No. 4, October 2004
Predictors of Dropouts From a San Diego Diabetes Program: A Case Control
Stephen R. Benoit, MD, MPH, Ming Ji, PhD, Regina Fleming, MD, MSPH, Athena
Suggested citation for this article: Benoit SR, Ji M, Fleming R,
Philis-Tsimikas A. Predictors of dropouts from a San Diego diabetes program: a
case control study. Prev Chronic Dis [serial online] 2004 Oct [date
cited]. Available from: URL:
The objective of this study was to determine the demographic, treatment,
clinical, and behavioral factors associated with dropping out of a nurse-based,
low-income, multiethnic San Diego diabetes program.
Data were collected during a 17-month period in 2000 and 2002 on patients
with type 2 diabetes from Project Dulce, a disease management program in San
Diego County designed to care for an underserved diabetic population. The study
sample included 69 cases and 504 controls representing a racial/ethnic mix of
53% Hispanic, 7% black, 16% Asian, 22% white, and 2% other. Logistic regression
was used to determine factors associated with patient dropout.
Patients who had high initial clinical indicators including blood pressure
and hemoglobin A1c and those who smoked currently or smoked in the past were
more likely to drop out of the diabetes program.
This study provides markers of patient dropout in a low-income, multiethnic,
type 2 diabetic population. Reasons for dropout in this program can be
investigated to prevent further cohort loss.
Back to top
More than 15 million people have diabetes in the United States, and it is now
the sixth leading cause of death (1,2). The United Kingdom Prospective Diabetes
Study and Kumamoto Study confirm that improved glucose control reduces the microvascular complications of type 2 diabetes such as retinopathy, nephropathy,
and neuropathy (3,4). Because of these findings, new standards of care and new
models of health care delivery have emerged to reduce the complications of
diabetes and thus improve quality of life (5).
Nonadherence to diabetes treatment strategies prevents patients from meeting
optimal standards of care. Jacobsen et al found that up to 34% of patients with
type 1 diabetes defaulted from care, and their glycemic control was
significantly worse than those who attended their visits (6). Evidence suggests
that patients with diabetes who are unable or unwilling to adhere to treatment
regimens suffer greater morbidity than those under regular medical supervision
Studies have investigated patients’ reasons for non-attendance in diabetes
care clinics. Reasons include financial or transportation issues, inability to
get time off work, forgetting about appointments, feeling too ill, crowded
clinic settings, administrative errors, and feeling the appointment was
unnecessary. Other studies have addressed nonadherence in diabetes clinic
settings and found significant factors to be smoking, poor education, living
long distances from the clinic, and dietary treatments (8).
The purpose of this study was to investigate the factors associated with
patients dropping out of Project Dulce, a nurse-based diabetes
disease-management system in San Diego, Calif (9). The factors studied included
demographic, treatment, clinical, and behavioral variables. The program, started
in 1998, is an initiative of Community Health Improvement Partners, the Council
of Community Clinics, and The Whittier Institute for Diabetes. Project Dulce
uses a multifaceted approach and focuses on providing care to racial/ethnic
groups that often lack access to medical services.
Back to top
Data study sample
Patients with diabetes are referred to Project Dulce by primary care
providers. After the patient is referred, a nurse educator conducts an initial
assessment and follows the American Diabetes Association (ADA) standards of
appropriate physical and laboratory exams and referrals to specialists (e.g.,
ophthalmologists, podiatrists). The nurse educator is the case manager and
follows up on missed patient appointments in addition to identifying individual
service and access needs of his or her panel of patients. The nurse also
communicates with the primary care physician regarding clinical-care issues.
Dieticians are also on staff at Project Dulce to meet with patients referred by
the nurse educators. The program is currently active in 17 sites, including
community clinics and hospital ambulatory care centers throughout San Diego
County. Project Dulce uses the same procedures at each site and tracks patients
with diabetes electronic management system (DEMS) software. The database
contains demographic, treatment, clinical, and behavioral factors for each
patient and collects the information over time. This study included data from
July 18, 2000 to October 7, 2002 and was approved by the Institutional Review
Board of San Diego State University.
Patients with type 2 diabetes were first selected from the database. This
reduced the population size to 1357 from 1728. Case-control methodology was used
in the analysis. We defined cases as patients who dropped out of the program and
selected them using the following inclusion criteria: the patient could not have
more than two Project Dulce visits; could not be in the program more than three
months; needed a baseline A1c value; and the last Project Dulce visit had to be
at least six months before October 7, 2002 (the last Project Dulce visit date in
the database). The control group consisted of active patients in Project Dulce,
and these patients were selected using the following criteria: each patient had
to have at least two A1c values at least six months apart, had to be in the
program for at least six months, and had to have at least three Project Dulce
visits. Eighteen cases (dropouts) and 69 controls were excluded due to missing
analytic variables, leaving 69 dropouts and 504 controls.
