Healthcare cost differences in the 1990s: the influence of metropolitan area marketplace dynamics.
The study begins with an exploration of several indicators of the underlying cost trends in general and commercial payment patterns in particular. Since no aggregate, comprehensive non-Medicare or Medicaid payment information exists at the MSA level, the study focuses on three specific representations: the commercial health maintenance organization (HMO) premium per member per month (PMPM), non-governmental (i.e., non-Medicare or Medicaid) payments to hospitals per non-elder, and payments made by the Federal Employee Health Benefits Plan per enrollee, for a common benefit plan, in each MSA.
To explore the costliness of healthcare for each of 20 MSAs in the 1990s--as well as whether these cost differences can be attributed systematically to shifting the shortfall in payments from Medicare onto commercial payers--the study uses regression analysis. As pointed out by Morrisey (1994) and Cutler (1998), among others, such cost shifting can only take place if either relative bargaining power shifts towards providers (and away from purchasers) or if healthcare providers who previously were not pricing at profit-maximizing levels choose to move in that direction.
Based on the analysis of 20 cities in the central portion of the US and data for 1990, 1995 and 2000, the study concludes that the relative bargaining power of providers in contrast with that of purchasers, especially HMOs, plays a central role in determining healthcare cost patterns in the 1990s. In contrast, cost shifting from Medicare onto private payers does not differentiate these MSAs from each other; however, Center for Medicare and Medicaid Services (CMS) payment information indicates that payments to MSA providers in the central portion of the country grew less rapidly in the 1990s and were lower in 2000 than the average payment rate per beneficiary in the US.
The rest of the paper is organized as follows: The next section provides a brief literature review as well as motivation for the cost drivers studied. The third section describes the methodology employed, as well as the data selected, to address empirically the role of specific cost drivers. The fourth section presents the results and posits several explanations for the wide variation in healthcare cost levels and trends across the 20 MSAs. A final section discusses these results and suggests how future research might better illuminate the lines of causation investigated in this study. A brief conclusion completes the paper.
In a recent Health Affairs web-exclusive article, Altman and Levitt (2002) portray real private health spending growth per capita in the US from 1961 to 2000. Their chart features a series of peaks and troughs with growth at the national level (in real per capita terms) in excess of eight percent in the mid-1960s, the late 1970s, the late 1980s and in recent years; these results are balanced with other years--such as the mid 1990s--when real per capita growth was negligible or negative. A study such as that by Strunk, Ginsburg, and Gabel (2002) confirms recent experience with double-digit increases in healthcare insurance premiums, much of which arise from increasing payments to the healthcare sector.
Martin et al. (2002) at the National Health Statistics Group within the Federal CMS recently analyzed state healthcare spending patterns for 1991-1998. They found that average healthcare spending in the US (not corrected for either general inflation or population growth) grew at an average annual rate of 7.0 percent with a range from 4.4 percent in Arizona to 9.4 percent in Texas. They attribute such variation to patient demographics, socioeconomic characteristics, concentration of healthcare resources, state and Federal spending policies, and other marketplace factors.
The issue of cost shifting generates substantial debate among policy analysts, especially evident at a recent Changes in Health Care Financing and Organization (HCFO) conference on the topic, sponsored by the Robert Wood Johnson Foundation, and also published as web-exclusive articles in Health Affairs (Morrisey, 2003). Michael Morrisey's (2003) presentation, consistent with the argument presented in his book-Cost Shifting in Health Care (1994), argues that for cost shifting to take place, either relative bargaining power must change (in the direction of providers and away from purchasers) or providers who previously were not pricing their services at what the market would bear, began to move in that direction. Others such as Paul Ginsburg (2002) note that profit-maximizing explanations may not apply to many of the participants in the hospital market; thus, he identifies additional circumstances under which cost-shifting exists. In a 1998 paper, Cutler determined that during the 1980s, hospitals shifted the cost burden of services from governmental payers onto commercial payers; however, in the 1990s, in response to reduced growth in governmental payments, hospitals tended to reduce resource use and the acquisition of new technology.
The incidence of healthcare costs at the metropolitan level differs from that experienced at the national level; thus, the cost of labor will also differ. The analysis parallels that for the property tax that features both national and local incidence. At the national level, most economists argue (see Pauly  for a detailed discussion) that healthcare costs are just one component of labor compensation. If workers are not likely to leave the national market or country to receive improved labor compensation, then they will be less responsive to changes in compensation than will their employer counterparts, and, thus, the cost of healthcare will be borne by laborers, even if they do not officially write the check (similar to property tax incidence--assuming that property does not leave the country). Furthermore, national evidence on inflation-adjusted benefit compensation growth from 1982-2003 (0.9% annual growth) in excess of wage and salary compensation growth (0.6% annual growth) is consistent with this claim. (See the Economic Report of the President 2004, Tables B-48 and B63).
