Costs of depression from claims data for medicare recipients in a population-based sample.
Alexandre, Pierre K. ; Hwang, Seungyoung ; Roth, Kimberly B. 等
ACKNOWLEDGEMENT
This study was funded by the National Institute on Drug Abuse (PI:
William W. Eaton, Ph.D.; R01 DA009897). No disclosures to report.
ABSTRACT
Background: Many persons with depressive disorder are not treated
and associated costs are not recorded.
Aims of the Study: To determine whether major depressive disorder
(MDD) is associated with higher medical cost among Medicare recipients.
Methods: Four waves of the Baltimore-Epidemiologic Catchment Area
(Baltimore ECA) Study conducted between 1981 and 2004 were linked to
Medicare claims data for the years 1999 to 2004 from the Centers for
Medicare and Medicaid Services (CMS). Generalized linear models
specified with a gamma distribution and log link function were used to
examine direct medical care costs associated with MDD.
Results: Medicare recipients with no history of MDD in either the
ECA or CMS data had mean six-year medical costs of US $40,670, compared
to $87,445 for Medicare recipients with MDD as recorded in CMS data and
$43,583 for those with MDD as recorded in Baltimore-ECA data.
Multivariable regressions found that compared to Medicare recipients
with no history of depression, those with depression identified in the
CMS data had significantly higher medical costs; about 1.87 times (95%
confidence interval (CI) 1.32 to 2.67) higher. Medicare recipients with
a history of depression identified in the ECA data were no more likely
to have higher costs than were Medicare recipients with no history of
depression (relative ratio 1.33, 95% CI 0.87 to 2.02).
Discussion: Medicare recipients with a history of depression
identified in claims data had significantly higher medical costs than
recipients with no history of depression. However, no significant
differences were found between Medicare recipients with depression in
the community-based Baltimore ECA data and those with no history of
depression. The results show that the source of diagnosis, in treatment
versus survey data, produces differences in results as regards costs.
Limitations: This study involved only Medicare recipients with
claims data over the six years 1999 to 2004. Many of the ECA respondents
were too young to qualify for Medicare.
Implications for Health Policy: Depressive disorder involves
substantial medical care costs. The findings provide information on the
economic burden of depression, an important but often omitted dimension
and perspective of the burden of mental illnesses.
INTRODUCTION
The most ambitious effort to evaluate the cost of illness (COI) of
diseases in terms of impairment and disability was the World Health
Organization--Global Burden of Disease (WHO-GBD) Study.(Murray &
Lopez, 1994) This study found that the category of neuropsychiatric
disorders, as a group, had the highest burden of disease in the world.
It is thus reasonable to expect higher medical costs associated with
these disorders, presumably stemming from long-lasting disability of
affected individuals. Direct medical costs of illness are expenditures
for medical services such as medications, doctor visits, and
hospitalizations. The primary objective of this study was to examine
direct medical costs associated with major depressive disorder (referred
simply as 'depression') for Medicare recipients who
participated in the Baltimore Epidemiologic Catchment Area (Baltimore
ECA) Study.
Luppa et al.(Luppa, Heinrich, Angermeyer, Konig, &
Riedel-Heller, 2007) in a systematic review of COI studies of depression
divided these studies into two categories: (1) per-capita cost studies
that estimated the cost of medical resources consumed by a patient; and
(2) national cost studies that estimated cost of depression by
multiplying national prevalence data (number of cases) with the relevant
unit cost of medical services published in national statistics.
Stoudemire et al.(Stoudemire, Frank, Hedemark, Kamlet, & Blazer,
1986) conducted one of the first studies on the costs of depressive
disorder in the adult US population using a range of statistics from
national data sources. They found that major depression was associated
with a total cost of $16.3 billion with an estimated annual medical cost
of $436 per person. Rice and Miller(Rice & Miller, 1995) estimated
the cost associated with affective disorders at $30.3 billion in 2000
dollars and an annual medical cost of depression equal to $877 in 1990
dollars. Greenberg and colleagues(Greenberg et al., 2003; Greenberg,
Kessler, & Nells, 1996; Greenberg, Stiglin, Finkelstein, &
Berndt, 1993) assessed the economic cost of depression at $43.7 billion
in 1990 dollars with annual medical cost of $640 per person and later in
2000 at $83.1 billion, with an estimated annual medical cost of $1,126
per case. Another study used the 1999 Medical Expenditures Panel Survey
and estimated total cost of depression at $5.2 billion in 1999 dollars;
annual medical cost per depression case was $5,871.(Trivedi, Lawrence,
Blake, Rappaport, & Feldhaus, 2004)
Simon et al.(Simon, VonKorff, & Barlow, 1995) used computerized
record systems of a large health maintenance organization (HMO) to
identify a sample of primary care patients with depression and a control
group with no depression diagnosis. They found that the annual cost for
adults with depression was about $4,246 compared to $2,371 for patients
with no depression (difference of $1,975) in 1993 dollars. In addition
to having significantly higher total medical care costs, depressed
patients had higher costs in every category of care. Other per-capita
studies examined medical cost of depression among individuals in their
later years of life.(Murray & Lopez, 1994) Unutzer et al.(Unutzer et
al., 1997) conducted a four-year prospective cohort study of Medicare
beneficiaries enrolled in a HMO in Seattle to examine whether patients
with depressive symptoms had higher costs of general medical services.