Measures and diagnostics
All variables came from the DEMS database and were grouped into five
clusters: demographic factors, disease duration, treatment factors, clinical
characteristics, and behavioral factors.
The five variables in the demographic cluster were sex, age, race/ethnicity,
primary language, and insurance. For purposes of this study, we created five
racial/ethnic categories: Hispanic, black, Asian (including Eastern Indian),
white, and other.
Most of the patients in Project Dulce have County Medical Services, an
insurance program funded by San Diego County to care for the medically indigent
adult (MIA) population. However, some patients who surpass maximum income limits
or who are not documented are uninsured and pay out-of-pocket to enroll in the
program. A smaller proportion of the patients have Medicare, Medicaid, or
private insurance. For purposes of this study, insurance status was categorized
as uninsured, County Medical Services (representing the MIA population), or
insured (including Medicare, Medicaid, or private insurance).
Disease duration was estimated by subtracting the diabetes diagnosis date
from the date of the patient’s initial Project Dulce visit. The treatment factor
cluster was represented by the type of medicine the patient was using at the
initial visit. The medicines used for glucose control (e.g., insulin,
sulfonylureas, metformin, glitazones, alpha glucosidase inhibitors, meglitinides)
were categorized into three levels: insulin alone or insulin with oral agents,
more than one oral agent but no insulin, and one oral medication or no
medication at all.
Clinical characteristics included baseline systolic (SBP) and diastolic (DBP) blood pressure, Body Mass Index, and
baseline A1c. Blood pressure was categorized according to American Diabetes
Association standards for optimal control: <130 mm Hg for SBP and <80 mm Hg
for DBP (5). The behavioral factor
in the model was smoking status. Variables left in the continuous form were
assessed for nonlinearity. Collinearity among independent variables was
assessed using tolerance values. All tolerance values were substantially greater
than 0.10, indicating that collinearity was not an issue.
Variable screening was done using univariate logistic regression with an
alpha significance level of 0.25. Those variables significant in univariate
analysis were placed in a multivariate model. Variables not significant at the
alpha level of 0.05 were assessed as confounders. Each potential confounder was
added to the exposures of interest one by one. Parameter estimate changes of
greater than 20% were considered significant. Sex, primary language, and
baseline DBP were considered confounders and were therefore included in the
Back to top
Table 1 shows the descriptive and univariate statistics. There appear to
have been fewer Asians and more whites in the dropout group compared with the
control group, while the percentage of Hispanics remained consistent among both
groups. More of the dropout population (58.6%) spoke English than the control
population (47.2%), and more of the dropout population was insured by County
Medical Service (60.9%) than the control population (43.5%). Among the dropouts,
high baseline blood pressures (defined here as ≥ 130 mm Hg for SBP and ≥ 80 mm
Hg for DBP) were more common compared to the control population: 46.1% of the
dropout population had high SBP compared to 29.5% of the control, and 26.9% of
the dropout population had high DBP compared to 11.7% of control. More of the
dropouts compared to controls were current smokers (20.3%, dropouts compared to
12.5%, controls) or past smokers (47.3%, dropouts compared to 32.0%, controls).
Table 2 presents the results of the final logistic model. Insurance status,
initial blood pressure, baseline A1c, and smoking habit were significant
predictors of dropout status. The odds of dropping out of Project Dulce were 95%
lower for a patient without insurance compared to a patient with insurance. The
odds of being a dropout were 1.8 times higher for a patient with a SBP ≥130 mm
Hg compared to a patient with a SBP <130 mm Hg and 2.3 times higher for a
patient with a DBP ≥80 mm Hg compared to a patient with DBP <80 mm Hg. For
every two-unit increase in baseline A1c, the odds of being a dropout increased
1.3 times. Compared with a patient who never smoked, the odds
of dropping out of the program were 3.7 times higher for a current smoker and
2.9 times higher for a past smoker. The variables mentioned above were
significant after controlling for all other variables in the model (sex, primary
language, insurance status, blood pressure, baseline A1c, and smoking status).
Back to top
Univariate associations between dropouts and controls showed that
race/ethnicity, primary language, insurance status, baseline blood pressure, and
smoking varied among the two groups. In multivariate analysis, insurance status,
baseline blood pressure, baseline A1c, and smoking habit were different between
dropouts and controls.
Similar to the findings of Graber et al (10) and Jacobson et al (6), sex and
age were not significant predictors of dropout. Other studies have found that
men and younger patients are less likely to adhere to treatment (11-13).
Race/ethnicity was a predictor of dropout in univariate analysis but lost its
controlling for other factors. Dove
and Schneider (14) and Goldman et al (11) also found this to be the case.
Primary language also lost its significance in multivariate analysis. Disease
duration and type of medicine used to treat diabetes did not predict patient
adherence to the program. This finding is consistent with Jacobson et al (6) and
Graber et al (10) but contradicts Irvine and Mitchell (15).
Insurance status was a significant predictor of patient adherence. The
uninsured were 95% less likely to drop out of the program compared to those with
insurance. The uninsured pay an enrollment fee to cover nurse visits, laboratory
measurements, and diabetes-related medications. They are not eligible for county
or federal health services because of documentation status or income limits. The
initial monetary investment and lack of affordable health care options are
clearly reasons that the uninsured were less likely to drop out. Those with
County Medical Services and Medicaid/Medicare do not pay to enroll in Project
Dulce and have other provider options for their diabetes care.
Those with high initial blood pressure or A1c were more likely to drop out of
the program. These clinical markers may indicate that these patients were more
ill and conceivably unable to keep their outpatient appointments. Perhaps their
time was occupied with other primary care or specialty visits to address other
medical problems. Regardless of the reasons for patient nonadherence,
these clinical markers are important. Higher A1c levels put the cases at
increased risk for developing microvascular disease, including retinopathy,
nephropathy, and neuropathy (3,4). Elevated initial blood pressure put the cases at increased risk of heart attacks, heart failure,
strokes, and kidney disease (16).
Smokers that have type 2 diabetes are at even greater risk of micro- and
macrovascular disease (17). This study showed that those who smoked or smoked in
the past were more likely to drop out of the program. Why these patients are leaving is not clear. Perhaps smokers are less interested in their health, which
would explain their lack of interest in Project Dulce. Another possible
explanation is that smokers feel rejected by providers that admonish their
behavior (10). Smoking could also be a surrogate for life stressors that prevent
appointment attendance due to unstable living or financial environments (18).
Regardless of the reason, the clinical and behavioral markers that predict nonadherence
leave these patients at high risk of morbidity and mortality
associated with microvascular and macrovascular disease (19).
This study has limitations. According to Griffin (8), adherence may be less
related to demographic factors and more to patients’ perceptions, beliefs, and
attitudes. Perhaps some patients may not have understood the importance of
consistent glucose control, had a different perception of the disease, or were
fearful of the multiple medications needed to treat diabetes. Others may have
experienced adverse side effects to medication, stopped treatment, and were
ashamed to return to their provider. Patient perceptions could explain
nonadherence, but none of this information was available for this study.
Similarly, other possible confounders, such as distance living from the
clinic, transportation issues, work schedules, whether they knew they had an
appointment, stability of living environment, alcohol and drug problems, other
comorbid medical/psychological problems, and patient satisfaction and ability to
talk with providers, were not available. These factors could also significantly
predict dropouts in Project Dulce. Some of the demographic, clinical, and
behavioral factors significant in this model may lose statistical significance
depending on the importance of the unmeasured predictors. Lastly, 18 dropouts
and 69 controls were excluded because of missing data, which could potentially
introduce a selection bias. Because patient identifiers were scrambled for
privacy prior to copying the dataset, medical records could not be reviewed to
recover this data.
In summary, with the available data, this predictor analysis found that
Project Dulce patients with higher initial blood pressure readings and A1c
values and those who smoked or smoked in the past were more likely to drop out
of the program. These results have clinical applicability in that those patients
in most need of care were more likely to drop out of the program. Investigating
reasons for dropping out would best be done with patient interviews. In this
way, patient perceptions, attitudes, and beliefs could be explored in addition
to other logistical reasons.
Project Dulce can use the results of this study to minimize future patient
loss to follow-up. Although all patients are case-managed by nurses, an
intensified approach would be beneficial in those patients with higher initial
A1c and blood pressure measurements and in those who smoke or smoked in
the past. Acquiring additional information from patients who drop out could help
explain the obstacles facing this low-income, multiethnic population. Addressing
these obstacles with interventions and/or organizational changes may lead to
better clinical outcomes.
Back to top
Corresponding author: Stephen R. Benoit, MD, MPH, UCSD-SDSU Preventive
Medicine Residency, University of California, San Diego Department of Family and
Preventive Medicine, La Jolla, Calif. Dr. Benoit is now with the Centers for
Disease Control and Prevention, 1600 Clifton Rd, Mail Stop E-68, Atlanta, GA
30333. E-mail: email@example.com.
Author affiliations: Ming Ji, PhD, San Diego State Graduate School of Public
Health, San Diego, Calif; Regina Fleming, MD, MSPH, UCSD-SDSU Preventive Medicine Residency;
Athena Philis-Tsimikas, MD, The Whittier Institute for Diabetes, La Jolla, Calif.
Back to top
- National Institute of Diabetes & Digestive & Kidney Diseases.
Diabetes Statistics fact sheet: general information and national estimates on
diabetes in the United States, 2000. Bethesda (MD): U.S. Department of Health and Human
Services, National Institutes of Health; 2002. NIH Publication No.:
- Arias E, Smith BL.
Deaths: preliminary data for 2001. Hyattsville (MD):
U.S. Department of Health and Human Services, National Center for Health Statistics,
Centers for Disease Control and Prevention; 2001. National vital statistics reports: vol. 51(5).
Intensive blood-glucose control
with sulphonylureas or insulin compared with conventional treatment and risk
of complications in patients with type 2 diabetes (UKPDS 33).
UK Prospective Diabetes Study (UKPDS) Group. Lancet 1998;
- Ohkubo Y, Kishikawa H, Araki E, Miyata T, Isami S, Motoyoshi S, et al.
Intensive insulin therapy prevents the progression of diabetic microvascular
complications in Japanese patients with non-insulin-dependent diabetes
mellitus: a randomized prospective 6-year study. Diabetes Res Clin Pract 1995;28:103-17.
- American Diabetes Association. Standards of medical care for patients
with diabetes mellitus. Diabetes Care 2003; 26:33-50.
- Jacobson AM, Adler AG, Derby L, Anderson BJ, Wofsdorf JI.
attendance and glycemic control. Study of contrasting groups of patients with
IDDM. Diabetes Care 1991;14:599-601.
- Hammersley MS, Holland MR, Walford S, Thorn PA.
What happens to
defaulters from a diabetic clinic? BMJ 1985; 291:1330-2.
- Griffin SJ.
Lost to follow-up: the problem of defaulters from diabetes
clinics. Diabet Med 1998;15(3):S14-S24.
- Philis-Tsimikas A, Walker C.
Improved care for diabetes in underserved
populations. J Ambul Care Manage 2001;24(1):39-43.
- Graber AL, Davidson P, Brown AW, McRae JR, Woolridge K.
relapse during diabetes care. Diabetes Care 1992;15:1477-83.
- Goldman L, Freidin R, Cook EF, Eigner J, Grich P.
approach to the prediction of no-show behavior in a primary care center. Arch
Intern Med 1982;142:563-7.
- Hurtado AV, Greenlick MR, Colombo TJ.
Determinants of medical care
utilization. Failure to keep appointments. Med Care 1973;11:189-98.
- Frankel S, Farrow A, West R.
Non-attendance or non-invitation? A
case-control study of failed outpatient appointments. BMJ 1989;298:1343-5.
- Dove HG, Schneider KC.
The usefulness of patients’ individual
characteristics in predicting no-shows in outpatient clinics. Med Care 1981;19:734-40.
- Irvine AA, Mitchell CM.
Impact of community-based diabetes education
on program attenders and nonattenders. Diabetes Educ 1992;18:29-33.
- National High Blood Pressure Education Program.
The Seventh Report of
the Joint National Committee on Prevention, Detection, Evaluation, and
Treatment of High Blood Pressure. Bethesda (MD): U.S. Department of Health and Human
Services, National Institutes of Health, National Heart, Lung, and Blood
Institute; 2003 NIH Publication No.: 03-5233.
- Eliasson B.
Cigarette smoking and diabetes. Prog Cardiovasc Dis 2003;45(5):405-13.
- McKee SA, Maciejewski PK, Falba T, Mazure CM.
Sex differences in the
effects of stressful life events on changes in smoking status. Addiction 2003;98(6):847-55.
- Bishop DB, Zimmerman BR, Roesler JS. Diabetes. In: Brownson RC,
Remington PL, Davis JR, eds. Chronic Disease Epidemiology and Control. 2nd
ed. Washington (DC): The American Public Health Association; 1998. p. 421-64.
Back to top