Effect of presentation format and number of blood glucose values on diabetologists' insulin dose recommendations

MISSING IMAGE

Material Information

Title:
Effect of presentation format and number of blood glucose values on diabetologists' insulin dose recommendations
Physical Description:
vii, 127 leaves : ill. ; 29 cm.
Language:
English
Creator:
Nurick, Michael Alan, 1961-
Publication Date:

Subjects

Subjects / Keywords:
Insulin -- administration & dosage   ( mesh )
Diabetes Mellitus, Type I -- drug therapy   ( mesh )
Decision Making   ( mesh )
Patient Care Team   ( mesh )
Department of Clinical and Health Psychology thesis Ph.D   ( mesh )
Dissertations, Academic -- Clinical and Health Psychology -- UF   ( mesh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1992.
Bibliography:
Includes bibliographical references (leaves 122-126).
Statement of Responsibility:
by Michael Alan Nurick.
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:
aleph - 001751963
notis - AJG4906
oclc - 27961521
System ID:
AA00009059:00001


This item is only available as the following downloads:


Full Text














EFFECT OF PRESENTATION FORMAT AND NUMBER OF BLOOD GLUCOSE
VALUES ON DIABETOLOGISTS' INSULIN DOSE RECOMMENDATIONS





By

MICHAEL ALAN NURICK


A 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


1992












ACKNOWLEDGMENTS


My utmost appreciation and love go to my parents, Ruth

and Mason Nurick. Through every facet of my life they have

been available to me with their emotional, as well as

financial, support. Through their encouragement, humor, and

patience they have allowed me to pursue this sometimes

frustrating educational endeavor.

I would like to thank Suzanne Johnson for all of her

support and assistance. She provided me with a 'stable' base,

which I felt comfortable to fall back on in times of

distress. Through Suzanne's patience, humor, clarity of

thought, and remarkable knack for being able to clarify my

sometimes muddled thoughts and writings, I was able to

persevere with my dissertation, as well as with all other

phases of my clinical training (therapy cases, classes,

qualifying exams, and internship).

Special thanks go to the 18 diabetologists who served as

the participants for this project. Their willingness to

complete the profiles made this investigation possible.

A very appreciative thank you goes to Becki Bucher for

her statistical assistance and willingness to curtail her

busy schedule to accommodate my needs. I would also like to

thank the Ames company for donation of the "Glucofacts Data










Management System" statistical software used to generate the

computer formats.

I would also like to thank my committee members, Sheila

Eyberg, Steve Boggs, Gary Geffken, and Bill Riley, for their

assistance. Sincere thanks go to Gary for his years of

support: his humor, insights, and willingness to serve as a

sounding board for my occasional tantrums.

Personal thanks go to my brother, Mitch Nurick, and many

friends and relatives who helped me through their love,

support, and time to rejuvenate me so I could continue

working on this project.











TABLE OF CONTENTS


page
ACKNOWLEDGMENTS ........................................... ii

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

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

Relationship of BG Self-Monitoring to Glycemic
Control ........ ...... .. .. ................... .... 4
Uncontrolled Studies ................................. 6
SMBG Was Not Associated with Improved Glycemic
Control .......................................... 6
SMBG Was Associated with Improved Glycemic Control.. 7
Controlled Studies ................................ 11
SMBG, Patient Compliance, Accuracy, and Diabetes
Control ................ ... ................ ... 18
SMBG and Regimen Adjustment ........................ 27
Patient Self-Adjustment of Insulin Dose ............. 28
Diabetologist Adjustment of Insulin Dose ............. 29
Current Investigation ................................ 33

METHODS.............. ............ .................. ..... 37

Participants ...o....................... .............. 37
Measures ................................. ..... ...... 39
Procedure ............................................ 41


RESULTS........... *............... ................. ....... 43

Overview .... ................. .......... .... ........ 43
Effect of Format and Number of BG Values on the
Decision to Change an Insulin Dose .............. 48
Effect of Format, Number of BG Values, and
Diabetologist Characteristics on the Decision to
Change an Insulin Dose .......................... 48
Effect of Format and Number of BG Values on the
Magnitude of Insulin Dose Change ................ 51
Effect of Format, Number of BG Values, and
Diabetologist Characteristics on the Magnitude of
Insulin Dose Change ..................... ....... 55
Effect of Format, Number of BG Values, and the
Mean Glycemic Control of the BG Profiles on the
Decision to Change an Insulin Dose .............. 62









Effect of Format, Number of BG Values, and Mean
Glycemic Control of the BG Profiles on the
Magnitude of Insulin Dose Change ................ 66
Effect of Format, Number of BG Values, and Mean
Glycemic Control of the BG Profiles on the
Magnitude of the Type and Time of Insulin Dose
Change ..................................... 66
Diabetologist Agreement............................. 103
Time to Complete the Task ......................... 103

DISCUSSION ................................................ 107

APPENDIX A--ALGORITHM FOR INSULIN DOSE ADJUSTMENT......... 118

REFERENCESI...... ........................................ 122

BIOGRAPHICAL SKETCH....................................... 127













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


EFFECT OF PRESENTATION FORMAT AND NUMBER OF BLOOD GLUCOSE
VALUES ON DIABETOLOGISTS' INSULIN DOSE RECOMMENDATIONS


By

Michael Alan Nurick


August 1992

Chairperson: Dr. Suzanne Bennett Johnson
Major Department: Clinical and Health Psychology


Eighteen diabetologists (8 physicians, 8 nurses, 1

physician assistant, and 1 dietician) reviewed 15-day records

of blood glucose values obtained from actual IDDM patients.

The records differed in format (logbook versus computer) and

number of blood glucose values (1 versus 3 values per day).

After reviewing each record, the diabetologist indicated

whether the patient's insulin dose should be changed and if

so, by how much. The format of presentation and number of BG

values were related to both the decision to change a dose and

the magnitude of change. When information was presented in

computer format versus logbook, more diabetologists

recommended a dose change and the magnitude of the change was










larger. The same pattern of findings occurred when

diabetologists were presented with three versus one blood

glucose value per day. No diabetologist characteristic was

related to the decision to change insulin doses. However, the

15-day record's mean blood glucose was associated with both

the decision to change the insulin dose and the magnitude of

change. More diabetologists recommended a dose change when

presented with records with high mean blood glucose compared

to records with low mean blood glucose; the magnitude of

recommended dose change was also greater. The

diabetologists' recommended greater change in long-acting

insulin than short-acting insulin, and greater change in the

morning insulin dose compared to the evening dose. However,

as a group, the diabetologists exhibited relatively poor

agreement as to whether to change an insulin dose. Among

those who recommended a change, there was poor agreement as

to what new dose should be prescribed.


vii
















INTRODUCTION


Insulin dependent diabetes mellitus (IDDM) is a chronic

disease that usually commences during childhood. It is

characterized by an inability of the pancreas to produce

endogenous insulin, resulting in inadequate utilization of

carbohydrates, fats, and proteins. Management of the disease

involves regulation of diet, exercise and insulin so as to

achieve as near to normal blood glucose (BG) levels as

possible (Edwards & Yates, 1985). It is believed that

improved metabolic control may prevent or forestall the long-

term complications of diabetes, including peripheral

neuropathy, nephropathy, and retinopathy (Skyler, 1979).

This assumption has yet to be empirically validated, although

longitudinal studies are currently underway through the

nationwide Diabetes Control and Complications Trial (DCCT,

1987).

Glycosylated hemoglobin (HbAlc and HbA1) levels are used

to assess the degree of metabolic control over periods of 2

to 4 months (Ziel & Davidson, 1987). Normal hemoglobin A

will combine with glucose to form hemoglobin Al (which is

comprised of hemoglobin Ala, Alb, and Alc). The proportion of

1










hemoglobin A that becomes glycosylated depends on the amount

of glucose in circulation in the blood. This binding of

glucose is irreversible and will last the life of the red

blood cell (90-120 days), and thus the measurement of HbA1 or

HbAlc reflects the level of blood glucose over time.

Glycosylated hemoglobin levels increase with higher glucose

intake, but do not decrease with lowered glucose intake

because of the irreversible nature of the compound. The

proportion of Alc hemoglobin in normal hemoglobin is about 4

to 8% (normal HbA1 is about 6 to 12%). In patients with

chronic hyperglycemia, the percentage of Alc hemoglobin is 10

to 15% or higher, while chronic hyperglycemia is reflected by

HbA1 values above 15% (Carney, Schechter, & Davis, 1983).

These ranges change slightly depending on the methodologies

used to establish glycemic values. Although diabetologists

generally consider normal BG (euglycemia) to be beneficial,

the long-term benefits of euglycemia have not been

empirically established. The acute deleterious effects of

both severe hypoglycemia (impaired cognitive functioning,

disorientation, confusion, and possible death, Cryer &

Gerich, 1985; Rovet, Ehrlich, & Hoppe, 1987; and Ryan, Vega,

& Drash, 1985) and hyperglycemia ketosiss with vomiting,

dehydration, and possible death) are well known and

empirically established. Although the literature indicated

that extreme BG excursions in either direction can have










serious physical as well as cognitive effects, there is no

agreed upon consensus as to what exactly the appropriate BG

target level should be.

Proponents of strict or "tight" BG control strive for BG

values between 80 mg/dl and 120 mg/dl, which is the usual BG

range in non-IDDM (normal) individuals. A strict regimen is

recommended, involving frequent measurement of BG, adherence

to a particular diet with modifications based upon resulting

BG values, a regular exercise regimen, and multiple daily

insulin injections or subcutaneous insulin infusion. In the

pursuit of tight BG control, many studies have noted that

hypoglycemic reactions are to be expected since exogenous

administration of insulin cannot mimic normal endogenous

release (Cryer & Gerich, 1985; Goldgewicht, Slama, Papoz, &

Tchobroutsky, 1983; and Miles & Jensen, 1986). Nevertheless,

proponents of strict BG control assert that maintenance of a

euglycemic profile is likely to prevent or delay the chronic

complications of IDDM (Leslie & Sperling, 1986).

Not all diabetologists favor tight BG control, since the

strict regimen required is so difficult and demanding,

especially for children and adolescents. Drash (1976)

advocated the following therapeutic objectives for children

with IDDM: (1) control of the symptoms of IDDM (including

polyuria, polydypsia, polyphagia, and avoidance of

hypoglycemia); (2) maintenance of normal growth and









maturation; (3) maintenance of normal blood lipids (including

cholesterol and triglyceride levels); (4) minimization of

urinary glucose loss; and (5) BG levels within the range of

80 to 180 mg/dl.

Whether or not a physician holds to a "tight" or "loose"

glycemic goal may be a function of the type of patients they

see. For example, Marteau and Baum (1984) studied

pediatricians and adult physicians with a particular interest

in diabetes. Pediatricians were more likely than adult

physicians to favor patient BG profiles that tended toward

hyperglycemia, although both groups of physicians held

similar views on the seriousness of diabetes as a disease

(ranking it second in seriousness only to tumors). However,

physicians treating adults estimated more complications and

early death due to diabetes than did the pediatricians. The

authors suggest that the adult physicians may have had more

exposure to the long-term complications of diabetes than did

the pediatricians and consequently may perceive a higher risk

of complications and death from this disease.


Relationship of BG Self-Monitoring to Glycemic Control


Many physicians recommend self-monitoring of blood

glucose (SMBG), with the assumption that the feedback from

SMBG at variable times during the day will improve glycemic







5

control. A number of studies have assessed the impact of

SMBG on glycemic control.

