You are on page 1of 11

Lauren Silver HPM 587 Fall 2009

PREDICTORS OF PARTICIPATION IN A
DIABETES DISEASE MANAGEMENT PROGRAM
INTRODUCTION
In a survey of medical directors of 65 health maintenance organizations, the three
diseases most often targeted by disease management programs were diabetes, asthma, and
congestive heart failure.
1
Diabetes disease management programs, in particular, are on the rise.
2

his should come as no surprise given that diabetes affects appro!imately 1" million #mericans
and contributes substantially to rising health care costs.
$
hus, most diabetes disease
management programs focus on how to improve diabetes outcomes, particularly by teaching
participants how to manage their own condition via self%care techni&ues.
'
Indeed, a recent
analysis of the impact of a national managed care organization(s diabetes disease management
program)entailing repeated telephone outreach by nurses, dietitians, and health educators as
well as web% and mail%based reminders and education)found that the program was associated
with 1* lower overall costs of care within one year at participating sites +22%$,- decrease in
hospitalizations*, 2* a slight decline in physician office visits, and $* improved diabetes%related
./DI0 +.ealth /mployer Data and Information 0et* and non%./DI0 &uality measures.
5

1iven the potential for diabetes disease management programs to improve diabetes
outcomes, this study see2s to identify predictors of participation in these programs)who is most
li2ely and least li2ely to participate3 4ore specifically, both demographic and clinical correlates
of participation in a diabetes disease management program are e!amined. 5y identifying
predictors of participation, health plan outreach activities can be tailored so as to increase the
1
Lauren Silver HPM 587 Fall 2009
rates of participation among those least li2ely to parta2e, thereby narrowing disparities in
diabetes health outcomes and improving outcomes, overall.
BACKGROUND
# number of demographic and clinical characteristics are associated with diabetes
patients( level of self%care, including race, age, health literacy, disease duration, and use of
medication, among others.
6rior research has demonstrated that diabetes patients with the highest ris2 for
complications are those least li2ely to self%monitor blood glucose levels.
6
In a study of ',565
adult, diabetic enrollees in an eastern 4assachusetts managed care organization, #dams et al.
found that the elderly, minorities, and those living in low socioeconomic neighborhoods were
significantly less li2ely to self%monitor than their younger, 7aucasian, and more well%to%do
counterparts. #dditionally, even among those ta2ing insulin, the effect of socioeconomic status
and age on self%monitoring persisted, however, minorities were more li2ely to self%monitor than
whites if ta2ing insulin.
8
It also has been established that health literacy influences the ability of those with chronic
conditions to self%monitor by posing barriers to educating patients about their condition. In a
survey of ',2 patients with hypertension and 11' patients with diabetes, 9illiams et al. found
that :almost half of the patients;had inade&uate functional health literacy, and these patients
had significantly less 2nowledge of their disease, important lifestyle modifications, and essential
self%management s2ills.<
"
hus, it follows that education level li2ely plays a significant role in
whether diabetics will foresee as well as reap the benefit of participating in a disease
management program. his study see2s to further confirm as well as e!pand on the literature
2
Lauren Silver HPM 587 Fall 2009
regarding predictors of diabetes self%management by e!amining participation in a diabetes
disease management program.
DATA SOURCE
Data for this study came from the 7enters for Disease 7ontrol and 6revention(s +7D7*
2,," 5ehavioral =is2 >actor 0urveillance 0ystem +5=>00* survey, a cross%sectional telephone
survey conducted by state health departments with technical assistance provided by the 7D7.
?nly one adult)1" years or older)per household is interviewed in all 5, states, the District of
7olumbia, 6uerto =ico, the @.0. Airgin Islands, and 1uam.. Information is collected on health
ris2 behaviors +e.g., diet and e!ercise*, preventive health practices, health care access, chronic
conditions, and sociodemographic characteristics. In the 2,," 5=>00, there was a total of
'1',5,B observations.
METHODS
Dependent Variable
he dependent variable e!amined in this study is whether a 5=>00 respondent attended a
class in managing diabetes, which was operationalized through the following survey &uestionC
:.ave you ever ta2en a course or class in how to manage your diabetes yourself3< >or the
present study, this variable is a binary categorical variable coded as , for :no< and 1 for :yes.<
In 2,,", a total of $2,8B5 respondents provided either a :yes< or :no< answer to this
&uestion)18,B$$ responded :yes< +5'.6"-* while 1',"62 responded :no< +'5.$2-*. hus, more
than half of the respondents to which diabetes%related &uestions applied reported ever attending a
diabetes disease management program in 2,,".
