In the U.S., there has been a rapid increase in emergency department visits, and studies indicate that up to 50% of emergency department visits are for nonurgent problems (1,2). Several studies have suggested that individuals with diabetes are significantly more likely to use the emergency department than their nondiabetic counterparts (35). However, most of these studies did not adequately control for key factors that influence health utilization, and many did not explore potential predictors of emergency department use in individuals with diabetes.

The objectives of this study were to use nationally representative data to 1) compare patterns of emergency department use in people with and without diabetes, 2) compare the odds of emergency department use in both groups while controlling for potential confounders, and 3) determine correlates of single and multiple emergency department use in people with diabetes. It was hypothesized that while controlling for known confounders, the odds of emergency department use would not differ significantly between people with and without diabetes.

Data from the sample adult core of 1999 National Health Interview Survey (NHIS) (6) were analyzed. The NHIS is a nationally representative household survey of U.S. adults aged ≥18 years. The sample is selected by a complex sampling design involving stratification, clustering, and multistage sampling with a nonzero probability of selection for each person. Final weights allow estimates from the NHIS to be generalized to the adult civilian population of the U.S. (7).

The diagnosis of diabetes was based on self-report. Emergency department use was defined as having one or more hospital emergency room visit in the previous 12 months. Emergency department use was further categorized according to the number of visits in the previous 12 months: none, one, two to three, and four or more. Confounding variables included age (18–34, 35–49, 50–64, and ≥65 years), race/ethnicity (white, black, Hispanic, and other), education (less than high school graduate and high school graduate or greater), employment (employed versus unemployed), household income (<$20,000 and ≥$20,000), having a single usual source of medical care in the previous 12 months, and perceived health status (defined as better or same versus worse compared with previous 12 months). Other variables included comorbidity (defined as self-reported diagnosis of coronary artery disease, heart failure, chronic obstructive pulmonary disease, stroke, end-stage renal failure, chronic liver disease, or cancer). An additional variable was created for individuals with diabetes to assess presence or absence of diabetes-related complications (defined as the presence of any of the following self-reported conditions: coronary artery disease, stroke, end-stage renal failure, macular degeneration, retinopathy, or blindness).

Statistical analysis was performed with Stata version 7.0 (8). Sample characteristics and patterns of emergency department use were compared among individuals with and without diabetes using χ2 statistics. Multiple logistic regression models were used to compare odds of emergency department use in people with and without diabetes while controlling for confounders. Confounding variables were selected for sequential entry into the model based on the behavioral model of health services use (9). Predisposing factors included age, sex, race/ethnicity, education, and employment. Enabling factors included household income and having a regular source of care. Need factors included perceived health status and comorbidity. Among people with diabetes, multiple logistic regression models were used to determine the independent correlates of single and multiple (two or more) emergency department visits while controlling for confounders.

Table 1 compares sample characteristics and patterns of emergency department use among people with and without diabetes. The prevalence of emergency department use was higher in people with diabetes than in people without diabetes: one visit (16 vs. 12%), two to three visits (8 vs. 4%), and four or more visits (3 vs. 1%).

In the full multivariate model comparing odds of emergency department use and controlled for known confounders, the odds ratio (OR) of emergency department use was not significantly different between people with and without diabetes (OR 0.92 [95% CI 0.78–1.08]). In models without need factors, individuals with diabetes had higher odds of emergency department use (predisposing alone: 1.83 [1.62–2.09]; enabling alone: 1.65 [1.45–1.88]; predisposing and enabling: 1.75 [1.53–2.00]). When only need factors were controlled for, there was no significant difference in odds of emergency department use (0.91 [0.78–1.05]), suggesting that need factors were primarily responsible for the differences in emergency department use.

Among people with diabetes, independent correlates of single emergency department use were age 18–34 years (OR 2.35 [95% CI 1.06–5.25]), perceived worsening of health (1.66 [1.12–2.44]), and having three or more chronic comorbid conditions (1.56 [1.06–2.30]). Correlates of multiple emergency department use included younger age (18–34 years: 8.69 [3.72–20.31]; 35–49 years: 2.87 [1.54–5.33]), unemployment (2.33 [1.39–3.90]), having a usual source of care (8.45 [1.08–65.76]), perceived worsening of health (2.77 [1.92–3.99]), having three or more chronic comorbid conditions (3.04 [1.40–6.58]), and having diabetes complications (2.11 [1.29–3.48]).

This study showsthat while controlling for factors known to predict emergency department use, the odds of emergency department use are not significantly different between individuals with and without diabetes. This demonstrates some of the pitfalls of analyzing health services use without using an appropriate conceptual framework or using multivariate models to adjust for known determinants of health services utilization. Prevalence estimates alone would suggest that individuals with diabetes have higher emergency department use than individuals without diabetes. However, once an appropriate conceptual framework is used to select confounding variables and these variables are included in multivariate models, the presumed higher use of emergency department services by individuals with diabetes is found to be nonexistent.

Among people with diabetes, there were differences in factors associated with single and multiple visits to the emergency department. Some of the factors have been reported previously (1,1012). However, in contrast to results of earlier studies, this study did not find significant associations between sex, educational level, and race/ethnicity and emergency department use.

The findings of this study are subject to some limitations. This study did not distinguish trauma-related visits from non-trauma-related visits or urgent from nonurgent visits, and diabetes and other chronic medical conditions were based on self-report.

Dr. Egede is supported by Grant 5K08HS11418 from the Agency for Health Care Research and Quality (Rockville, MD).

The contents of this article are solely the responsibility of the author and do not necessarily represent the official views of the Agency for Health Care Research and Quality or the Centers for Disease Control and Prevention.

1
Burt CW, McCaig LF: Trends in hospital emergency department utilization: UnitedStates, 1992–99.
Vital Health Stat
13 Sep
:
1
–34,
2001
2
Liu T, Sayre MR, Carleton SC: Emergency medical care: types, trends, and factors related to nonurgent visits.
Acad Emerg Med
6
:
1147
–1152,
1999
3
Laditka SB, Mastanduno MP, Laditka JN: Health care use of individuals with diabetes in an employer-based insurance population.
Arch Intern Med
161
:
1301
–1308,
2001
4
Krop JS, Saudek CD, Weller WE, Powe NR, Shaffer T, Anderson GF: Predicting expenditures for Medicare beneficiaries with diabetes: a prospective cohort study from 1994 to 1996.
Diabetes Care
22
:
1660
–1666,
1999
5
McCusker J, Cardin S, Bellavance F, Belzile E: Return to the emergency department among elders: patterns and predictors.
Acad Emerg Med
7
:
249
–259,
2000
6
National Center for Health Statistics:
Dataset documentation, National Health Interview Survey, 1999 (machine readable data file and documentation)
. Hyattsville, MD, National Center for Health Statistics,
2002
7
National Center for Health Statistics:
NHIS survey description, National Health Interview Survey, 1999 (machine readable documentation)
. Hyattsville, MD, National Center for Health Statistics,
2002
8
Stata:
Stata statistical software: release 7.0
. College Station, TX, Stata,
2001
9
Andersen RM: Revisiting the behavioral model and access to medical care: does it matter?
J Health Soc Behav
36
:
1
–10,
1995
10
Hopper SV, Schechtman KB: Factor associated with diabetic control and utilization patterns in a low-income, older adult population.
Patient Educ Couns
7
:
275
–288,
1985
11
Chin MH, Zhang JX, Merrell K: Diabetes in the African-American Medicare population: morbidity, quality of care, and resource utilization.
Diabetes Care
21
:
1090
–1095,
1998
12
Bazargan M, Johnson KH, Stein JA: Emergency department utilization among Hispanic and African-American under-served patients with type 2 diabetes.
Ethn Dis
13
:
369
–375,
2003

A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.