At the local level, however, the compensation elasticity for laborers is surely much higher; that is, if employers (or healthcare providers) shift payment onto labor, more laborers will leave or be less inclined to move to a particular geographic area. Differences in healthcare costs across metropolitan areas, for similar quality healthcare, can be treated as excise taxes related to living in particular areas. Over time, workers--especially those with significant human capital and mobility--will leave high healthcare cost areas in search of areas with lower costs, all else equal (similar to responses to a local property tax rate higher than the national average). Of course, these cost differences must be of sufficient magnitude to make such movement worthwhile. This study does not attempt to evaluate this argument but simply uses the argument as motivation to study the differences in healthcare costs across metropolitan areas.
In the April 2001 Part II issue of Health Services Research devoted to assessing the data needs for studies of healthcare competition, Baker (2001) identifies five areas for the measurement of competition that deserve analytical scrutiny: product and competitors of interest, geographic market areas, measures of competition, changes in competitive dynamics, and the role of managed care. This study contributes to the first category by defining a measure of hospital costliness for commercial payers. It uses metropolitan statistical area definitions, consistent with much existent literature, but acknowledges that service area measures such as defined by Makuc et al. (1991) or hospital referral regions as defined in the Dartmouth Atlas (2004) project would be an improvement. Lack of data availability, however, limits implementation of either definition. To measure competition, the study uses both Baker's recommended Herfindahl Hirschman indicator (HHI) of competition and the Wholey et al. (1995) measure for HMOs. The regression results reported below focus on the levels of various indicators for three years--1990, 1995 and 2000; thus, competitive dynamics are reflected by differences in the demographic, hospital sector and HMO sector across the 20 MSAs. Various indicators of the market role played by HMOs represent the role of managed care. In short, this study has attempted to make both empirical and conceptual progress on all of the areas identified by Baker.
SPECIFICATION AND DATA SOURCES
This study explores several cost indicators and regresses each on a hypothesized set of cost drivers for a panel of data from three years: 1990, 1995 and 2000. Since no comprehensive representation of healthcare costs exists at the metropolitan level, the study examines several specific segments of localized health markets. In contrast, the academic literature focuses either on the US or on state level analysis, while proprietary studies by large consulting firms, such as Mercer Human Resource Consulting, Hewitt Associates, or Milliman USA, tend to examine a segment of the employer base in terms of either available claims data or responses to surveys.
Since funding for this study came from Milwaukee-based Cobalt Corporation, a BlueCross and BlueShield (BCBS) Association member, the MSAs were selected from the same part of the country as Milwaukee. The 20 selected MSAs feature at least 400,000 residents in 2000 and are within 625 miles of Milwaukee. They stretch as far east as Pittsburgh, PA, as far south as Memphis, TN, as far west as Omaha, NE, and as far north as Minneapolis-St. Paul, MN.
One comprehensive representation of healthcare cost for those enrolled in HMOs is the PMPM premium. Of course, HMO plans differ within and across markets in their benefit structures and network breadth. Furthermore, the HMO penetration rate varies widely across markets (in 2000, penetration ranged from 11 percent in Memphis, TN to 61 percent in Madison, WI); so, HMO premiums provide only a partial portrait of healthcare cost differences across MSAs. These data are available from InterStudy for each year studied, and Wholey et al. (1995) has prorated HMO enrollment by county to generate MSA specific indicators.
As a second cost indicator, the study examines the payments received by hospitals from sources other than Medicare or Medicaid per non-elder, and refers to them as commercial payments to hospitals per non-elder. For the most part, the numerator represents payments received from commercial payers and self-insured employers for inpatient and outpatient services. These data were provided by the American Hospital Association (AHA) at the metropolitan level (for all but three data points among the 60 observations) and are derived from subtracting allowances, discounts, and the aforementioned governmental payments from billed charges. The total population aged less than 65 (the non-elderly) serves as the denominator. Although many Medicaid recipients, and even a few Medicare beneficiaries, would be counted among those less than 65 years of age, the non-elderly as a group are predominantly served by commercial insurance.
Finally, payments for healthcare services PMPM by the Federal Employee Health Benefits Plan (FEHBP) constitute a third cost indicator. A Preferred Provider Organization (PPO) plan with similar benefits across markets is considered, though network size and membership levels differ markedly across the 20 MSAs. These payments are corrected for differences in case-mix (by diagnostic cost group) but are only available for 2000-2002, as provided by the BCBS Association of America. Since these data are not available for 1990 or 1995, insufficient observations exist on which one can perform regression analysis; however, the distribution has been included with other descriptive statistics in Exhibit 1.