$
Lauren Silver HPM 587 Fall 2009
Independent Variables
he independent variables included in the model +able 1* were chosen based on findings
from the literature regarding the association between diabetes self%care and demographic and
clinical characteristics. 4ore specifically, predictors were chosen based on whether research to
date has demonstrated a consistent association with participating in a disease management
program and whether research to date has not fully established a relationship between an
e!planatory variable and participation +e.g., marital status as a pro!y for social support*. 5elow
is the model e!amined.
DIABEDU
i
= f(RACE
i
, INC
i
, EDU
i
, AGE
i
, SEX
i
, EYE
i,
MAR
i,
INS
i,
HP
i
) + , where
=#7/i D a five%category dummy variable representing the ith respondent(s reported preferred
race +:5lac2 or #frican #merican,< :#sian,< :Eative .awaiian or other 6acific Islander,<
:#merican Indian or #las2an Eative,< compared to :9hite only< as the reference group*.
IE7i D an eight%category dummy variable representing the ith respondent(s reported annual
income +:less than F1,,,,,,< :F1,,,,, to less than F15,,,,,< :F15,,,, to less than F2,,,,,,<
F2,,,,, to less than F25,,,,,< :F25,,,, to less than F$5,,,,,< :F$5,,,, to less than F5,,,,,,<
and :F5,,,,, to less than F85,,,,< compared to :greater than or e&ual to F85,,,,< as the
reference group*
/D@i D a four%category dummy variable representing the ith respondent(s reported education
level +:did not graduate high school,< :graduated high school,< and :attended college or
technical school,< compared to :graduated college or technical school< as the reference group*
#1/i D a si!%category dummy variable representing the ith respondent(s age in years +:1" to 2',<
:25 to $',< :$5 to '',< :'5 to 5',< and :55 to 6'< compared to :65 or older< as the reference
group*
0/Gi D a dummy variable e&ual to 1 if the ith respondent(s se! was male, , otherwise, with male
as the reference group
/H/i D a dummy variable e&ual to 1 if the ith respondent reported ever being told by a doctor
that diabetes has affected his or her eyes or that he or she had retinopathy, , otherwise, with no as
the reference group
'
Lauren Silver HPM 587 Fall 2009
4#=i D a dummy variable representing the ith respondent(s marital status +:divorced,<
:widowed,< :separated,< :never married,< and :member of unmarried couple< compared to
:married< as the reference group*
IE0i D a dummy variable e&ual to 1 if the ith respondent reported he or she currently is ta2ing
insulin, , otherwise, with not ta2ing insulin as the reference group
.6i D a dummy variable e&ual to 1 if the ith respondent reported having current health insurance
coverage, , otherwise, with having health insurance coverage as the reference group
Statistical Analysis
1iven that the outcome variable of interest is binary +,,1*, a probit analysis was
performed using the dprobit command in 0## in order to calculate averageIpartial effects for
each of the categorical independent variables. hus, partial effects were calculated for each
independent variable while all other independent variables were held at their sample mean.
Hypothesized Signs of the Coefficients
6rior to conducting the analysis, the directions of the estimated coefficients for each
independent variable were hypothesized. >or the demographic characteristics, it was
hypothesized that females, whites, and those who are married or younger would be more li2ely
to participate in a diabetes disease management program. hus, relative to their respective
reference groups, a negative coefficient should appear for the various categories of =#7/i and
4#=i while a positive coefficient should appear for #1/i and 0/Gi. >urthermore, higher levels
of income, higher levels of education, and health insurance coverage were e!pected to
correspond to a higher li2elihood of participation. =elative to their respective references groups,
/D@i, IE7i and .6i should e!hibit negative coefficients. >or the clinical characteristics, those
ta2ing insulin and those who had been told by a doctor that their diabetes has affected their eyes
or contributed to retinopathy were e!pected to participate more. hus, relative to their respective
reference groups, positive coefficients should emerge for /H/i and IE0i.
5
Lauren Silver HPM 587 Fall 2009
RESULTS
Descriptive Statistics
able 1 presents the fre&uency and percentage of respondents falling into the various
categories of the independent variables e!amined.
Table 1 De!"#i$%i&e S%a%i!%i"! f'#
De(')#a$*i" a+, Cli+i"al P#e,i"%'#!, BR-SS .//0
Independent Variable (n) Frequency (%)
Race (31.252)
White 24,886 79.63
Black or African American 4,658 14.90
Asian 504 1.61
ati!e "a#aiian or other $acific %slan&er 194 0.62
American %n&ian or Alaskan ati!e 1,010 3.23
%ncome (28,105)
'(10,000 2,972 10.57
(10,000 to '(15,000 3,034 10.8
(15,000 to '(20,000 3,371 11.99
(20,000 to '(25,000 3,748 13.34
(25,000 to '(35,000 4,068 14.47
(35,000 to '(50,000 4,063 14.46
(50,000 to '(75,000 3,342 11.89
)(75,000 3,507 12.48
*&+cation (32,681)
,i& not -ra&+ate hi-h school 5,811 17.78
"i-h school -ra&+ate 11,427 34.97
Atten&e& colle-e or technical school 8,554 26.17
.ra&+ate& colle-e or technical school 6,889 21.08
/e0 (32,795)
1ale 13,123 40.02
2emale 19,672 59.98
"ealth 3lan co!era-e (32,736)
4es 30,050 91.79
o 2,686 8.21
1arital stat+s (32,671)
1arrie& 15,992 48.95
,i!orce& 5,360 16.41
Wi&o#e& 7,308 22.37
/e3arate& 843 2.58
e!er marrie& 2,810 8.60
1em5er of +nmarrie& co+3le 358 1.10
6akin- ins+lin (32,767)
4es 9,024 27.54
o 23,743 72.46
*!er tol& 57 &octor &ia5etes has affecte& 7o+r
e7es8retino3ath79 (32,411)
4es 6,580 20.30
o 25,831 79.70
6
Lauren Silver HPM 587 Fall 2009
?nly those respondents who answered either :yes< or :no< to the survey &uestion regarding
participation in a diabetes disease management program were included in the analysis. Eearly ",
percent of the respondents were white, appro!imately 6, percent were female, nine out of ten
had health insurance coverage, and close to half had at least some college or technical school
education. #bout one in four respondents were ta2ing insulin and about one in five indicated that
they had been told by a doctor at some point that diabetes has affected their eyes.
Statistical Analysis Results
#mong the demographic predictors e!amined, all demonstrate a statistically significant
relationship with participation in a diabetes disease management program. 5lac2 or #frican
#merican and #merican Indian or #las2an Eative respondents were significantly more li2ely to
report attending a diabetes disease management program than whites)the probability that a
5lac2 or #frican #merican or an #merican Indian or #las2an Eative participated was ,.,$1 and
,.,'$ higher than for whites +able 2*. In contrast, #sian and 6acific Islander or Eative .awaiian
respondents were significantly less li2ely to participate than whites. >or income, respondents
earning less than F$5,,,, were significantly less li2ely to participate in a diabetes disease
management program than those earning greater than F85,,,, per year. 5y way of e!ample, the
probability that those earning less than F1,,,,, per year participated in a diabetes disease
management program was ,.126 less than those earning more than F85,,,,. Ji2ewise,
respondents who did not graduate from high school or who graduated from high school but did
not attend college or technical school were significantly less li2ely to participate than those who
graduated from college or technical school. 4ore specifically, the probability that a respondent
who did not graduate from high school participated in a diabetes disease management program
was ,.1B2 less than for those who graduated from college or technical school.
8
Lauren Silver HPM 587 Fall 2009
Table . T*e I($a"% 'f De(')#a$*i" a+, Cli+i"al C*a#a"%e#i!%i"! '+
Pa#%i"i$a%i'+ i+ a Diabe%e! Di!ea!e Ma+a)e(e+% P#')#a(, BR-SS .//0
Independent Variable
(n=26,494)
dF/dx
SE !"alue
Race
Black or African American .031 .009 0.001:::
Asian ;.086 .024 0.000:::
$acific %slan&er or ati!e "a#aiian ;.045 .038 0.244
American %n&ian or Alaskan ati!e .043 .017 0.014::
%ncome
'(10,000 ;.126 .015 0.000:::
(10,000 to '(15,000 ;.069 .015 0.000:::
(15,000 to '(20,000 ;.051 .014 0.000:::
(20,000 to '(25,000 ;.025 .013 0.058:
(25,000 to '(35,000 ;.021 .013 0.097:
(35,000 to '(50,000 ;.001 .012 0.958
(50,000 to '(75,000 ;.004 .013 0.767
*&+cation
,i& not -ra&+ate hi-h school ;.192 .011 0.000:::
"i-h school -ra&+ate ;.076 .009 0.000:::
Atten&e& colle-e or technical school ;.014 .009 0.136
A-e in 7ears
18 to 24 .046 .060 0.446
25 to 34 .018 .025 0.475
35 to 44 .049 .014 0.001:::
45 to 54 .035 .010 0.000:::
55 to 64 .037 .008 0.000:::
/e0
2emale .073 .007 0.000:::
"ealth 3lan co!era-e (32,736)
o ;.066 .012 0.000:::
1arital stat+s
,i!orce& .006 .009 0.509
Wi&o#e& ;.042 .009 0.000:::
/e3arate& ;.017 .021 0.413
e!er marrie& ;.036 .012 0.003::
1em5er of +nmarrie& co+3le ;.015 .030 0.606
6akin- ins+lin
4es .162 .007 0.000:::
*!er tol& 57 &octor &ia5etes has affecte& 7o+r
e7es8retino3ath79
4es .052 .008 0.000:::
:::/i-nificant at <=0.001> ::/i-nificant at <=0.05> :/i-nificant at <=0.10
hose aged $5 to 6' years old were significantly more li2ely to participate than those 65
years old or older and females were significantly more li2ely to participate than males. >inally,
those with health insurance coverage were significantly more li2ely to report having participated
"
Lauren Silver HPM 587 Fall 2009
in a diabetes disease management program while those who were widowed or never married
were significantly less li2ely to report participating.
5oth clinical predictors e!amined)whether respondents currently were ta2ing insulin to
manage their diabetes and whether respondents had ever been told by a physician that diabetes
has affected their eyes or contributed to retinopathy)were statistically significantly associated
with a higher probability of participating in a diabetes disease management program. he
probability of participation for respondents ta2ing insulin was ,.162 higher than for those not
ta2ing insulin while the probability of participation for those who had been told diabetes affected
their eyes was ,.,52 higher than for those who had not been told diabetes affected their eyes.
CONCLUSIONS
his study has corroborated some findings in the literature while also providing conflicting
results that merit further research. 9hile minorities have been shown to participate less than
whites in diabetes disease management programs, results from this study indicate that 5lac2 or
#frican #merican respondents were more li2ely to participate. hus, further research is needed
to clarify the impact of race on participation in disease management programs. Jow
socioeconomic status and a low level of educational attainment clearly impose barriers to
participating in a diabetes disease management program)whether these barriers are more
strongly related to access problems versus health literacy should be e!amined. >inally, the
clinical predictors e!amined in this study provide evidence that patients with a more complicated
case of diabetes)i.e., re&uiring insulin and e!periencing eye problems)are more li2ely to
participate in programs focused on how to better manage their own care. hese findings may be
useful for health plans offering diabetes disease management programs by clarifying the
B
Lauren Silver HPM 587 Fall 2009
characteristics of diabetic enrollees who are most and least li2ely to participate in these
programs, thereby identifying potential targets for outreach.
REFERENCES
1,
1
9elch 96, 5ergsten 7, 7utler 7, 5occhino 7, 0mith =I. Disease management practices of health plans. he #merican
Kournal of 4anaged 7are. 2,,2 #prL"+'*C$5$%61.
2
Aillagra A1 and #hmed, . /ffectiveness of a disease management program for patients with diabetes. .ealth #ffairs.
2,,' Kul%#ugL2$+'*C255%66.
$
Ibid.
'
9elch 96, 5ergsten 7, 7utler 7, 5occhino 7, 0mith =I. Disease management practices of health plans. he #merican
Kournal of 4anaged 7are. 2,,2 #prL"+'*C$5$%61.
5
Aillagra A1 and #hmed, . /ffectiveness of a disease management program for patients with diabetes. .ealth #ffairs.
2,,' Kul%#ugL2$+'*C255%66.
6
#dams #0, 4ah 7, 0oumerai 05, Mhang >, 5arton 45, =oss%Degnan, D. 5arriers to self%monitoring of blood glucose
among adults with diabetes in an .4?C a cross%sectional study. 547 .ealth 0ervices =esearch. 2,,$ 4ar 1BL$+1*C6.
8
Ibid.
"
9illiams 4A, 5a2er D9, 6ar2er =4, Eurss, K=. =elationship of functional health literacy to patients( 2nowledge of
their chronic disease. # study of patients with hypertension and diabetes. #rchives of Internal 4edicine. 1BB" Kan
26L15"+2*C166%82.

You might also like