The first section of studies reviewed did not use a

control group design (Wing, Lamparski, Zaslow, Betschart,

Siminerio, & Becker, 1985; Hirsch, Matthews, Rawlings,

Broughton, Breyfogle, Simonds, Kossoy, England, Wiedmeyer,

Little, & Goldstein, 1983; Belmonte, Schiffrin, Dufresne,

Suissa, Goldman, & Polychronakos, 1988; Geffner, Kaplan,

Lippe, & Scott, 1983; Dupuis, Jones, & Peterson, 1980; Wilson

& Endres, 1986; and Marrero, Kronz, Golden, Wright, Orr, &

Fineberg, 1989). Four of these studies found an association

between SMBG and improved glycemic control (Geffner et al.

1983; Dupuis et al. 1980; Wilson & Endres, 1986; and Marrero

et al. 1989). The remaining studies found no such

association.

The second section of studies did use a control group

design. All of these studies found no association between

SMBG and improved glycemic control (Daneman, Siminerio,

Transue, Betschart, Drash, & Becker, 1985; Mann, Noronha, &

Johnston, 1984; Worth, Home, Johnston, Anderson, Ashworth,

Burrin, Appleton, Bender, & Alberti, 1982; Miller, Stratton,

& Tripp, 1983; Gonder-Frederick, Julian, Cox, Clarke, &

Carter, 1988; and Mazze, Pasmantier, Murphy, & Shamoon,

1985).










Uncontrolled Studies


SMBG Was Not Associated with Improved Glycemic Control


Wing, Lamparski, Zaslow, Betschart, Siminerio, and

Becker (1985) instructed children and adolescents, who were
generally in adequate glycemic control (mean HbA1=11.93 +

.13% at study initiation), to complete a questionnaire

regarding SMBG practice and to alter their insulin dose

and/or food intake according to their BG readings. The

authors found no improvement in glycemic control as a

function of frequency of SMBG per day.

Hirsch, Matthews, Rawlings, Broughton, Breyfogle,

Simonds, Kossoy, England, Wiedmeyer, Little, and Goldstein

(1983) investigated a group of children and adolescents in

adequate glycemic control (mean HbAlc=8.8% at study

initiation) who had expressed a desire to SMBG at home. They

subsequently underwent intensive individual instruction and

reinforcement for SMBG and were instructed to make insulin

dose changes based upon their BG test results. This study

found no significant improvement in glycemic control after

one year of study participation.

Belmonte, Schiffrin, Dufresne, Suissa, Goldman, and

Polychronako (1988) also used a group of children and

adolescents in generally adequate glycemic control (mean









HbAl=11.4 + 1.8% at study initiation). They were instructed

to SMBG two to three times daily and to adjust their short-

acting insulin using an algorithm and their BG test results.

The study duration was three years; no significant

differences in glycemic control occurred.

These three uncontrolled studies indicated that the

addition of SMBG to a person's diabetes regimen did not

improve glycemic control, even when coupled with insulin dose

changes based upon the resulting BG values.


SMBG Was Associated with Improved Glycemic Levels



Geffner, Kaplan, Lippe, and Scott (1983) reported

improvement in glycemic control in a very poorly controlled

group of youngsters with IDDM. The study involved changing

the number of injections per day from one to two or three,

along with SMBG up to six times per day, and intensive

changes in insulin dose using algorithms based upon the

resulting BG values. The improvement in glycemic control

could be explained by a number of factors other than SMBG.

The patient population was so poorly controlled (mean
HbA1=14.63 + .42%) that without a no-treatment control group,

the results could be explained by regression to the mean. In

addition, the change in the patients' insulin regimen was










much more intense than in the other studies; this alone could

have accounted for the improved glycemic control.

Dupuis, Jones, and Peterson (1980) also found improved

glycemic control in a very poorly controlled group of young

adults with IDDM (all participants had experienced ketotic

episodes, four patients had proliferative retinopathy, and

eight patients had evidence of peripheral neuropathy, and the

sample's mean HbAlc was 10.3% with great variability from

6.5% to 15.0%). Patients were encouraged to SMBG pre- and

postprandial for all meals. The patients were also

administered a psychiatric interview at study initiation and

termination. Results indicated significant improvement in

HbAlc levels as well as marked improvements in depression

scores. It is not clear if the significantly improved

glycemic control was due to improvement in the worst

controlled patients, or if all the patients improved. The

authors speculated that SMBG may have alleviated feelings of

depression because it gave patients an active role in their

diabetes management as well as allowing them to control the

fluctuation of BG values.

Wilson and Endres (1986) also found a positive

relationship between SMBG and improved glycemic control. In

this study, adolescents with adequately controlled diabetes
(mean HbAl=11.3 + 2.7% at study initiation) were requested to

SMBG four times a day and record their results in a logbook.










During the first 6-week phase of the study, unbeknownest to

the patients, the BG results were stored in the BG monitor's

memory. During the second 6-week phase, the patients were

given the same BG monitor that now appeared different (due to

removal of a metal strip) and were told of the meter's memory

ability. They were again requested to SMBG four times a day

and record their results in a logbook. Analysis of HbA1

values from study initiation to study termination indicated a

significant improvement in HbA1 values (mean HbA1 at study

termination=10.5 + 2.6%, p=.03). The reason for the improved

glycemic control was unclear; there was no mention of any

insulin dose adjustments, nor did the authors speculate as to

why the patient's glycemic control changed.

Although these three studies did find improvement in

glycemic control, it is unclear what may have accounted for

this improvement. The Geffner et al. (1983) and Dupuis et

al. (1980) patient populations were in worse glycemic control

than participants in the studies described previously.

Perhaps poorly controlled patients benefit more from such a

procedure. Other factors that may have improved glycemic

control include intensive changes in insulin dose in the

Geffner et al. (1983) study, or alleviation of depression in

the Dupuis et al. (1980) study.

A study by Marrero, Kronz, Golden, Wright, Orr, and

Fineberg (1989) evaluated the effectiveness of SMBG in two










groups of adolescent patients. One group, the "experimental"

group, used a BG meter with memory capabilities (Glucometer

M). The "control" group used a meter without memory. Both

groups engaged in SMBG and acted upon this information with

the advice of their physician in order to improve glycemic

control. Both groups were in good metabolic control at

initiation of the study (HbA1=10.3 + 1.8% vs 10.8 + 2.0%,

respectively) and both were trained in the use of their

respective meters. The duration of the study was 4 months,

the patients returned for clinic visits at 2-month intervals

when their physician reviewed their SMBG data, and possible

solutions to any identified problems were discussed. Both

groups showed a significant (p<.001) decrease in HbA1 levels

from initiation of the study to its termination four months

later (HbA1=9.3 + 1.7% vs 9.8 + 1.7%, respectively, at

termination). However, there were no between group

differences. The authors noted that it was not clear why

both groups' HbAl values improved during the study period.

Although physicians spent a greater amount of time discussing

SMBG results with the experimental group, the experimental

group did not exhibit a greater decrease in HbA1 levels

compared to the control group.

In order to better investigate the association between

SMBG and glycemic control, studies incorporating a no

treatment control group design were conducted.












Controlled Studies


A loose definition of "control" was used in selecting

studies for review. Included in this section are no-

treatment control studies, as well as studies that compared

SMBG to either no testing or urine testing, studies using

crossover designs, and studies using pre-post designs. This

section represents the great majority of studies assessing

the relationship between SMBG and glycemic control. The

reader should note that in these "controlled" studies, the

term "control" has two different meanings. It may refer to

design methodology or the regulation of glycemic metabolism.

The first study is prototypical and is presented in

detail. Daneman, Siminerio, Transue, Betschart, Drash, and

Becker (1985) assessed the impact of SMBG on glycemic control

in 16 children with IDDM (mean age=i3.1 years, range=9.8 to

17.2 years; mean duration of IDDM=4.1 years, range=l to 14

years; 11 girls, 5 boys) who had no prior SMBG experience but

had received extensive diabetes education. The children were

divided into two groups and underwent a 26-week double-

crossover design comparing glycemic control with and without

the addition of SMBG to their usual management regimen. The

two groups were similar in age, duration of IDDM, and

percentage of glycosylated hemoglobin (10.5 + .6% and 9.5 +










.3% in Groups 1 and 2, respectively). The study design was

as follows: Group 1 (n=9), period 1: (weeks 1-13) urine

testing plus SMBG; period 2: (weeks 14-26) urine testing

only. Group 2 (n=7), period 1: urine testing only; period 2:

urine testing plus SMBG. The BG target during the SMBG

period was between 80 and 180 mg/dl, and the children and

their families were trained in reading the test strips

accurately. The families were encouraged to make insulin

dosage adjustments based upon BG and/or urine test results,

whereby increments or decrements of 10% of the daily insulin

dosage were advised except when severe hypoglycemia occurred,

at which time a 20% decrease was advised. The families were

to adjust insulin once or twice weekly depending on resulting

glucose values from the previous 3-7 days. Frequent phone

contact was implemented during the experimental periods (6-12

calls per period) in order to optimize dosage adjustments.

Glycemic control was assessed by measures of HbAlc at

initiation of the study and at the end of each experimental

period. Weekly glycemic control was assessed by calculating

mean BG during the periods of SMBG. Results of the study

showed that HbAlc levels did not improve with the

introduction of SMBG between the groups, nor within either

group. No significant decrease in mean BG occurred for

either group, although the authors noted a trend toward

significance in both groups (no p value was provided); Group










1 mean BG during the first week=201 + 18 mg/dl and 182 + 23

mg/dl during the last week of period 1; while Group 2 had a

mean BG of 166 + 7 mg/dl and 133 + 10 mg/dl during the first

and last weeks, respectively, during period 2. Although a

significant improvement in glycemic control was not

illustrated in this study, the authors stated that use of

SMBG confirmed symptoms of hypoglycemia in all children, and

allowed them to detect asymptomatic hypoglycemia (BG <40

mg/dl) in 11 children. In addition, 69% of the children

preferred SMBG to urine testing and no complications of SMBG

were noted. The authors concluded that SMBG is an acceptable

part of routine diabetes care in children. Although, SMBG did

not lead to improved metabolic control. There are a number of

possible explanations as to why glycemic control did not

improve. The study participants' initial HbAlc levels were

relatively good (10.5% and 9.5% for groups 1 and 2,

respectively), making any significant change in glycemic

control difficult. Although the patients were encouraged to

adjust their insulin dose based upon the SMBG results and to

phone their diabetologist for assistance in dose adjustments,

the study did not assess whether this actually occurred for

each patient. The authors did report that insulin dose

changes ranged from -30% to +57% over the course of the

study. The small sample size (n=16) also made detection of

differences difficult, and the duration of the study may not










have been long enough to produce a significant change in

glycemic control.

Mann, Noronha, and Johnston (1984) investigated the

possible benefits of long-term SMBG in children with poorly

controlled IDDM. The design incorporated an 18-month

prospective time frame using two groups of children. One

group (mean HbAlc values at study initiation=12.7 + 2.0%)

received intensive diabetes education, while the second group

(mean HbAlc values at study initiation=14.1 + 1.3%) also

received diabetes education in addition to being requested to

SMBG 2 days a week, four times daily as well as extra

postprandial tests during periods of BG instability. This

second group was also requested to adjust their insulin dose

using an algorithm developed by Skyler, Skyler, Seigler, and

O'Sullivan (1981) [see Appendix A]. Mean HbAlc values for

both groups were not significantly different from each other

at baseline or at the 18-month termination point. Even

though the group that was requested to SMBG was in extremely

poor glycemic control, their HbAlc values did not improve

with the addition of SMBG as well as an algorithm for insulin

dose adjustment. Perhaps SMBG only 2 days per week was

insufficient to have a meaningful impact.

