This Article Abstract Submit a response to this article Services E-mail this article to a friend Alert me to new issues of the journal

H E A L T H  T R A C K I N G
T R E N D S
W E B  E X C L U S I V E   

9 June 2004  
  

Tracking Health Care Costs: 
Trends Turn Downward In 2003

Although health care costs grew more slowly than in 2002,
their growth still outpaced growth in the overall U.S. economy in 2003.

By Bradley C. Strunk and Paul B. Ginsburg

ABSTRACT:

Health care spending per privately insured person increased 7.4 percent in 2003. While lower than the 2002 increase, it still outpaced growth in the overall economy by a margin that exceeds the historical average. The trend for drug spending decelerated the most. Meanwhile, hospital spending grew 9 percent in 2003—1.8 percentage points less than the 2002 increase. This reflected a sharp deceleration in growth of hospital use, while growth in hospital prices accelerated for the sixth year in a row. The trend for health insurance premiums fell in 2004. Employers raised patient cost sharing for the third year in a row.



As premiums for private health insurance continue to increase at a double-digit rate, the issue of health care costs has come to the forefront of current health policy debates.¹ The term costs is used in many different contexts; in this paper we focus on a specific subset of health care costs: the cost of services covered by private insurance.² In other words, we examine trends in the health care spending that underlies private health insurance premiums. Underlying cost trends are the critical long-term determinant of premium trends.

One year ago we reported that trends in costs that underlie private insurance leveled out after more than a half-decade of steady acceleration.³ This paper updates that analysis and reports that the turnaround that began a year ago has continued. Growth in health care spending decelerated in 2003 as a result of slowing trends for all of the major categories of services that are typically covered by private insurance, particularly for prescription drugs. Moreover, this decline in the spending trend appears to be slowing growth in health insurance premiums, which is welcome news for those who pay for coverage. The burden of rising costs continued to shift to those who use services, as patient cost sharing increased for the third year in a row.

Data Sources And Methods

Cost trend data. We used the Milliman USA Health Cost Index (HCI) to examine recent trends in health care spending underlying private health insurance premiums. Milliman USA constructs this index from both publicly available and proprietary data on provider revenues (a proxy for spending on services) gathered through various provider surveys. The index is designed to reflect the claims expenses experienced by private insurers for a typical policy.⁴ As such, it measures spending on only health services that tend to be insured: inpatient and outpatient hospital services, physician services, and drugs. Although Milliman USA removes spending by Medicare from these series, its inability to remove spending by Medicaid and uninsured patients is an important limitation of the HCI’s ability to track spending trends underlying private insurance. Nevertheless, a past comparison of the HCI and the National Health Accounts (NHA), compiled by the Centers for Medicare and Medicaid Services (CMS), indicated that the HCI is a good measure of trends in private health care costs.⁵ We use the HCI instead of the NHA because the HCI is available with a shorter time lag.

To gain insight into the factors driving growth in spending on hospital care, we broke down the spending trend into its price and quantity components. We used the “all other payers” series of the Bureau of Labor Statistics (BLS) Producer Price Index (PPI) for general medical and surgical hospitals to measure changes in hospital prices for privately insured patients. This series (hereafter referred to as the “hospital PPI”) reflects negotiated payments rather than billed charges.⁶

Changes in hospital quantity (that is, utilization, length-of-stay, and resource intensity per case) were calculated indirectly as the residual of the HCI for hospital services (inpatient and outpatient combined) and the hospital PPI. This quantity index is subject to the same limitation as the HCI: that is, it includes changes in prices and use for Medicaid and uninsured patients. Data for one of these factors—hospital prices for Medicaid patients— are available, and we have estimated that this factor has exerted a small downward bias (about one percentage point or less in any given year) on hospital spending and use trends in recent years. No data are available to measure, and therefore remove, different trends in use and length-of-stay for Medicaid patients, but we expect any differences to have only small effects on the measure.