To assess whether Medicare cost-shifting exists, two regression equations are specified. To evaluate the results for hospitals, the study uses the aforementioned commercial payments to hospitals per non-elder as the dependent variable and Medicare Part A fee-for-service payment per beneficiary as the cost-shift factor. Controls for work-force age distribution, HMO competition, commercial admission share, hospital admissions per capita, hospital market power, and physician presence are included. For HMOs, the PMPM premium is regressed against total Medicare payment (Part A plus Part B), with similar controls included.
Areas with different demographic characteristics should experience different healthcare cost burdens. To determine whether demographic characteristics matter, two indicators are studied. First, the percent of the population age 65 or older represents Medicare beneficiaries. Second, two non-elderly working age groups were considered: those between age 20-34 and those between age 45-64. These groupings match those reported by the Centers on Disease Control. Since those in the latter category (of non-elderly) incur expenses two to three times those in the former group, as the ratio of the older group to the younger group rises as a percentage of the population, so should healthcare expenditures. This indicator varies markedly across the metropolitan areas, and is incorporated as an explanatory variable in the hospital cost and HMO regression equations. Some specifications also included the percentage of population age 65 and older; insignificant results, however, warranted exclusion from the preferred specification.
Finally, the study attempts to understand the role of relative bargaining power of providers and purchasers. For hospitals, a Herfindahl index based on the commercial admission share for each hospital system constitutes the primary indicator. Individual hospital data on commercial admissions have been aggregated to the system level, prior to calculation of the Herfindahl index. Substantial variation in the Herfindahl index exists both across the 20 MSAs and across time, as mergers and dissolutions took place.
To understand the competitive effects of HMOs, the study adopts the approach used by Wholey et al. (1995). They argue that the Herfindahl index is not an appropriate indicator of competition since HMOs produce differentiated products and may incorporate cooperative behavior among contracted healthcare providers. The number of HMOs provides some measure of competition but says little about the extent of the commercial market covered; thus, this paper accepts Wholey et al.'s argument that the interaction of the number of HMOs and the market penetration rate best represents the competitive effect of HMOs on market prices.
Supplier-induced demand theory suggests that the cost of healthcare rises directly with the number of suppliers. To address these effects, either the total number of physicians actively practicing medicine per thousand residents or the number of non-primary care specialists per thousand residents has been included in each regression specification. Typically, the latter group (specialists) order and often perform the most intensive, and thus the most expensive, medical services. Unfortunately, the Area Resource File does not contain specialty breakdowns for 1990; thus, the specification only includes the total number of practicing physicians per 1,000 residents.
Exhibit 1 contains the descriptive statistics for both the dependent and independent variables used in the study (Panel A) and their growth rates (Panel B). For the year 2000, the dependent variables feature a maximum value roughly 50 percent above the minimum value for the first two indicators (HMO PMPM premium and non-governmental payments to hospitals per non-elder), and a two-fold difference for the Federal Employees Health Benefits Plan (FEHBP) enrollees. For the decade from 1990 to 2000, the growth rates for the first two dependent variables feature much more variation than characterized by the observations for the year 2000; some regression to the mean seems to exist.
The second segment of Exhibit 1 provides descriptive statistics for Medicare Part A, Part B, and total payments to the medical community by metropolitan area. A (supplemental) regional analysis of the data from 2000 indicated Total Medicare payments ranging from a minimum of $346.63 to a maximum of $555.65 per beneficiary. On average for ali Medicare beneficiaries across the US, total payment grew from $274 per enrollee (compared to a 1990 regional mean of $302) to $464 per enrollee (compared to a 2000 regional mean of $434). National Medicare payments per beneficiary grew by 69 percent, whereas mean payments to residents in the areas studied grew by only 45 percent. The differences are particularly pronounced for Part A for which national payments grew at 66 percent while the region studied grew at only 37 percent. A smaller growth rate difference existed for Part B Medicare payments: national growth at 74 percent and regional growth at 59 percent. These patterns are consistent with concerns raised by politicians within the region (and even have included law suits) that horizontal inequities in Medicare payments across different regions of the country should be addressed.
Income and demographic characteristics occupy the third segment in the Exhibit. Income levels differed modestly across the 20 MSAs, and, with the exception of St. Louis (30%), the MSAs experienced quite similar income per capita growth (between 45 and 56%) in the 1990s. The urban areas studied differed markedly in their demographic characteristics and also aged at quite different rates across the decade. In 2000, the area with the highest ratio of older to younger work eligible population (Pittsburgh) had a 61 percent higher rate than that with the lowest ratio (Memphis). Growth rates also differed markedly (23% to 81%) across the decade. Almost 18 percent of Pittsburgh population were 65 and older in 2000 while, at the other extreme, fewer than 10 percent of Minneapolis-St. Paul residents were senior citizens. For the decade, the share of population 65 and older varied modestly around an average of non-change.