Two other controlled studies compared SMBG to urine

testing for glucose; their results duplicated the Daneman et

al. (1985) and Mann et al. (1984) findings. Worth, Home,









Johnston, Anderson, Ashworth, Burrin, Appleton, Binder, and

Alberti (1982) requested 46 young adults with IDDM to urine

glucose test for 6 months during which time they received

diabetes education. Thirty-eight of the 46 patients then

proceeded to a 9-month cross-over study in which they

continued to urine test daily plus were requested to SMBG 2

days per week, 7 times each day and were told to adjust their

insulin dose using Skyler's algorithm (Skyler, 1981).

Results indicated no improvement in glycemic control during

the 9-month SMBG period as compared to the previous 6-month

urine testing only period. Like Mann et al. (1984), SMBG

twice a week may not have been sufficient to produce a

glycemic change.

A similar study was conducted by Miller, Stratton, and

Tripp (1983) using 19 children with IDDM. Self-monitoring of

blood glucose was compared to urine glucose testing. The

patients were randomly assigned to an initial period of 5

months' blood or urine testing, followed by a similar period

using the alternative form of testing. Urine testing was

conducted twice daily, while SMBG was conducted twice daily

before meals. Patients were seen monthly throughout the

study in an outpatient clinic, at school, or in their homes.

Patients were not instructed to change their insulin dose but

physicians did make changes in dosages and types of insulin

based on the previous month's glycemic pattern. Although the










patients in this study engaged in SMBG twice daily and the

duration of the study was 5 months, the results indicated no

significant difference in blood glucose values at the end of

the study between the SMBG patients (mean BG=167 mg/dl) and

those who urine tested only (mean BG=199 mg/dl; p=NS).

Gonder-Frederick, Julian, Cox, Clarke, and Carter (1988)

used a within-subject design with 30 young IDDM adults. The

patients were given BG monitors to use for 2 weeks; they did

not know the monitors had memory capabilities. They were

told that the purpose of the research project was to gather

information concerning their daily BG fluctuations. They

were told not to change their usual self-measurement

schedules but to continue their usual routine, recording all

test results in BG logbooks. Results of the study indicated

no relationship between frequency of SMBG and glycemic

control. Since the patients were not told to adjust their

insulin dosage, nor did their physicians adjust their dosage,

this may have been the reason for the lack of relationship

between SMBG frequency and glycemic control.

Mazze, Pasmantier, Murphy, and Shamoon (1985) studied 20

young IDDM adults all of whom were receiving two or more

injections of insulin per day. As part of their standard

treatment regimen, before the study began, the patients were

instructed to record BG values, insulin type and dose, diet,

exercise, and any symptoms of hypo- or hyperglycemia in a










logbook. At the initiation of the study the patients were

told to maintain their current SMBG regimen but were

instructed in the use of the memory meter if they had not

already been familiar with its use, and were told about the

meter's memory capability. The study duration was 6 weeks,

and at three 2-week intervals the patients returned to the

clinic with their logbooks and memory meters so the meter

data could be offloaded onto a computer. No feedback

concerning the consistency between their logbooks' and memory

meter data was given. These 20 patients' data were then

contrasted with an earlier group of patients who were using

the same memory meter, but were unaware of or "blind" to its

memory capabilities. Of the 20 patients participating in the

current study, 13 also took part in the earlier 2-week study.

In order to assess any change in glycemic control, Mazze et

al. (1985) calculated mean BG values from the patients'

meters for the three 2-week clinic visit intervals that

comprised the study period. The values obtained (162 mg/dl,

168 mg/dl, and 164 mg/dl, respectively) were not

significantly different from each other nor from the mean BG

level (165 mg/dl) for the entire study period. The authors

also compared the mean BG levels of the 13 patients who

participated in the earlier "blind" study to their mean BG

during the current study. Although an overall decrease of

mean BG from 179 + 16 mg/dl to 167 16 mg/dl occurred, this










decrease was not significant. Further, no significant

pattern of improvement in daily mean BG levels or the

variance in BG was seen as the study progressed. The authors

also examined the timing and dosage of insulin administration

and found no significant alterations in these parameters.

Despite the study's negative findings, the authors concluded

that "this form of self-generated glucose data promises to

provide information that would be of great use to the health

care practitioner" (p. 212).


SMBG, Patient Compliance, Accuracy, and Diabetes Control



A number of studies indicated that patients with IDDM

are not compliant with physician requests to SMBG, or when

patients do test, reported values are often inaccurate.

Patient noncompliance may underlie the difficulty many

investigators have experienced in their efforts to document

positive effects of SMBG on diabetes control.

In a study by Ziegler, Kolopp, Got, Genton, Debry, and

Drouin (1989) logbook generated BG values were compared to BG

values stored in a meter with memory capabilities unbeknownst

to the sample of 14 young adult patients. Perfect numerical

agreement between logbooks and meter values occurred in only

3 of the 14 patients, 9 showed underreporting (failing to

report an average of 1 in 6 values); and 8 overreported










(inventing an average of 1 out of 3 BG values). Patients

minimized the occurrence of hypo- and/or hyperglycemic

excursions in their logbook by either omitting the value or

by reporting better (more euglycemic) BG values. This study

suggests that logbook data are a poor source of comparison in

any study.

Wilson and Endres (1986), in one of the few studies to

document a positive association between SMBG and improved

glycemic control, also reported that patient SMBG test

results are of questionable accuracy. In their study,

compliance was assessed by comparing SMBG frequency when the

patients did not know and when they did know of the meter's

memory abilities. Although there was no difference in

frequency of SMBG during these two periods, a large

percentage of the patients fabricated BG test results (i.e.

values were recorded in the logbook that were not recorded by

the meter). When the patients were not aware of the meter's

memory ability, the mean percentage of fabricated test

results was 39.8%, while this mean percentage significantly

dropped (p<.003) to 16.3% during the period when the patients

knew of the meter's memory capability. When the patients

were confronted with their fabricated test reports, they had

no adequate explanation for the bogus entries, although many

speculated that the meter may have malfunctioned.










Daneman et al. (1985) briefly noted that patient

compliance to SMBG varied greatly in both the experimental

and control groups (reported SMBG frequency mean=12.8

tests/week; range=8.8-17.2 tests/week).

There are a number of possible reasons why a patient

would present bogus BG values. By presenting in better

glycemic control, the patients may avoid possible ridicule or

proselytizing by the patient's physician and/or parent, more

frequent BG testing, or more intensive treatment, such as

inpatient hospitalization. The patient and physician may

differ as to what optimal BG control is, so the patient may

write in values preferred by the physician in order to

appease him/her. Patients may also fabricate BG values so as

to achieve some goal, such as eating more or less than is

recommended.

For example, a study by Delameter, Kurtz, White, and

Santiago (1988) explored the social demand placed upon

patients by physicians and subsequent patient reports of BG

values. Nineteen adolescents were randomly assigned to a low

demand group (given instructions emphasizing honest,

anonymous self-monitoring with no expectation for high

adherence and no further experimental contact) or high demand

group (asked to keep logbooks with the expectation that their

records would be reviewed by a diabetologist and that they

would be contacted by phone after their records were reviewed










to discuss how well they were doing with their diabetes

management). Both groups had similar BG values for the week

prior to the intervention (low demand group mean=158 mg/dl,

high demand group mean=150 mg/dl; p=NS) but analysis of post-

intervention BG values indicated that the low demand group

reported significantly higher BG values (mean=172 mg/dl) as

compared to pre-intervention and to the high demand group at

post-intervention (mean=140 mg/dl). These findings suggest

that SMBG and reporting of BG values is a social behavior

affected by the demand characteristics of the

diabetologist/patient relationship.

Wing et al. (1985) found that among their 209

adolescents who had been taught to SMBG, 12% indicated that

they were not using SMBG at all, 25% reported using it less

than once a day, 37% reported using it 1-2 times a day, and

26% reported using SMBG three or more times a day.

Unfortunately, the design of the study did not allow any

means of varifying the "truth" or accuracy of these self-

reports of SMBG, and reported frequency was not associated

with glycemic control. Belmonte et al. (1988) instructed 219

children and adolescents with IDDM to SMBG twice to three

times daily. Compliance to the SMBG requests was assessed by

inspecting the number of fingertip punctures in a subset of

86 randomly chosen patients who had recorded two to three BG

tests per day in their logbooks. Of these 86 patients, 23%










showed no evidence whatsoever of puncture marks, 23% had less

than 10 puncture marks, while the remaining 54% had 10 or

more puncture marks. The authors speculated that lack of

improvement in glycemic control may be a consequence of

failure to perform SMBG in a consistent fashion. In the

1983 sample of Hirsch et al. of 98 adolescents who had

expressed a desire to SMBG, at 1-year follow-up, 83% of the

adolescents had begun to SMBG, but only 14% were testing more

than five times per week, and 42% patients were testing less

than one time per week. Patient noncompliance could explain

the authors' failure to document significant changes in

metabolic control as a consequence of SMBG. Unfortunately,

comparisons between compliant and noncompliant subjects were

not conducted.

Mazze et al. (1985) assessed the relationship between

SMBG and glycemic control as well as assessing compliance and

accuracy of SMBG. Thirteen patients participated in a 2-week

study that required recording of BG test results in a

logbook. Unbeknownst to the participants, the results were

recorded in a memory meter. Later, the patients were told of

the meter's memory. Regular BG testing continued for another

6-weeks. The patients' performance when they were aware of

the meter's memory (AWARE) was compared to their BG

recordings when they were not aware of the meter's memory

(UNAWARE). Patients were consistently more accurate in the










AWARE versus UNAWARE condition: underreporting (failure to

report a value in the logbook that was recorded in the

meter), AWARE=7% vs UNAWARE=10% (p=NS); overreporting

(addition of a value in the logbook that was not recorded in

the meter), AWARE=1% vs UNAWARE=34% (p=.0027); precision (the

degree to which the patient accurately recorded the test

result at a time corresponding to the test), AWARE=99% vs

UNAWARE=72% (p=.0037). This study indicated that patients

can be compliant with physician recommendations to SMBG when

they are aware that the meter they are using has memory

abilities. Yet SMBG associated with informed use of a memory

meter did not produce any significant change in glycemic

control.

In a similar study by Mazze, Shamoon, Pasmantier,

Lucido, Murphy, Hartmann, Kuykendall, and Lopatin (1984), 19

young adults received instructions and practiced SMBG while

recording their test results in a logbook. The patients were

then given BG meters that looked identical to the ones they

had been using, but unbeknownst to them had memory

capability. The patients were told to use the "new" meters

for 12 to 14 days while continuing to record their test

results in a logbook; they were told not to alter their usual

frequency or method of testing. Comparisons between the

memory meters and logbook entries revealed a significantly

lower (E<.0001) mean BG level reported in the logbooks (mean










BG=169 mg/dl) than recorded in the meters (mean BG=195

mg/dl). Analysis of the specific errors made revealed that

they were mainly omissions of high BG values and addition of

values lower than the true (memory meter) values. To a

lesser degree, hypoglycemic values (<50 mg/dl) were also

omitted. The authors noted that for a substantial number of

patients (74% in their study), logbook data could result in

misleading clinical impressions, which called into question

the relative value of logbook recordings for therapeutic

decision making. Many studies examining the effects of SMBG

have used logbooks which may explain the difficulty

investigators have had documenting positive effects for SMBG.

