Health Affairs, 22, no. 2 (2003): 129-138 doi: 10.1377/hlthaff.22.2.129 © 2003 by Project HOPE	New Online   Pay Cuts For Medicare Docs   Access To Care Woes   Public Coverage More Efficient   Empowering Consumers

This Article Abstract Reprint (PDF) Submit a response to this article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to Citation Manager Reprints & Permissions Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (38) Citing Articles via Google Scholar Google Scholar Articles by Cohen, J. W. Articles by Krauss, N. A. Search for Related Content PubMed PubMed Citation Articles by Cohen, J. W. Articles by Krauss, N. A. Related Collections Chronic Care Mental Health/Substance Abuse Quality Of Care Health Spending

Research Challenge

Spending And Service Use Among People With The Fifteen Most Costly Medical Conditions, 1997

Abstract

 
This study addresses the Institute of Medicine’s recommendation that AHRQ use MEPS data to identify a set of priority conditions to inform efforts at improving quality of care. Using MEPS data we identify the fifteen most expensive conditions in the U.S. in 1997: chronic diseases such as heart disease, cancer, and diabetes, and acute conditions such as trauma, pneumonia, and infectious disease. Comorbidities were also associated with increased expenses. Type-of-service and source-of-payment distributions varied considerably across this set of conditions. Our findings highlight some of the challenges likely to be encountered in efforts to reform the current system.

Although the set of conditions developed for this analysis is based on expenditures that are directly related to services provided for the condition in question, the information gained from estimating the total direct cost for a specific medical condition is limited insofar as it underestimates the resources consumed by people with the condition of interest. Previous research has shown that a sizable proportion of the population with chronic conditions has more than one such condition and that a sizable proportion of people with chronic conditions have complications resulting in higher resource consumption.³ If comorbidity and other complicating factors are not included in cost estimates, then complete expenditures related to medical conditions are not captured. Consequently, this study examines both direct health care costs for the fifteen most costly conditions and the total costs for all medical care incurred by people with these conditions. We also describe the spending distribution across types of services and the impact of coexisting conditions on spending for people with these conditions.

Study Methods

Top Study Methods Study Results Discussion And Policy... NOTES

 
The MEPS Household Component (MEPS HC), conducted by the Agency for Healthcare Research and Quality (AHRQ), is designed to provide estimates of health care use, spending, sources of payment, and insurance coverage for the U.S. civilian noninstitutionalized population, and to provide a comprehensive database for supporting behavioral research related to issues of health policy.⁴ The MEPS HC uses an overlapping panel design in which data are collected through a preliminary contact followed by a series of five rounds of interviews over a two-and-a-half-year period to collect self-reports of health care use and spending. Data for this paper were derived from the 1997 Full Year Public Use File (HC-020), which includes information collected for the second year of the 1996 Panel, the first year of the 1997 panel, and the 1997 Medical Conditions File (HC-018).⁵ The 1997 full-year MEPS HC comprises a sample of 33,000 people and oversamples several categories of people who were likely to have higher-than-average use of medical care.⁶ Thus, although more recent years of MEPS data are available, the size and composition of the 1997 sample make it best suited for this analysis.

Condition data are collected from household respondents during every round as verbatim text and coded by professional coders using the International Classification of Diseases, Ninth Revision (ICD-9). Because they are self-reports, conditions identified by household respondents may not conform perfectly to diagnoses made by physicians. A study by AHRQ staff, however, indicated that at the three-digit ICD-9 code level, there was agreement between household- and provider-reported conditions in the overwhelming majority of cases.⁷ Condition categories for this study were constructed using AHRQ’s Clinical Classification Software (CCS), which aggregates ICD-9 codes into clinically meaningful categories.⁸ During the review process for this analysis, some categories were collapsed when appropriate. Decisions about collapsing CCS categories were based on retaining the clinical significance of categories, the ability of household respondents to accurately report the condition, and the frequency of reporting the condition.⁹

The MEPS HC also collects detailed information about health care use and spending during every round of data collection. Payment information is based primarily on a survey of the specific providers used by sampled people, which collects information about diagnoses, charges, and payments for their individual medical events.¹⁰ For this analysis medical events were classified and enumerated into mutually exclusive categories as hospitalizations, emergency room visits, outpatient hospital visits, office-based visits, dental visits, home health care, and the purchase of prescribed medicines. We then classified all office-based, outpatient hospital, and emergency department visits as ambulatory care.¹¹