We decomposed the trend in spending on prescription drugs in similar fashion. We used the Consumer Price Index (CPI) for prescription drugs and medical supplies to measure changes in drug prices. Like the hospital PPI, the prescription drug CPI reflects actual transaction prices. It accounts for price changes for the existing roster of prescription drugs, as well as changes in prices paid by consumers from the increased use of generic substitutes.⁷ We also calculated the residual of the HCI and CPI for prescription drugs. This residual captures the change in the total number of prescriptions written per person (that is, drug usage), which itself reflects changes in the proportion of the population that receives at least one prescription during the year and the average number of prescriptions per person among all people that receive at least one (that is, intensity). It also reflects changes in the therapeutic mix of drugs prescribed and the introduction of new brand-name drugs. When drugs move to over-the-counter status, the HCI, and hence our residual, reflects the reduced prescription drug spending but does not have an offset for what is spent over the counter.

Finally, we used data on payroll costs for hospitals to understand changes in hospitals’ largest operating cost factor. These data, compiled monthly by the BLS through its Current Employment Statistics survey, are useful for their reliability and very short time lag. We report BLS payroll data per capita because this makes them directly comparable to the HCI and to data on premiums (what is charged to cover an individual or family) and therefore the most relevant measure for policymakers.

Premiums and cost sharing. We examined findings from two employer surveys, the Towers Perrin 2004 Health Care Cost Survey and the Ninth Annual National Business Group on Health (NBGH)/Watson Wyatt Survey 2004, to glean insights into health insurance premium trends for 2004.⁸ Both of these surveys rely on convenience samples. We regard a convenience sample as less problematic for estimating the trend in premiums than for estimating the level of premiums. The Towers Perrin survey included responses from 311 employers representing approximately 4.1 million employees and retirees. Participating employers were asked to report their 2004 per capita premium costs for insured health plans and premium equivalents for self-insured plans. The NBGH/Watson Wyatt survey included responses from nearly 450 employers with 1,000 or more full-time employees, representing more than eight million employees.

The principal limitation of these surveys for our purposes is that the samples do not include small employers. For this reason, we focus the most attention on how the 2004 increase compared with the 2003 increase. Since premium increases for small employers have been larger than those for large employers in recent years, the estimates cited here probably underestimate the increase across employers of all sizes.⁹ Also, neither survey provided an estimate of the role of increased cost sharing and reduced benefits (“benefit buy-down”) in the premium trends. To fill this gap, we obtained information from Smith Barney Equity Research on actual buy-down percentages reported by the major publicly traded insurers.¹⁰

Health Care Spending Trends

Total health care spending per privately insured person rose 7.4 percent in 2003 (Exhibit 1). This increase was 2.1 percentage points lower than the 2002 increase of 9.5 percent, which was itself slightly lower than the 2001 increase of 10 percent. Therefore, it is now clear that we have entered a period of decelerating cost trends following a steep acceleration during 1996–2001. Nevertheless, the cost trend remained high by historical standards and continued to outpace U.S. economic growth by a sizable margin. With per capita gross domestic product (GDP) increasing 3.8 percent in 2003, the gap between it and the trend in health care spending remained larger than the average 2.5-percentage-point gap over the past thirty years.

The deceleration of the overall spending trend was the result of decelerating trends in all four spending categories. We examine each category in turn.

Hospital spending. Spending on hospital inpatient services per privately insured person rose 6.5 percent in 2003. This increase was 1.9 percentage points lower than the 2002 increase of 8.4 percent, signaling a turnaround from a remarkable period of rapidly accelerating growth in inpatient spending that began in 1998 following a 5.3 percent decline in 1997. Spending on hospital outpatient care per privately insured person rose 11 percent and was the fastest-growing spending category. This increase was 1.9 percentage points lower than the 2002 increase of 12.9 percent and 3.6 percentage points lower than the increase in 2001, when the trend for outpatient spending peaked at 14.6 percent. Taken together, growth in spending on inpatient and outpatient hospital care accounted for 53 percent of the total increase in health care spending.