Excess annual costs due to depression were estimated at $789 in 1995
dollars. Medical cost for older adults with depressive symptoms enrolled
in that same HMO for the year 1999 was estimated at $6,494 compared to
$4,231 for the non-depressed patient, i.e., excess cost of depression
estimated at $2,263 per year.(Katon, Lin, Russo, & Unutzer, 2003)
Luber et al.(Luber et al., 2001) analyzed health services utilization
and costs for a sample of 3,481indivuals aged 65 and older seen in a
primary care setting over 12 months. Ambulatory and inpatient costs were
calculated for the year 1994 for patients with a diagnosis of depression
and a control group with no depression. Annual costs were estimated at
$1,556 per depressed patients compared to $927 for non-depressed
patients; or an excess of $629.
Table 1 presents a summary of studies on the costs of depression
published for the US. For comparison purposes, these costs were adjusted
to 2004 US dollars using the US Gross Domestic Product (GDP) deflator.
An important issue in analyzing costs of depression is whether the focus
is on treated depression in a particular setting, or depression in the
general population regardless of the treatment. This study is the first
to examine these two issues in a single paper. In addition to a control
group with no history of major depressive disorder, this paper analyzed
Medicare recipients who were treated for depression as reported in the
Medicare claims data, and Medicare recipients with a history of
depression identified in the community-based data that might or might
not have been in treatment. Another significant contribution of this
paper is that it focuses on the elderly, the most important demographic
group when it comes to controlling the rise of medical costs in the
United States.
METHODS
Study Sample
The study sample was constructed by linking two data sources: the
Baltimore Epidemiologic Catchment Area (Baltimore ECA) Study conducted
between 1981 and 2004 and Medicare claims data for the years 1999 to
2004. The Baltimore ECA was part of the larger Epidemiologic Catchment
Area (ECA) study and the first cohorts interviewed were among the
earliest in the nation sampled from the general population to include a
wide range of psychopathology according to diagnostic criteria.(Eaton,
Regier, Locke, & Taube, 1981) Of the five ECA sites (Baltimore, MD.;
New Haven, CT.; Durham, NC; St Louis, MO; and Los Angeles, CA), the
Baltimore ECA Study was the only site to designate its entire baseline
sample of household respondents interviewed in 1981 for follow-ups. A
total of 175,211 household residents of Eastern Baltimore were sampled
probabilistically in 1981 for participation in the Baltimore- ECA
study.(Eaton, Kessler, & (eds), 1985; Regier, Myers, & Kramer,
1985) Persons aged 65 and older were oversampled by selecting all
elderly members of the household for interview, as well as whomever in
the household was randomly selected. Four thousand two hundred and
thirty-eight (4,238) persons were designated for the baseline sample and
3,481 completed the interview.(Eaton et al., 1984) One year later, 2,768
of these participants were re-interviewed. From the end of 1993 through
early 1996, 1,920 of those interviewed in 1981 were interviewed
again(Eaton et al., 1997) (abbreviated below to "1994" to
designate this wave of the study cohort because most of the survey
occurred in 1994). From 2004 to 2005, 1,071 of those interviewed in 1994
were followed-up with an additional interview (abbreviated below to
"2004").(Eaton, Kalaydjian, Scharfstein, Mezuk, & Ding,
2007) The research was approved by the Johns Hopkins Bloomberg School of
Public Health Institutional Review Board.
For this analysis we requested Medicare data for the years 1999 to
2004 from the Centers for Medicare and Medicaid Services (CMS) to be
linked to Baltimore ECA study participants that were interviewed during
the '1994' wave. Medicare research identifiable files (RIF)
that contain person-specific data on Medicare providers, beneficiaries,
and recipients were not available for years prior to 1999. Because only
study participants with a listed social security number could be linked
to Medicare RIF data, we requested data only for study participants for
whom social security numbers were available (about 90%). The study
sample included 63% females, 62% Caucasians, 35% African-Americans and
3% other racial groups, which was nearly identical to the composition of
the original cohort. Participants in the follow-up interviews did not
differ from the survivors with respect to age, race, and gender.(Badawi,
Eaton, Myllyluoma, Weimer, & Gallo, 1999) We submitted a finder file
containing social security number, date of birth, and gender for 1,702
participants that were linked to 1999-2004 Medicare data (at a cost of
about $140,000). The linked file included 581 (34.1%) study participants
that were eligible to enroll in Medicare during 1999-2004 because of
age, disability, and/or end-stage renal disease. We excluded 15 persons
that were enrolled in a Medicare health maintenance organization who did
not have claims data, leaving 566 persons for our analysis.
Cost Assessment
To estimate direct medical costs we used final reimbursements from
all Medicare claims to reflect actual resource use. Total direct medical
costs were calculated by adding all payments made by beneficiaries
(i.e., deductibles, coinsurance), Medicare and health care providers.
All costs were adjusted in 2004 dollars using 3% annual inflation rate
as recommended by the Panel on Cost Effectiveness in Health and
Medicine.(Weinstein, Siegel, Gold, Kamlet, & Russell, 1996)
Diagnosis of Major Depressive Disorder (MDD)
Two groups of study participants with a history of major depressive
disorder were constructed. The first group included participants with
depression status assessed in the Baltimore ECA data via the Diagnostic
Interview Schedule (DIS). The DIS is a structured diagnostic interview
that can be administered by trained non-clinicians to elicit information
for assessment according to the Diagnostic and Statistical Manual of
Mental Disorders (DSM).(Robins, Helzer, Croughan, Williams, &
Spitzer, 1981) The DIS was used in all four waves in the Baltimore ECA
study.(Robins, Helzer, Cottler, & Goldring, 1989; Robins, Helzer,
Croughan, & Ratcliff, 1981; Robins, Helzer, Croughan, Williams, et
al., 1981) A total of 45 respondents were considered positive for a
diagnosis of depression because they met criteria for depressive
disorder at any time during the survey years.
The second group of study participants included individuals with
diagnosis of depression assessed in the Medicare claims data using the
International Classification of Diseases, 9th edition, Clinical
Modification (ICD-9-CM) codes associated with each claim. Individuals
were identified as suffering from depression if they had ICD-9-CM codes
of 296.2 (depression, single episode) or 296.3 (depression, recurrent
episode) in any of the Medicare files obtained from CMS for
institutional inpatient and outpatient care (Part A), physician care
(Part B), skilled nursing facilities, and home health care.
Assessment of Participant Characteristics
The last wave of the Baltimore ECA was used to obtain the most
recent data on demographic and socioeconomic characteristics of study
participants, including age, gender, race, educational attainment level,
marital status, and household income. If personal information was
missing for that wave, data were taken from prior waves. Otherwise, we
imputed the missing values through Markov Chain Monte Carlo (MCMC) and
logistic regression methods using the PROC MI procedure in SAS. Missing
values on household income were imputed for 14 subjects (2.5%) in the
dataset. The Medicare data files were used to obtain information on
respondents' comorbid medical conditions, including myocardial
infarction, congestive heart failure, peripheral vascular disease,
cerebrovascular disease, chronic pulmonary disease, peptic ulcer
disease, mild liver disease, diabetes, hemiplegia or paraplegia, renal
disease, and cancer.
Statistical Analysis
Using the group with no indication of MDD in either of the two
datasets as controls, we compared patient demographic, socioeconomic,
and clinical characteristics across depression status using t-test and
Fisher's exact test at the 0.05 significance level.(Allison, 2012)
In multivariable analysis, we examined direct medical costs for these
groups using generalized linear models (GLM) specified with a gamma
distribution and log link function to deal with skewedness of medical
cost measures in the upper tail.(Barber & Thompson, 2004) Models
were adjusted for differences in demographic, socioeconomic, and
clinical variables that could confound the relationship between
depression and medical costs. The covariates included age, gender (male
vs. female), race (nonwhite vs. white), education (less than high
school, high school, beyond high school), marital status (never married
vs. married), household income (less than $25,000, $25,000 to 49,999;
$50,000 or higher). To control for comorbid medical conditions we
included a weighted Charlson comorbidity index based on the 11 medical
conditions from the Medicare data as discussed earlier.(Charlson,
Pompei, Ales, & MacKenzie, 1987; Quan et al., 2005) We did not
control for differential treatment and survival according to depression
status; rather our approach mirrors that of cost-effectiveness analysis
in which costs are estimated in parallel with effects of treatment and
survival but not adjusted for.(Gold, Siegel, Russel, & Weinstein,
1996) All analyses were conducted using Statistical Analysis System
(SAS), Version 9.4 (SAS Institute Inc., Cary, NC, USA).