The Gonder-Frederick et al. (1988) study is one of the

few that investigated patient compliance, accuracy of

reported BG values, and the association between SMBG and

glycemic control. Compliance to physician requests to SMBG

was assessed in four different ways: (1) actual frequency as

determined by memory meter values; (2) self-reported

frequency from BG logbooks; (3) patient recall of physicians'

recommended frequency; and (4) physicians' reported

frequency. The actual frequency of SMBG ranged from less

than one to more than four times per day. A total of 47%

(n=14) of the patients averaged three or more self-

measurements per day, while 23% of the patients averaged one

daily self-measurement, and 20% averaged two daily self-










measurements. Only three patients averaged less than one

self-measurement per day. These findings indicate that

patients can be compliant with physician requests to SMBG.

However, analysis of the discrepancies between the memory

meter and logbook values indicated that there were

significantly more omissions than additions in the logbooks,

but the omissions were not always eliminations of

unacceptable and addition of acceptable BG values, which

would result in a more positive glycemic profile. Patients

may be compliant with SMBG but may be inaccurate as to the BG

values they enter in logbooks. In the study, there was no

relationship between compliance with SMBG and glycemic

control.

Although many studies found noncompliance and/or

inaccuracies in SMBG, the inaccuracies between patient

reports of BG values and actual BG values may not always be

the result of the patient's desire to purposely present a

more positive picture of diabetes control. As Gonder-

Frederick et al. (1988) pointed out, there are other types of

errors that contribute to patient-generated versus memory-

meter BG discrepancies, including poor record keeping due to

forgetting or inaccurate transcription of BG values. In

their study, patients were not told to change their self-

monitoring routine and fabricated values were rare. They

suggest that fabricated BG values may be a function of SMBG










regimen requirements. Gonder-Frederick et al. (1988) stated

"it is not at all surprising that patients expected to adhere

to such a demanding SMBG regimen would be more likely to add

fabricated values to glucose diaries" (p. 584).

In contrast, Marrero et al. (1989) reported no

significant differences in either frequency (compliance) or

actual BG values (accuracy) between meter-memory and logbook

values in their patients. There were also no significant

differences between the memory-meter and no memory-meter

groups in average frequency of SMBG per day. Although

Marrero et al. (1989) found that patients using meters with

and without memory capabilities were both compliant and

accurate with SMBG requests, HbA1 improvement was not

associated with frequency of SMBG.

The great majority of studies that have investigated the

relationship between SMBG and glycemic control have not found

an association between the two. Some have argued that

noncompliance to the recommended SMBG regimen and/or

inaccurate record keeping explains the literature's failure

to document an association between SMBG and diabetes control.

However, this explanation does not fully account for the lack

of a positive association. For example, Gonder-Frederick et

al. (1988) found their patients to be compliant with SMBG yet

this was not associated with improved glycemic control, while










Wilson and Endres (1986) found their patients to be

noncompliant with SMBG; yet their glycemic control improved.


SMBG and Regimen Adjustment



Theoretically, SMBG information needs to be acted upon

in order to produce glycemic improvement; adjustments in

insulin dose, diet, and/or exercise need to occur based upon

available SMBG values. When the diabetologist requests the

patient to SMBG a certain number of times and/or at certain

times of the day, there is an underlying assumption that the

patient or diabetologist uses this information to make

changes in the patient's insulin dose, or perhaps in other

aspects of the diabetes regimen, such as diet or exercise.

Indeed, the Wing et al. (1985) study found that the best

predictor of compliance to SMBG was the patient's perception

of improved glycemic control through BG monitoring. Analysis

of the patients' responses to the questionnaires revealed

that those who believed that SMBG improved glycemic control

tended to monitor most frequently. The authors postulated

that the key to improving glycemic control may be the use of

the BG information to regulate diet, exercise, and/or insulin

dosage. As the authors pointed out, "even if SMBG is done

accurately and frequently, it can be effective in lowering

blood sugar only if it is used to make adjustments in other










aspects of the regimen" (p. 218). These authors raise a very

important point, simply monitoring BG does not alter glycemic

control; the information must be acted upon in order to

produce a change. Unfortunately, as the authors state;

"there appear to be no data to indicate what percentage of

patients actually use the information to make changes in

their regimen and what types of changes are made" (p. 218).


Patient Self-Adjustment of Insulin Dose


A number of the previously described studies that did

not find improvement in glycemic control associated with SMBG

also requested their patients to adjust their insulin based

upon the resulting BG values. Daneman et al. (1985)

encouraged patients to adjust their insulin dose once or

twice weekly based upon the previous 3-7 days values.

Patients were also encouraged to phone their diabetologist in

order to optimize dose adjustments. The study did not

explain how the patients or diabetologist used the BG

information to make insulin dose changes, and no comparisons

were made between those who self-adjusted and those who did

not. Similarly, Wing et al. (1985) instructed their

patients to adjust their insulin dose based upon resulting BG

values but did not assess whether insulin adjustments

actually occurred. Hirsch et al. (1983) also instructed










their patients to adjust their insulin dose but found that

patients did not comply with this request.


Diabetologist Adjustment of Insulin Dose


Diabetologist insulin dose prescription behavior has not

been examined. If variability between diabetologists exists,

this may be an additional variable underlying the difficulty

investigators have had documenting positive SMBG effects on

diabetes control.

Although actual insulin dose prescriptions made by

practicing diabetologists have not been studied, general

prescriptions have been formulated. In a standard textbook,

Galloway (1988) described the initial insulin prescription

for a patient with IDDM. The following is a general

guideline of ranges of insulin dosages in various types of

patients with IDDM:


Type of Patient Total Daily Dosage

(Unit/Ku)
IDDM (overall) .52 to .75

Before Puberty .5 to .7

During Puberty .8 to 1.2

During "Honeymoon" Phase <0.5










The prescription begins with the premise that the regimen

should produce a serum insulin profile most like that of a

normal individual (euglycemia). He asserted that normal

insulin secretion is characterized by two modes: basal and

meal-stimulated secretion. The function of basal insulin

secretion is to restrain glucose output from the liver

overnight and between meals, while the function of meal-

stimulated insulin secretion is to promote the disposal of

ingested nutrients, primarily glucose. Starting with these

basal and meal-stimulated goals, the diabetologist should

modify the insulin prescription "according to the

pharmacokinetics of particular insulin preparations and other

factors to arrive at the regimen best suited to the patient"

(Galloway, 1988 p.110). These "other" factors include goals

for glycemic control; the patient's ability to detect and

counterregulate hypoglycemia; the patient's capacity to

comply with diet, exercise, SMBG, and insulin regimen

guidelines; the patient's educational level, lifestyle, and

home support system; the patient's endogenous insulin

secretary status, including type of diabetes (IDDM or non-

IDDM); the patient's age and weight; the patient's concurrent

medical conditions, including pregnancy; the patient's

pharmacodynamic response to or ability to use specific types

of insulin or insulin prescriptions; and the availability of

competent professional supervision.










In describing general treatment regimens Galloway (1988)

described two general insulin regimens: (1) a single daily

dosage of intermediate-acting insulin and (2) a split/mix

regimen that uses an intermediate-acting (such as NPH or

Ultralente) insulin to provide the basal insulin supply and a

short-acting (Regular) insulin to provide the meal-stimulated

supply administered before breakfast and dinner. The single

dose regimen will not be discussed further since the

split/mix regimen is more common and more relevant to the

focus of this investigation.

The split/mix regimen is as follows: The basal insulin

requirement of normal weight IDDM patients is usually .4 to

.5 U/Kg/day (28 to 35 units). The obese patient usually

requires more insulin. To control postprandial

hyperglycemia, Regular insulin may be given 20 to 30 minutes

before the main meals. The amount of each dose of Regular

insulin will vary in most cases from 4 to 20 units, depending

on the severity of the diabetes and on the diet and exercise

program of the patient.

Holman and Turner (1988) also provided a formula for

establishing the initial dosage of Ultralente insulin in

normal weight patients with IDDM based upon fasting BG (FBG)

values:











Estimated Dose = FBG 50

10


Example: FBG = 260 mg/dl


Estimated Dose = 260 50

10 = 21 units


Holman and Turner's equation may be considered a standard

when only Ultralente is used, while Galloway's insulin dose

regimen may be considered a standard for a mixture of short

and intermediate-acting insulins, as can Skyler et al.'s

(1981) algorithm. The acceptance and use by diabetologists

of any "standard" insulin dose is purely speculative.

The previous insulin dose "standards" reflect how

variable insulin dose determination can be. How

diabetologists actually determine initial insulin dosages and

how they adjust insulin dosages once an initial dosage has

been determined has not been studied.

The American Diabetes Association, Centers for Disease

Control, Food and Drug Administration, and the National

Institute of Diabetes and Digestive and Kidney Diseases

convened and published a "Consensus statement on self-

monitoring of blood glucose" (1987). Nine issues concerning










SMBG were discussed including "How are the data generated by

SMBG used in patient-care management" (p.95). Particularly

noteworthy is a recent American Diabetes Association survey

indicating that many health professionals do not review the

patient generated SMBG data at all. In such cases it seems

clear that SMBG data are being generated by the patient but

not effectively used by the diabetologist.

If one were to view SMBG behavior on a continuum, with

patient behavior at one end and diabetologist behavior at the

other, it becomes clear that the great majority of SMBG

studies place the burden of responsibility on the patient.

No studies have investigated diabetologist behavior. Whether

or not the diabetologist uses the algorithm of Skyler et al.

(1981), a modification of the algorithm, some other standard

insulin adjustment formula, or clinical experience, when

adjusting insulin doses is completely unknown. Not only is

within diabetologist consistency of insulin dose

recommendations unknown, but consistency of insulin dose

recommendations across different diabetologists when

presented with the same BG data is unknown.



Current Investigation


Diabetologists routinely request their patients to SMBG

as a means of altering insulin levels so as to approximate










euglycemia. The literature has clearly indicated that

increased frequency of SMBG does not lead directly to

improved glycemic control; this patient generated data must

be acted upon in order to produce glycemic changes. It is

not clear whether diabetologists use the BG information in a

consistent fashion when adjusting insulin doses.

Among the three aspects of diabetes control (diet,

exercise, and insulin dose) that the patient is able to

manipulate either by himself or by diabetologist

recommendations, insulin dose manipulations are probably the

most powerful means to achieve euglycemia (Davidson, 1986).

Yet, there are no published studies that have looked at the

consistency of insulin dose prescriptions that diabetologists

recommend to patients.

The current study investigated whether different

presentation formats (logbook and computer generated output)

influence diabetologists' decisions to change an insulin dose

and the magnitude of insulin dose change. Marrero et al.

(1989) found that physicians spent significantly more time

discussing BG test results when presented with computer-

generated data as opposed to logbook presented data. The

patients who used the memory meter that produced the

computer-generated data also differed significantly from

those patients presenting their data in logbook format on a

number of parameters involved in diabetes care, including an










increase in perceived quality of interactions with their
physician (p<.001), perceived understanding of their diabetes

and its treatment (E=.002), and the perceived importance of

BG testing (L=.006). The authors speculated that one benefit

of the computer format is that it allows the diabetologist to

spend less time organizing or transforming data and more time

interpreting the results and discussing therapeutic options

with patients.

The print-out generated from the Glucometer M memory BG

monitor includes a mean profile by time of day, by day of the

week, and a preset BG target range (from 80 mg/dl to 120

mg/dl) that displays the percentage of BG values within,

below, and above the target range. This information is

provided in histogram form, scatter plot form, as well as a

listing of all BG values and their corresponding day and

time. In contrast, the logbook provides columns for BG

entries by time of day and a space for comments.

