|
|
| |||||||||||||||||||||||||||||||||||||||
Background The national volume-performance standard recently implemented by Medicare does not account for geographic variation in expenditures for physicians' services. To study this variation, we examined expenditures for physicians' services in all metropolitan areas in the United States.
Methods We used Medicare claims data for 1989 to measure rates of service use for beneficiaries living in the 317 U.S. metropolitan statistical areas (MSAs). The variables investigated were rates of admission to the hospital, payments to physicians for inpatient care per admission and per beneficiary, payments to physicians for outpatient care per beneficiary, and overall payments to physicians per beneficiary. Expenditures were measured in terms of allowed charges as adjusted to reflect prevailing charges in each MSA. Rates of use were adjusted for age and sex, with the exception of the variable for payments to physicians for inpatient care per admission, which was adjusted for case mix.
Results Expenditures for the delivery of physicians' services to Medicare beneficiaries varied markedly among MSAs, with those for the areas with the lowest and the highest rates differing at least twofold on each measure. The measures for specific areas varied in parallel: areas with high rates of admission tended to have high levels of payment to physicians for inpatient care per admission, and areas with high payments for inpatient services tended to have high payments for outpatient services. Expenditures were not related to the number of physicians per capita but were lower in MSAs with a high proportion of primary care practitioners. The variation persisted when the 25 largest MSAs were examined; for total payments to physicians per beneficiary, there was a twofold difference between the area with the lowest rate and that with the highest, San Francisco ($872) and Miami ($1,874). The states with the highest overall payments to physicians per beneficiary were Florida, Louisiana, and Michigan.
Conclusions The marked variation among metropolitan areas in payments to physicians underscores the lack of consensus among physicians about which services are required. Moreover, the practice style in a given community appears to be influenced not by the aggregate supply of physicians but rather by the mixture of primary care physicians and specialists.
Two decades of research have documented persistent and substantial geographic variations in use of services3,4,5,6,7,8,9. These investigations have generally focused on rates of admission to the hospital and of procedures, such as hysterectomy or coronary-artery bypass surgery. Although variations in expenditures for physicians' services have been documented in several states,10,11 the magnitude of the variation across the United States has not been evaluated systematically.
To examine the extent of this variation, we analyzed 1989 Medicare data for the 317 metropolitan statistical areas (MSAs) in the United States. We considered all physicians' services charged to Medicare, classifying them according to whether they were for inpatient or outpatient care. Inpatient services were subclassified to reflect two types of fundamental clinical decisions: decisions to hospitalize patients and decisions about how many services to deliver to them during hospitalization.
Methods
Overview
To explore variation among large geographic areas, we analyzed 1989 Medicare data on admission rates and expenditures for physicians' services. The unit of analysis was the MSA. With the exception of Providence, Rhode Island (dropped from the study because of incomplete billing data), all 317 MSAs were included in the analysis. These areas contained a resident population of 23.1 million Medicare beneficiaries.
We first calculated population-based rates of service use, assigning hospital admissions and expenditures for physicians' services to the geographic area where the beneficiary resided, rather than to the place where the service was delivered. Thus, the admission of a New York City resident to a Miami hospital was attributed to the New York City MSA.
The method of calculating the five variables related to service use that were studied in each MSA is summarized in Table 1. Detailed below are the methods used to calculate the three fundamental variables: the rate of hospital admissions, payments to physicians for inpatient care per admission, and payments to physicians for outpatient care per beneficiary. In addition, two derivative variables were calculated. The variable for payments for inpatient care per beneficiary was the product of the admission rate and payments to physicians for inpatient care per admission; and the variable for overall payments to physicians was the sum of inpatient and outpatient expenditures per beneficiary. Both summary results (Table 2) and results for selected MSAs (Table 3) are reported for all five variables.
|
|
|
The Medicare Provider Analysis and Review (MedPAR) file was the source of the data on admissions. This file includes all 1989 Medicare admissions except those of beneficiaries in risk-contract health maintenance organizations. The numerator includes approximately 10 million hospital admissions, or about one third of all admissions in the nation12. Indirect-standardization techniques were used to adjust the data for differences among MSAs with respect to the distribution of age and sex.
Payments to Physicians for Inpatient Care per Admission
Information on Medicare charges for physicians' services in 1989 was obtained from the Part B Medicare Annual Data file, which contained all payments to physicians corresponding to a random sample comprising 5 percent of beneficiaries. This allowed the measurement of payments to physicians for inpatient care for about 500,000 admissions. Allowed charges, which include copayments made by beneficiaries, were used rather than charges reimbursed by Medicare,13 because the latter were affected by the size of the copayments.