The decline in the hospital spending trend in 2003 reflects a continuing and substantial deceleration in the hospital utilization trend (Exhibit 2). Hospital utilization (as measured by our residual hospital quantity index) increased 0.9 percent in 2003, representing a sharp decline from the 5.3 percent increase in 2002. We believe that the completion of the adjustment to more loosely managed care played an important role in this decline. The sharp utilization increase in 2001 was likely the result of the managed care industry’s efforts to respond to the consumer backlash by reducing its use of tools to tightly manage enrollees’ care.¹¹ With this transition apparently complete, we would expect the utilization trend to return to its long-term rate of increase. Patient cost sharing also increased, but we would not expect a major impact on the utilization trend, since the magnitude of the change in patient responsibility was similar to that in 2002.

In contrast to the decelerating utilization trend, hospital price increases continued to accelerate for the sixth year in a row. Hospital prices (for both inpatient and outpatient care) increased 8.0 percent in 2003, much larger than the 2002 increase of 5.2 percent and the largest one-year increase since the BLS began tracking negotiated prices for hospital care through the PPI program in 1993.¹² This trend is consistent with qualitative research, which has showed that many hospitals solidified their negotiating leverage over plans during 2002 and 2003 and continued to use their formidable power to demand large payment rate increases.¹³ Hospital price increases accounted for most of the overall increase in hospital spending in 2003.

One of the important underlying drivers of accelerating hospital price trends in recent years has been a shortage of hospital workers, particularly nurses. This shortage has driven up wage rates for hospital workers, and hospitals have sought to pass these increases on to payers through higher prices. However, this pressure on hospitals is waning (Exhibit 3). Although it continued to outpace wage rate growth in the overall economy, the increase in wage rates for hospital workers decelerated for the second year in a row, growing 4.2 percent in 2003 compared to 5.2 percent in 2002 and 6.1 percent in 2001. If this trend is sustained, hospitals will likely temper their demands for large rate increases, and the hospital price trend will begin to slow in the near future. Despite the deceleration in wage rate growth, the 2003 increase in hospital payroll costs (6.5 percent) changed little because of a small increase in the trend for total hours worked.

Another factor that could influence hospital prices is cost shifting between public and private payers. Some have argued that hospitals raise prices to private payers to offset payment rate reductions by Medicare and Medicaid.¹⁴ However, the most recent Medicare payment rate changes have been two rounds of “givebacks” following sizable Medicare payment rate cuts made by the Balanced Budget Act (BBA) of 1997, so cost shifting could be restraining hospital price trends at this point in time. The recently mandated higher rate increases for hospitals could exert additional downward pressure on hospital prices in the near future. The widely noted movement toward tiered provider networks has probably had little aggregate effect on these trends, as only about 6 percent of preferred provider organization (PPO), health maintenance organization (HMO), and point-of-service (POS) enrollees are subject to such designs.¹⁵

Prescription drugs. Spending on prescription drugs per privately insured person grew by 9.1 percent in 2003.¹⁶ This increase was much lower than the 2002 increase of 13.2 percent and less than half the 1999 increase, when the drug spending trend peaked at 18.4 percent. The trend has now decelerated for four straight years after becoming a major source of cost growth during the second half of the 1990s. The share of total spending growth attributable to increased drug spending has now declined to 20 percent.

This remarkable turnaround reflects both price and utilization components (Exhibit 4). After holding steady at a little over 5 percent, drug price inflation declined to 3.1 percent in 2003. This likely reflects, in part, a shift to generic drugs. A number of major brand-name drugs have recently lost patent protection and now face competition from cheaper generic versions. Also, three-tier drug copayments have become more prevalent in recent years, and these designs create incentives for patients to favor generic over brand-name drugs. By encouraging use of generic and preferred brand-name drugs, three-tier designs may also be creating pressure on drug companies to hold down price increases for brand-name drugs.

Meanwhile, the trend for our residual index for prescription drugs slowed as well, growing 5.8 percent in 2003 compared with 7.6 percent in 2002 and a high of 12 percent in 1999. In addition to its effect on the mix of generic versus brand-name drugs that patients use and on drug prices, greater patient cost sharing has likely slowed growth in prescription drug use. Other factors include the recent slowdown in the number of new drugs, particularly so-called blockbuster drugs, that have been approved for sale and the recent reclassification of some major prescription drugs to over-the-counter status (such as the nonsedating antihistamine Claritin).¹⁷

Physician care. Spending on physician care per privately insured person increased 5.1 percent in 2003, compared with 6.5 percent in 2002, and again was the slowest-growing category of health spending. This slowdown is likely another reflection of the completion of the transition to looser forms of managed care. For example, if fewer authorization requirements to get access to specialists led to increased use of specialty services in 2001 and 2002, the stability of this model would lead us to expect a smaller increase in usage. Indeed, a decomposition of physician spending comparable to that in Exhibit 2 for hospital spending shows a sharp decline in the rate of growth of physician use and a slight increase in growth of prices for physician services.

Health Insurance Premium Trends

Findings from the Towers Perrin and NBGH/Watson Wyatt employer surveys suggest that premiums for employment-based insurance continue to grow at a double-digit rate but that the trend turned downward in 2004. The Towers Perrin survey reports that premiums for active employees increased 12 percent on average this year, which was three percentage points lower than the 2003 increase of 15 percent. Similarly, the NBGH/ Watson Wyatt survey found that large employers expect a median premium increase of 12 percent this year, whereas last year’s respondents predicted that premiums would rise by a median rate of 15 percent in 2003.¹⁸ Since premium growth continues to outstrip growth in underlying costs, these trends also suggest that health insurers have yet to engage in the kind of vigorous price competition that would characterize a turn in the insurance underwriting cycle, the industry’s interdependent pattern of profitability and pricing.

Implications For Consumers

According to insurers’ reports to Wall Street, consumers appear to be facing a third round of sizable cost-sharing increases in 2004. Most publicly traded health insurers that provided such information reported buy- downs in the range of 2.5–3.5 percent.¹⁹ This increase in cost sharing comes on the heels of buy-downs of similar magnitude in both 2002 and 2003.²⁰

While employers increased cost sharing to control rising premiums, they made little change to the proportion of the total premium that employees pay. According to the Towers Perrin employer survey, active employees are paying 19 percent of the cost of single coverage in 2004, compared with 18 percent in 2003 and 19 percent in 2002. In addition, active employees are paying 22 percent of family coverage in 2004, compared with 21 percent in 2003 and 22 percent in 2002.²¹,Employers may be opting for increased patient cost sharing rather than increasing employees’ premium contribution percentages to encourage more judicious use of health services while maintaining employees’ takeup of health insurance.

Discussion And Outlook

After a long period of steady acceleration, the cost trend finally turned downward in 2003. Underlying costs are a critical long-term determinant of trends in health insurance premiums, and this slowdown has already initiated a decline in the premium trend. Moreover, if financially healthy managed care companies begin to engage in fierce price competition to build market share, the underwriting cycle could turn and drive the premium trend below the trend for underlying costs. A turn in the underwriting cycle alone would not, however, slow the premium trend to the very low levels we experienced in the mid-1990s; only a sustained and substantial decline in the cost trend would do that.

It is clear that the main tool health plans and employers are using to control costs is increased patient cost sharing.²² However, we are skeptical that increased cost sharing can have a major impact on the cost trend going forward. Standard actuarial assumptions suggest that health care demand is fairly inelastic—that is, a 1 percent increase in cost sharing decreases use of hospital and physician services by about 0.05 percent and 0.15 percent, respectively, and use of prescription drugs by about 0.3 percent.²³ Even though the 2004 increase in cost sharing is large by historical standards, it is similar to last year’s increase, which suggests that it would not necessarily lower the cost trend any further in 2004. With demand for health care relatively inelastic, those who pay premiums will realize most of their savings by shifting costs to those who use services rather than by inducing less usage.

Despite these considerations, we raise the possibility that an environment in which many patients have extensive cost sharing and the information it stimulates could lead to larger effects on usage than the actuarial assumptions outlined above predict and that these effects could grow over time. With greater cost-sharing responsibilities borne by many patients, patients are likely to be offered more information from insurers, their doctors, and the media and become more skilled at economizing on service use. Benefit structures might become more sophisticated, emphasizing incentives to use efficient providers or more effective therapeutic alternatives. Indeed, this could even affect development of new medical technologies over time if patient cost sharing causes the health services market to become less accepting of new technologies with low or uncertain medical value.

Nevertheless, many factors make it less likely that increased patient cost sharing alone will have a large impact on cost trends over time. For one thing, because a large portion of medical spending is for patients who use large amounts of services—and would meet even large deductibles—large portions of medical care might be immune to traditional cost-sharing incentives. Second, research is clear that the key long-term driver of cost trends is technological advancement in medical care. Much new technology will have compelling reasons for widespread adoption, and Americans continue to demand the latest technology. A financing system that facilitates the rapid diffusion of costly new technologies by paying most of their cost—even in the absence of careful consideration of their clinical effectiveness—has probably been a factor behind the aggregate cost impact of new technologies. Fundamental change in this dynamic would require support for improved and more frequent evaluation of new technologies prior to coverage decisions, as well as carefully differentiated incentives built into the financing system that encourage both providers and patients to evaluate the clinical effectiveness of a given course of treatment against its cost.

Costs—and therefore premiums—will likely continue to outstrip growth in the overall U.S. economy by a sizable margin for the foreseeable future. Although some argue that more spending for medical care is a good thing, it clearly will stress the mechanisms that the United States uses to finance care. It makes insurance unaffordable to more employers and consumers. Unless policymakers become more willing to increase public revenues more rapidly than incomes, increased spending for medical care will crowd out other public spending priorities. These financing pressures are behind the recent urgency about how to contain costs and ultimately will lead to a willingness to discuss taboo subjects, such as rationing health care.

The authors are grateful to John Cookson of Milliman USA for permission to use the Health Cost Index. Comments from anonymous reviewers contributed important insights. The authors gratefully acknowledge the Robert Wood Johnson Foundation for its financial support.

NOTES

1. J. Gabel et al., “Health Benefits in 2003: Premiums Reach Thirteen-Year High as Employers Adopt New Forms of Cost Sharing,” Health Affairs 22, no. 5 (2003): 117–126.
2. Another term that we and others use interchangeably with “cost” is “spending.” Conceptually, the primary interest is in costs, which reflect the resources devoted to health care that are not available to produce other goods and services. Practically, most available data, including these used here, reflect spending, or what is paid for health services by those who purchase them (or received by those who provide them).
3. B.C. Strunk and P.B. Ginsburg, “Tracking Health Care Costs: Trends Stabilize but Remain High in 2002,” Health Affairs, 11 June 2003, content.healthaffairs.org/cgi/content/abstract/hlthaff.w3.266 (17 May 2004).
4. The index that Milliman USA provides to its clients is intended to assist insurers in forecasting their claims payments and comparing them with those of others. It simulates trends in claims for a “standard” private health insurance policy with a $250 deductible. The trend in such an index would slightly overstate the actual trend in spending because the standard policy would pay for a slightly higher proportion of expenditures each year. To avoid this problem, Milliman USA has provided us with a version of the index that reflects a hypothetical policy with no deductible.
5. B.C. Strunk, P.B. Ginsburg, and J.R. Gabel, “Tracking Health Care Costs: Growth Accelerates Again in 2001,” Health Affairs, 25 September 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.299 (12 May 2004). One important difference between the HCI and the NHA is the classification of ambulatory facilities that are not owned by hospitals. The HCI includes such facilities in its hospital outpatient category, while the NHA includes them with physicians’ services. Since spending in ambulatory facilities is growing rapidly, we would expect the HCI to have a higher rate of increase for outpatient hospital services and a lower rate of increase for physicians’ services than the NHA does.
6. For more information about the hospital PPI’s methodology, see B. Catron and B. Murphy, “Hospital Price Inflation: What Does the New PPI Tell Us?” Monthly Labor Review 119, no. 7 (1996): 24–31. We considered the hospital portion of the Consumer Price Index (CPI) as an alternative to the hospital PPI, but the hospital CPI is less useful for our purposes because it uses billed charges for hospitals not responding to its survey. The PPI, in contrast, drops such non-responders.
7. For more information, see U.S. Department of Labor, Bureau of Labor Statistics, “Measuring Price Change for Medical Care in the CPI,” stats.bls.gov/cpi/cpifact4.htm (30 March 2004).
8. Towers Perrin, 2004 Health Care Cost Survey, 2004, www.towersperrin.com/hrservices/us/default_us.htm (31 March 2004); and National Business Group on Health and Watson Wyatt Worldwide, New Reality. New Choices. Ninth Annual National Business Group on Health/Watson Wyatt Survey 2004, 2004, http://www.watsonwyatt.com/research/resrender.asp?id=w-736&page=1 (7 June 2004).
9. According to the Henry J. Kaiser Family Foundation/Health Research and Educational Trust Survey of Employer Health Benefits, premium increases among small firms (3–199 workers) averaged 15.5 percent in 2003, compared with 13.2 percent among large firms; moreover, premium increases among small firms have consistently outpaced those among larger firm for at least the past decade. See Kaiser/HRET, Employer Health Benefits: 2003 Annual Survey (Menlo Park, Calif.: Henry J. Kaiser Family Foundation, September 2003), and previous versions of this report.
10. Charles Boorady, Smith Barney Equity Research, personal communication, 7 May 2004.
11. D. Draper et al., “The Changing Face of Managed Care,” Health Affairs 21, no. 1 (2002): 11–23.
12. In contrast to trends for the hospital PPI, the hospital portion of the Consumer Price Index decelerated slightly in 2003 compared with 2002, probably because of the different treatment of nonresponders (see Note 6). Hospitals have recently come under fire for billed charges’ being much higher than costs; the deceleration in the hospital CPI in 2003 may reflect hospitals’ response to such criticism.
13. J. White, R.E. Hurley, and B.C. Strunk, Getting Along or Going Along? Health Plan–Provider Contract Showdowns Subside, Issue Brief no. 74 (Washington: Center for Studying Health System Change, January 2004).
14. P.B. Ginsburg, “Can Hospitals and Physicians Shift the Effects of Cuts in Medicare Reimbursement to Private Payers?” Health Affairs, 8 October 2003, content.healthaffairs.org/cgi/content/abstract/hlthaff.w3.472 (4 May 2004).
15. Kaiser/HRET, Employer Health Benefits: 2003 Annual Survey.
16. The prescription drug portion of the HCI does not include spending on specialty pharmaceutical products administered by physicians, such as injectibles. As a result, the HCI may slightly understate the actual increases in prescription drug spending in recent years.
17. See U.S. Food and Drug Administration, Center for Drug Evaluation and Research, “NDAs Approved in Calendar Years 1990–2003 by Therapeutic Potentials and Chemical Types,”
www.fda.gov/cder/rdmt/pstable.htm (3 May 2004).
18. The 2004 NBGH/Watson Wyatt survey reported that the actual median premium increase in 2003 was 13 percent. However, since the 2004 estimate from this survey represents a projection and projections can be biased, we felt a comparison of projections in both years provided better insights into trends.
19. A buy-down of, for example, 2 percent means that insurance premiums would have increased 2 percent more than they did if employers had made no changes to their benefit structures.
20. Strunk and Ginsburg, “Tracking Health Care Costs: Trends Stabilize but Remain High in 2002.”
21. The Towers Perrin survey includes responses from mostly large employers, and large employers usually pay a larger portion of the cost of family coverage than small employers pay. This explains why the Towers Perrin estimate of the employee share of family coverage is lower than other employer surveys that include all employers.
22. L. Regopoulos and S. Trude, Employers Shift Rising Health Care Costs to Workers: No Long-Term Solution in Sight, Issue Brief no. 83 (Washington: HSC, May 2004).
23. James Baumgardner, Congressional Budget Office, personal communication, 6 May 2004.

Bradley Strunk is a health research analyst at the Center for Studying Health System Change in Washington, D.C. Paul Ginsburg (PGinsburg{at}hschange.org) is the center's president.

DOI: 10.1377/hlthaff.W4.354
©2004 Project HOPE–The People-to-People Health Foundation, Inc.