All estimates presented in this study were weighted to represent the U.S. civilian noninstitutionalized population.¹² The sampling weights also were designed to adjust for potential survey nonresponse bias. SUDAAN was used to account for the complex sample design in the calculation of the standard errors for the estimates presented. All differences discussed in the text are significant at the .05 level or better.¹³

Study Results

Top Study Methods Study Results Discussion And Policy... NOTES

 
The fifteen most expensive conditions. Exhibit 1 examines spending for the top fifteen conditions for calendar year 1997, ranked according to the total amounts spent for care. The data reflect spending for only medical care that was directly related to the condition in question—that is, for events that were at least in part or entirely associated with care for that condition, not all spending for people with the condition. Note also that because a provider visit may occur for multiple reasons, spending associated with specific conditions is not mutually exclusive. Because the primary condition associated with a visit is not identified in MEPS, we could not assign a primary condition to each event.¹⁴ Consequently, spending estimates in Exhibit 1 summed over conditions will double-count some expenditures.

With respect to the size of the population reporting each condition, pulmonary conditions affected the largest number of people (forty-one million), followed by trauma (thirty-seven million), and then hypertension (twenty-seven million). There were also large differences in mean expenditures per person across conditions, as well as in the distributions by type of service. Examining the combination of the size of the population affected by a condition and the mean expenditures per person for services resulting from that condition suggests that both population size and the cost of providing care determine whether a condition is among the most expensive. Across all conditions, neither of these two factors appeared to dominate as a reason for ranking high on the list.

In general, higher mean expenditures were associated with the extent to which spending was related to inpatient hospitalization. Inpatient care accounted for approximately two-thirds of expenditures for cerebrovascular disease and cancer, the top two conditions in terms of mean expenditures, but only about one-quarter of expenditures for infectious diseases and skin disorders, the bottom two. Inpatient expenditures were not always associated with higher average expenses, however. Kidney disease had the third-highest mean but showed only a little more than a third of spending associated with inpatient care. More than half of spending for kidney disease was accounted for by ambulatory visits, most likely related to dialysis. Interestingly, there also appears to be an overall inverse relationship between spending for prescription drugs and inpatient spending, which is consistent with previous research on the effects of drug use on hospitalizations.¹⁶

Medical expenditures are highly concentrated, with a small proportion of the population accounting for a large proportion of expenses, creating a highly skewed distribution.¹⁷ Examining median compared with mean expenditures illustrates that the overall skewness of medical expenditures is also apparent when looking at expenditures within conditions. For all fifteen conditions the medians are much lower than the means, which indicates that within these conditions, there are a small number of very-high-cost cases contributing disproportionately to the mean, the extent of which varies by condition. In general, it appears that there is much more variability in the mean expenditures than the medians, which again is indicative of a high degree of skewness, particularly among conditions with higher mean expenditures.

Sources of payment. Private insurance and Medicare were the primary sources of payment for most of the top fifteen conditions (Exhibit 2). Among the top five conditions (heart disease, cancer, trauma, mental disorders, and pulmonary conditions), private insurance accounted for at least 35 percent of all directly related expenses for each; it exceeded 50 percent for cancer. Medicare accounted for at least 30 percent of the total for seven conditions and never covered less than 16 percent. Hypertension was associated with the highest out-of-pocket proportion (31 percent). Medicaid did not cover a high proportion of these expenditures.

As was true for condition-related-only spending, mean spending for people with the fifteen most costly conditions was highest for those with cerebrovascular disease, kidney disease, and cancer. There were still large differences between the mean and median expenditures, again indicating a skewed distribution. As was also true for the directly related expenditures, the distribution of expenditures across types of service at the person level showed a heavy concentration of inpatient expenditures, although in some cases the distributions were very different after expenditures for comorbidities were included. For example, although only 10 percent of direct spending for hypertension was for inpatient care, at the person level 43 percent of spending for people with hypertension was for inpatient hospitalization. This indicates that people with hypertension are likely to incur inpatient expenses but that those expenses are in large part not directly related to that condition.

Expenditures by number of conditions. Most people with at least one of the top fifteen conditions have comorbidities (Exhibit 4). These data illustrate that spending increases dramatically as the number of conditions rises. For many conditions, the differences were dramatic, with as much as eight- to elevenfold differences between per capita spending for people with only one versus three or more conditions. The spending distribution across types of services, however, did not appear to vary a great deal according to the number of conditions present.

Top Study Methods Study Results Discussion And Policy... NOTES

There were also substantial differences in the distributions of services and sources of payment for these conditions; these differences have implications for reform efforts. For example, although spending for many of the conditions was concentrated in inpatient care, this was not always the case. In addition, although prescription drug spending has been one of the fastest-growing components of health care spending in recent years, it approaches half of the total spending level for only one of these top fifteen conditions (hypertension).²⁰ This variability suggests that financial incentives and care processes will have to be tailored to fit a wide variety of circumstances.

The source-of-payment distributions among these high-cost conditions also suggest that there will have to be coordination across payers to properly align the financial incentives of the various payment systems used by private and public insurers to encourage quality improvement. For example, while the private insurance system in recent years has moved heavily toward managed care, Medicare has remained primarily a fee-for-service system. To the extent that providers organize care to respond to the incentives of their primary payers, differences in those incentives across payers will make it difficult to ensure that the alignment of incentives is proper for all types of patients.

Informing efforts to reform the U.S. health care system is no simple task. This analysis has been a first attempt to satisfy the IOM’s request for AHRQ to examine a set of priority conditions. Although we have successfully completed this task, further research is needed to determine the specific characteristics of people with these conditions and how those characteristics affect spending levels, and to sort out the independent effects of comorbidities on spending for health care. The data presented here may represent a starting point, but much work remains to be done.

Editor's Notes

 
Joel Cohen directs the Division of Social and Economic Research, Center for Cost and Financing Studies, Agency for Healthcare Research and Quality, in Rockville, Maryland. Nancy Krauss is a survey statistician in that center.

The views expressed in this paper are those of the authors, and no official endorsement by the Agency for Healthcare Research and Quality or the Department of Health and Human Services is intended or should be inferred.

NOTES

Top Study Methods Study Results Discussion And Policy... NOTES

 

1. Institute of Medicine, Crossing the Quality Chasm: A New Health System for the Twenty-first Century (Washington: National Academy Press, 2001).
2. See, for example, B.G. Druss et al., "Comparing the National Economic Burden of Five Chronic Conditions," Health Affairs (Nov/Dec 2001): 233–241; and B.G. Druss et al., "The Most Expensive Medical Conditions in America," Health Affairs (July/Aug 2002): 105–111.
3. C. Hoffman et al., "Persons with Chronic Conditions: Their Prevalence and Costs," Journal of the American Medical Association 276, no. 18 (1996): 1473–1479; [Abstract]T. Baker et al., "Challenges to Identifying Actual Case-Mix Complexity and Total Treatment Charges," Disease Management 10, no. 4 (1997): 91–96; E. Yellin et al., "A National Study of Medical Care Expenditures for Respiratory Conditions," European Respiratory Journal 19, no. 3 (2002): 414–421; and [Abstract/Free Full Text]T.A. Hodgson and A.J. Cohen, "Medical Care Expenditures for Diabetes, Its Chronic Complications, and Its Comorbidities," Preventive Medicine 29, no. 3 (1999): 173–186.[Medline]
4. For further information on the survey, see J.W. Cohen et al., "The Medical Expenditure Panel Survey: A National Health Information Resource," Inquiry (Winter/Spring 1996–1997): 373–389.
5. Full documentation for Public Use File (PUF) HC-020 and PUF HC-018 is available from Agency for Healthcare Research and Quality, "MEPS HC-020: 1997 Full Year Consolidated Data File," www.meps.ahrq.gov/Puf/PufDetail.asp?ID=36 (9 December 2002).
6. Using the 1997 MEPS files provided the largest relevant sample sizes for this type of analysis. In addition to having oversamples of people with impairments and those predicted to have high medical expenditures, the overall 1997 sample size was approximately 50 percent larger than for the 1996 or 1998 MEPS. For a detailed description of the 1997 MEPS household sampling methodology, see S.B. Cohen, Sample Design of the 1997 Medical Expenditure Panel Survey Household Component, MEPS Methodology Report no. 11, AHRQ Pub. no. 01-0001 (Rockville, Md.: AHRQ, 2000).
7. N. Krauss and B. Kass, "Comparison of Household and Medical Provider Reports of Medical Conditions" (Paper presented at the Joint Statistical Meetings, Indianapolis, Indiana, August 2000).
8. Clinical Classification Software was formerly called Clinical Classification for Health Policy Research. See A. Elixhauser et al., Clinical Classifications for Health Policy Research: Hospital Inpatient Statistics, 1996, Health Care Utilization Project, HCUP-3 Research Note, Pub. no. 98-0049 (Rockville, Md.: AHRQ, 1998).
9. Collapsed CCS codes were reviewed by Healthcare Cost and Utilization Project (HCUP) staff at AHRQ. Codes were aggregated into the following categories: heart disease, 96, 97, 100–108; cancer, 11–45; trauma, 225–236, 239, 240, 244; mental disorders, 65–75; pulmonary conditions, 127–134; diabetes, 49, 50; hypertension, 98, 99; cerebrovascular disease, 109–113; arthritis, 201–204; pneumonia, 122; kidney disease, 156–158, 160, 161; endocrine disorders, 48, 51, 52, 54–58; skin disorders, 197–200; back problems, 205; and infectious diseases, 1–9.
10. For a detailed description of the MEPS Medical Provider Component, see S. Machlin and A. Taylor, Design, Methods, and Field Results of the 1996 Medical Expenditure Panel Survey Medical Provider Component, MEPS Methodology Report no. 9, Pub. no. 00-0028 (Rockville, Md.: AHRQ, 2000).
11. In cases where emergency department visits were immediately followed by an inpatient hospital stay, the emergency department expenses are included with inpatient hospital expenses.
12. Note that the population covered by the survey excludes people in institutional living arrangements, such as nursing homes. Thus, not all types of health care spending are represented, and some payment sources, such as Medicaid, which covers a large portion of nursing home expenditures, may be underrepresented as well.
13. Standard errors for all of the estimates presented are available from the authors on request. Send e-mail to Joel Cohen, jcohen{at}ahrq.gov.
14. Although spending estimates in Exhibit 1 represent some double-counting as a result of the inability to identify the primary reason for a provider visit, analyses limited to principal diagnoses would produce an underestimate of diagnoses that tend to be secondary such as hypertension or dementia. See D.S. May et al., "Surveillance of Major Causes of Hospitalization among the Elderly, 1988," Morbidity and Mortality Weekly Report 40 (SS-1) (1991): 7–17.
15. Medical care spending for the U.S. civilian noninstitutionalized population totaled $553 billion in 1997. This is less than the amount reported in the Centers for Medicare and Medicaid Services’ (CMS’s) National Health Accounts, largely because of differences in the scope of the populations and services included. For a thorough analysis of these differences, see T. Selden et al., "Reconciling Medical Expenditure Estimates from the MEPS and the NHA, 1996," Health Care Financing Review 23, no. 1 (2001): 161–178.[Medline]
16. F.R. Lichtenberg, "Are the Benefits of Newer Drugs Worth Their Cost? Evidence from the 1996 MEPS," Health Affairs (Sep/Oct 2001): 241–251.
17. M.L. Berk and A.C. Monheit, "The Concentration of Health Care Expenditures, Revisited," Health Affairs (Mar/Apr 2001): 9–18.
18. As noted previously, these spending estimates are not mutually exclusive, and spending for care of people with multiple conditions is included in each condition category.
19. Note that this ranking is based on annual spending. A ranking by lifetime spending might introduce some rare conditions that are not prevalent enough to be adequately represented in the survey and could change the rankings.
20. K. Levit et al., "Inflation Spurs Health Spending in 2000," Health Affairs (Jan/Feb 2002): 172–181.

                      What's this?

This article has been cited by other articles:

				J. S. Banthin and D. M. Bernard Changes in Financial Burdens for Health Care: National Estimates for the Population Younger Than 65 Years, 1996 to 2003 JAMA, December 13, 2006; 296(22): 2712 - 2719. [Abstract] [Full Text] [PDF]

				R. L. Nahin Identifying and pursuing research priorities at the National Center for Complementary and Alternative Medicine FASEB J, August 1, 2005; 19(10): 1209 - 1215. [Abstract] [Full Text] [PDF]

				T. Bodenheimer and A. Fernandez High and Rising Health Care Costs. Part 4: Can Costs Be Controlled While Preserving Quality? Ann Intern Med, July 5, 2005; 143(1): 26 - 31. [Abstract] [Full Text] [PDF]

				P. J. Neumann, A. B. Rosen, D. Greenberg, N. V. Olchanski, R. Pande, R. H. Chapman, P. W. Stone, S. Ondategui-Parra, J. Nadai, J. E. Siegel, et al. Can We Better Prioritize Resources for Cost-Utility Research? Med Decis Making, July 1, 2005; 25(4): 429 - 436. [Abstract] [PDF]