The study also investigated whether the number of BG

values presented influences the decision to change an insulin

dose as well as the magnitude of insulin dose change.

Presumably, the amount of information presented could

influence diabetologists' decision-making behaviors.

Diabetologist characteristics of employment setting,

years of experience working with IDDM patients, number of

patient contacts per week, and professional training were










tested as possible determinants of insulin dose decisions. A
study by Spevack, Johnson, Harkavy, Silverstein, Shuster,

Rosenbloom, and Malone (1987) found that when presented with

different indices of diabetes control, diabetologists in two

different university affiliated hospitals used the same

indices, except for one discrepant measure (total

cholesterol) when rating the diabetes control of patients.

In explaining the cross-setting differences, the authors

speculated that the diabetologists that came from the same

institution, "may have developed an unwritten consensus of

what "excellent" or "poor" diabetes control means" (p.222).

This unwritten consensus of a shared setting, as well as the

other diabetologist characteristics, may influence how one

group of diabetologists recommend insulin doses as compared

to another group.

The BG profile characteristic, mean BG levels was

studied to investigate its influence on the recommended

insulin dose decisions. Profiles with higher mean BG values

were expected to induce greater insulin dose change

recommendations than profiles with lower mean BG levels.

Once the decision to change the insulin dose was made,

an analysis of the specific type, amount, and timing of the

insulin changes were also conducted.















METHODS


Participants



There were a total of 18 diabetologists who participated

in the study (see Table 1). They were recruited through

personal and/or phone contact. Five additional physicians were

solicited but chose not to participate due to time constraints

or lack of interest. An additional nurse's data were discarded

because she completed only half of the required patient BG

profiles.

The diabetologists included 8 physicians, 8 nurses, 1

physician assistant, and 1 dietician all specializing in the

care of patients with IDDM. Although not licensed to make

insulin dose adjustments, the dietician was included because

she had many years experience working with IDDM patients. She

worked in a diabetes clinic where she made insulin dose

recommendations which were then confirmed by a licensed

physician. Further, statistical analyses indicated that her

insulin dose decisions did not differ significantly from the

other diabetologists.

All participants were from either Shands Teaching Hospital

at the University of Florida (Shands, n=6), University of South
37
















(1) C rllq ouUl Ln o L n n )\nO
leq (N H
S'^~gS-iiS-n\o


1 >1 4>1


* 9


SmmmmmMOO
WWem0






U0U U 4 'H .
Soomga g

kk









Florida Hospital (USF, n=7), or private practice in Florida

(private, n=5).

The diabetologists all had at least two years experience

reviewing and recommending insulin doses with IDDM patients

(mean=10.2 years; range=2-20 years) and all had contact with

patients at least twice a week (mean=16.1/wk; range= 2-60/wk).

The diabetologists at Shands had a mean of 9.7 years experience

working with IDDM patients, those at USF had a mean of 8 years,

and those in private practice had a mean of 13.8 years; these

differences were not significant. The diabetologists at Shands

had a mean of 14.6 contacts per week with IDDM patients, those

at USF had a mean of 12.3 contacts, and those in private

practice had a mean of 23 contacts; there were no significant

between setting differences.

There were setting differences in type of diabetologist

participant. While all five of the participants in private

practice were physicians, only two of the six Shands

participants were physicians, and only one of the seven USF

participants was a physician.


Measures


The BG records used to create the logbook and computer

profiles were obtained from actual adolescent patients housed

in a residential treatment program at UF's Shands Hospital.

During their hospitalization, the adolescents were







40
metabolically regulated while they and their families were

provided psychological treatment and their families were

educated on diabetes management. All patients had been

diagnosed with IDDM for at least one year prior to their

admission. As part of patients' management, BG values were

routinely obtained at three times per day at variable

intervals: preprandial, postprandial, at bedtime and upon

arising. All BG tests were monitored by the treatment staff.

All records selected for use in the current study met the

following criteria:

(1) 15 consecutive days of BG tests were recorded

(2) at least 3 tests per day were recorded

(3) no insulin changes were made during the 15 day period

Fifteen records were identified. From each record, two

profiles were constructed. One profile presented results of

three BG tests per day. The other presented only one BG test

per day. The mean glycemic control of each profile was within

+ 10 mg/dl of the other profile.

On each profile, the insulin dose prescribed for that 15

day period was provided. At the end of the record, the

diabetologist was asked whether the insulin dose should be

changed and if so, what the new dose should be.

The 30 profiles (15 with 1 BG test per day and 15 with 3

BG tests per day) were presented in two formats: logbook and

computer generated. The computerized format was generated from







41
a Glucometer M software program. The logbook format mimicked

that of a patient diary.

Half of the diabetologists completed the logbook profiles

first and half completed the computer generated profiles first.

The presentation of records consisting of 3 BG values per day

versus 1 BG value per day was randomized within each format,

with the following stipulation: BG profiles with one versus

three BG values generated from the same 15-day record had to be

separated by at least five intervening profiles. This

decreased the likelihood that the diabetologists would

recognize that the 1 BG value profiles were a subset of the 3

BG values profiles.

Using a 2 (format) X 2 (number of blood glucose tests)

repeated measures design, each of the 18 diabetologists

completed a total of 60 patient BG profiles (15 logbook

profiles with 1 BG value per day, 15 logbook profiles with 3 BG

values per day; 15 computer profiles with 1 BG value per day,

and 15 computer profiles with 3 BG values per day).



Procedure


Through written instructions, the diabetologists were

informed that all of the patient BG profiles came from actual

patients with IDDM, all patients had diabetes for at least one

year, no patient was receiving more than 1.5 U/Kg of insulin

per day, the BG values were known to be accurate because the







42

patients were observed at the time of BG testing, and all

patients were receiving human insulin twice daily (pre-

breakfast and pre-dinner). They were also told that some of

the profiles were from the same patient but from different 15-

day periods. This information was included to provide a

rationale for the apparent similarity in insulin dose across

some profiles, reducing the probability that diabetologists

would notice that the same BG profile was repeated on four

different occasions (varying in format and number of BG tests).

The participants were also told that in the interest of time,

they should make an insulin dose decision and move on to the

next profile; reconsideration of prior insulin dose decisions

was discouraged. Written and oral questioning of the

diabetologists upon the completion of all profiles indicated

that none were aware that the same patient BG profiles were

repeated.
















RESULTS


The results are divided into six sections: (1) an

overview of the dependent variables, (2) the effect of

diabetologist characteristics on the dependent variables and

statistical analyses conducted, (3) the effect of BG profile

characteristics on the dependent variables, (4) the effect BG

profile characteristics on the type and time of insulin dose

changes, (5) the degree of diabetologist agreement on the

recommended insulin dose, and (6) the time to complete the

task.



Overview


Univariate repeated measures analyses of variance

(ANOVA) procedures were conducted to determine what

characteristics of the diabetologists and BG profiles

contributed to the insulin dose decisions.

The format of presentation (FORMAT) and number of BG

values per day (#BGVALUES) served as within subject variables

in all ANOVA procedures. The effect of FORMAT and #BGVALUES

will be described once for each dependent variable, although

43








44

it was included in all ANOVA procedures. The FORMAT and

#BGVALUES interaction was conducted in all ANOVA procedures

and was never found to be significantly related to any of the

dependent variables. Consequently, the interaction was

excluded. All ANOVA models initially analyzed interactions

between the within subject variables (FORMAT and #BGVALUES)

and the between subject variables (diabetologist

characteristics and the BG profile characteristic of the mean

glycemic control); the interaction models were retained only

if significance was found. Main effects models were analyzed

when significant interactions did not occur.

The 18 diabetologists (DIABETOLOGIST) and the 15 BG

profiles (PROFILE) were included as random variables. The

PROFILE variable was expected to be significantly related to

the insulin dose decisions. Since we were not interested in

the individual PROFILE effects, the results section will not

elucidate these findings. However, the effect of including

the DIABETOLOGIST variable in the ANOVA models was not known

beforehand.

There were two dependent measures analyzed: (1) the

percentage of the "yes/no" decisions for which a change in

the insulin dose was recommended (%CHANGE), this variable was

the percentage of all the profiles recommended for an insulin

dose change (n=1080); (2) among those insulin doses that were

changed, the magnitude of change was calculated. Difference








45
(DIFF) and absolute change (ABSDIFF) change scores were

calculated by subtracting the initial insulin dose from the

recommended dose. ABSDIFF was the absolute value of the DIFF

score. DIFF takes into consideration the direction of the

insulin dose change, while ABSDIFF takes into consideration

only the magnitude of the change.

The DIABETOLOGIST variable was included in all ANOVA

procedures investigating %CHANGE. Both DIABETOLOGIST and

PROFILE were included in all ANOVA procedures investigating

DIFF and ABSDIFF.

In view of the large number of analyses conducted, the

p-value for all ANOVA tests of significance were set at .01.



In order to investigate the effect of diabetologist

characteristics on insulin dose decisions, the following

diabetologist characteristics were tested: the professional

setting in which the diabetologist was employed (SETTING),

Shands Hospital at the University of Florida, n=6 vs

University of South Florida Hospital [USF], n=7 vs private

practice, n=5; the diabetologist's professional education

(OCCUPATION), physician, n=8 vs other, n=10, [nurses,

physician assistant, dietician]; years experience working

with IDDM patients, as determined by the median experience

being 10 years, thereby dividing into greater than and less

than 10 years experience (EXPERIENCE), 1-9 years, n=8 vs > 10








46
years, n=10; and the number of times per week each

diabetologist had contact with IDDM patients, as determined

by a median split (CONTACTS), 1-10 times per week, n=9 vs >10

times per week, n=9.

In order to investigate the effects of BG profile

characteristic on diabetologist insulin dose decisions, the

mean glycemic control (MEANBGCONTROL) of each 15 day record

was calculated. Each record was then classified as high (mean

BG >180 mg/dl, n=18) or low (mean BG <180 mg/dl, n=12). These

categorizations were previously determined by Skyler (1979)

to be clinically meaningful.

The effect of the variability of glycemic values over

the 15 day record on the dependent variables was also

attempted, but could not be analyzed because of confounding

with MEANBGCONTROL. The BG profiles with a low mean almost

always had a small variability of BG values, while those with

a high mean had large variability. The analysis of the

variability of the BG profiles was attempted two ways: (1) By

a median split of the standard deviation of the profiles, and

(2) by partitioning the profiles according to the percentage

falling above a BG target range (>120 mg/dl) and below the

target range (<80 mg/dl). Both of these ANOVA procedures

lead to confoundings with MEANBGCONTROL that rendered any

result uninterpretable.








47
To further explore diabetologists' insulin dose

decisions among the BG profiles that were changed, the effect

of FORMAT, #BGVALUES, and MEANBGCONTROL on DIFF and ABSDIFF

were examined by the type of insulin administered (Regular vs

NPH) and times of administration (morning vs evening).

Regular insulin has a short-acting time course that starts to

work 30 minutes to 1 hour after injection, peaks in 2 to 3

hours, and is used-up after 4 to 6 hours. NPH insulin has a

intermediate-acting time course that starts to work 1 to 1.5

hours after injection, peaks in 8 to 12 hours, and is used-up

after 20 to 24 hours (Travis, 1988). The morning

administration of a mixture of Regular and NPH insulins are

designed to cover the rise in glucose following breakfast,

lunch, and the pre-dinner snack. The evening administration

of a mixture of Regular and NPH insulins is designed to cover

the rise in glucose following dinner and the pre-bed snack,

as well as to maintain glucose levels within acceptable

limits overnight and in the early morning (Travis, 1988).