RESULTS
Sample Characteristics
Table 2 compares the demographic, socioeconomic, and clinical
characteristics for the two groups of Medicare recipients with MDD
relative to Medicare recipients with no history of MDD. Among the 566
Medicare recipients in the study, we found 472 (83.4%) had no history of
depression, 59 (10.4%) had a history of depression in the CMS data, and
45 (8.0%) had a history of depression in the ECA data. These two groups
overlapped, in that ten respondents with a record of depression in the
CMS data also had depression assessed in the ECA data. Table 3 shows the
agreement between the two sources of diagnosis, in which the kappa
statistic is estimated to be 0.11, indicating poor agreement. This
reinforces the importance of studying individuals with depression
treated in a medical setting as well as those with depression assessed
in the community.
The analysis presented on Table 2 indicates that a significantly
higher percentage of Medicare recipients with depression in the ECA data
were females (80.0%) compared to recipients with no depression (65.0%).
Regardless of the data sources, Medicare recipients with depression were
younger. With regards to medical conditions, except for cancer there
were no significant differences between Medicare recipients with
depression in the ECA data relative to recipients with no depression. On
the other hand, a significantly higher percentage of Medicare recipients
in the CMS data (column 3 in Table 2) had a variety of comorbid medical
conditions, including myocardial infarction, peripheral vascular
disease, cerebrovascular disease, peptic ulcer disease,
hemiplegia/paraplegia, and renal disease.
Mean Six-year Medical Costs for Medicare Recipients
Medicare recipients with MDD had higher mean six-year medical costs
relative to with no history of MDD, regardless of data sources used to
indicate depression status (Table 4). Medicare recipients with no
depression had mean six-year medical costs of $40,670 (95% CI [$35,227,
$46,112]), compared to $87,445 (95% CI [$61,655, $113,235]) for
recipients with depression in the CMS data and $43,583 (95% CI [$17,949,
$69,218]) for recipients with depression identified in ECA data.
For easy interpretations, relative cost ratios associated with
depression were calculated from the estimates of the adjusted least
squares means. For the depression variable for example, the relative
cost ratio represents the ratio of the expected cost between a person
with depression and a person with no depression. After adjustment for
demographic, socioeconomic, and clinical characteristics, participants
with depression in CMS data had significantly higher six-year total
medical costs than those with no depression (relative cost ratio = 1.87,
95% CI 1.32 to 2.67, P < 0.001). Participants with depression in ECA
data had 1.33 times (95% CI: 0.87 to 2.02) higher costs than those
without a history of depression, but the depression was not significant
associated with medical costs (P = 0.1899). On an annual basis, the
estimated medical cost of depression in this study were $7,796 =
($87,445-$40,670) / 6 for those with depression in CMS data and $486 =
($43,583-$40,670) / 6 for those with depression in ECA data. These
unadjusted costs for depression assessed in the CMS were slightly higher
than unadjusted estimates obtained from previous studies (see Table 1),
but the costs associated with depression assessed in ECA data were lower
than those in previous studies.
DISCUSSION
Medicare recipients with a history of depression had higher medical
costs than Medicare recipients with no history of depression. The
medical cost difference existed regardless of the source of the
diagnosis of depression, although the difference was much larger and
only statistically significant when the source of the diagnosis of
depression was the Medicare record. There are several plausible
explanations for the higher medical costs for Medicare recipients
identified in the CMS data relative to those with a DIS diagnosis in the
Baltimore ECA data. Depression often goes without treatment.(Kessler et
al., 2003; Mojtabai, Eaton, & Maulik, 2012) Individuals with
depression who are not in treatment as might be the case for many
individuals with a history of depression in the Baltimore ECA data might
have less severe forms of depressive disorder. It is worth nothing that
most cases of depression in the ECA group occurred earlier than the
period 1999-2004, whereas the occurrence of depression recorded in the
CMS is constrained to have occurred during that period. A third
explanation is that the persons with depression recorded in the ECA
interview who are not in the Medicare data are avoiding medical
treatment in general, reducing their costs. It could also be that, among
individuals with depression, those who seek treatment for medical
conditions are more likely to have their depression identified. It is
also important to consider that cases of depression that go untreated
might eventually generate higher medical costs associated with
later-occurring consequences of failure to treat depression, because
depressive disorder is a risk factor for important and other costly
medical conditions such as diabetes, cerebrovascular disease, heart
attack, and dementia. (Eaton et al., 2012) These latter conditions have
onset later in life, and higher costs associated with these conditions
which are consequent to depressive disorder will be more prominent in
elderly populations--perhaps older than the sample in this study.
Another consideration is the small amount of overlap between the two
estimates of depression: 10 respondents with depression recorded in both
data sources. Part of the discrepancy is due to the lifetime record in
the ECA versus the six year record in the CMS. But there are likely many
other sources of variation, as it has been shown repeatedly in comparing
estimates of psychiatric diagnosis with different methods of
assessment.(Eaton, Hall, Macdonald, & McKibben, 2007) This lack of
overlap indicates that the cost estimates for depression indeed depend
on the source of the diagnosis: methods of estimation that apply cost
data from treatment or medical record sources to estimates of the
prevalence of depression from population-based surveys are likely to
produce inflated estimates of cost.
This study is unusual in its combination of a population-based
sample with cost data based on medical claims records. But it has
important limitations. First, many of the ECA respondents were too young
to qualify for Medicare. Second, this study involved only Medicare
recipients with claims data over the six years 1999 to 2004. If
individuals with no claims at all were more likely to be in the "no
depression" group, the costs associated with depression might be
underestimated. Moreover, we assigned missing indicators to 14
recipients with no information on household income. But sensitivity
analyses, in which we performed analyses with and without adjustment for
household income limited to recipients without missing information on
these variables, revealed similar results for these variables. Despite
these limitations, the direct medical costs estimated in this study
provide information on the economic burden of depression in our society,
an important but often omitted dimension and perspective of the burden
of mental illnesses. It is worth nothing in this context that direct
medical costs represent only 30 percent of the total costs of depression
to society.(Greenberg et al., 2003) Although there is a range of
estimates depending on the methods used, all of them demonstrate that
depression represents a serious public health problem and an economic
burden of major proportions for society.
REFERENCES
Allison, P. D. (2012). Logistic Regression Using SAS: Theory and
Application (2 ed.). Cary, NC: SAS Institute.
Badawi, M. A., Eaton, W. W., Myllyluoma, J., Weimer, L. G., &
Gallo, J. (1999). Psychopathology and attrition in the Baltimore ECA
15-year follow-up 1981-1996. Soc Psychiatry Psychiatr Epidemiol, 34(2),
91-98.
Barber, J., & Thompson, S. (2004). Multiple regression of cost
data: use of generalised linear models. J Health Serv Res Policy, 9(4),
197-204. doi: 10.1258/1355819042250249
Charlson, M. E., Pompei, P., Ales, K. L., & MacKenzie, C. R.
(1987). A new method of classifying prognostic comorbidity in
longitudinal studies: development and validation. J Chronic Dis, 40(5),
373-383.
Eaton, W. W., Alexandre, P., Kessler, R. C., Martins, S. S.,
Mortensen, P. B., Rebok, G. W.,... Roth, K. (2012). The Population
Dynamics of Mental Disorders. In s. Eaton WW and the faculty, and
fellows of the Department of Mental Health, Bloomberg School of Public
Health (Ed.), Public Mental Health (pp. 125-150). New York, NY: Oxford
University Press.
Eaton, W. W., Anthony, J. C., Gallo, J., Cai, G., Tien, A.,
Romanoski, A.,... Chen, L. S. (1997). Natural history of Diagnostic
Interview Schedule/DSM-IV major depression. The Baltimore Epidemiologic
Catchment Area follow-up. Arch Gen Psychiatry, 54(11), 993-999.
Eaton, W. W., Hall, A. L., Macdonald, R., & McKibben, J.
(2007). Case identification in psychiatric epidemiology: a review. Int
Rev Psychiatry, 19(5), 497-507. doi: 10.1080/09540260701564906
Eaton, W. W., Holzer, C. E., 3rd, Von Korff, M., Anthony, J. C.,
Helzer, J. E., George, L.,... Locke, B. Z. (1984). The design of the
Epidemiologic Catchment Area surveys. The control and measurement of
error. Arch Gen Psychiatry, 41(10), 942-948.
Eaton, W. W., Kalaydjian, A., Scharfstein, D. O., Mezuk, B., &
Ding, Y. (2007). Prevalence and incidence of depressive disorder: the
Baltimore ECA follow-up, 1981-2004. Acta Psychiatr Scand, 116(3),
182-188. doi: 10.1111/j.1600-0447.2007.01017.x
Eaton, W. W., Kessler, L. G., & (eds). (1985). Epidemiologic
Field Methods in Psychiatry: The NIMH Epidemiologic Catchment Area
Program. Orlando, FL: Academic Press Inc.
Eaton, W. W., Regier, D. A., Locke, B. Z., & Taube, C. A.
(1981). The Epidemiologic Catchment Area Program of the National
Institute of Mental Health. Public Health Rep, 96(4), 319-325.
Gold, M. R., Siegel, J. E., Russel, L. B., & Weinstein, M. C.
(1996). Cost-Effectiveness in Health and Medicine (1 ed.). New York:
Oxford University Press.
Greenberg, P. E., Kessler, R. C., Birnbaum, H. G., Leong, S. A.,
Lowe, S. W., Berglund, P. A., & Corey-Lisle, P. K. (2003). The
economic burden of depression in the United States: how did it change
between 1990 and 2000? J Clin Psychiatry, 64(12), 1465-1475.
Greenberg, P. E., Kessler, R. C., & Nells, T. L. (1996).
Depression in the workplace: an economic perspective. In J. P. Feighner
& W. F. Boyer (Eds.), Selective Serotonin Re-uptake Inhibitors:
Advances in basic Research and Clinical Practice (2 ed.). New York, NY:
John Wiley & Sons.
Greenberg, P. E., Stiglin, L. E., Finkelstein, S. N., & Berndt,
E. R. (1993). The economic burden of depression in 1990. J Clin
Psychiatry, 54(11), 405-418.
Katon, W. J., Lin, E., Russo, J., & Unutzer, J. (2003).
Increased medical costs of a population-based sample of depressed
elderly patients. Arch Gen Psychiatry, 60(9), 897-903. doi:
10.1001/archpsyc.60.9.897
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Koretz, D.,
Merikangas, K. R.,... Wang, P. S. (2003). The epidemiology of major
depressive disorder: results from the National Comorbidity Survey
Replication (NCS-R). JAMA, 289(23), 3095-3105. doi:
10.1001/jama.289.23.3095
Luber, M. P., Meyers, B. S., Williams-Russo, P. G., Hollenberg, J.
P., DiDomenico, T. N., Charlson, M. E., & Alexopoulos, G. S. (2001).
Depression and service utilization in elderly primary care patients. Am
J Geriatr Psychiatry, 9(2), 169-176.
Luppa, M., Heinrich, S., Angermeyer, M. C., Konig, H. H., &
Riedel-Heller, S. G. (2007). Cost-of-illness studies of depression: a
systematic review. J Affect Disord, 98(1-2), 29-43. doi:
10.1016/j.jad.2006.07.017
Mojtabai, R., Eaton, W. W., & Maulik, P. K. (2012). Pathways to
Care: Need, Attitudes, Barriers. In s. Eaton WW and the faculty, and
fellows of the Department of Mental Health, Bloomberg School of Public
Health (Ed.), Public Mental Health (pp. 415-456). New York, NY: Oxford
University Press.
Murray, C. J. L., & Lopez, A. D. (1994). Global Comparative
Assessments in the Health Sector: Disease Burden, Health Expenditures,
and Intervention Packages Global Health in Transition: A Synthesis:
Perspectives from International Organizations (pp. 44-50). Geneva: World
Health Organization (WHO).
Quan, H., Sundararajan, V., Halfon, P., Fong, A., Burnand, B.,
Luthi, J. C.,... Ghali, W. A. (2005). Coding algorithms for defining
comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care,
43(11), 1130-1139.
Regier, D. A., Myers, J. K., & Kramer, M. (1985). Historical
context, major objectives, and study design. In: Eaton WW, Kessler LG,
eds. Epidemiologic field methods in psychiatry: the NIMH Epidemiologic
Catchment Area Program. Orlando, FL: Academic Press Inc.
Rice, D. P., & Miller, L. S. (1995). The economic burden of
affective disorders. Br J Psychiatry Suppl(27), 34-42.
Robins, L. N., Helzer, J. E., Cottler, L., & Goldring, E.
(1989). NIMHDiagnostic Interview Schedule: Version III Revised
(DIS-III-R). St. Louis: Washington University School of Medicine.
Robins, L. N., Helzer, J. E., Croughan, J., & Ratcliff, K. S.
(1981). National Institute of Mental Health Diagnostic Interview
Schedule. Its history, characteristics, and validity. Arch Gen
Psychiatry, 38(4), 381-389.
Robins, L. N., Helzer, J. E., Croughan, J., Williams, J. B. W.,
& Spitzer, R. L. (1981). NIMH Diagnostic Interview Schedule: Version
III. St Louis: Washington University School of Medicine.
Simon, G. E., VonKorff, M., & Barlow, W. (1995). Health care
costs of primary care patients with recognized depression. Arch Gen
Psychiatry, 52(10), 850-856.
Stoudemire, A., Frank, R., Hedemark, N., Kamlet, M., & Blazer,
D. (1986). The economic burden of depression. Gen Hosp Psychiatry, 8(6),
387-394.
Trivedi, D. N., Lawrence, L. W., Blake, S. G., Rappaport, H. M.,
& Feldhaus, J. B. (2004). Study of the economic burden of
depression. Journal of Pharmaceutical Finance, Economics, and Policy,
13(4), 51-66.
Unutzer, J., Patrick, D. L., Simon, G., Grembowski, D., Walker, E.,
Rutter, C., & Katon, W. (1997). Depressive symptoms and the cost of
health services in HMO patients aged 65 years and older. A 4-year
prospective study. JAMA, 277(20), 1618-1623.
Weinstein, M. C., Siegel, J. E., Gold, M. R., Kamlet, M. S., &
Russell, L. B. (1996). Recommendations of the Panel on
Cost-effectiveness in Health and Medicine. JAMA, 276(15), 1253-1258.
PIERRE K. ALEXANDRE Florida Atlantic University
SEUNGYOUNG HWANG KIMBERLY B. ROTH JOSEPH J. GALLO WILLIAM W. EATON
Johns Hopkins University
Table 1
Annual medical costs of depression from existing studies in 2004
dollars (increasing order)
Study Sample size (a) Data source Age
Luber 3,481 Primary 65-97
2001 (246) care (b)
Stoudemire National ECA [less than or equal to]18
1986 estimates Prevalence (c)
Greenberg National ECA [less than or equal to]18
1993 estimates Prevalence (c)
Unutzer 2,558 Primary [less than or equal to]65
1997 (353) care (b)
Rice National National [less than or equal to]18
1995 estimates Statistics (c)
Greenberg National ECA [less than or equal to]18
2003 estimates Prevalence (c)
Simon 12,514 Primary [less than or equal to]18
1995 (6,257) care (b)
Katon 9,001 Primary [less than or equal to]60
2003 (1,736) care (b)
Trivedi 24,617 MEPS (d) All
2004 (703)
Annual medical
cost of
Study Diagnosis depression
Luber Depressive $391
2001 disorder
Stoudemire Major $435
1986 depression
Greenberg Affective $853
1993 disorders
Unutzer Subclinical $933
1997 depression
Rice Affective $1,162
1995 disorders
Greenberg Affective $1,226
2003 disorders
Simon Depressive $2,270
1995 disorder
Katon Subclinical $2,519
2003 depression
Trivedi Depressive $6,535
2004 disorder
(a) Number in parenthesis indicates cases of depression.
(b) Cost estimation based on health care consumption of individual
patients using database of health care providers.
(c) Cost estimation based on national statistics.
(d) This dataset consists of all the required healthcare expenditures
needed for the study.
Notes. This table was adapted from Luppa et al. 2007. PC, Primary Care;
ECA, Epidemiological Catchment Area; MEPS, Medical Expenditure Panel
Survey.
Table 2
Description of Medicare recipients with and without a history of major
depressive disorder (MDD)
No MDD in both
Medicare and ECA
(N=472)
Personal characteristics
Age, years 77.5 [+ or -] 11.6
Female 307 (65.0)
White 334 (70.8)
Education
Eleventh grade or less 257 (54.4)
High school 127 (26.9)
Beyond high school 88 (18.6)
Married 204 (43.2)
Household income
Low (<$25K) 345 (73.1)
Medium ($25K to $50K) 89 (18.9)
High
([greater than or equal to]$50K) 38 (8.1)
Medical conditions
Myocardial infarction 91 (19.3)
Congestive heart failure 171 (36.2)
Peripheral vascular disease 192 (40.7)
Cerebrovascular disease 166 (35.2)
Chronic pulmonary disease 179 (37.9)
Peptic ulcer disease 39 (8.3)
Mild liver disease 53 (11.2)
Any diabetes 218 (46.2)
Hemiplegia or paraplegia 73 (15.5)
Renal disease 21 (4.4)
Any cancer 105 (22.2)
MDD
in Medicare
(N=59)
Personal characteristics
Age, years 73.5 [+ or -] 14.2 (*)
Female 45 (76.3)
White 43 (72.9)
Education
Eleventh grade or less 37 (62.7)
High school 13 (22.0)
Beyond high school 9 (15.3)
Married 19 (32.2)
Household income
Low (<$25K) 46 (78.0)
Medium ($25K to $50K) 10 (16.9)
High
([greater than or equal to]$50K) 3 (5.1)
Medical conditions
Myocardial infarction 23 (39.0) (**)
Congestive heart failure 28 (47.5)
Peripheral vascular disease 35 (59.3) (**)
Cerebrovascular disease 39 (66.1) (***)
Chronic pulmonary disease 30 (50.8)
Peptic ulcer disease 12 (20.3) (**)
Mild liver disease 12 (20.3)
Any diabetes 35 (59.3)
Hemiplegia or paraplegia 16 (27.1) (*)
Renal disease 7 (11.9) (*)
Any cancer 9 (15.3)
MDD
in ECA
(N=45)
Personal characteristics
Age, years 67.1 [+ or -] 12.1 (***)
Female 36 (80.0) (*)
White 29 (64.4)
Education
Eleventh grade or less 18 (40.0)
High school 19 (42.2)
Beyond high school 8 (17.8)
Married 18 (40.0)
Household income
Low (<$25K) 30 (66.7)
Medium ($25K to $50K) 8 (17.8)
High
([greater than or equal to]$50K) 7 (15.6)
Medical conditions
Myocardial infarction 10 (22.2)
Congestive heart failure 17 (37.8)
Peripheral vascular disease 19 (42.2)
Cerebrovascular disease 18 (40.0)
Chronic pulmonary disease 20 (44.4)
Peptic ulcer disease 4 (8.9)
Mild liver disease 5 (11.1)
Any diabetes 23 (51.1)
Hemiplegia or paraplegia 10 (22.2)
Renal disease 4 (8.9)
Any cancer 2 (4.4) (**)
Source. Baltimore Epidemiologic Catchment Area (ECA) Follow-up Study,
1981-2004 and from the Medicare, 1999-2004.
(*) P<0.05; (**) P<0.01; (***) P<0.001 for comparison between Medicare
recipients with no MDD in both Medicare and ECA as a reference group
and Medicare recipients with MDD in either Medicare or ECA using t-test
for continuous variables and Fisher's exact test for categorical
variables; all other comparisons were not significant (P>0.05).
Table 3
Relationship between depression based on ECA data and claim of
depression in Medicare data
No depression Any depression
based on Medicare data based on Medicare data Totals
No depression 472 49 521
based on ECA
data
Any depression 35 10 45
based on ECA
data
Totals 507 59 566
Kappa = 0.1122, 95% CI = (0.0043, 0.2201).
Source. Baltimore Epidemiologic Catchment Area (ECA) Follow-up Study,
1981-2004 and from the Medicare, 1999-2004.
Table 4
Six-year average Medicare costs for Medicare recipients with and
without a history of major depressive disorder (MDD)
Unadjusted Adjusted
Mean Cost Relative Cost Ratio
(95% CI) (95% CI)
No MDD in both
Medicare and ECA $40,670 1.00
(N=472) ($35,227-$46,112)
MDD in Medicare $87,445 (***) 1.87 (1.32-2.67) (***)
(N=59) ($61,655-$113,235)
MDD in ECA $43,583 1.33 (0.87-2.02)
(N=45) ($17,949-$69,218)
Source. Baltimore Epidemiologic Catchment Area (ECA) Follow-up Study,
1981-2004 and from the Medicare, 1999-2004.
Notes. Control variables in the multivariable models included terms for
depression status, age, gender, ethnicity, education, marital status,
household income, and Charlson comorbidity index.
(*) P<0.05; (**) P<0.01; (***) P<0.001.
CI: Confidence Interval.