The bottom half of Exhibit 1 portrays the healthcare sector. On average the number of hospitals fell, but some areas with few hospitals in 1990 grew rapidly. On average, largely due to mergers and consolidation, the Hospital Herfindahl index (HHI) grew markedly over the decade. In 2000, 12 MSAs featured HHI values in excess of 2,000, an FTC benchmark for potential market power. In 1990, only five MSAs featured such concentration. In 2000, HHI measures exhibited a 10-fold range with very competitive Chicago (416) on one end and Cincinnati, Fort Wayne, and Des Moines, near 4,000 at the other extreme.
Physician availability also differed widely across the 20 MSAs. Since the Madison, WI service area is about one-third wider than its MSA (based on Mekuc's  Health Service Area definition) and given the medical school at the University of Wisconsin, its abundance of physicians (3.9 per 1,000 population) is not surprising. Without Madison, the number of physicians per thousand ranged from 1.6 to 3.9 with modest growth over the decade.
Hospital admissions per 1,000 residents, one measure of practice style differences, indicate a wide range of practices across the 20 MSAs. In 2000, admissions per thousand residents averaged (median) 123.7 (124.9) with a standard deviation of 18.1. Admissions ranged from 86.3 per thousand Kansas City residents to 165.8 admissions per thousand for Pittsburgh residents. During the 1990s, hospital admissions per 1,000 fell on average by 5.0 percent with wide variation across the 20 MSAs studied.
The role of HMOs constitutes the final segment of metropolitan area health sector dynamics. Over the decade of the 1990s, the percentage of residents served by commercial HMOs (HMO Penetration Rate) rose by almost 200 percent on average. In 1990, the HMO penetration rate was below 10 percent for all but six of the 20 urban areas, and no area had more than a 20 percent share. By 2000, all of the areas featured double digit HMO commercial sector representation; five MSAs had penetration rates in excess of 30 percent; and only five MSAs had penetration rates below 20 percent. HMO competitiveness as represented by the product of the number of HMOs offering products in the market and the overall HMO penetration rate also grew rapidly during the 1990s with triple digit growth rates for all but two of the 20 MSAs (Chicago and Minneapolis-St. Paul). Wide variation in competitiveness existed in 2000 with the maximum value more than five times the minimum value.
Finally, Exhibit 1 displays the influence of HMO capitation payment policy for specialists. Use of this strategy varied widely across MSAs and across time in the 20 MSAs studied. Between 1995 and 2000, 11 areas featured marked increases in capitated payments to specialists, while 9 areas either significantly decreased capitation payment share or virtually abandoned the use of capitation. Since these data are not available for 1990, this variable has not been entered in the regression specifications described in the next section.
With the description of metropolitan healthcare dynamics in Exhibit 1 complete, we now turn to analysis of the combined influence of these forces on two cost indicators: commercial payments to hospitals per non-elder and HMO PMPM premiums.
Exhibit 2 presents the regression analysis of metropolitan area healthcare factors on commercial payments (i.e., non-Medicare and Medicaid) to hospitals per non-elder. This regression includes 57 observations taken from 1990, 1995 and 2000, and accounts for over 78 percent of the variation in commercial payments, after adjusting for the available 47 degrees of freedom. A number of specifications were tried, but those variables identified as statistically significant in Exhibit 2 stand out.
* Costs increase with the passage of time (Year). For each year after 1990, commercial payment per non-elder increased by $51.48.
* Hospital admission practice also plays a very strong role in explaining differences in commercial payments. A one standard deviation (18.1) increase in the number of admits per thousand yields a $79 increase (11%) in the expected commercial payment per non-elder.
* Those areas that featured the strongest effect from the rise in HMO competition in the commercial insurance rate (HMO*Penetration), have lower than average payments to hospitals per non-elder. This result, however, is only significant at the 90 percent confidence level. An area with HMO competition one standard deviation (1.86) above the mean would, on average, experience a $45 (6%) lower payment per non-elder; competitive HMOs appear to have some bargaining influence on hospital payments.
* Hospital payments are also significantly related (above the 90% confidence level) to the number of physicians (MDs) per thousand residents. At one standard deviation (0.6) above the mean in 2000 (3 MDs per 1,000), hospital payments would be, on average, $36 (4.8%) above the mean. In a separate regression that included the number of non-primary care specialists per 1,000, a similar influence results. For the year 2000, at one standard deviation (0.4) above the mean (2 specialists MDs per 1,000), hospital payments per non-elder would be $78 (8.5%) above the mean.
* The coefficient on Medicare Part A (hospital) payments is negative (-0.117), as expected if cost shifting takes place, but the results are far from statistically significant.
* The coefficient on the Old/Young ratio has a sign different from that expected, and is significantly above the 90 percent Confidence level. The average aging of the population would be accounted for by the time trend term (Year), but differences across MSAs appear not to further contribute as expected to the explanation of the variation in the commercial payment per non-elder.
* The degree of competition among hospitals for commercial admission market share (Hospital Herfindahl) negatively but insignificantly affects commercial payment. The negative coefficient may be related to scale economies in bargaining with hospital systems, as indicated below.
* Inclusion of the percentage of the population 65 and older had no influence on the regression results.
The study of a parallel set of influences helps us to understand the variation in HMO PMPM premiums. The regression analysis portrayed in Exhibit 3, adjusted for degrees of freedom, accounts for 82.8 percent of the variation in HMO PMPM premiums across the 20 MSAs and the three time periods. In addition to the intercept, only three of the variables entered show statistical significance at the 90 percent level or higher.
* Not surprisingly, the time trend term (Year) enters strongly. For each year after 1990, HMO premiums (PMPM) rose on average by $6.13 (about 5% of the mean).
* The concentration of hospitals as reflected by a Herfindahl Index of commercial admissions market share entered significantly (at the 99% confidence level) but with a negative sign in the regression model. This confounds a simplistic interpretation that correlates more market power with higher prices (and, thus, insurance premiums). In metropolitan healthcare markets, however, the stow is more complex. If hospitals join together to become systems (and contracting entities) and if such systems provide wide geographic access within a metropolitan area, then HMOs might choose to contract with a selected subset of such systems and offer a patient base for those systems' hospitals in exchange for a volume discount. In short, contracting efficiencies may arise from hospital concentration. To fully test this claim, one would need to know to what degree HMOs selectively contract with particular hospital systems in each MSA as well as whether hospital systems cover an entire MSA or just divide the MSA into service regions with little competition among them. Such data are not readily available. The regression analysis (described below) of the Hospital Herfindahl index on the HMO penetration rate and on HMO competitiveness confirms the positive influence of HMOs on hospital market concentration.
* The regression results in Exhibit 3 suggest that HMO penetration positively influences premiums above the 95 percent confidence level. This result conflicts with the suggestion that an increased HMO share of the market should lead to increased buyer purchasing power and, thus, lower prices. One might argue that reverse causality applies here; namely, markets with relatively high premiums tend to draw increased HMO participation. Two interactive terms that contain the HMO penetration rate--HMO*Penetration (the number of HMOs times the penetration rate) and also Penetration*Cap (the HMO penetration rate times the percentage of specialist payments from HMOs received as capitation)--were entered in some of the model specifications, but neither demonstrated statistical significance.
The two commonly cited cost drivers--Medicare cost shifting and an aging commercial population--did not enter any of the posited regressions with a significant result. In Exhibit 3, AAPCC (average annual Medicare per beneficiary payment) enters with an unexpected positive sign, but is marginally significant at the 92 percent level. Additionally, Old/Young (the ratio of residents 45-64 to those 20-34) enters with a negative sign and has a rejection probability of 20 percent. Since one would expect cities with a large portion of relatively older residents to feature both higher costs and higher insurance premiums, the negative sign might result from an HMO tendency to enroll a greater number of younger workers as the relative proportion of older residents rises; that is, HMOs segment the health insurance market in search of relatively young enrollees. However, tests of this hypothesis would require information about the age distribution for each HMO's enrollees, and such data were not available.
The analysis of healthcare cost levels and growth rates by metropolitan area reveals substantial differences across MSAs in the 1990s. Regression analysis of non-Medicare and Medicaid payments to hospitals indicates that practice style (hospital admits per 1,000 residents) plays a strong role in accounting for differences in payment. No other factor beside a time trend is significant at the 95 percent level; however, the competitiveness of HMOs does have a negative influence identifiable with 90 percent confidence. A high relative share for the population aged 45 to 64 in comparison with that aged 20 to 34 yields a surprising negative influence on commercial payments. Clearly, some re-specification is in order. With more observations, either a fixed effect term for each metropolitan area could be included or first differences might be checked to see if the negative coefficient remains.
The HMO PMPM premium differences are largely accounted ([R.sup.2] = 0.828) for in the regression specification presented in Exhibit 3. In addition to the time trend, the most significant indicator (at the 99 percent confidence level) is the concentration of hospitals. More concentration was related to a lower premium. As noted above, this result may come from selective contracting and bargaining economies. These effects deserve further exploration. No other factor plays a statistically significant role, even at the 90 percent level.