The degree to which diabetologists agreed on insulin

dose decisions was further explored by calculating the

interexaminer correlation coefficient (Fleiss, 1986).

The diabetologists were requested to indicate how many

minutes it took to complete the logbook and computer profiles

so that possible differences in time to review profiles and

make insulin dose decisions could be explored.










Effect of Format and Number of BG Values on the Decision to
Change an Insulin Dose

There were 18 diabetologists who each made 60 "yes"/"no"

decisions whether or not to change an insulin dose, for a

total of 1080 decisions. Of all the decisions made, 69.7%

(753/1080) were to change the insulin dose.

Table 2 provides the percentage of the 15 BG profiles

that were given insulin dose changes by each of the 18

diabetologists within each of the four levels of FORMAT and

#BGVALUES.

Table 3 indicates that DIABETOLOGIST, FORMAT, and

#BGVALUES were all significantly related to %CHANGE. More of

the BG profiles were changed when presented in computer

format, 76.48% (SD=18.55,range=20 to 100%), than when

presented in logbook format, 62.97% (SD=25.12,range=0 to

100%). In addition, more of the BG profiles were changed when

the profile consisted of 3 BG values per day, 79.08%

(SD=15.94,range=26.7 to 100%), compared to only 1 BG value

per day, 60.37% (SD=25.19,range=0 to 93.3%).

As Table 2 indicates, there were wide differences

between diabetologists' recommendations for an insulin dose

change. Diabetologist #8 changed only 42% of the BG profiles,

while diabetologist #2 changed 92% of the profiles.


Effect of Format, Number of BG Values, and Diabetologist
Characteristics on the Decision to Change an Insulin Dose















4.


r-U
0


d


U
0











Qf
U



4m





0
)4






u
a)








O
03



12


dp







Wa o r 0 o % % % o o w %0 en % o N


*P

o



-4 (1 m -w in %o 0 C h4 M 0 '0 N 0
0 0t
Q E-


1

1
1


o r M 0 m 0 0 o 0 0 o m N C ) m ) 0 0 r-






00 C 0 M %0 0 %.0 0 O N- 00 N w r- m w C


r-4





a

, n t'

to

00


8 U






H






















,-4 il *' r-I
0 0 00
P 0 0 00
0 0 00
*


















ri *
8 -----









ON
1I ----^









N0 .0
43
p 4 C r- 2
D H *4 11
rz4 a- >4 um









C-
0 H



pa **rl .J
4 1 4 r .





0w)





i- l
-








51
The next ANOVA procedure analyzed main effects of the
within subject variables (FORMAT and #BGVALUES) along with

diabetologist characteristics (SETTING, OCCUPATION,

EXPERIENCE, and CONTACTS). The interactions between each

within subject variable and each diabetologist characteristic

were not significant. Interactions between diabetologist

characteristics were not estimable because of confounding

between these variables.

As Table 4 indicates, none of the diabetologist

characteristics were statistically significant.


Effect of Format and Number of BG Values
on the Magnitude of Insulin Dose Change

Among the insulin doses changed (n=753), there was a

mean increase of 2.57 units in the total daily insulin dose,

with a SD of 2.09 and a range from an 18 unit daily increase

to 12 unit daily decrease.

Table 5 provides the mean insulin dose change across the

BG profiles for which diabetologists' recommended an insulin

dose change; Table 6 provides the same data using the ABSDIFF

score. Diabetologist #4 recommended no change in insulin dose

for BG profiles presented in logbook format with 1 BG value

per day, while Diabetologist #8 changed only one profile,

whereby the total change was '0'.

For the dependent variable DIFF, there were significant

main effects for DIABETOLOGIST, PROFILE, and FORMAT (see


























O
4.)

r-l
0
O








U
idp

M

D )
0 W
go

4-W





A-U
E-I

U
S0
0 u
w-


+J


r- (V r-4
ON m Ln r-
00 0 r
N
r-4
II



C'04
0-4 0 0
* U')
0000 N
N



N N r '



M


VN

)-l






U.
I-H %D
o l r. o













4-)
u ur V 0M







0w p d
I^S-













Sv r r- co ( o w 00 w N 0 %
<- 4 3' in ^: SD '. 4 N ko V -q %o
+ + co

V^ 'A0 N 0 V 0 (h N Nf- 0 0 N CM (w I M o

1- 4 N '11 N to 4 1 qv N 4 4 1 4 N N. w4 N. ID






N o 0 (' (') 4 0 + + + + N
*0 I .me4m; w e
Im D 0 0 ('0' N I .- '- N


1-1






U
c0



H
4 4
o "
0







Wa











P
H 0



02

0 l
n m






02q







>a,


wM


.. . .. .
+ + + + + + + + + + + + + + + + +
I I I I I I I I I I I I I I I I I
I++ + + I + I I I I I + I +++ I


mC '-4 1 o 0 r- i N o o N Nw N. 0' N No A
S. . .


v w t- v i o %o ODn t o n o a


+ + + + + + + + + + + + + + + + + +

+ + + + + + + + + .. ..
II I I I I I I I I +I I I I +
S -4 I' N fn N 0h 1 wD N 'o '4 no N -4

t + g + + I I7 j + + +

S A 01 C CD fn ID ODoo w0 I D Cl 0 r'A w m m rN

N 0 1 m CD) N N (4 m m 01 ID m 0' 4'i N N CD


v %0 w % In w 0 %V OD % t-%tm

g T .' .-
N'ommomA00AwomA0 0N) C





I CD N N N 0 1 A o o 'A N' C. Cl

* 0 I CDo o o oo oI CD ID N
S+ + + +











0N 0' -4 m-o N "-
C O '0 IM t-v 0 + 0 t
0N



0 I ,\ 1 ') 0 0 N N 9 0 D N 1 n '
N -4 Z 0- N C; N m m 14 N NelN N I 1

o-


0< 0 N cD 01 0' ID 0 0 N~ 'A N~ N^ .-4 N. ('4 c') C


~ n '" m o e m > N ^ ^ M w pi n n i ric
CI ID ID -4 N I N NM CD n 0'f 0i C) l 'A 'A '- l ID



* S .. ,
'a Clo '00 ON m C sON
+++ + + + + + + + + + + + + +



















U)

,-
5 a




H H

0





a,





0 )

-I
o a,
O H
d)


-'






0
D I



(



1 a



CU
0






S '-

a a



(1)


o 0 0 ( 0 0N I I o0 I I
e1i CO oO o N 0 O (' 0

-40N m t
S0 0 o l o 0 ( 0 N 0 o o 0n wn
o .n .. ..- (' -. n r4 %i O (N .







S-'o4 N M4 4 0oN o 4 in r( o4 N 0 r o 4 NM 4 in
W)n %D 0 rN 4 v4-4 r4 t- M 0 iO D Co ID 00 N Ch
0 n. n N



N 10 0) 0 N N m CO m C! 10 r C O I9
I I I I I I I I I I I4 I I I I I I



C N v 10 m o 0% Crt4 n o c ( n or







S In I In 4 I C CI I 0 I I W!



5 o N o % r 1 0 ) 0 w w (4 I N In to m (
0 C. .O 0. 1
C m p N o o in oo o o Inm m O q) (







I I I I I i I i I i i i i I I I I I
N (' '' 0 ( v m m (' v v fn 4 m v N- 4 C4







. .











o0 0 0 0 0 0 4 0 ) 0 .4 n
0 C) 0 1 0 0 I 0 N0 N CO In t- 0 t N N






_ a m m m ooo Cooo mninm ooo moo o





S0 oo n o o eo o





N NN n o vm m N o
0
' 0 0 N In % 1 1 CO S 1 In In N0 % 0 1
























*S1
.Q __ (1 ___I_0 CO_0_('__In 0 N C








55

Table 7). The insulin doses were increased more when

presented in computer format, +2.83 units (SD=2.98,range=-8

to +11), than when presented in logbook format, +2.14 units

(SD=3.31,range=-12 to +18). There were large differences

between diabetologists in their magnitude of insulin dose

change. For example, diabetologist #5 mean increase of

insulin was only 0.64 units, while diabetologist #3 increased

insulin doses an average of 4.26 units.

For the dependent variable ABSDIFF, there were

significant main effects for DIABETOLOGIST, PROFILE, FORMAT,

and #BGVALUES (see Table 8). Similar to the DIFF findings,

the absolute change of the insulin dose was greater when

presented in computer format, 3.60 units (SD=1.98,range=0 to

11), than when presented in logbook format, 3.31 units

(SD=2.14,range=0 to 18). Whether the dependent variable

analyzed was DIFF or ABSDIFF, diabetologists increased

insulin doses more when information was presented in computer

format compared to logbook. The absolute change in the

insulin was more when 3 BG values per day were presented,

3.72 units (SD=2.13,range=0 to 18), compared to 1 BG value

per day, 3.14 units (SD=1.91,range=0 to 9). Including the

direction of change obscured these effects (#BGVALUES for

DIFF was not significant).


Effect of Format, Number of BG Values, and Diabetologist
Characteristics on the Magnitude of Insulin Dose Change


















r-4 '-4w r- 4%
0 00 00






o o o )
H










0 -h "


E- m Eco 0 m .






4 0 .4 t-
E o 0 -r t

r C rQ i-l



o 'tr H v- 4 v
1 | .4!-
+I -----


a ') I
O-4 4 V


M t

H M H

5a a8






















-f r-1 r-I 1 r-l
0 00 0w
P4 0 00 00
0 00 00








n II
0 P4 0 0
*4 r








Ln a% m r'-4
H __ r- Cha_


Nv C4 ko
4 m ---------










0 004 0
'- '-I 4VI
r-l 144. l






a0)
pO I $4 r(
M a, *r--









0 C0 -Ha





*









In order to evaluate whether diabetologist

characteristics may have influenced the magnitude of their

insulin dose changes, an ANOVA was performed examining the

effect of DIABETOLOGIST, PROFILE, FORMAT, #BGVALUES, SETTING,

OCCUPATION, EXPERIENCE, and CONTACTS, as well as the

interactions between each within subject variable and each

diabetologist characteristic. The estimable interactions

between diabetologist characteristics were not significant.

Results of the DIFF analysis revealed main effects for

DIABETOLOGIST, PROFILE, FORMAT, and a significant #BGVALUES

and CONTACTS interaction (see Table 9). Table 10 provides the

mean insulin dose increase, SD, and Tukey tests of

significance for the #BGVALUES and CONTACTS interaction. The

largest increase of insulin occurred when the diabetologists

with the fewest contacts with patients were presented with 3

BG values per day. Table 9 indicates that the OCCUPATION

variable approached significance. The physician's mean

insulin dose increase was 2.05 units (SD=3.29,range=-9 to

+10), while the other diabetologist's mean increase was 2.86

units (SD=3.01,range=-12 to +18).

Results of the ABSDIFF analysis revealed main effects

for DIABETOLOGIST, PROFILE, FORMAT and #BGVALUES, but no

effect associated with diabetologist characteristics (see

Table 11).
























