) (.0001)
carbohydrate

Concentrated .69(63) .73(20) .80(22) .58(21)
sweets (.0001) (.0003) (.0001) (.0054)

Eating .83(63) .80(20) .77(22) .90(21)
frequency (.0001) (.0001) (.0001) (.0001)

Exercise .63(63) .68(20) .43(22) .86(21)
duration (.0001) (.0010) (.0457) (.0001)

Exercise .70(63) .75(20) .60(22) .73(21)
type (.0001) (.0001) (.0033) (.0002)

Exercise .76(63) .81(20) .63(22) .86(21)
frequency (.0001) (.0001) (.0016) (.0001)

Glucose .92(63) .86(20) .98(22) .84(21)
testing (.0001) (.0001) (.0001) (.0001)
frequency


* indicates significant differences by age group.








highest parent/child agreement (r=.86) on the exercise

duration measure, while the 10- to 11.4-year olds exhibited

the weakest correlation (r=.43). On the glucose testing

frequency measure, all groups' reports correlated highly

with their parents, however there was a significant dif-

ference between the 10- to 11.4-year olds' (r=.98) and the

11.5- to 12.6-year olds' (r=.84) parent/child correlations.

The correlations between parent and child report were

then collapsed across the age groups so that the parent/

child correlations for the total sample could be examined

across the four time periods of the study. A test of

compound symmetry was used to test for the equality of these

dependent correlations (Steiger, 1980). Results of these

analyses (Table 8) indicated that agreement between parent

and child did not differ significantly across time for 8 of

the 13 adherence measures: calories consumed; percent

calories: fat; percent calories: carbohydrates; eating

frequency; exercise duration; exercise type; exercise

frequency; and testing frequency. On measures of injection

regularity, injection interval, injection-meal timing,

regularity of injection-meal timing, and concentrated sweets

there were significant differences in the parent/child

correlations across two or more time periods. These

differences exhibited no consistent pattern. Injection

regularity, for example, exhibited the best parent/child

agreement at Time 5 (r=.81) and the worst at Time 3 (r=.54),

while concentrated sweets exhibited the highest parent/child







TABLE 8
Parent-Child Correlations for Total Sample
Over Four Time Periods: Pre and Post-Camp


Time 1 Time 3 Time 4 Time 5
Pre-Camp Post-Camp 6 wks Post-Camp 3 Mos Post-Camp
r(n) r(n) r(n) r(n)
pAdhere< p< p< pleasure
Adherence Measure


Injection *
regularity

Injection *
interval

Injection- *
meal timing

Regularity of *
injection-
meal timing

Calories
consumed

Percentage
calories-fat


.65(60)
(.0001)

.75(59)
(.0001)

.54(62)
(.0001)

.09(59)
(.5010)


.73(62)
(.0001)

.78(63)
(.0001)


.54(55)
(.0001)

.73(56)
(.0001)

.68(57)
(.0001)

.44(54)
(.0008)


.62(59)
(.0001)

.74(60)
(.0001)


.64(53)
(.0001)

.66(53)
(.0001)

.76(56)
(.0001)

.26(52)
(.0615)


.69(56)
(.0001)

.86(57)
(.0001)


.81(55)
(.0001)

.89(52)
(.0001)

.77(60)
(.0001)

.25(54)
(.0665)


.69(60)
(.0001)

.66(61)
(.0001)






TABLE 8
(continued)
Time 1 Time 3 Time 4 Time 5
Pre-Camp Post-Camp 6 wks Post-Camp 3 Mos Post-Camp
r(n) r(n) r(n) r(n)
P < p < p < p <
Adherence Measure

Percentage .72(63) .75(60) .83(57) .65(61)
calories- (.0001) (.0001) (.0001) (.0001)
carbohydrate

Concentrated .63(63) .59(60) .68(57) .43(61)
sweets (.0001) (.0001) (.0001) (.0005)

Eating .72(63) .67(60) .74(57) .80(61)
frequency (.0001) (.0001) (.0001) (.0001)
Exercise .29(63) .21(60) .30(57) .66(61)
duration (.0211) (.1106) (.0235) (.0001)

Exercise .64(63) .64(60) .56(57) .68(61)
type (.0001) (.0001) (.0001) (.0001)

Exercise .68(63) .68(60) .58(57) .71(61)
frequency (.0001) (.0001) (.0001) (.0001)

Glucose .90(63) .78(60) .91(57) .91(61)
testing (.0001) (.0001) (.0001) (.0001)
frequency


* indicates significant differences between correlations
across time periods.








agreement at Time 4 (r=.68) and the poorest concordance at

Time 5 (r=.43). Only the injection-meal timing measure

exhibited any evidence of a linear trend toward improvement

in parent/child agreement across time (see Table 8).

However, by the end of the study (Time 5) parents and

children were significantly correlated on 12 of the 13 ad-

herence measures (p< .0005). Only parent/child agreement

for the regularity of injection-meal timing remained weak.

Differences between parent and child report were

further analyzed with repeated measures ANOVA on each of the

13 adherence measures collapsed across all parent/child

interviews using two between subject factors: Age group (7-

to 9-year olds, 10- to 11.4-year olds, and 11.5- to 12.6-

year olds) and Sex, and one within subject factor: Respon-

dent (parent, child). Results indicated that there were few

differences between parent and child report. Simple main

effects for Respondent emerged on three measures. On the

injection-meal timing measure, parents reported an average

of 14 minutes while children reported an average of 19

minutes between injection and meal, F(1,56)=5.98, p < .02.

On the eating frequency measure parents reported that

children ate an average of 5.25 meals per day while children

reported eating an average of 5 times per day, F(1,56)=

25.29, p< .0001. On calories consumed parents reported that

their children ate approximately 92 calories more than the

recommended amount per day while children reported ingesting






59
approximately 13 calories less than the recommended amount

per day, F(1,55)=6.83, p < .01.

On measures of exercise type and exercise frequency a

Respondent x Sex interaction emerged F(1,56)=9.37 p < .003

and F(1,56)=7.28, p < .009 respectively. To determine the

source of the Respondent x Sex interactions, the data were

divided by sex and repeated measures ANOVA were conducted on

each of the two exercise measures using one within subject

factor (Respondent). For both exercise type and exercise

frequency, boys' reports did not significantly differ from

their parents' reports. In contrast, girls reported

participating in more strenuous and more frequent episodes

of exercise than did their parents, F(1,29)=11.99 p < .002,

and F(1,29)=5.90 p < .02 respectively. These interactions

were further examined using T tests to explore possible

differences between parents reports of boys' versus girls'

exercise type and exercise frequency; parents reported

significantly more strenuous and more frequent episodes of

exercise for boys compared to girls, t(62)=-3.35, p < .001.

In contrast, the children themselves did not report the same

male/female differences found in the parent data.

In summary, the children's reports correlated with

their parents reports in the expected direction, both across

time and across age groups. In addition, even when

differences between parent and child report did exist, they

were relatively small and of questionable clinical sig-

nificance. Since the youngsters' self-report data were






60
generally reliable, replicating reliability data reported in

a previous study (Johnson et al., 1986), and since the

counselor interview data were unusable, it was decided to

use only the children's report in the succeeding analyses.


Effect of Diabetes Camp on Adherence

To test for possible changes in children's management

behaviors during camp, a repeated measures ANOVA was

conducted on each of the 13 adherence measures using two

between subject factors--Age Group (7-9 years, 10-11 years,

11.5-12.6 years) and Sex (M,F) and one within subjects

factor--Time (Time 1: before camp, Time 2: during camp, Time

3: immediately after camp, Time 4: 6 weeks after camp, and

Time 5: 3 months after camp). Significant Time main effects

or interactions with Time emerged for 9 of the 13 adherence

measures (i.e., injection regularity, injection interval,

injection-meal timing, calories consumed, eating frequency,

exercise duration, exercise type, exercise frequency, and

glucose testing frequency). Duncan's Multiple Range Tests

performed on these measures revealed that in all cases

adherence behaviors at camp (Time 2) differed significantly

from one or more of the pre- or post-camp periods (Time 1,

Time 3, Time 4, and Time 5). These results are presented in

Table 9.

Simple main effects for Time emerged for behaviors as-

sociated with injection regularity F(4,164)=6.04, p < .0001,

injection-meal timing F(4,168)=2.83, p < .03, eating







TABLE 9
Duncan's Multiple Range Tests (p< .05) on the Adherence Measures
Demonstrating Within Subject Effects


Time Periods Compared
Pre-camp vs Post-camp Camp vs Pre- and Post-camp Post-camp
1 3 1 4 1 5 2 1 2 3 2 4 2 5 3 4 3 5 4 5
Adherence Measure


Injection
regularity

Injection 7-9yrs
interval 10-11.4yrs
11.5-12.6yrs


* *




* *

* *


Injection-
meal timing

Regularity of
injection-
meal timing


Calories 7-9yrs
consumed 10-11.4yrs
11.5-12.6yrs


* *
* *
* *


Percentage
calories-fat






TABLE 9
(continued)


Time Periods Compared
Pre-camp vs Post-camp Camp vs Pre- and Post-camp Post-camp
13 14 15 21 23 24 25 34 35 45
Adherence Measure

Percentage
calories-
carbohydrate

Concentrated
sweets

Eating *
frequency

Exercise # # *
duration

Exercise #* *
type

Exercise 7-9yrs # *
frequency 10-11.4yrs *
11.5-12.6yrs # # # *

Glucose # # *
testing
frequency


* indicates significant improvement.
# indicates significant deterioration.








frequency F(4,200)=26.15, p < .0001, exercise duration

F(4,200)=5.97, p < .0001, exercise type F(4,200)=101.08,

p < .0001, and testing frequency, F(4,200)=8.11 p < .0001.

In every instance children exhibited more compliant behavior

during camp (Table 10).

Repeated measures ANOVA revealed Time main effects as

well as Time x Age group interaction effects on 3 additional

adherence measures (Table 11): injection interval (Time

effect, F(4,140)=2.45, p < .05; Time x Age group interac-

tion, F(8,140) = 2.21, p < .03), calories consumed (Time

effect, F(4,196)=7.96, p < .0001; Time x Age group interac-

tion, F(8,196)=2.39, p < .02), and exercise frequency (Time

effect, F(4,200)=179, p < .0001; Time x Age group interac-

tion, F(8,200)=2.27, p < .02). However, subsequent analyses

indicated that injection interval was the only measure to

evidence differential effects at camp for different age

groups. That is, the oldest children (11.5-12.6 years)

reported significantly improved compliance during camp,

while the 7- to 9-year olds and the 10- to 11.4-year olds

reported no change on these measures between home and camp.

Time x Age group interactions for the calories consumed and

exercise frequency measures were due to differences between

the age groups at one of the five time periods rather than

to a differential effect of camp on the age groups. In

fact, all age groups exhibited a significant increase on

calories consumed during camp and all youngsters reported a






TABLE 10
Means and Standard Deviations for Adherence
Measures Demonstrating Simple Main Effects for Time


Time Period
Time 1 Time 2 Time 3 Time 4 Time 5
Means (SD) Means (SD) Means (SD) Means (SD) Means (SD)
(Interp) (Interp) (Interp) (Interp) (Interp)
Adherence Measure

Injection .47(.38) .27(.27) .50(.35) .46(.31) .68(.55)
regularity (minutes) (28.02) (15.92) (30.06) (27.60) (40.59)

Injection- .70(.50) .49(.44) .68(.37) .66(.47) .71(.46)
meal timing (minutes) (42.14) (29.65) (40.81) (39.52) (42.74)

Eating 15.35(12.62) 2.58(3.54) 15.31(13.55) 17.33(13.93) 16.26(14.67)
frequency (per day) (5.02) (5.86) (5.07) (4.97) (5.02)

Exercise .09(.14) .03(.01) .16(.23) .17(.24) .12(.18)
duration (minutes (20.03) (42.41) (14.44) (15.35) (14.07)
per occasion)

Exercise .97(.01) .94(.01) .98(.01) .97(..02) .97(.01)
type (.03) (.06) (.02) (.03) (.03)
(kilocalories/min)

Glucose 51.23(24.41) 45.80(12.78) 54.09(24.50) 59.18(26.09) 60.80(25.47)
testing (1.88) (2.17) (1.79) (1.58) (1.53)
frequency
(per day)






TABLE 11
Means and Standard Deviations for Adherence
Measures with Age x Time Interaction at the 5 Time Periods


Time Period
Time 1 Time 2 Time 3 Time 4 Time 5
Means(SD) Means(SD) Means(SD) Means(SD) Means(SD)
(Interp) (Interp) (Interp) (Interp) (Interp)
Adherence Measure

Injection 7-9yrs .43(.41) .46(.54) .63(.65) .98(.66) .45(.34)
interval (25.6) (27.6) (37.8) (58.8) (26.7)
(minutes) 10-11.4yrs .91(.87) 1.05(1.9) .94(.73) .79(.77) 1.31(1.2)
(54.8) (63.3) (56.4) (47.4) (78.5)
11.5-12.6yrs .79(.50) .24(.23) 1.28(1.0) .84(.42) 1.06(.94)
(47.1) (14.5) (76.54) (50.3) (63.5)
Calories 7-9yrs -7.4(489) 164.9(962) -4.1(619) 18.1(721) 277.1(480)
consumed
10-11.4yrs 100.0(843) 598.9(794) 228.5(610) 111.5(693) 297.4(726)

11.5-12.6yrs -307.8(612) 382.9(874) -327.9(533) -325.3(649) -396.9(550)

Exercise 7-9yrs 72.9(14.8) 30.8(11.7) 80.9(12.0) 74.8(12.0) 67.0(13.4)
frequency (1.8) (4.5) (1.3) (1.6) (2.0)
10-11.4yrs 73.7(11.1) 23.4(9.7) 74.2(19.5) 71.1(16.3) 74.6(13.8)
(1.6) (4.2) (1.6) (1.8) (.07)
11.5-12.6yrs 69.3(14.3) 28.4(16.7) 80.4(12.5) 81.2(11.1) 78.4(11.2)
(.10) (4.3) (.04) (04) (1.3)







significant increase in exercise frequency associated with

the camp experience (see Table 9 and 11).

Five Measures exhibited no significant change related

to the camp experience: regularity of injection-meal timing,

percentage calories: fat, percentage calories: carbohydrat-

es, and concentrated sweets (Table 12).

On the regularity of injection-meal timing measure,

there was also an Age group x Sex interaction, F(2,38)=3.41,

p < .04). Subsequent LSMEANS procedure revealed that the 10-

to 11.4-year old girls were significantly less adherent than

the 11.5- to 12.6-year old girls or the 10- to 11.4-year old

boys. No other main or interaction effects for sex emerged

for any other adherence measure.

Although it is clear that significant changes occurred

during camp for 9 of the 13 adherence behaviors, examination

of Table 9 demonstrates that these changes were not main-

tained. That is, Time 2 (camp) exhibited consistent

significant differences from Times 1, 3, 4, and 5. In

contrast there were fewer significant differences between

Time 1, 3, 4, and 5. When differences between these times

did occur, they were most often in the direction of poorer

adherence (see Table 9 and interpretations in Tables 10, 11,

and 12).



Effect of Diabetes Camp on Glycemic Control

This project used two measures to assess glycemic

control. Glycosylated hemoglobin (HbAlc) which provides an






TABLE 12
Means and Standard Deviations for Adherence
Measures that Remained Stable Over the 5 Time Periods


Time Period
Time 1 Time 2 Time 3 Time 4 Time 5
Means(SD) Means(SD) Means(SD) Means(SD) Means(SD)
(Interp) (Interp) (Interp) (Interp) (Interp)
Adherence Measure

Regularity of 18.68(19.3) 18.10(13.5) 19.49(18.2) 12.02(10.4) 18.20(19.22)
injection- (minutes)
meal timing

Percentage 23.67(6.9) 22.04(5.7) 24.53(7.2) 23.99(9.0) 22.73(6.5)
calories-fat (48.7) (47.0) (49.5) (48.9) (47.7)

Percentage 24.33(7.2) 23.35(6.0) 25.44(7.5) 24.76(8.8) 23.35(6.7)
calories- (35.7) (36.7) (34.6) (35.2) (36.7)
carbohydrate

Concentrated 1.52(1.9) 1.24(1.1) 1.33(1.8) 1.38(1.7) 1.56(1.7)
sweets (per day)






68
average estimation of glycemic control for a period of 2 to

4 months, and glycosylated serum protein (GSP) which

provides a measure of glycemic.control over a 10 to 14 day

period. In order to support the reliability of the laborat-

ory assays, Pearson product moment correlations were

performed (Table 13). It was expected that the highest GSP/

HbAlc correlation would occur between the pre-camp GSP

measure and the pre-camp HbAlc measure since they were drawn

at the same time and reflect some overlap in time periods.

This was found to be the case (r=.71, p < .0001, Table 13).

Further, it was expected that the pre-camp GSP measure would

correlate less well with the 3 month followup HbAlc since

these measures reflect entirely different time periods. In

fact, the correlation between pre-camp GSP and followup

HbAlc was markedly lower (r=.52, p < .0001). The lowest

GSP/Hbalc correlations were expected between post-camp GSP

and both pre-camp HbAlc and followup HbAlc, taken at 3

months after camp, since the GSP and HbAlc measures reflect

completely different time periods and completely different

settings (camp versus home). In fact, these correlations

(r=.43; r=.39) represent the lowest correlations in the

matrix depicted in Table 13. Overall, the pattern of

results depicted in Table 13 supports the reliability of our

glycemic control measures.

To assess possible changes in glycemic control due to

the camp experience,a repeated measures ANOVA was performed

on the GSP pre- and post-camp measures using 2 between











TABLE 13
Correlations Between Glycosylated Hemoglobins
and Serum Proteins


HbAIC GSP

3 months
pre-camp post-camp pre-camp post-camp
r(n) r(n) r(n) r(n)
p< p < p < p <


HbAlc :

pre-camp 1.00(63)
(.0000)

3 months .62(61) 1.00(62)
post-camp (.0001) (.0000)


GSP

pre-camp .71(53) .52(52) 1.00(54)
(.0001) (.0001) (.0000)

post-camp .43(60) .39(59) .78(54) 1.00(61)
(.0007) (.0020) (.0001) (.0000)






70
subjects factors (Age group and Sex), and 1 within subjects

factor (Time: pre- and post-camp). A significant main

effect for Time, F(1,50)=11.00, p < .0017, and a significant

main effect for Age group, F(2,50)=3.31, p < .0447 emerged.

Subsequent LSMEANS procedures revealed that the overall mean

GSP values increased from 5.63% to 6.44% and that the 7- to

9-year olds (5.40%) and the 11- to 11.4-year old youngsters

(5.23%) had significantly lower mean GSP values than the

11.5- to 12.6-year old youngsters (6.82%).

It was then hypothesized that children arriving at camp

at different levels of glycemic control may have been dif-

ferentially affected by the camp experience. Therefore,

the youngsters were assigned to good (GSP less than 4.5%),

moderate (GSP between 4.5 and 7.0%), and poor control (GSP

greater than 7.0%) categories based on groupings apparent

from a univariate procedure on the pre-camp GSP values.

Repeated measures ANOVA on the GSP values pre- and post-

camp was conducted using one between subjects factor (GSP

control group) and one within subjects factor (Time: pre-

and post-camp). Results indicated a significant main effect

for Time, F(1,51)=12.45, p < .0009, and a significant main

effect for GSP control group, F(2,51)=55.91, p < .0001.

However, the Time x GSP control group interaction was not

significant, indicating that the various control groups did

not react differentially to the camp experience.

To assess changes in HbAlc, a repeated measures

analysis of variance (ANOVA) was conducted on the HbAlc pre-








camp and at followup using 2 between subjects factors (Age

group and Sex), and 1 within subjects factor (Time). There

were no significant main effects for Time and no significant

interactions, indicating that the camp experience had no

perceivable lasting effect. However, it was hypothesized

that the youngsters may have demonstrated differences

depending on their glycemic level of control before camp.

Therefore, based on a univariate procedure performed on

HbAlc drawn at the beginning of camp, the children were

divided into good (HbAlc less than 8.4%), moderate (HbAlc

between 8.4 and 10.4%), and poor (HbAlc greater than 10.4%)

HbAlc control groups. A repeated measures ANOVA procedure

was conducted on the HbAlc (pre-camp and at followup) using

one between subjects factor (HbAlc control group) and one

within subjects factor (Time). Results demonstrated a

between subjects effect of HbAlc control group, F(2,58)-

=61.75 p < .0001, and a Time x HbAlc control group

interaction F(2,58)=4.81, p < .0117, suggesting that not all

three HbAlc control groups were equally stable on the HbAlc

measure at followup. Therefore, repeated measures ANOVA

procedures were conducted with each of the HbAlc control

groups separately using one within subjects factor (Time:

pre-camp and followup). Results from these analyses

indicated that children in good and poor control remained

stable, while youngsters in moderate control demonstrated a

significant time effect F(1,26)=11.09, p < 0026. Examina-

tion of the mean HbAlc values pre-camp and at followup for






72
each of these groups indicated that while the means for the

good and poor control groups remained stable (7.45 to 7.80%

and 11.36 to 10.72% respectively), the moderate control

group experienced a significant decrease going from 9.51% to

8.80%.

To determine if these changes in HbAlc levels were

associated with camp, a repeated measures ANOVA was con-

ducted on the GSP pre- and post-camp measures using one

between subjects factor (the HbAlc control groups) and one

within subject factor (Time). The result demonstrated a

significant main effect for Time, F(1,51)=8.31, p < .0058,

and a significant between subjects effect of HbAlc control

group, F(2,51)=7.51, p < .0014) but no HbAlc control group x

Time interaction. These results suggest that the improve-

ments noted in the moderate HbAlc control group from pre-

camp to 3 months post-camp did not appear to be the result

of changes associated with the camp experience.



Relationship between Adherence and Diabetic Control

This study explored the relationship between adherence

and diabetic control in two ways. First adherence behaviors

during camp were employed in hierarchical regression models

to predict to post-camp GSP, and adherence behaviors during

the 3 months following camp were used in hierarchical

regression models to predict to the followup HbAlc. Second,

categorical analyses, which involve the division of data into

logical groupings, were utilized to explore possible








associations between adherence and diabetic control for

differing categories of campers (e.g., those in good versus

moderate versus poor diabetic control).

Adherence/GSP Relationships

Hierarchical regression analyses

In order to explore possible relationships between

adherence behaviors and diabetic control as measured by GSP,

it was hypothesized that adherence measures during camp

would predict GSP values post-camp. Since pre- and post-

camp GSP measures were highly correlated (r=.78), pre-camp

GSP was entered first (R2=.69). To determine whether simple

patient characteristics would predict post-camp glycemic

control, age and duration of disease were added to the

model. Since no significant increase in the R2 occurred,

age and duration of disease were dropped from subsequent

models. Since increases in GSP levels as well as increases

in calories consumed were reported for all youngsters during

camp, calories consumed was added next to the model.

However, no improvement in the model's predictive power

occurred. Next, calories consumed before coming to camp was

added to control for pre-camp calorie consumption. The

addition of pre-camp caloric consumption did not enhance the

model's predictive power. Finally, it was hypothesized that

calories consumed during camp may affect post-camp GSP at

varying levels of pre-camp GSP control. Therefore a model

including pre-camp GSP, calories consumed during camp, and

the interaction of pre-camp GSP x calories consumed during








camp was tested. This model did not significantly enhance

the R2 above that offered by the original model using pre-

camp GSP as the sole predictor.variable.

Since the obvious hypothesis that increases in calories

consumed at camp was responsible for the post-camp GSP

increases was not confirmed, further analyses were conducted

to assess whether any of the other adherence measures may

have related to post-camp GSP. To do so, the data were

combined into adherence factors found by Johnson et al.,

(1986). However, Johnson et al.'s factors were based on a

combination of child and parent report whereas the data in

this study were based exclusively on child report. Accord-

ingly, to support the combination of the child report data

into factors, it was necessary to determine whether the

measures (based on child report) within each of the Johnson,

et al. factors were more highly correlated than measures

between factors. The intercorrelation matrix of child

reported pre-camp adherence measures was examined. The

correlations of measures within factors were averaged

(excluding the 1.0s) and compared to the average correlation

of measures between factors. Results indicated that similar

to Johnson et al.'s results, average correlations of

measures within factors were larger (ranging from .22 to

.96) than the average correlations of measures between

factors (ranging from .01 to .21, see Table 14). These

results supported the combination of the measures. There-

fore, measures were standardized to pre-camp measures and











TABLE 14
Correlations Between Adherence Measures
Within and Between Factors


Factor: Exercise Injection Diet Eat/Test Diet
Type Frequency Amount


Exercise .66

Injection .02 .22

Diet Type .12 .01 .96

Eat/Test -.03 .10 .21 .27
Frequency

Diet Amount -.11 -.06 .01 .01 .27






76
five factors were created based on the Johnson et al. (1986)

model. Factor 1 contained the three exercise measures,

Factor 2 was made up of all four injection measures, Factor

3 consisted of measures of diet type (percentage of calori-

es: carbohydrates and percentage calories: fat), Factor 4

contained measures of eating frequency and glucose testing

frequency, and Factor 5 included measures of calories

consumed and concentrated sweets. A factor score was

calculated by averaging the standardized scores of measures

loading highly on that factor. It was expected that the

average correlation between factors would be low. This was

found to be the case (Table 15) with correlations ranging

from r=.01 (between the injection factor and the diet type

factor) to r=.26 (between the diet type factor and the

eating/testing frequency factor). These correlations were

considered to support the relative independence of the

factors and justify their use in further exploration of the

adherence/control issue.

Hierarchical regression analyses were again used to

determine whether adherence behaviors during camp would

predict post-camp GSP. Therefore, each of the five

adherence factors reflecting camp behavior was added

separately to the original model which contained the pre-

camp GSP measure. None contributed significant additional

variance. Next, it was decided to control for adherence

behavior before coming to camp. Therefore, adherence before

camp and adherence during camp were added to the original












TABLE 15
Average Correlations between Adherence Factors


Factor: Exercise Injection Diet Eat/Test Diet
Type Frequency Amount


Exercise 1.00

Injection .03 1.00

Diet Type .14 .01 1.00

Eat/Test -.05 .19 .26 1.00
Frequency

Diet Amount -.16 -.14 -.02 .03 1.00








model. For example, the model controlling for exercise

behavior before coming to camp included pre-camp GSP,

exercise factor (before camp) and exercise factor (during

camp). None of these five models significantly enhanced the

predictive power of the pre-camp GSP prediction model.

Finally, it was hypothesized that insulin dosage during

camp might predict post-camp glycemic control. Repeated

measures ANOVA on the pre- and during-camp insulin dosage

with one within subject factor (Time) indicated that there

was a significant difference between dosage before camp

compared to dosage during camp F(1,57)=6.31, p < .01. The

average dose before camp was 32.2 units while the average

dose during camp was 27.2 units. Adding insulin dosage

during camp to pre-camp GSP did not increase the variance

accounted for by pre-camp GSP alone. A model which included

pre-camp GSP, dosage during camp, and insulin dosage before

camp was also tested. No significant increase in the R2 was

achieved. To determine whether children entering camp at

different levels of glycemic control would be differentially

affected, a model containing pre-camp GSP, insulin dosage

during camp, and an interaction of these two measures was

tested. This model did not strengthen the predictive power

of the original pre-camp GSP prediction model.

Categorical analyses

Categorical analysis involves the division of data into

categories based on logical groupings. In this study, the

data were grouped based on levels of glycemic control and








levels of adherence as well as change in glycemic control

and change in adherence.

Accordingly, it was first hypothesized that children

arriving at camp at different levels of diabetic control

would report different levels of adherence prior to attend-

ing camp. Therefore, based on a univariate procedure on the

pre-camp GSP measure, the youngsters were divided into three

GSP control categories. Pre-camp GSP of less than or equal

to 4.5 was considered good control, greater than 4.5 and

less than or equal to 7.0 was considered moderate control,

and greater than 7.0 was considered poor control. ANOVAs

were conducted on each of the five pre-camp adherence

factors using one between subject factor (GSP control

group). Results indicated no significant differences

between the groups on any of the adherence factors.

Next, it was postulated that the adherence/GSP control

relationship might be detected if youngsters were categori-

zed on their degree of adherence before coming to camp.

Theoretically, those children who had been more adherent

before coming to camp would also exhibit lower GSP values at

the beginning of camp. Therefore, based on the pre-camp

median score for each of the five adherence factors,

youngsters were assigned a score of 1 for being below the

median (adherent) and a score of 0 for being above the

median (nonadherent). Their scores for the five adherence

factors were summed and children were placed into three

adherence categories (0-1 was poor adherence, 2-3 was








moderate adherence, and 4-5 was good adherence). An ANOVA

was performed on the pre-camp GSP measure using one between

subject factor (Adherence group). Results indicated no

significant difference between the adherence groups on the

pre-camp GSP measure.

It was conjectured that an adherence/control relation-

ship might be better detected if the youngsters were divided

into categories of metabolic change. Presumably, those

children who exhibited an improvement in diabetic control

pre- to post-camp, might show an improvement in adherence

behaviors from pre- to during-camp. Such improvement in

adherence behaviors would not be expected for children whose

diabetic control stayed the same or deteriorated over the

camp experience. Consequently, post-camp GSP values were

subtracted from the pre-camp GSP values and based on

univariate analysis of the resulting difference, the

youngsters were divided into children who improved (dif-

ference greater or equal to .5), did not change (difference

less than .5 and greater or equal to -.4), and got worse

(difference less than -.4). A repeated measures ANOVA was

conducted on each of the pre- and during-camp adherence

factors as well as the calories consumed measure using one

between subject factor (GSP difference group) and one within

subject factor (Time). Results of these analyses indicated

only simple Time main effects for the exercise factor

F(1,59)=155.34, p < .0001, the injection factor F(1,55)=








15.51, p < .0002, the eating/testing frequency factor

F(1,59)=39.71, p < .0001, and the calories consumed measure

F(1,58)=10.93, p < .001. There was no Time x GSP difference

group interaction; youngsters who exhibited differential

changes in GSP pre- to post-camp did not exhibit differen-

tial changes in adherence pre- to during-camp.

Finally, it was conjectured that changes in adherence

might clarify the association between adherence and control.

It would seem that children who improved in adherence during

camp would exhibit a concomitant improvement in GSP control.

while these who did not improve were expected to exhibit

stable GSP values pre- to post-camp. Therefore, the

children were divided into change groups. Post-camp

adherence factors were subtracted from pre-camp adherence

factors to produce two groups (those who improved in

adherence and those who did not improve in adherence).

Repeated measures ANOVA was then performed on the GSP

measures using one within subject factor (Time, pre- and

post-camp) and one between subject factor (Change group).

Results revealed a simple Time main effect. The interaction

term was nonsignificant.



Adherence/HbAlc Relationships

Hierarchical regression analyses

To assess possible relationships between adherence and

glycemic control after camp, it was hypothesized that

adherence behaviors after camp would predict the followup








HbAlc. Since pre-camp and followup HbAlc values were

moderately correlated (r=.62), the pre-camp HbAlc was

entered into the hierarchical regression model first

(R2=.39). To determine whether simple patient charac-

teristics would enhance prediction of followup HbAlc, age

was added to this model. Age significantly increased the R2

to .47 and was retained in all subsequent models. Next,

duration of disease was added to this model, but did not

enhance the predictive power of the model and was dropped

from all subsequent analyses. Further analyses were

conducted to assess whether any of the adherence factors

were related to followup glycemic control. To do so, the

post-camp adherence measures were standardized to the pre-

camp adherence measures and combined into the five adherence

factors as previously described. Each of the five post-

camp adherence factors was then added to the model in five

separate regressions. None of the factors contributed sig-

nificant additional variance. It was conjectured that

controlling for adherence before camp might enhance the

predictive power of the model. Therefore, each adherence

factor pre-camp was added to the model containing pre-camp

HbAlc, age, and the same post-camp adherence factor. The

addition of the pre-camp adherence factor did not sig-

nificantly increase the R2 in any of the five models tested.

It was then postulated that the adherence factors post-camp

might predict post-camp HbAlc for various levels of glycemic

control at the outset of the study. Therefore, an








interaction term (adherence factor x pre-camp HbAlc) was

tested for each of the five adherence factors. Only the

model that included pre-camp HbAlc, age, post-camp injection

factor, and the pre-camp HbAlc x post-camp injection factor

interaction term proved to increase the R2 significantly

(R2=.51 see Table 16). The interaction between the pre-camp

HbAlc and injection was interpreted by calculating the

nonstandardized Beta weights for injection for varying pre-

camp HbAlc levels from 5.8 to 13.1 (the ranges of pre-camp

HbAlc found in this study's sample) using the equation from

Table 16 (as suggested by Cohen and Cohen, 1983). As is

apparent from Table 17, injection behaviors in children with

low pre-camp HbAlc (e.g., 5.8 to 7.0) was negatively

associated with post-camp HbAlc: less compliant behavior

(higher injection scores) was associated with lower post-

camp HbAlc (i.e., better diabetic control). At pre-camp

HbAlc levels between 8.0 and 9.0 the relationship between

injection behavior and post-camp HbAlc diminishes to zero.

For pre-camp HbAlc levels between 11.0 and 13.1 (poor

control) there was a positive association between injection

behaviors and post-camp diabetic control, i.e., less

adherent behaviors were associated with poorer control.

Table 18 depicts the characteristics of three groups divided

on the basis of the injection factor Beta weights depicted

in Table 17. Noteworthy was a relatively large increase in

insulin dosage for both groups where the Injection Beta

weights were large (good and poor control youngsters). Due










TABLE 16
Predicting Post-camp HbA1c by Pre-camp HbAlc,
Age, Post-camp Injection Factor,
and Pre-camp HbAlc x Post-camp Injection Interaction


Variables in model Beta weights p< R2 = .51


Pre-camp HbAlc .47445345 .0017

Age .28755174 .0142

Post-camp -4.07128891 .0510
Injection Factor

Pre-camp HbAlc x .45447440 .0487
Post-camp
Injection Factor












TABLE 17
Predicting Post-camp HbAlc:
Nonstandardized Injection
Beta Weights at Varying Levels
of Pre-camp HbAlc


Levels of
Pre-camp HbAlc


Injection Factor (Post-camp)
Beta Weights


5.8
7.0
8.0
9.0
10.0
11.0
12.0
13.1


-1.435
- .890
- .435
- .021
.473
.928
1.382
1.882


* Y= 1.33 + .47 (Pre-camp HbAlc) + .29 (Age)
4.07 (Post-camp Injection factor)
+ .45 (Pre-camp HbAlc x Post-camp Injection factor)











TABLE 18
Pre-camp HbAlc x Post-camp Injection Factor:
Descriptive Characteristics


Pre-camp HbAlc 5.8-7.0 8.0-10.0 11.0-13.0
Injection B Wts -1.4 to -.9 -.4 to .5 .9 to 1.9


N 8 37 5

Pre-camp HbAlc 6.5 8.9 11.6

Post-camp HbAlc 7.4 8.5 11.0

Age 9.9 10.5 11.2

Duration of disease 3.8 4.0 3.5

Injection factor .10 .01 .18
pre-camp

Injection factor .20 .34 .59
post-camp

Insulin dose 25.9 30.9 33.0
pre-camp

Insulin dose 21.8 27.9 29.4
during camp

Insulin dose 34.5 29.5 40.7
at followup

Change in insulin 8.6 1.4 7.7
dose between pre-
camp and followup








to this finding it seemed appropriate to consider whether

insulin dosage and change in insulin dose would enhance the

predictive power of the model that contained pre-camp HbAlc

and age. Insulin dosage at camp did not increase the R2.

To control for pre-camp insulin dose, dosage before camp was

added to the model with no increase in R2. Keeping in mind

the role of injection which had earlier been established, it

was added to the model. No significant increase of R2

resulted. It was then hypothesized that the changes in

injection behavior at various changes in insulin dose from

pre-camp to followup might differentially predict post-camp

HbAlc. Change in insulin dosage was calculated by sub-

tracting the total insulin dosage reported by the child on

the last interview (3 months post-camp) from total dosage

reported by the child during the last interview before

coming to camp. A model including pre-camp HbAlc, age, dif-

ference in dosage, injection factor, and difference in

dosage x injection factor interaction term was tested. This

model significantly increased the variance, R2=.57 (see

Table 19). This interaction was interpreted by calculating

the nonstandardized Beta weights for injection for various

units of change in dosage from -40 to +40 (the range of

total units change found in this study's sample) based on

the equation from Table 19. As is apparent from Table 20,

the greater the decrease in insulin dose, the higher the

injection factor Beta weight. At the insulin dose change of

-5 to +10 the relationship between injection and post-camp











TABLE 19
Predicting Post-camp HbAlc by
Pre-camp HbAlc, Age, Insulin Dose Change,
Post-camp Injection Factor, and
Insulin Dose Change x Post-camp Injection Interaction


Variables in model Beta weights p< R2 = .57


Pre-camp HbAlc .63400558 .0001

Age .24956679 .0321

Change in insulin .02784116 .0212
dose from pre-camp
to followup

Post-camp .21885475 .4495
Injection factor

Change in insulin x .06829848 .0065
Post-camp
Injection factor











TABLE 20
Predicting Post-camp HbAlc:
Nonstandardized Injection
Beta Weights at Varying Levels
of Change in Insulin Dose from
Pre-camp to 3 Months Post-camp


Levels of Change in Injection Factor (Post-camp)
Insulin Dose from Beta Weights
Pre-camp to Followup


-40 2.951
-30 2.268
-20 1.585
-10 .901
- 5 .560
0 .219
5 .123
10 .464
15 .806
20 1.147
30 1.830
40 2.513


* Y= .26 + .63 (Pre-camp HbAlc) + .25 (Age) + .03 (Insulin
Dose Change) + .22 (Post-camp Injection Factor)
.07 (Insulin Dose Change x Post-camp Injection Factor)








HbAlc approached zero. Increases in insulin dose were as-

sociated negatively with the injection factor Beta weight.

Table 21 depicts the characteristics of groups divided based

on the Beta weights for injection at varying levels of

change in insulin dose (Group 1 decreased insulin, Group 2

stayed the same, and Group 3 increased insulin dose). A

number of characteristics are noteworthy for the group of

children who reported the largest increases in insulin dose

during camp: children in this group were in the best HbAlc

control before camp; this group was the youngest of the

three groups; children in this group reported the greatest

post-camp compliance; this group was the only one that ex-

perienced an increase in HbAlc while the other two groups

demonstrated slight improvement in glycemic control at

followup.

The other four adherence factors were considered in

similar models to determine if their interaction with

insulin dose change would offer similar predictive power.

This was not the case.


Categorical analyses

Based on a univariate procedure, the pre-camp HbAlc

values were divided into control categories (less than

8.4=good, greater than 8.4 and less than 10.4=moderate, and

greater than 10.4=poor control). ANOVAs were conducted on

each of the five adherence factors using one between subject

factor (HbAlc control group). A group effect was detected

for the injection factor with the differences between the











TABLE 21
Pre- Post-camp Change in Insulin Dose x Post-camp
Injection Factor: Descriptive Characteristics


Change in Dose(CH) Ch <= -10 -10 < Ch < 15 Ch => 15
Injection B Wts 3.0 to .9 .6 to -.5 -.8 to -.2.5


N 13 28 9

Pre-camp HbAlc 9.3 8.7 8.1

Post-camp HbAlc 9.0 8.5 8.5

Age 10.6 10.7 9.9

Duration of disease 3.5 4.2 4.5

Injection factor .14 .18 .14
pre-camp

Injection factor .40 .47 .21
post-camp

Insulin dose 44.1 29.6 13.3
pre-camp

Insulin dose 25.5 30.3 24.4
during camp

Insulin dose 21.2 32.4 42.6
at followup

Change in insulin -23.7 2.8 29.3
dose between pre-
camp and followup








good and poor control groups F(2,57)=3.75, p < .03.

However, the group effect was in the opposite direction of

what was expected. That is, the good control group was

actually less adherent on the injection factor (.28) than

the poor control group (-.24).

The youngsters were then divided into control groups

based on a univariate procedure on the followup HbAlc (less

than 8.4=good, greater than 8.4 and less than or equal to

10.4 moderate, and greater than 10.4=poor). ANOVAS on the

adherence post-camp adherence factors using one between

subject factor (followup HbAlc control group) revealed no

differences. In other words, the pre-camp findings were not

replicated with the post-camp data.

Next, the adherence/control relationship was examined

by assessing whether control was associated with degrees of

adherence. Based on the median score on each of the five

adherence factors, youngsters were assigned a score of 1 for

being below the median of the pre-camp adherence factor

(adherent) and a score of 0 for being above the median of

the pre-camp adherence factor (nonadherent). Their scores

for the five adherence factors were summed and children were

paced into three adherence categories (0-1 was poor ad-

herence, 2-3 was moderate adherence, and 4-5 was good

adherence). Separate ANOVA procedures using one between

subject factor (Adherence group) on the pre-camp and the

followup HbAlc measures revealed that these groups did not




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EDIID8NTB_6IJGAN INGEST_TIME 2012-12-07T18:23:23Z PACKAGE AA00012894_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES