Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Ment Health Policy Econ. Author manuscript; available in PMC 2009 August 16.
Published in final edited form as:
PMCID: PMC2727632

Differences in Medical Care Expenditures for Adults with Depression Compared to Adults with Major Chronic Conditions

Ithai Z. Lurie, Ph.D.,1,* Larry M. Manheim, Ph.D.,2 and Dorothy D. Dunlop, Ph.D.3



Approximately 17.1 million adults report having a major depressive episode in 2004 which represents 8% of the adult population in the U.S. Of these, more than one-third did not seek treatment. In spite of the large and extensive literature on the cost of mental health, we know very little about the differences in out-of-pocket expenditures between adults with depression and adults with other major chronic disease and the sources of those expenditures.


For persons under age 65, compare total and out-of-pocket expenditures of those with depression to non-depressed individuals who have another major chronic disease.


This study uses two linked, nationally representative surveys, the 1999 National Health Interview Survey (NHIS) and the 2000 Medical Expenditure Panel Survey (MEPS), to identify the population of interest. Depression was systematically assessed using a short form of the World Health Organization's (WHO) Composite International Diagnostic Interview – Short Form (CIDI-SF). To control for differences from potentially confounding factors, we matched depressed cases to controls using propensity score matching.


We estimate that persons with depression have about the same out-of-pocket expenditures while having 11.8% less total medical expenditures (not a statistically significant difference) compared to non-depressed individuals with at least one chronic disease.


High out-of-pocket expenditures are a concern for individuals with chronic diseases. Our study shows that those with depression have comparable out-of-pocket expenses to those with other chronic diseases, but given their lower income levels, this may result in a more substantial financial burden.

Implication for Policy:

High out-of-pocket expenditures are a concern for individuals with depression and other chronic diseases. For both depressed individuals and non-depressed individuals with other chronic diseases, prescription drug expenditures contribute most to out-of-pocket expenses. Given the important role medications play in treatment of depression, high copayment rates are a concern for limiting compliance with appropriate treatment.


Approximately 17.1 million adults report having a major depressive episode in 2004 which represents 8% of the adult population in the U.S. [1]. Of these, more than one-third did not seek treatment [1]. Without treatment, symptoms of depression can persist for months or years. Depression affects the productivity of workers in terms of higher rates of absenteeism and reduced on-the-job output [2-5] and can lead to disability [6]. Lost productivity may account for more than 60% of the total social economic burden of depression in the US in 2000, which was estimated at $52.9 billion. The direct costs of depression are estimated to account for 31% of the total costs of depression [7].

If individuals with depression have higher medical expenditures, they are likely to also face higher out-of-pocket costs, a situation that would be exacerbated if individuals with depression pay higher copayments for the services they receive. This could occur either because they pay higher copays for the same type of services or because they have a different health services mix.

This paper estimates total and out-of-pocket expenditure differences between those with depression and those with at least one other major chronic condition, after controlling for differences in demographics, other chronic conditions, and economic resources. We use a special one-time 1999 National Health Interview Survey to obtain a systematic measure of depression independent of health services utilization. This information was linked to detailed information on subsequent medical expenditures. This allows us to obtain nationally representative measures of total and out-of-pocket medical expenditures for those with and without depression.


Estimating the out-of-pocket expenditures associated with depression is difficult because depressed patients frequently suffer from other medical conditions or they may have symptoms that are easily attributed to physical illness [8]. Alternatively, medical conditions may precipitate a depression episode.

Prior literature has estimated out-of-pocket expenditures of individuals with depression or psychological distress. Harman [9], focusing and those over age 65, found that the mean out-of-pocket expenditures of individuals with a diagnosis of depression were greater than for elderly persons with arthritis and hypertension, but similar to individuals with diabetes and heart disease. Harman [9] also found that only 8% of the total out-of-pocket expenditures by those with depression were for depression-specific services. Ringel and Sturm [10] calculated the share of out-of-pocket expenditures relative to family income among individuals who report “high psychological distress” or utilized specialty mental health services in the previous year. This study found that out-of-pocket expenditures were less than 10% of income for most individuals. Zuvekas [11] estimated that total out-of-pocket expenditures represented 23 percent of the total mental health cost. However, the first study lacks an adequate comparison group and the latter two studies based estimates only on users of the mental health system.