4-)
Un
0
H
0
r4.
P






rl
a








.a)
M
0







-4
0









44)



a)
E
"0




pi-
*Hi
c


o- r0- 0r-4 -4 M O h 0
o co o m r- mt n
a 00 Q Co r_4 r- CM lw r_4
0 00 00 OCM'0o 0
S S S S









S 5 33rl~da a












*



0






n S 0 S S
*ld P o E-4 Pi u









00
r( r O (










Ue Qt r '-

w 4' u'E:


.,

o II
II




*..M
aN








.o
0 ||
oq n





II I
0 *
r-1



I Wo




'44-i HI
*d'











NU (N %0
a) N rN








V 0 0 0
H l W)"







m n
U M H
*d M
M 1k i



















O
++ ++
00 O0
4- 4- 4


CI




** *


r-4 00
* +
+ +


U)
Qlc
.


a,
0i




a,
I



m
a
n









Sa
*<






i-











a,
0U









a,
H0
't-




Os

'0
C
H


o *
C1 e


% -41
* *
++


C -

*(M
-^ 02



4 p
m p




u
0
0 A-
vi


M (t

g,
00) 02
043 i



0
OP ^
0U 0

-4 m
VAo
6-i n
























4.)
0)
0
"r-

0


.1
ri





rz4


itU)


%14


E-Ia
0
m
9Q V





me








'0 -
'-I
wO











a,
(X)
^ 0


r-4 r-4 r-4 tnr-i t- Un CM n
o 00 v 0 ON 0 r-
0o o0o 00o n N t r-
O 00 00 O#NftD
















w-4 ChI NC4v4

ir- n u<








o a








4 4r. H 0 H. 0 _E-4 P u
0 1041 ri m U






*dc* WQLMHap


o
m



*4 4
00



NO

) 9





H- a
V3 0
no
o-q w w
0 01M
*- ^
QMM



'-r-- '0
U *O
w t-Il II




r -
'4-M *o



40 ) 10
0)
P 0 0 0









* r M i
*
1- -1n i








62
The significant finding for DIABETOLOGIST in all of the

ANOVA's indicated that the diabetologist's were inconsistent

in their insulin dose decisions.


Effect of Format. Number of BG Values, and the Mean Glycemic
Control of the BG Profiles on the Decision to Change an
Insulin Dose

Each BG profile was categorized into high (_180 mg/dl)

and low (<180 mg/dl) by the mean of the 15-day record

(MEANBGCONTROL). Table 12 provides each BG profiles mean and

categorization. The effect of DIABETOLOGIST, FORMAT,

#BGVALUES, and MEANBGCONTROL on the percentage of insulin

doses changed (%CHANGE) was examined. The interactions

between the within subject variables and MEANBGCONTROL were

not significant. All main effects were significant (see Table

13). When the BG profiles were presented in computer format,

73.99% of the profiles (SD=24.02,range=16.67 to 100%) were

changed, compared to 61.03% (SD=29.46,range=0 to 100%) when

presented in logbook format. When the BG profiles were

presented with 3 BG values per day, 76.85% of the profiles

(SD=22.55,range=0 to 100%) were changed, compared to 58.18%

(SD=29.07,range=0 to 100%) changed when 1 BG value per day

was presented. When the BG profiles' mean glycemic control

was high, 78.55% of the profiles (SD=23.32,range=0 to 100%)

were changed, compared to 56.48% (SD=27.20,range=0 to 100%)

changed when the BG profiles' mean glycemic control was low.

Table 14 provides the percentage of the 15 BG profiles that








































1 -1 ( W 0 ,- 0 0 0 0 0
14 ta to -1 H **3




0 tr ON 1 ON fi tp ON -lr




S o 8 m S > e l c
6 i *d 0i *d *4 *.4 .9 .9 *d -. 3 0
Smm m m m m N m


%0 W V Wc4 %0 o cV m') 0 (r) OD Wi (4 %o
0 0 w w 0 w C- v r-4 w m 1m w


r'-
*^ Q Y ^ d O ~1 N rv r
_&_______________________







64









r r ( rr-4 r-
0 O 00 0
4 Q0 0 00 0
C 0 00 0





O
0 0o
.n


0





N
in -o- r
rq( } mo






'44 44

VW

V U -r4
4 -



0) r-4 E-4
fl tl 0 wm0

0 to




a4 0 | |















0
*r

0
r4-









0 m
Ur
14



oa m
< d






) t








S 0g




E-A











ON
S-4
in 3 &
















O

'4-4
0
a






P4
saI,
424p


0 0 to4 w' 0 D w 0 t- w 0 0 N C' O ('4 C' wD %0
0 ~ u O', 0r uO U) CD aO 1 D CD U) cD IA n U C 1


CP

0 0w 000 0 0 0 0
0 N 0 000 0



_0 0






a O
0~~~~~% w r D N C





O r- 0


to S !o 0 w m in w; w w m; w m %0 m w w
0 ) 00
CD %D 00 -1 %0



a __
0 ) N Q D 0Q~ O~D 0) 0 0 C N 0
0 0 CDc N D V 0 C D D U) N 13






0p '-4 1-- -'--41
* ^ ^



a 0 0 1



10 CD CD CD N i o 0 ) CD C 0^ N^ 0 CD
*C C 'i ^U)i 0 0 '0 C C ( 1 C ')




0 0I 100 'V C 0 U) C) I

11 -4 N C' U) "I V N C 0 .- N C' '0)~ a cy U) 10 C
0 -' "'''








66

were given insulin dose changes by each of the 18

diabetologists within each of the eight levels of FORMAT,

#BGVALUES, and MEANBGCONTROL.


Effect of Format, Number of BG Values, and Mean Glycemic
Control of the BG Profiles on the Magnitude of Insulin Dose
Change

For Diff, there were significant main effects for

DIABETOLOGIST, PROFILE, FORMAT, and MEANBGCONTROL (see Table

15). The interactions were not significant. The insulin dose

was increased more, 3.42 units (SD=2.74, range=-6 to +18)

when the BG profile's mean glycemic control was high,

compared to when it was low, 0.64 units (SD=3.13, range=-12

to +6).

For ABSDIFF, there were significant main effects for

DIABETOLOGIST, PROFILE, FORMAT, and #BGVALUES, while the

MEANBGCONTROL variable approached significance (see Table

16). The insulin dose was changed more, 3.84 units

(SD+2.11,range=0 to 18) when the BG profile's mean glycemic

control was high, compared to low, 2.70 units

(SD=1.70,range=0 to 12).

The slight discrepancy between the DIFF and ABSDIFF

findings suggest that insulin doses were frequently decreased

when the BG profiles' mean glycemic control was low.


Effect of Format, Number of BG Values, and Mean Glycemic
Control of the BG Profiles on the Magnitude of the Type and
Time of Insulin Dose Change






















r( r- 4 r4
O 00 Q0 V) r
0 00 00 0







S* *
. *
Sm. r-4 r






r-4 V-4 g-4 4 -
ai 0-r-
So ai



0 r- Ch




4 M t- mo
V Irn r-4r-


-* 4
W L
H *
8Ha


p

a

ul


0






toN

on
*o
4-)





UN





.4 t





4 r-

44




.a
4 0





.^l
0)Q
ap
01O









68










H r.- rl -4 Ln tr- m
0 00 aO '.
40o oo oo o
O 00 00 m
0* 0




0*


0
o II
So m V- m


r- ko r L o *




O *



V C40
(4 M n--4J M_











O q E- 0 0 0 fq
> ------------






CO

H 4

4 1-4 j 8UT N 4 H N
> >^> D04 r



EM -H


rVl
-








69
DIABETOLOGIST, PROFILE, FORMAT, #BGVALUES, and

MEANBGCONTROL were included in all models. Interactions were

also tested. The DIFF and ABSDIFF of insulin dose in the

total daily administration of Regular and NPH insulins, as

well as the morning and evening administrations of insulin

served as dependent variables. The DIABETOLOGIST and PROFILE

variables were highly significant in all of the ANOVA models

(p<.0001) and will not be discussed further. Table 17

provides the DIFF and ABSDIFF mean insulin dose change, SD,

and range by FORMAT, #BGVALUES, and MEANBGCONTROL for the

Regular and NPH insulin dose changes. Table 18 provides the

DIFF and ABSDIFF mean insulin dose change, SD, and range by

FORMAT, #BGVALUES, and MEANBGCONTROL for the morning and

evening insulin dose changes. Table 19 provides the DIFF and

ABSDIFF mean insulin dose change, SD, and range by FORMAT,

#BGVALUES, and MEANBGCONTROL for the morning Regular and

evening Regular insulin dose changes. Table 20 provides the

DIFF and ABSDIFF mean insulin dose change, SD, and range by

FORMAT, #BGVALUES, and MEANBGCONTROL for the morning NPH and

evening NPH insulin dose changes.


Regular insulin dose change. For Regular insulin

administration DIFF, there was a significant main effect for

FORMAT, while the #BGVALUES and MEANBGCONTROL interaction

approached significance (see Table 21). As expected, the

























0

























-,-

,--4
0











r r
E-






















O

1+



0

a





ar


S-* | -- .-4 -





S0 00
t + + q + r-I
0 I I I I I I
.O O O









r4 C4 -n N 'n ('O









r- C O r- -O OD r- a0
aQ Q imum oa









v4 v c4 ( .

mz -D r- OD
(a r
M %0















0 0- 00 0 0 0(
0 P0





,-4 4 4 -4 -4
0Q '0 0' ('M tl '0' 0















H
L3 ---------












*NNCN o N N








4 m
o 1 o0 0 0' c-













I llll-4
jHra~rlla^a
"

__________^ '
















0






















OD


.




14













4I
0k











I

0
m
C



d H

p q



















Q




E W











44
00










gP4


0 1-4 1-4

C % 0 t o 0





( N 0 0 %o


U U) 0 O o
S (' .i r-. .- 0 0 .



C + + I + +



0 o m in 0% 0
% 0 001
N N .- N N N








So% oo


* .









0-i i I 1i


N N r1 NM 2
0 0 00 0 0 0
U) r-t r- l r r


















+ + + + + + +
4 0 '0
ON








I ?I
~Q~Q)~00-






Xir~crcc uur-4

19c i,




























0o



a











a
C:
I
' I





01











m











8
0)




















0)
01


H U)
0 0 0o 0 m a0 0



>I~4 t- _
W CU
- I .4 % .


+ + + + + ++
H t o Ir I rI -
m g

Q)~~~ --------------4
v I- m I I % I





N0 C N t Nl ON N
g* .







H r- O O r- O 0


0 '. ;0 N C; COW
O* C U)' +
0 w


oo 0 or 0o

V 5 a 1 u U
SE o I I I I I o

0, 0 _0 _
S 00oo 0 0o
1:14 Q E1 O'i 0 Q '.00 (
oa .







00
'-4




_3 -
k M


































rTH













ag
w







rz











0











,.


OD V f CD COD OD


0 0 0 0 0 0 0 0
.-I C 10 M 0 1 0
C) Cl .-4 0 C r-i .-1 r



V0 t 0a 0 0- N l0
a .

* 0 OD C 00 c4 0r a)
S+ + + + + + +




m *


0 C C! Ci Cl -4 CD
S00 C 000 0 0
+ + + + + + +


C I I I I I I I
S0 0 oo oo o0






( aa 000100



0 1 .


a *
0 Pi N Ny 1 in U







* WO O 0 0 IA 0
+ + + + I + +
C I I












0 8m
allc ~S ^
v l

ti S ti ^ ^ "
da tt~~-rl^
*
rOT'O3 S






























I



am
H













IQ
*0







'44

.C.


'M4


-4
11-



U*






1
tn
rin








a .
0
o




4J-4

(-


CO




r4







cu
V C









administration of Regular insulin was increased significantly

more when the BG profiles were presented in computer format,

+0.87 units (SD=1.86,range=-8 to +7), than when presented in

logbook format, +0.42 units (SD=1.99,range=-7 to +7). Table

22 provides the mean insulin dose increase, SD, and Tukey

tests of significance for the #BGVALUES and MEANBGCONTROL

interaction. The smallest increase of Regular insulin

occurred when there were 3 BG values per day presented and

the BG profiles' mean glycemic control was low.

For the dependent variable of ABSDIFF, only the

#BGVALUES variable approached significance (see Table 23).

The absolute change of Regular insulin was slightly greater

when 3 BG values per day were presented, 1.59 units

(SD=1.51,range=0 to 8) than when 1 BG value per day was

presented, 1.32 units (SD=1.28,range=0 to 7). Table 17

provides the DIFF and ABSDIFF insulin dose changes for the

Regular insulin.

The significant finding for FORMAT with DIFF, but not

with ABSDIFF indicated that the Regular insulin was increased

more when presented in computer format, compared to logbook.


NPH insulin dose change. For NPH insulin administration

DIFF, there were significant main effects for #BGVALUES, and

MEANBGCONTROL, while the FORMAT variable approached

significance (see Table 24). The administration of NPH






76








SO ++ ++
a 00 00o
oo oo




( l U 0 1(4
g e4-1 0
144
Oll co.4



m *c"

0D 4





0 d
0 o
0 0 00 0 0r|





0 m




0 o V
0 ri C: l


r i 1 a .


> Ii

s I g^$ ^
o X&:
c~t V d *
H10 5 i04
3 gM






















r-4 r-4,- rU) mc
P0 00 h Or-4 Un
o oo (M o 0 00 NO w









0e 0 5
r- o 4\O 0
i i- C v-4w N
















14 4 r 4 oM ,-M
i M r- r-4 r-4 -


Q)
4,




0






0











r
44
0

P4
H






0
'44
0
r(


4' 4J
E -
H *







a -
0


U,














H4j
II
a








II
4-)

SO




-4
U
00
n



uw
C'


U)






a '




41 -
m M
0)


*

























0
oG













-44
0






IE 0










C/I



4-




w




a)
O
o
-H
w


m 0
.4W
(a 0
C
H w
r-
S' -








79
insulin was increased significantly more when 3 BG values per

day were presented, +1.96 units (SD=2.27,range=-7 to +12)

than when 1 BG value per day was presented, +1.71 units

(SD=2.12,range=-6 to +8). The NPH insulin was also increased

more when the BG profiles' mean glycemic control was high,

+2.64 units (SD=1.82,range=-4 to +12) than when it was low,

+2.09 units (SD=2.04,range=-7 to +5). The NPH insulin was

increased more when information was presented in computer

format, +1.97 units (SD=2.13,range=-5 to +8) than when

presented in logbook format, +1.72 units (SD=2.30,range=-7 to

+12).

For the dependent variable of ABSDIFF, there was a

significant main effect for #BGVALUES, while the

MEANBGCONTROL variable approached significance (see Table

25). The absolute change of NPH insulin was more when 3 BG

values per day were presented, 2.39 units (SD=1.82,range=0 to

12) than when 1 BG value per day was presented, 2.20 units

(SD=1.61,range=0 to 8). The NPH insulin was changed more when

the BG profiles' mean glycemic control was high, 2.71 units

(SD=1.72,range=0 to 12) than when it was low, 1.46 units

(SD=1.44,range=0 to 7). Table 17 provides the DIFF and

ABSDIFF insulin dose changes for the NPH insulin.

The consistent significant findings for #BGVALUES with

DIFF and ABSDIFF indicated that diabetologists' increased the

NPH insulin much more when 3 BG values per day were





















4I

0

























0


0
a)
w!
r








81

presented, compared to 1 BG value per day. The consistent

findings for the MEANBGCONTROL variable indicated that the

NPH insulin was increased more when the BG profiles' mean

glycemic control was high, compared to when it was low.


Morning insulin dose change. For morning insulin dose

administration DIFF, there were significant main effects for

FORMAT, #BGVALUES, and MEANBGCONTROL (see Table 26). The

morning administration of insulin was increased significantly

more when information was presented in computer format, +1.91

units (SD=2.04,range=-5 to +9) than when presented in logbook

format, +1.39 units (SD=2.30,range=-16 to +7). The morning

insulin was also increased more when 3 BG values per day were

presented, +1.88 units (SD=2.36,range=-16 to +9) than when 1

BG value per day was presented, +1.42 units (SD=1.88,range=-4

to +7). In addition, the morning insulin was increased much

more when the BG profiles' mean glycemic control was high,

+2.26 units (SD=1.63,range=-2 to +9) than when it was low,

+0.48 units (SD=2.63,range=-16 to +5).

For the dependent variable of ABSDIFF, there was a

significant main effect for #BGVALUES, as well as the FORMAT

and MEANBGCONTROL interaction (see Table 27). The absolute

change of the morning insulin dose was greater when 3 BG

values per day were presented, 2.50 units (SD=1.68,range=0 to

16) than when 1 BG value per day was presented, 1.87 units
























0
-p


0





a0r
gc










H









520

0



144
'44


*


r,

r-4
II
*O






.4.

*0

o0








II
)
ar-







IM
CM






U
)r4















*
*-1




















,-4 --4r-9 r- v-4 0 C
O 00 -40 CO 0
o oo wr o r" o
0 00 0 r- 0









1 W! < 1 a a
N m* i r Qn
.
o o












r-4 ,-4 (N



-' r-r -
r 0N


c,

0










-aI





m P
H










HP t
0
0 3









V)
.01






S44









E-4


* 4-)

S00






rH .
0 R4


r4

11






r-4



-4
0






$4
*O
Vm





0)
* P



II
U-<

0
O











* .
*dl k
aO
C:
M -
c-








84

(SD=1.44,range=0 to 7). Table 28 provides the mean absolute

change in the morning insulin dose, SD, and Tukey tests of

significance for the FORMAT and MEANBGCONTROL interaction. An

inspection of Table 28 indicates that the insulin dose was

changed significantly less when information was presented in

computer format and when the BG profiles' mean glycemic

control was low, compared to when presented in computer

format and the mean glycemic control was high. The

diabetologists' dose changes were not significantly different

when information was presented in logbook format, suggesting

that diabetologists' made greater insulin dose changes based

upon the BG profiles' mean glycemic control when information

was presented in computer format. Table 18 provides the DIFF

and ABSDIFF insulin dose changes for the morning insulin

administration.

The consistent findings for #BGVALUES with DIFF and

ABSDIFF indicated that the morning insulin dose

administration was increased more when 3 BG values per day

were presented, compared to 1 BG value per day.


Evening insulin dose change. For evening insulin dose

administration DIFF, there were no significant main or

interaction effects (see Table 29).

For the dependent variable of ABSDIFF, there was a

significant main effect for MEANBGCONTROL (see Table 30). The







85








0 0o 00
8 0 o


0 a 0


O o 0I
G 4I




.H
S(44 -4 N
V -H H
g, a,



0 U



CI

* 3W






II

0
.4.)0 0
H.^ ^ 8
H |8 g l




















-I -i i-4 in in r-
O 00 % N( r-I
04 o 00 Ln o r
0o oo -4 o o
*






m t- N r- 00 to
04 r o in o ua









4NJ M *

13 1 m V- '-4
(V ,-4,
00(3'" O


4,





r.
4.1



0


I






II
0 V




a

|-4










'44
ri








a1


* 4.1
E .
M
H i*r


0\o

4-
II



14





43
0



*

CM





* -
w-4









01
a,
a,


'-4--.C






'-4
C tl



















a,



go

4)
0
Q 9I
1 0




*4a



9C







O1


0)




E4-4
0)








88
absolute change of the evening insulin dose was greater when

the BG profiles' mean glycemic control was high, 1.90 units

(SD=1.50,range=0 to 11) than when it was low, 1.08 units

(SD=1.30,range=0 to 6). Table 18 provides the DIFF and

ABSDIFF insulin dose changes for the evening insulin

administration.

The inconsistent significant finding for MEANBGCONTROL

with ABSDIFF (p<.0011), but not with with DIFF (E<.0417)

indicated that the evening insulin dose administration was

frequently decreased when the BG profiles' mean glycemic

control was low.


Morning Regular insulin dose change. For morning

administration of the Regular insulin DIFF, there was a

significant main effect for FORMAT, as well as a #BGVALUES

and MEANBGCONTROL interaction (see Table 31). The morning

administration of Regular insulin was increased more when the

BG profiles' were presented in computer format, +0.79 units

(SD=1.19,range=-5 to +7) than when presented in logbook

format, +0.51 units (SD=1.36,range=-7 to +5). Table 32

provides the mean insulin dose increase, SD, and Tukey tests

of significance for the #BGVALUES and MEANBGCONTROL

interaction. An inspection of Table 32 suggests that the

morning administration of Regular insulin was increased more

when the BG profiles' mean.glycemic control was high,
























09
m
SOH





Co
0 r-I



0


CO 0





ri




4-)







VQ
(0)



'C
















O OO OO
Ln wf t~
0 00 00
0) 4) VP V V
r$4 4) 00 00r
Ho
.^0 tn o' t.-o
0 ++ ++
UU
H 00 0 0 43
0
fO
I -- dI

V014 00 c-4 a 4

rol
O '
C ; 4


o to tl
14 .m n 4 m >
-4 (d (n \ ON ow -I




Co o
.i r -l 0n r- in1


q4 C O;



C H -- --------- (v

en 43 V o
63 r, ri0

r C: to 3 r-* M *

E 4 Vs U 0
r4 ++ ++





>~ m
U

0 ,H 0

oi 4 >g8 >1>5
1 0
0) V 0 0





Ev i go:ro 4
Co E O m
a u d r




(D U
g|p S- : S1 u
c c R aS I




_* H
M








91
compared to low, but the increase was significantly greater

only when 3 BG values per day were presented.

For the dependent variable of ABSDIFF, there was a

significant main effect for #BGVALUES, as well as the FORMAT

and MEANBGCONTROL interaction (see Table 33). The absolute

change in the morning administration of Regular insulin was

greater when 3 BG values per day were presented, 1.09 units

(SD=1.14,range=0 to 7) than when 1 BG value per day was

presented, 0.78 units (SD=0.97,range=0 to 5). Table 34

provides the the absolute change in the morning

administration of Regular insulin, SD, and Tukey tests of

significance for the FORMAT and MEANBGCONTROL interaction. An

inspection of Table 34 indicates that the morning

administration of Regular insulin was changed the most when

information was presented in logbook format and the BG

profiles' mean glycemic control was low, with both increases

and decreases of the dose. The direction of the insulin dose

change was a more sensitive measure than the absolute

magnitude of change. The insulin dose was increased the least

when information was presented in logbook format and the BG

profiles' mean glycemic control was low. Table 19 provides

the DIFF and ABSDIFF insulin dose changes for the morning

Regular insulin administration.

















0



II

'r4
0







40
"I
4-)








U
2c,

I3
m

<0



'44




44rz
wVn
I!





E-4(
&I


0





U
IRI
4 *




'O
N




U
(
00




4r-4
VH

00




4a
0
m'-



&
















H
*S c

H0




$18
0
wO
044

4-)

Ug



C13
'I44
0V
.d








V P4
0
m
43

H

1^
V







0r~
go
wo
4

tp
*s

2

u^
r. V

r1 0
0 D




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 EQNXMM7YN_DK993Q INGEST_TIME 2012-02-29T17:44:58Z PACKAGE AA00009059_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES