- Open Access
Diabetes, minor depression and health care utilization and expenditures: a retrospective database study
© Nichols et al; licensee BioMed Central Ltd. 2007
Received: 28 August 2006
Accepted: 18 April 2007
Published: 18 April 2007
To estimate the prevalence of minor depression among US adults with diabetes, health care resource utilization, and expenditures by people with diabetes with and without minor depression.
Among adult 2003 Medical Expenditure Panel Survey respondents, diabetes was identified by diagnosis code and self-report. Depression was identified by diagnosis code plus ≥ one antidepressant prescription. Odds of having depression was estimated in people with diabetes and the general population, adjusted for sociodemographic variables (e.g., age, sex, race/ethnicity). Multivariate regressions evaluated factors associated with utilization and log-transformed expenditures for ambulatory care, hospitalizations, emergency visits, and prescriptions.
In 2003, 1932 respondents had diabetes, 435/1932 had diabetes and minor depression. Adults with diabetes were more likely than the general population to have depression (adjusted OR 1.81, 95% CI 1.56, 2.09). People with diabetes with versus without comorbid depression were more likely to be women, have lower incomes and health status, and more diabetes complications (all p < 0.05). In unadjusted analyses, ambulatory care visits were higher for those with versus without depression (17.9 vs. 11.4, p = 0.04), as were prescriptions (60.7 vs. 38.1, p = 0.05). In adjusted analyses, depression was not associated with increased resource use or higher expenditures in any category. Increased number of comorbid conditions was associated with increased resource use in all categories, and increased expenditures for ambulatory care and prescriptions.
People with diabetes are twice as likely to have depression as the general population. Screening for and treatment of depression is warranted, as is additional research into a causal relationship between diabetes and depression.
Diabetes and depression are costly chronic medical conditions, suffered by millions of Americans every year. The American Diabetes Association estimates that 20.8 million people had diabetes in 2005. In 2002, total expenditures for people with diabetes in the U.S. (including all inpatient care, outpatient care and pharmaceuticals for all health conditions – not only diabetes care) were estimated to be nearly $132 billion. In 2000 the prevalence of depression in the United States was between 5 and 10.3 percent of the population, affecting more than 19 million Americans each year. The economic burden of depression in 2002 was estimated at $83.1 billion.[4, 5]
Co-occurrence of these two illnesses is prevalent in the United States. People with diabetes are approximately 2.5 times more likely to be diagnosed with depression than those without diabetes. [6–8] Temporal or causal relationships between the two conditions remain unclear but under study.
Health care expenditures and utilization among individuals with diabetes and comorbid depression warrant additional research for a number of reasons. Depressed patients with a chronic illness are less likely to follow a recommended health or medication regimen. Depressive symptoms are associated with a perception of more impaired physical health and may mimic worsening diabetes symptoms. Poor glucose control is associated with depression and effective treatment of depression may improve diabetes control.
Previous studies evaluating health care expenditures and utilization associated with diabetes and comorbid depression report that in regional or selected populations, comorbid diabetes and depression are associated with more health care utilization and expenditures than diabetes without depression.[9, 11, 12] The most recent study estimating odds of diagnosed depression in individuals with diabetes and relationships between depression and health care utilization and expenditures in a nationally representative database was performed using 1996 data from the Medical Expenditure Panel Survey (MEPS). We propose to update that analysis, and hypothesize, based on recent drug developments in both diabetes management and treatment of depression, that individuals with both diabetes and minor depression will have higher rates of health care utilization in four categories: ambulatory care, inpatient care, emergency department visits, and prescription medications. While an increase in health care utilization often predicts a proportionate increase in expenditures, this may not hold true due to improved overall health resulting from effective treatment. Therefore, expenditures in all four categories will also be examined for US adults with diabetes with and without minor depression.
Study subjects were respondents in the 2003 MEPS database. MEPS is cosponsored by the Agency for Healthcare Research and Quality (AHRQ) and the National Center for Health Statistics and is representative of the US non-institutionalized civilian population. MEPS uses sampling weights reflecting adjustments for survey non-response and population totals from the Current Population Survey. As a result, MEPS weights were applied in all analyses to obtain nationally representative estimates.
Case Definition for Study Subjects
The study cases were defined as individuals with diabetes with and without comorbid depression, identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. MEPS uses three digit (less specific) ICD-9-CM codes rather than five digit codes in order to protect respondent confidentiality. ICD-9-CM code 250, Diabetes Mellitus, was used to identify people with diabetes. Diabetes was also identified using patient self-report in answer to survey questions asking if respondents had ever been diagnosed with or told they have diabetes. The case definition for minor depression required a record of ICD-9 code 311 (Depression NOS), plus a prescription for an antidepressant medication. When determining the odds of having comorbid depression, the cohort of individuals with diabetes was compared with all of the MEPS survey respondents for the entire year.
Definition of Health Care Utilization and Expenditures
Four categories of health care utilization and expenditures were examined: 1) ambulatory care (including visits with a medical provider visits in an office or outpatient setting), 2) inpatient care (number of hospital admissions including zero-day stays), 3) emergency department visits, and 4) prescription medications received. Expenditures in MEPS are defined as the sum of payments for care, not charges or resource costs. Expenditure data include out-of-pocket payments, third-party payer payments (including Medicare, Medicaid and private insurance) and amounts for services rendered by public providers (including Department of Veterans Affairs Medical Centers). Alternative care services and over-the-counter medications are not included in MEPS total expenditures. In cases where expenditure data are missing or where care is provided under a capitated reimbursement arrangement, MEPS uses an imputation process to estimate expenditures. The process, known as the "hot-deck" method, involves matching the medical event for which expenditure data are missing with an event that has similar characteristics. The amount of expenditures for the similar event is then assigned to the missing data.
Population characteristics were examined for differences between individuals with diabetes with and without depression using chi-square or t-tests as appropriate. Characteristics included age, sex, race/ethnicity, level of education, health insurance status, number of comorbidities (based on number of unique ICD-9 codes), poverty level, physical and mental health status, and diabetes severity. Physical and mental health status were based on the Short-Form 12 survey's Physical and Mental Component Summary scores (SF-12 PCS and SF-12 MCS). Proxy measures for diabetes severity were the presence of coronary heart disease (CHD), diabetes-related eye problems, or diabetes-related kidney problems, or use of insulin.
The prevalence of having depression in individuals with diabetes versus the general population was determined using unadjusted chi-square analysis as well as analyses adjusting for age, sex, race/ethnicity and marital status. Mean, unadjusted annual utilization rates and expenditures for ambulatory care visits, inpatient admissions, emergency department visits, and prescription medications were estimated in univariate analyses. Due to their skewed nature, expenditure data were log transformed prior to testing for statistical significance.
Multivariate regression methods for survey data were used to identify those factors that were significantly associated with resource use and log-transformed expenditures in each category. All expenditure data were expressed in 2003 US$. Resource utilization data are discrete data in that an individual can have one or two office visits, but not 1.5. In addition, many respondents have a low number of visits (e.g., zero or one), while a few people have many visits. Due to this over-dispersion of data and the fact that the data are discrete, negative binomial regression methods were used to evaluate predictors of health care utilization. Linear regression methods were used to evaluate predictors of log-transformed expenditures. For both utilization and expenditures, the predictor variable of interest was the presence or absence of diagnosed minor depression. Covariates in both models included age, sex, years of education, race/ethnicity, income level, insurance status, number of comorbid conditions, presence of diabetes complications, use of insulin, having a usual source of care, and physical and mental health status were included in the models. STATA Statistical Software was used in order to account for the MEPS survey weights and to provide nationally representative estimates. All differences between estimates were considered statistically significant at the p = 0.05 level. The Colorado Multiple Institutional Review Board approved this study.
A total of 1932 US adults with diabetes were identified, 435 with minor depression and 1497 without minor depression. Individuals with diabetes were significantly more likely to have depression than were people without diabetes (OR 2.47, 95% CI 2.17, 2.81). After adjusting for age, sex, race/ethnicity and marital status, individuals with diabetes remained almost twice as likely as individuals without diabetes to have depression (OR 1.81, C.I. 1.56–2.09).
Demographic and clinical characteristics of Adults with diabetes with and without comorbid depression. (United States, 2003)
n = 435
n = 1497
Age (mean, SE)
Sex (% male, SE)
Education (mean years, SE)
Race/Ethnicity (%, SE)
Black, not Hispanic
Asian, not Hispanic
White and other
Income level (%, SE)
Poor (< 100% federal poverty level (fpl))
Near poor (100–124% fpl)
Low income (125–199% fpl)
Middle income (200–399% fpl)
High income (> 400% fpl)
Insurance coverage (%, SE)
Having a Usual Source of Care (%, SE)
# of Comorbid conditions (mean, SE)
Diabetes complications (% yes, SE)
Insulin use (% yes, SE)
SF-12 PCS (mean, SE)
SF-12 MCS (mean, SE)
Mean unadjusted annual health care utilization and expenditures among US adults with diabetes with and without minor depression
Emergency Department Visits
Factors associated with health care utilization by category by US adults with diabetes, 2003
Years of Education
0.03, 0.04 †
Black, not Hispanic
Asian, not Hispanic
White/Other, not Hispanic
Have usual source of care
# of Comorbid conditions
0.05, 0.12 †
Factors associated with health care expenditures by adults with diabetes in 2003.*
Years of Education
Black, not Hispanic
Asian, not Hispanic
White/Other, not Hispanic
Have usual source of care
# of Comorbid conditions
US adults with diabetes were nearly twice as likely to have minor depression compared with the general population in 2003. These results are consistent with previous reports of increased depression prevalence among those with diabetes. In a meta-analysis of 42 eligible studies, 18 included a non-diabetic control group and provided data sufficient to calculate a pooled odds ratio. Results of that study indicated that among people with diabetes, the odds of having depression were twice that of those without diabetes. The finding in the current study of an increased prevalence of minor depression among people with diabetes in a nationally representative database reinforces the need to consider comorbid depression when treating patients with diabetes for a number of reasons. First, comorbid depression in people with diabetes has been associated with poor glycemic control. Lustman and colleagues report, also based on a meta-analysis, that adults with diabetes and depression were more likely to have elevated glycosylated hemoglobin (HbA1c) than those without depression. Results of increased HbA1c for diabetes with compared to without depression were consistent for both type1 and 2 diabetes, and for depression assessed by self-report and diagnosis codes. A number of studies suggest that the mechanism for poorer glycemic control among those with diabetes with versus without depression may be related to the impact of depression on self-care activities. Depression and diabetes is associated with poorer adherence to medication, diet, and exercise regimens.[9, 17, 18]
Secondly, an association between comorbid depression and diabetes and increased mortality has been proposed in a number of studies. Katon, et al, reported that younger age is associated with major depression among people with diabetes receiving care through a managed care organization in Washington state. Similar results have been reported for patients with minor depression and diabetes using data from the Medical Expenditure Panel Survey. Finally, Zhang demonstrated an increased risk for mortality among people with versus without major depressive symptoms based on analyses using the National Health and Nutrition Examination Survey. Together, these results present a compelling need to screen for, diagnose, and effectively treat depression among people with diabetes. Screening for depressive symptoms could be incorporated into patient care activities with screening tools. One such instrument is the PHQ-2, a validated 2 question screen that can identify patients with diabetes that may benefit from further evaluation for comorbid depression. [22–24]
In adjusted analyses, no significant associations exist between comorbid depression and total expenditures, expenditures in any category, or resource utilization in any category. An increased number of comorbid conditions were associated with increased utilization in all four categories. Health status was associated with hospitalizations, emergency visits, and prescription use. Higher levels of physical functioning were associated with decreased use of emergency visits and prescriptions, while higher mental functioning was associated with decreased use in all categories except ambulatory care visits. Increased prescription use was associated with being white (not Hispanic) compared to Hispanic ethnicity. Being near poor or having high income were both associated with decreased use of prescription medications compared to the lowest income category. Because use of prescription medications requires access to physician services, the potential exists that those in the near poor category have difficulty either seeing a prescribing health provider and/or affording the out-of-pocket costs of medications. Those in the highest income category may be in overall better health. These issues warrant further study.
In adjusted analyses evaluating factors associated with expenditures, an increased number of comorbid conditions was associated with increased expenditures for ambulatory visits and prescription medications. Having public only or no insurance was associated with lower expenditures for inpatient and emergency care. No other consistent associations were observed between covariates and expenditures in more than one category.
These results differ from those previously reported from the 1996 MEPS data, perhaps due to the fact that our analyses adjusted for more covariates. The earlier reported analyses adjusted for age, sex, race/ethnicity, health insurance, and number (ranging from one to seven) of comorbid chronic conditions. Multivariate regressions in the current study adjusted for additional variables including education, income, the presence of diabetes-related complications, and physical and mental health status. Variables relating to diabetes-related complications and health status were not available in the 1996 MEPS database. Factors that were significant predictors of total costs for US adults with diabetes in 2003 were insurance status, number of comorbidities, physical health status, and insulin use.
US adults with diabetes are twice as likely to have comorbid minor depression as US adults without diabetes. As discussed earlier, comorbid depression has a deleterious impact on adherence to medication and medical regimens, and is associated with poor glycemic control. Ongoing research is necessary to elucidate both the causal relationship between diabetes and depression, as well as identify interventions to treat depression effectively in adults with diabetes. The American Diabetes Association recommends that all factors, clinical as well as psychosocial, should be considered when developing treatment plans for people with diabetes. Screening for and treating depression in the diabetes population is an important component of care for the whole patient.
Another area warranting additional research is the potential that the association between depression and glycemic control is mediated through the deleterious impact of depression on self-care behaviors. As discussed earlier, depression among people with diabetes is associated with poor self-care behaviors. Improving self-care behaviors has been shown to improve glycemic control, supporting the need for ongoing research to confirm the links between depression, diabetes self-management, and glycemic control, as well as research to investigate whether interventions treating depression can improve self-care and glycemic control.
In addition to providing a basis for future research, there are several strengths of this study. The MEPS database is a large, nationally representative sample of the United States civilian non-institutionalized population, with an over-sampling of black and Hispanic subpopulations. The results are generalizable and the study was able to incorporate appropriate adjustments for demographic characteristics, such as age, sex, and race/ethnicity, as well as clinical characteristics, such as number of comorbidities.
The potential exists for underreporting of either diabetes or depression in the MEPS database, either of which would have limited the sample size of this study. A diagnosis of depression is particularly vulnerable to this limitation due to under-diagnosing and potential stigma that may be attached to mental disorders. The case definition for minor depression used in the current study required a diagnosis code and use of a prescription antidepressant, cases included here are likely to be accurately assessed as being depressed. A potential limitation of this restrictive case definition is that individuals with undiagnosed depression, as well as those with a diagnosis but without treatment with prescription medication will be misclassified as having diabetes without depression. The 2003 MEPS database does not include data obtained by screening respondents for depressive symptoms, which may be a limitation of these analyses. However, we report prevalence estimates for comorbid depression consistent with published studies, despite the potential for misclassification. A further source of misclassification in this study is the inclusion of adults with diabetes with major depression in the non-depressed (i.e., depression -) group. Based on a post-hoc analysis, only 19 individuals had diagnosis codes for diabetes and major depression in the 2003 MEPS database. Therefore, the impact of any misclassification of respondents with major depression is likely to be minimal. Finally, MEPS includes only 3-digit ICD-9 codes, preventing a classification of Type 1 or Type 2 diabetes. Due to the relatively low prevalence of Type 1 diabetes (5–10% of all people with diabetes), the impact of this limitation is expected to be minimal.
People with diabetes are twice as likely to be depressed as the general population, a factor to be considered when developing diabetes treatment programs. Screening for and treatment of depression is appropriate, as is additional research into a causal relationship between diabetes and depression.
The authors wish to acknowledge and thank Weiming Zhang for his contributions to the data analyses conducted for this study. Funding was provided in part by a grant from the American College of Clinical Pharmacy (MM). The funding agency had no role in study design or conduct.
- Centers for Disease Control and Prevention: National diabetes fact sheet: general information and national estimates on diabetes in the United States. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2005.Google Scholar
- American Diabetes Association. Economic costs of diabetes in the U.S. in 2002 Diabetes Care 2003, 26: 917–932. 10.2337/diacare.26.3.917Google Scholar
- National Institute of Mental Health: The Numbers Count. NIMH Publication No. 01(4584) 2001, 1–4.Google Scholar
- Greenberg P, Kessler R, Birnbaum H, Leong S, Lowe S, Berglund P, Corey-Lisle P: The Economic Burden of Depression in the United States: How Did It Change Between 1990 and 2000? J Clin Psychiatr 2003, 64: 1465–1474.View ArticleGoogle Scholar
- Wang P, Simon G, Kessler R: The economic burden of depression and the cost-effectiveness of treatment. Int J Methods Psychiatr Res 2003,12(1):22–33. 10.1002/mpr.139PubMedView ArticleGoogle Scholar
- Egede L, Zheng D, Simpson K: Comorbid Depression is Associated With Increased Health Care Use and Expenditures in Individuals With Diabetes. Diabetes Care 2002,25(3):464–470. 10.2337/diacare.25.3.464PubMedView ArticleGoogle Scholar
- Gavard J, Lustman P, Clouse R: Prevalence of Depression in Adults With Diabetes. Diabetes Care 1993,16(8):1167–1177. 10.2337/diacare.16.8.1167PubMedView ArticleGoogle Scholar
- Anderson S, Freedland K, Clouse R, Lustman P: The Prevalence of Comorbid Depression in Adults With Diabetes. Diabetes Care 2001,24(6):1069–1077. 10.2337/diacare.24.6.1069PubMedView ArticleGoogle Scholar
- Ciechanowski P, Katon W, Russo J: Depression and diabetes: Impact of depressive symptoms on adherence, function and costs. Arch Int Med 2000, 160: 3278–3285. 10.1001/archinte.160.21.3278View ArticleGoogle Scholar
- Lustman P, Anderson R, Freedland K, Groot Md, Carney R, Clause R: Depression and Poor Glycemic Control: A Meta-Analytic Review of the Literature. Diabetes Care 2000,23(7):934–942. 10.2337/diacare.23.7.934PubMedView ArticleGoogle Scholar
- Husaini B, Hull P, Sherkat D, Emerson J, Overton M, Craun C, Cain V, Levine R: Diabetes, Depression, and Healthcare Utilization Among African Americans in Primary Care. J Natl Med Assoc 2004,96(4):476–484.PubMed CentralPubMedGoogle Scholar
- Finkelstein E, Bray J, Chen H, Larson M: Prevalence and Costs of Major Depression Among Elderly Claimants With Diabetes. Diabetes Care 2003,26(2):415–420. 10.2337/diacare.26.2.415PubMedView ArticleGoogle Scholar
- Machlin S, Yu W, Zodet M: Computing Standard Errors for MEPS Estimates. Agency for Healthcare Research and Quality, Rockville, Md; 2005. [http://www.meps.ahrq.gov/mepsweb/survey_comp/standard_errors.jsp]Google Scholar
- Machlin SR, Dougherty DD: Overview of Methodology for Imputing Missing Expenditure Data in the Medical Expenditure Panel Survey. Methodology Report No.19 Agency for Healthcare Research and Quality, Rockville, Md; 2007. [http://www.meps.ahrq.gov/mepsweb/data_files/publications/mr19/mr19.pdf]Google Scholar
- Schneeweiss S, Seeger J, Maclure M, Wang P, Avorn J, Glynn R: Performance of comorbidity scores to control for confounding in epidemiologic studies using claims data. Am J Epidemiol 2001,154(9):854–864. 10.1093/aje/154.9.854PubMedView ArticleGoogle Scholar
- Ware J, Kosinski M, Keller S: A 12-item short-form health survey: Construction of scales and preliminary tests of reliability and validity. Med Care 1996, 34: 220–233. 10.1097/00005650-199603000-00003PubMedView ArticleGoogle Scholar
- Ciechanowski P, Katon W, Russo J, Hirsch I: The relationship of depressive symptoms to symptom reporting, self-care, and glucose control in diabetes. Gen Hosp Psych 2003, 25: 246–252. 10.1016/S0163-8343(03)00055-0View ArticleGoogle Scholar
- Park H, Hong Y, Lee H, Ha E, Sung Y: Individuals with type 2 diabetes and depressive symptoms exhibited lower adherence with self-care. J Clin Epidemiol 2004, 57: 978–984. 10.1016/j.jclinepi.2004.01.015PubMedView ArticleGoogle Scholar
- Katon W, Korff MV, Ciechanowski P, Russo J, Lin E, Simon G, Ludman E, Walker E, Bush T, Young B: Behavioral and clinical factors associated with depression among individuals with diabetes. Diabetes Care 2004, 27: 914–920. 10.2337/diacare.27.4.914PubMedView ArticleGoogle Scholar
- McCollum M, Ellis S, Regensteiner J, Zhang W, Sullivan P: Minor depression and health status among US adults with diabetes mellitus. American Journal of Managed Care 2007,13(2):65–72.PubMedGoogle Scholar
- Zhang X, Norris S, Gregg E, Cheng Y, Beckles G, Kahn H: Depressive symptoms and mortality among persons with and without diabetes. Am J Epidemiol 2005, 161: 652–660. 10.1093/aje/kwi089PubMedView ArticleGoogle Scholar
- Arroll B, Khin N, Kerse N: Screening for depression in primary care with two verbally asked questions: cross sectional study. BMJ 2003, 327: 1144–1146. 10.1136/bmj.327.7424.1144PubMed CentralPubMedView ArticleGoogle Scholar
- Kroenke K, Spitzer R, Williams J: The Patient Health Questionnaire-2: Validity of a Two-Item Depression Screener. Medical Care 2003,41(11):1284–1292. 10.1097/01.MLR.0000093487.78664.3CPubMedView ArticleGoogle Scholar
- Price D: How should we screen patients for major depression? The Permanente Journal 2004,8(4):24–26.Google Scholar
- American Diabetes Association: Standards of medical care in diabetes. Diabetes Care 2007, 30: S4–41. 10.2337/dc07-S004View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.