Many studies focus on patients with a specific diagnosis like HIV [12] or diabetes [13] and estimate the difference in expenditures between depressed and non-depressed individuals for the specific diagnosis. Other studies estimate total and out-of-pocket expenditures that can be linked directly to mental health use either by diagnosis codes or by type of treatments [10]. Results from the former approach cannot be generalized to the entire population. The latter approach does not account for individuals with depression who forgo mental health treatment, but have greater utilization of non-mental health related services, nor does it control for differences in characteristics of depressed and non-depressed individuals that may influence patterns of medical expenditures.

This study adds to the literature in three ways. First, a systematic validated measure of depression, the Composite International Diagnostic Interview – Short Form (CIDI-SF) [14] is used to identify individuals with depression. Second, since our identification of depression does not depend on utilization of medical services (e.g., diagnostic data) and prospective data are analyzed, we are able to avoid the problem of only including depressed individuals who have already accessed the medical system. Finally, our estimation is based on propensity score matching to identify a comparison group (with at least one major chronic condition but without depression) with similar sociodemographic, and health characteristics as the target group (individuals with depression). This allows estimation of the effect of a diagnosis of depression, based on the CIDI-SF, after accounting for potential confounding effects of sociodemographic characteristics, and co-morbidities.

Data and Methods

Data Sample

This study uses two linked, nationally representative surveys, the 1999 National Health Interview Survey (NHIS) and the 2000 Medical Expenditure Panel Survey (MEPS), to identify the population of interest. The NHIS is a national representative sample of the U.S. civilian non-institutionalized population, with over-sampling of Hispanics and African-Americans. In 1999 NHIS included a mental health supplement for a subgroup of respondents. This one time NHIS supplement provided a systematic assessment of depression, although on a limited sample. The MEPS is a sub-sample of the NHIS. The MEPS household component collects information at the person level on medical utilization, expenditures, demographic characteristics, health factors, and economic information. Our NHIS/MEPS study sample linked the 1999 NHIS mental health supplement sample of adults 18-64 with the 2000 MEPS sample. Person-weights, stratum, and sampling error codes for the 1999 NHIS allow population estimates. Our analyses are restricted to 2,578 persons in the 2000 MEPS sample ages 18-64 who responded to the 1999 NHIS mental health supplement subsample of 24,799 persons. In order to examine whether those with depression had similar medical expenditures to individuals with other major chronic conditions, the sample was limited to individuals that had depression or at least one of the following chronic conditions: arthritis, asthma, cardiovascular disease, diabetes, or stroke. For analytic purposes, we also excluded 25 individuals with incomplete baseline data to obtain an analysis sample of 969 individuals with NHIS mental health data and MEPS expenditure data.

Outcome Variables

MEPS provides expenditures for each individual by type of service. Total health care expenditures (defined here as the sum of expenditures on ambulatory care, physician office visits, emergency room visits, hospital inpatient, and prescription drugs) were obtained from the MEPS 2000 interview, which followed the baseline NHIS interview. Those data are collected by the MEPS for each sampled person and are then summarized to provide annual utilization and expenditure data [15].

Independent Variables

Depression status was determined from the 1999 NHIS which included supplemental questions designed to collect information on mental disorders in the U.S. adult population. The supplemental questions were asked of one adult age 18 and older selected randomly from each responding household. Depression refers to a depressive episode [16] within the last 12 months and was measured using a short form of the World Health Organization's (WHO) Composite International Diagnostic Interview – Short Form (CIDI-SF) [14]. The full WHO-CIDI has excellent validity and reliability [17, 18] among individuals of different nationalities and is currently the most widely accepted method to determine the prevalence of psychiatric disorders in the United States using lay interviewers [19]. The CIDI-SF retains only those questions needed to ascertain diagnoses defined in the Diagnostic and Statistical Manual of Mental Disorders of the American Psychiatric Association, third edition revised (DSM-III-R), uses a twelve month time frame to capture current depression, and re-organizes questions to minimize interview burden. The CIDI-SF is a valid and reliable assessment instrument with an accuracy of 93% for major depressive disorder [20]. A major depressive episode is attributed to a CIDI-SF score of 3 or more symptoms from a 0 to 7 scale following recommended guidelines [19].

Sociodemographic variables

NHIS collects detailed demographic information on a personal level. For our analysis we used the demographic information collected in 1999 from the NHIS. Ethnicity/race information was used to classify people into two mutually exclusive groups: non-Hispanic White and Other. NHIS also provides information on gender, age, marital status, income relative to Federal Poverty Line (FPL), and education. Educational attainment was used to classify people into three mutually exclusive groups (high school graduate or less, some college, and college degree). MEPS information on health insurance status concurrent with medical utilization was categorized: some private coverage during the year, Medicare coverage only (a disabled population for those under age 65), public coverage only, or uninsured.

Other Covariates

Individuals are in this sample if they have depression, arthritis, asthma, cardiovascular disease, diabetes, or stroke. We included a covariate to indicate the number of other chronic diseases (beyond depression or, for the comparison group, beyond one) individuals have to further control for health status, we included an indicator variable for cancer and an indicator variable if individuals had at least one of the following: kidney stones, ulcer or liver problems. Presence of these chronic conditions is derived from NHIS questions about the presence of chronic conditions.

Data Analytic Procedures

Descriptive analyses based on NHIS/MEPS subsample (rather than matched case-control clusters) reflect the U.S. population. Respondents in the NHIS and MEPS subsamples are handled as an additional sampling stages to readjust sampling weights using standard sampling methodology [21]. The probability of being in the NHIS/MEPS mental health subsample compared to those not selected was estimated for each individual as a function of race, age, gender, marriage, education and the report of any additional chronic disease. The sampling weight for the NHIS/MEPS mental health subsample equals the 1999 NHIS mental health supplement sampling weight multiplied by the inverted probability of selection given these characteristics; that probability is estimated using logistic regression. These weights are used to provide unadjusted population estimates for the different subgroups.

The comparison of medical expenditures of persons with and without depression used a matched case-control design. This approach recognizes that the MEPS and NHIS are observational studies, persons with and without depression may differ in characteristics that also influence the outcome (medical expenditures) and confound analytical findings. To control for potentially confounding factors, we used propensity score matching of cases (depression), following the methodology of Rosenbaum and Robin [22].

Propensity score matching facilitates the selection of controls, which are similar to cases over a profile of characteristics. Since it is not feasible to match individuals using all their characteristics, propensity score matching summarizes baseline characteristics into a single-index variable (the propensity score) to facilitate matching. For all 969 individuals in our study sample, we estimate the likelihood of having depression based on an individual's characteristics. This likelihood is estimated from a probit regression model.

Each depression case is matched with a person from the control group (without depression) who has the closest propensity score to that of the case's score using the nearest neighbor matching with common support [23]. The match is deemed successful if the estimation balances the “pre-assignment” covariates for the case and control conditional on the propensity score. Balance is assessed using Dehejia's methodology [24]. Once cases are matched to controls, the effect of depression on medical expenditures is estimated from matched case/control pairs by examining the average difference in expenditures between case/control pairs. Standard errors of the differences are estimated by bootstrapping methods [23]. Because population weights are no longer meaningful in the context of matched samples (i.e., only those non-depressed control individuals are selected who look most like cases with depression), the results of the matched case-control analyses are unweighted.


Table 1 shows population estimates of baseline characteristics based on depression status for the entire NHIS/MEPS sample (n=969). Individuals with depression represent about 15.9% of the total sample. Persons with depression compared to those with another chronic condition are more likely to be female (72.1% versus 55.7%), single (56.8% versus 37.7%), attain no more than a high school education (52.5% versus 46.0%), are less likely to have some private coverage (55.3% verses 76.8%) and have lower income (50.7% with income below 300% FPL versus 35.3%).

Table 1
Population Weighted Distribution of Sociodemographic Characteristics for Individuals with Depression Compared to Non-depressed Individuals with Another Chronic Disease

Table 2 shows the population mean, minimum, maximum, median, and interquartile range (25th to 75th percentile range) for out-of-pocket and total expenditures for all sampled NHIS/MEPS adults with depression (n=151) and without depression (n=818). The distribution of expenditures are skewed (both total and out-of-pocket), which is common in health expenditures data. On average, total expenditures are $718.9 lower and out-of-pocket expenditures are $116.8 higher for individuals with versus without depression. The share of expenditures that are paid out-of-pocket is 27.8% ($665/$2,387) for individuals with depression and 17.6% ($547/$3,105) for individuals without.

Table 2
Population Weighted Distribution of Total, and Out-of-Pocket Expenditures for Individuals with Depression Compared to Non-depressed Individuals with Another Chronic Disease

Recognizing that differences in socio-demographic, insurance, and health characteristics may confound observed medical expenditures differences related to depression, we identified matched controls. Following Becker and Ichino, [23]) the propensity score is used to select the control subject with the closest propensity score to each of the 151 depression cases, whether or not that matched control subject was chosen for another depressed individual. This led to 117 matches, less than one per depressed individual. The propensity score estimation is based on a probit model, which is provided in the appendix. Dehejia's [24] balance property for a successful match was satisfied. Table 3 shows unweighted baseline characteristics based on depression status for the entire NHIS/MEPS sample (n=969) and for the matched sample. Sample characteristics in Table 3 show that the matched sample characteristics from the propensity score matching model are similar to the depressed sample, confirming a satisfactory match.

Table 3
Sociodemographic Characteristics (Un-weighted) of Individuals With Depression Matched to Non-Depressed Controls with Another Chronic Disease

Table 4 shows, the means of total expenditures, the means of out-of-pocket expenditures and the mean differences in expenditures for the depressed group and the matched non-depressed chronic disease group by type of expenditures. Total expenditures are 11.8% lower for the depressed group, compared to their matched chronic disease controls, which is not a statistically significant difference. On the other hand, total out-of-pocket expenditures for the depressed group are similar to their matched non-depressed chronic disease group ($676 versus $682).

Table 4
Medical Care Expenditures Differences between Individuals with Depression and Matched non-Depression Controls Who Have Another Chronic Disease


Our findings do not show a statistically significant difference in total expenditures or out-of-pocket expenditures between those who are depressed and those who are non-depressed but have another chronic condition. This study establishes these findings using the methodologically strong platform of a systematic assessment of depression in contrast to assessment based on utilization, a national sample, and the use of a matched control group to account for potential confounders.

The greatest contributor to out-of-pocket expenses is prescription medications, which have similar but very high co-payments relative to other services for both depressed and not depressed persons with other chronic conditions. Ambulatory/ physician visits also shows large total expenditures, but these services play a much smaller role in the total out-of-pocket expenditures because of lower co-payments for those services relative to prescription medication. For most services, persons with depression generally pay similar proportions out-of-pocket as their not depressed peers with another chronic condition. There is a difference in inpatient out-of-pocket cost in this group with chronic disease for persons with and without depression. However, we cannot tell if this result is meaningful because of the small sample of individuals with inpatient stays.

Because prescription medication plays a large role in the treatment of depression, [25] the high co-payment rate for prescription medication may serve as a barrier to adhering to appropriate care for depression. High out-of-pocket costs for prescription medication are the largest contributor to the substantially greater economic burden (out-of-pocket expenditures relative to income) that individuals with depression face compared to those without depression [26]. High out-of-pocket costs may cause depressed persons to forgo medication treatment or depression treatment as a whole which can contribute to poor adherence [27-29]. Ambulatory/ physician visits services, although important for depression treatment, are less of a barrier for treatment in terms of expenditures because of the low contribution to the economic burden it provides relative to prescription medication.

Further, in the general population a larger proportion of depressed individuals reported income below 200% of the poverty level, compared to those with another chronic disease, but without depression (Table 1). Therefore, this higher out-of-pocket expenditure represents an even greater percent of their income, resulting in a higher financial burden.

This study is limited by the modest number of cases of depression in the sample. The modest number is due to the administration of the CIDI-SF to only a subsample of NHIS respondents. However, the use a systematic assessment of depression using the CIDI-SF is valuable for a valid epidemiologic assessment because it provides a systematic ascertainment of depression that is unrelated to utilization of medical services. The major area where the small sample size is problematic is the estimate of hospital expenditures, where it is not clear whether the observed out-of-pocket differences (not statistically significant) reflect true differences (not significant due to a lack of power) or simply represent random variation.

Other limitations of the study should be noted. The presence of depression and other diseases are ascertained in 1999 while expenditures are taken from 2000. It is possible that some people will no longer have a disease in the following year, though the focus on chronic depression assessed by the CIDI-SF and chronic conditions should limit this problem. Because the other chronic diseases are measured by patient response as to the their presence while depression is systematically measured,, our comparison group may include people with more severe forms of the other diseases – those with less symptoms may not have been diagnosed. This will tend to overstate costs of the other chronic diseases relative to depression. Finally, the data this study is based on is now almost ten years old and it is possible that expenditures today have a different pattern than in 2000.

Harmon [9] found that for those over age 65, mean out-of-pocket expenditures of people with a diagnosis of depression were greater than for people with arthritis or hypertension but similar to individuals with diabetes and heart disease. This paper compares out-of-pocket costs for those under age 65 and also finds them similar. Zuvekas [11] found that out-of-pocket expenditures represented 23 percent of the total mental health cost, similar to our finding that the share of all medical expenditures for those that are paid out-of-pocket is 27.8% for individuals with depression. Ringel and Sturm [10] found out-of-pocket expenditures were less than 10% of overall income for most people either reporting psychological distress or using mental health services in the last year. While we do not have a comparable data, we do find that at the 75th percentile (table 2) out-of-pocket expenditures are just $477 for those with depression. However, 28.5% of those with depression under age 65 had income less than twice the poverty level and mean out-of-pocket expenditures were $665, suggesting the out-of-pocket costs are extremely skewed with a low percentage of people with very high out-of-pocket expenditures (maximum amount = $11,028).

In summary, high out-of-pocket expenditures appear to be an equally large concern for individuals with depression and for non-depressed individuals with other chronic diseases. Because prescription medications play a large role in the treatment of depression and other chronic diseases, and because those with depression, on average, have lower income, this high co-payment rate may serve as a barrier to appropriate care of depression, at least for a small subset of those with depression.


This study is supported funding from NIH/National Institute for Arthritis and Musculoskeletal Diseases P60-AR48098, and NIH/ National Center for Medical Rehabilitation Research R01-HD45412.


Probit Estimation Used for Propensity Score Matching

Propensity Score Estimation
Covariates\Specification (N=969)
(standard Deviation)
[Marginal Effect]

Number of Other Chronic Diseases *0.521
Age squared−0.001
Race: Other (relative to non-Hispanic White)−0.233
Some College (Relative to High School diploma)−0.067
College Diploma (Relative to High School diploma)−0.018
Medicare Only (Relative to any private)0.214
Other Public Only (Relative to any private)0.492
Uninsured all year (Relative to some private)0.382
Above 300% FPL Income (Relative to below 300% FPL)−0.045
Do Not Know Income (Relative to below 300% FPL)−0.001
Kidney Stones/Liver/Ulcer0.301

Note: Bold coefficients are significant at a nominal 5% level

*Includes arthritis, asthma, cardiovascular disease, diabetes, or stroke


1. Office of Applied Studies (OAS) RTI International U.S.D.O.H.H. SERVICES, editor. The NSDUH Report: Depression among Adults. 2005
2. Simon GE, Barber C, Birnbaum HG, Frank RG, Greenberg PE, Rose RM, Wang PS, Kessler RC. Depression and work productivity: the comparative costs of treatment versus nontreatment. J Occup Environ Med. 2001;43(1):2–9. [PubMed]
3. Burton WN, Conti DJ. The real measure of productivity. Bus Health. 1999;17(11):34–6. [PubMed]
4. Kessler RC, Barber C, Birnbaum HG, Frank RG, Greenberg PE, Rose RM, Simon GE, Wang P. Depression in the workplace: effects on short-term disability. Health Affairs. 1999;18(5):163–71. [PubMed]
5. Kessler RC, Frank RG. The impact of psychiatric disorders on work loss days. Psychol Med. 1997;27(4):861–73. [PubMed]
6. Dunlop DD, Manheim LM, Song J, Lyons JS, Chang RW. Incidence of disability among preretirement adults: the impact of depression. Am J Public Health. 2005;95(11):2003–8. [PubMed]
7. Greenberg PE, Kessler RC, Birnbaum HG, Leong SA, Lowe SW, Berglund PA, Corey-Lisle PK. The economic burden of depression in the United States: how did it change between 1990 and 2000? J Clin Psychiatry. 2003;64(12):1465–75. [PubMed]
8. Dunlop DD, Lyons JS, Manheim LM, Song J, Chang RW. Arthritis and heart disease as risk factors for major depression: the role of functional limitation. Medcal Care. 2004;42(6):502–11. [PubMed]
9. Harman JS, Kelleher KJ, Reynolds CF, Pincus HA. Out-of-pocket healthcare expenditures of older Americans with depression. J Am Geriatr Soc. 2004;52(5):809–13. [PubMed]
10. Ringel JS, Sturm R. Financial Burden and Out-of-Pocket Expenditures for Mental Health Across Different Socioeconomic Groups: Results from HealthCare for Communities. J Ment Health Policy Econ. 2001;4(3):141–150. [PubMed]
11. Zuvekas SH. Trends in mental health services use and spending, 1987-1996. Health Aff (Millwood) 2001;20(2):214–24. [PubMed]
12. Joyce GF, Chan KS, Orlando M, Burnam MA. Mental health status and use of general medical services for persons with human immunodeficiency virus. Med Care. 2005;43(8):834–9. [PubMed]
13. Egede LE, Zheng D, Simpson K. Comorbid depression is associated with increased health care use and expenditures in individuals with diabetes. Diabetes Care. 2002;25(3):464–70. [PubMed]
14. Wittchen H-U. Reliability and validity studies of the WHO-Composite International Diagnostic Interview (CIDI): A critical review. Journal of Psychiatric Research. 1994;28(1):57–84. [PubMed]
15. Center for Financing Access and Cost Trends Agency for Healthcare Research and Quality MEPS HC-038: 1999 Full Year Consolidated Data File - Utilization. Oct, 2002. 2002 [cited 2004 October 28th]; Available from:
16. United States Department of Health and Human Services . Mental Health: A Report of the Surgeon General. U.S. Department of Health and Human Services, Substance Abuse and Mental Health Services Administration, Center for Mental Health Services, National Institutes of Health, National Institute of Mental Health; Rockville, MD: 1999.
17. Tacchini G, Coppola MT, Musazzi A, Altamura AC, Invernizzi G. Multinational validation of the Composite International Diagnostic Interview (CIDI) Minerva Psichiatr. 1994;35(2):63–80. [PubMed]
18. Wittchen HU. Reliability and validity studies of the WHO--Composite International Diagnostic Interview (CIDI): a critical review. J Psychiatr Res. 1994;28(1):57–84. [PubMed]
19. Walters EE, Kessler RC, Nelson CB, Mroczek D. Scoring the World Health Organization's. Composite International Diagnostic Interview Short Form (CIDI-SF) 2002;1.1
20. Kessler RC,AG, Mroczek D, Utsun TB, Wittchen HU. The World Health Organization's Composite International Diagnostic Interview Short-Form (CIDI SF. International journal of methods in psychiatric research. 1997;7:171–185.
21. Korn E, Graubard BI. Analysis of Health Surveys. John Wiley and Sons, Inc; New York: 1999.
22. Rosenbaum P, Rubin D. The Central Role of the Propensity Score in Obsrvational Studies for Causal Effects. Biometrika. 1983;70(1):41–55.
23. Becker SO, Ichino A. Estimation of average treatment effects based on propensity scores. The Stata Journal. 2002;2(4):358–377.
24. Dehejia R. Practical Propensity Score Matching: A Reply to Smith and Todd. Journal of econometrics. 2005;125(12):355–64.
25. (NIMH), N.I.o.M.H. Depression. In: D.o.H.a.H. Services, editor. NIH Publication. NIH Publication; 2000.
26. Barry CL, Gabel JR, Frank RG, Hawkins S, Whitmore HH, Pickreign JD. Design of mental health benefits: still unequal after all these years. Health Aff (Millwood) 2003;22(5):127–37. [PubMed]
27. Cantrell CR, Eaddy MT, Shah MB, Regan TS, Sokol MC. Methods for evaluating patient adherence to antidepressant therapy: a real-world comparison of adherence and economic outcomes. Med Care. 2006;44(4):300–3. [PubMed]
28. Robinson RL, Long SR, Chang S, Able S, Baser O, Obenchain RL, Swindle RW. Higher costs and therapeutic factors associated with adherence to NCQA HEDIS antidepressant medication management measures: analysis of administrative claims. J Manag Care Pharm. 2006;12(1):43–54. [PubMed]
29. Steinman MA, Sands LP, Covinsky KE. Self-restriction of medications due to cost in seniors without prescription coverage. J Gen Intern Med. 2001;16(12):793–9. [PMC free article] [PubMed]