Medicare cost-shifting was investigated for both the hospital commercial revenue and HMO payment specifications. In neither case did the Medicare term (Part A payments for the first specification and Part A plus Part B payments for the second one) enter the regression estimates in a statistically significant way. Furthermore, Medicare cost shifting did not show up in separate analysis that used InterStudy data on all metropolitan areas in the US including observations from each year between 1985 and 2001, and a more comprehensive regression specification based on Wholey et al. (1995). Evidence of cost-shifting seems to exist in some markets, but the specification and data in this study are not rich enough to distinguish these effects from the effects of significant healthcare provider market power and scale economies.
Ideally, cost shifting from Medicare onto commercial payments for physician services should also be investigated. Medicare Part B payments per enrollee are readily available, but no comprehensive and validated physician payment indicator exists for each metropolitan area. Physicians per 1,000 residents is available for all metropolitan areas from the Area Resource File, at both the aggregate and individual specialty level, but survey pricing data only exists for some metropolitan areas and limited information on the quantity or intensity of service can be obtained at the local level. Furthermore, physician bargaining power--as reflected, for example, by participation in group practices--can be readily obtained at the state level, but not for the metropolitan level. More definitive results clearly depend on more in-depth study using a larger database.
Demographic differences across metropolitan areas seem to play a limited role in explaining the variation in commercial hospital payments per non-elder or HMO PMPM premiums. Alternative specifications of the demographic structure might be tried to better capture the relationship between cost and age distribution. For the HMO regression, one would need to know HMO enrollee characteristics relative to market averages to determine if selectivity bias (younger and less costly residents enrolling in HMOs) exists at the metropolitan level. Such data are not readily available at the metropolitan level. Furthermore, even if these data were available at the health plan level, one would need to allocate membership by age to the metropolitan area. Such an allocation decision, however made, may not accurately represent actual enrollment patterns.
Given the panel data set used for the regression analysis, techniques such as fixed effects variables and first differences might add more precision to the results. Alternatively, other cost differentiating factors could be included. The CMS annually collects hospital wage date and aggregates it to the metropolitan area level. Based on the March 2004 report (CMS 2004), wage rates range from a low of $23.07 per hour to a high of $29.16 with a standard deviation of $1.54. This amount of variation falls well short of that for healthcare costs indicated in Exhibit 1 and, thus, is unlikely to be a significant regressor.
Finally, the relative bargaining power of healthcare providers, in contrast with that of purchasers, deserves further scrutiny. This study highlights a number of such representations, but the regression estimates only begin to reveal information about the influence of bargaining strength. For commercial payments to hospitals, HMO competition (as reflected by the joint product of the number of HMOs and the penetration rate) does suggest some downward influence, as expected. Hospital concentration, based on a Herfindahl index of commercial admission shares for each hospital system, does not, however, influence the resulting payments. Ways to differentiate scale economies from bargaining power are needed. For HMO premiums, the Herfindahl index for hospitals enters with a negative sign, which suggests transactions scale economies. Analysts should devote additional study to these influences as well as to the physician services sector, a particularly important component of the healthcare market for those under the age of 65.
This study, motivated by the potential effects on employment and burden sharing of significant differences in healthcare costs across metropolitan areas, finds marked differences in healthcare costs and trends among the 20 MSAs located in the central portion of the US. These differences seem to be unrelated to differences in either Medicare payments or population demographics among the 20 MSAs. Representations of MSA market dynamics reveal some influence of relative market power, scale economies, and competition in explaining differences in healthcare costs, but further study of these dynamics and their influences is clearly needed. Such study should begin with attempts to find both more complete and comprehensive measures of healthcare cost at the metropolitan level and better representations of the bargaining power of healthcare providers and purchasers. Expansion of the number of MSAs studied would also increase confidence in inferences that can be drawn from the methods used in this study.
Altman, D.E., & Levitt, L. (2002). The sad history of health care cost containment as told in one chart. Health Affairs, web exclusive article (Jan. 23): www.healthaffairs.org.
Baker, L.C. (2001). Measuring competition in health care markets. Health Services Research, (Apr., Part II). 36(1): 223-51.
Center for Medicare & Medicaid Services. (2004). FY 2005 Proposed Public Use Files. www.cms.hhs.gov/providers/hipps/ippswage.asp. Posted Mar. 2, 2004.
Cutler, D.M. (1998). Cost shifting or cost cutting?: the incidence of reductions in Medicare payments. In NBER/Tax Policy and the Economy, 12(1).
Dartmouth Atlas of Health Care. (2004). www.dartmouthatlas.org. Accessed Jun. 1, 2004.
Economic Report of the President (2004). Transmitted to The Congress, Feb. 2004. US Government Printing Office. Washington, DC.
Ginsburg, P.B. (2002). Hospital and physician cost shifting: a conceptual framework. Presented at conference, When public payment declines, does cost-shifting occur? Hospital and Physician Responses, (Nov. 13): www.hcfo.net/costshiftingslides.htm.
Makuc, D.M., Haglund, B., Ingrain, D.D., Kleinman, J.C., & Feldman, J.J. (1991). Health service areas for the United States. Vital Health Statistics, (Nov.) 112(2): 1-102.
Martin, A., Whittle, L., Levit, K., Won, G., & Hinman, L. (2002). Health care spending during 1991-1998: a fifty state review. Health Affairs, (Jul./Aug.) 21 (4): 206-214.
Morrisey, M.A. (1994). Cost Shifting in Health Care: Separating Evidence from Rhetoric, Washington, DC: AEI Press. 1994.
Morrisey, M.A. (2003). New Myths, Old Confusion, and Enduring Reality. Health Affairs, web exclusive article (Oct.): www.healthaffairs.org.
Pauly, M.V. (1997). Health Benefits at Work: An Economic and Political Analysis of Employment-Based Health Insurance, University of Michigan Press.
Strunk, B.C., Ginsburg, P.B., & Gabel, J.R. (2002). Tracking health care costs: growth accelerates again in 2001, Health Affairs, web Exclusive article (Sept.): www.healthaffairs.org.
Wholey, D., Feldman, R., & Christianson, J.B. (1995). The effect of market structure on HMO premiums. Journal of Health Economics, (May), 14(1): 81-105.
Merton D. Finkler
Address for correspondence: Merton D. Finkler, Lawrence University, PO Box 599, Appleton, WI 54912-0599 USA, email@example.com.
EXHIBIT 1 DESCRIPTIVE STATISTICS PANEL A: 2000 Variable Mean Median Std. Dev. HMO PMPM $149.06 $146.96 $14.70 NGH$/NE (a) $916.39 $897.74 $150.10 FEHBP (b) $151.99 $141.85 $31.79 Medicare Part A $262.81 $260.31 $38.79 Medicare Part B $171.07 $171.61 $18.24 Total Medicare $433.88 $431.92 $52.18 Per Capita Income $28,738 $28,081 $1,728 Old/Young (c) 105.3% 105.5% 13.5% Age 65 or Older Share 11.9% 11.6% 1.9% HEALTHCARE SECTOR No. of Hospitals 20 12 18 Hospital HHI (d) 2,428 2,132 1,038 Hospital Admits/1000 123.7 124.9 18.1 Physicians/1000 2.4 2.4 0.6 Specialists/1000 1.6 0.16 0.4 HMO Penetration Rate 26.3% 24.0% 11.4% No. of HMOs 13.7 14 4.1 HMO Competitiveness 3.61 3.39 1.86 Capitation/Spec. Rev. 26.3% 13.9% 19.1% Variable Min. Max. HMO PMPM $123.14 $178.73 NGH$/NE (a) $745.28 $1,164.47 FEHBP (b) $113.77 $227.52 Medicare Part A $195.34 $352.62 Medicare Part B $139.70 $206.03 Total Medicare $346.63 $555.65 Per Caodta Income $26,877 $32,550 Old/Young (c) 84.1% 135.0% Age 65 or Older Share 9.6% 17.5% HEALTHCARE SECTOR No. of Hospitals 4 84 Hospital HHI (d) 416 4,265 Hospital Admits/1000 86.3 165.8 Physicians/1000 1.6 3.9 Specialists/1000 1.0 2.6 HMO Penetration Rate 10.9% 60.6% No. of HMOs 6.0 22.0 HMO Competitiveness 1.42 7.27 Capitation/Spec. Rev. 0.0% 67.4% Notes: (a) NGH$/NE = Non-Governmental Hospital Payments per Non-Elder (B) FEHBP = Federal Employee Health Benefit Program (c) Old/Young = Ratio of those age 20-34 to those age 45-64 (d) Hospital HHI = Hospital Herfindahl Hirschman Index PANEL B: 1990-2000 GROWTH RATE (e) Mean Median Std. Variable Growth Growth Dev. HMO PMPM 62.0% 58.8% 17.1% NGH$/NE (a) 58.2% 49.6% 16.3% FEHBP (b) n/a n/a n/a Medicare Part A 37.2% 34.3% 9.5% Medicare Part B 58.7% 63.0% 20.0% Total Medicare 44.6% 44.4% 9.4% Per Capita Income 49.5% 50.0% 5.3% ld/Youn (c) 43.7% 43.5% 12.6% Aged 65 or Older Share 0.0% 1.0% 6.8% HEALTHCARE SECTOR No. of Hospitals -3.9% -8.6% 20.2% Hospital HHId 80.7% 53.9% 82.0% Hospital Admits/1000 -5.0% -7.5% -22.0% Physicians/1 000 12.6% 14.4% 6.1% Specialists/1 000 n/a n/a n/a HMO Penetration Rate 191.1% 190.0% 164.8% No. of HMOs 26.3% 61.1% 53.9% HMO Competitiveness 459.1% 362.8% 451.4% Capitation/Spec. Rev. 26.3% 61.1% 53.9% Variable Min. Max. HMO PMPM 33.2% 96.7% NGH$/NE (a) 18.3% 161.1% FEHBP (b) n/a n/a Medicare Part A 22.3% 57.4% Medicare Part B 10.7% 95.6% Total Medicare 22.6% 57.4% Per Capita Income 30.0% 56.2% ld/Youn (c) 22.6% 80.6% Aged 65 or Older Share -14.2% 15.2% HEALTHCARE SECTOR No. of Hospitals -27.3% 60.0% Hospital HHId -4.2% 288.2% Hospital Admits/1000 -20.0% 19.7% Physicians/1 000 -0.3% 23.8% Specialists/1 000 n/a n/a HMO Penetration Rate 49.6% 704.8% No. of HMOs -24.0% 160.0% HMO Competitiveness 49.6% 199.2% Capitation/Spec. Rev. -100.0% 12012.5% Notes: (a) NGH$/NE = Non-Governmental Hospital Payments per Non-Elder (B) FEHBP = Federal Employee Health Benefit Program (c) Old/Young = Ratio of those age 20-34 to those age 45-64 (d) Hospital HHI = Hospital Herfindahl Hirschman Index (e) Growth for Capitation as a percentage of HMO payments to specialists for 1990-2000 EXHIBIT 2 REGRESSION ANALYSIS OF METROPOLITAN AREA HEALTHCARE FACTORS ON COMMERCIAL PAYMENTS DEPENDENT VARIABLE Non-Medicare & Medicaid Mean $744.97 Payments to Hospitals Std. Dev. $194.24 per on-Elder Adjusted [R.sup.2] 0.781 Std. Error $90.86 Observations 57 -Statistic 23.21 Significance 1.91 * [10.sup.-14] INDEPENDENT VARIABLES Variable Coefficient Std. Error Intercept -48.83 339.87 Medicare Part A -0.117 0.507 Old/Young -301.95 171.67 Year 51.48 7.288 Commercial Share 488.2 405.92 HMO * Penetration -24.36 14.35 Hospital Herfindahl -0.051 0.017 Ds/1000 60.16 33.39 HMO Penetration Rate 94.51 232.85 Hospital Admits/ 1,000 4.318 0.725 INDEPENDENT VARIABLES Variable t-Statistic p-Value Intercept -0.14 0.89 Medicare Part A -0.23 0.82 Old/Young -1.76 0.09 Year 7.06 6.56 * [10.sup.-09] Commercial Share 1.20 0.24 HMO * Penetration -1.7 0.10 Hospital Herfindahl -1.19 0.23 Ds/1000 1.80 0.08 HMO Penetration Rate 0.41 0.69 Hospital Admits/ 1,000 5.95 3.17 * [10.sup.-07] EXHIBIT 3 REGRESSION ANALYSIS ADJUSTED FOR DEGREES OF FREEDOM VARIATION IN HMO PMPM PREMIUMS ACROSS 20 MSAs AND 3 TIME PERIODS DEPENDENT VARIABLE Mean $119.99 Std. Dev. $26.05 HMO premiums dusted [R.sup.2] 0.828 per member Std. Error $10.80 per month Observations 60 F-Statistic 41.6 Significance 3.97 * [10.sup.-19] INDEPENDENT VARIABLES Variable Coefficient Std. Error Intercept 111.47 14.84 Medicare AAPCC 0.0038 0.038 Old/Young -20.3 15.70 Year 6.13 0.788 HMO Penetration 50.6 23.52 HMO * Penetration -1.209 1.485 Hospital Herfindahl -0.00515 0.00195 MDs/1000 0.665 2.209 Variable t-Statistic p -Value Intercept 7.51 7.58 * [10.sup.-10] Medicare AAPCC 0.10 0.92 Old/Young -1.29 0.20 Year 7.78 2.86 * [10.sup.-10] HMO Penetration 2.15 0.04 HMO * Penetration -0.81 0.42 Hospital Herfindahl -2.65 0.01 MDs/1000 -0.30 0.77
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Health maintenance organizations cost cutting|
|Author:||Finkler, Merton D.|
|Publication:||Research in Healthcare Financial Management|
|Date:||Jan 1, 2005|
|Previous Article:||Cost analysis of computerized physician order entry using value stream analysis: a case study.|
|Next Article:||An examination of the financial structure of not-for-profit hospitals engaging in joint ventures: is tax-exempt status in jeopardy?|