Using the beneficiary's identification number and the dates of service, we identified payments to all physicians involved with each admission, and we linked the hospital claims with the physicians' claims. Payments to physicians for services provided between admission and discharge were totaled to obtain the overall payments for each hospital stay. In a related study, we found that including services delivered during the 28 days before and the 28 days after the hospital stay did not materially affect the relative payments to physicians among hospitals14.
Payments per admission were then adjusted to control for both prices and case mix. To adjust for price differences among the MSAs, payments were deflated according to the prevailing charge index for the Medicare locality15. To adjust for case-mix differences in admitting diagnoses among MSAs, payments were divided by the relative weighted payment for the relevant diagnosis-related group (DRG). Although data on weighted DRG costs are available for hospital services, weights for payments to physicians had to be calculated from the Medicare Part B data. Elsewhere, we have computed the average payment to physicians during a hospitalization (after adjustment for price) for each DRG14. The weighted payment to physicians for each DRG was calculated as the ratio of the average payment for that DRG to the average payment nationally for all DRGs. The adjusted payments were then assigned to the MSA where the beneficiary lived and were averaged.
Payments to Physicians for Outpatient Care per Beneficiary
Allowed charges for physicians' services for outpatient care per beneficiary were considered to include all payments to physicians for services rendered to a beneficiary when that person was not hospitalized. These data were then adjusted to reflect both prices and the distribution of the population with regard to age and sex, as described above.
Statistical Analysis
To illustrate the precision available from such a large data base, we selected payments to physicians for inpatient care per admission because that variable had the lowest coefficient of variation (Table 2) and was based on the fewest observations (i.e., only beneficiaries who were admitted to the hospital). Standard errors were used to calculate 95 percent confidence intervals for each MSA. For 62 percent of the MSAs, payments to physicians for inpatient care per admission differed significantly from the national mean, whereas with random variation alone, only 5 percent of the MSAs would have differed from the national mean.
Physician Supply
To examine the relation between the five variables for use of services and the supply of physicians, we obtained data on the number of physicians and the proportion who were engaged in primary care within each MSA. These data (counts of physicians engaged in patient care in 1989 who were not federal employees) were obtained from the American Medical Association Physician Masterfile. Two variables were calculated: the number of physicians per capita and the number of primary care physicians as a percentage of all physicians. Primary care was considered to include general practice, family practice, and internal medicine.
Results
Although the MSAs studied contained relatively large populations, payments for physicians' services varied markedly among them. The average MSA included more than 70,000 Medicare beneficiaries, and 303 of the 317 MSAs had at least 10,000 beneficiaries. In addition, when payments to physicians for inpatient care per admission and payments for outpatient care were calculated, each MSA had at least 150 admissions and 500 beneficiaries included in the 5 percent sample.
Table 2 shows the mean, range, coefficient of variation, and other summary statistics for the five variables, and Table 3 shows these five measures for the largest metropolitan area in each state, as well as for the 25 largest MSAs in the nation. For each measure, the highest and lowest results among the 317 MSAs differed at least twofold. After adjustment for age and sex, the admission rate per 1000 beneficiaries (mean, 304) ranged from 219 in Boise, Idaho (Table 3), to 533 in Pascagoula, Mississippi. After adjustment for price and case mix, payments to physicians for inpatient care per admission (mean, $1,180) ranged from $677 in Vancouver, Washington, to $1,580 in Miami. Payments to physicians for inpatient care per beneficiary (mean, $360), the product of the admission rate and payments to physicians for inpatient care per admission, ranged from $212 in Vancouver to $700 in Pascagoula. After adjustment for price, age, and sex, payments to physicians for outpatient care per beneficiary (mean, $641) ranged from $388 in Hagerstown, Maryland, to $1,388 in Miami. Overall payments to physicians, calculated as the sum of payments for inpatient and outpatient care (mean, $1,001), ranged from $655 in Sheboygan, Wisconsin, to $1,874 in Miami.
The relations between one measure of use and another were examined to establish the extent to which variations might cancel one another. For example, a propensity to admit patients might be associated with a relatively healthy inpatient population and consequently with lower payments to physicians per admission. MSAs with high admission rates, however, tended to have high levels of payments to physicians for inpatient care per admission (r = 0.24, P<0.001). Also, we found no evidence of the substitution of outpatient care for inpatient services. In fact, the opposite was true: MSAs with high levels of payment for inpatient care tended also to have high levels of payment for outpatient care (r = 0.25, P<0.001). A stronger correlation was found between payments to physicians for inpatient care per admission and payments for outpatient care (r = 0.42, P<0.001).
We also examined the 25 largest MSAs separately. They ranged in size from 188,000 beneficiaries in San Francisco to more than 1 million in New York City. The correlations described previously were similar. Figure 1 shows that the variation in payments to physicians persisted among these large metropolitan areas and demonstrates a twofold difference in overall physician expenditures between San Francisco ($872) and Miami ($1,874).
|
|
|
Geographic variation in medical care is well documented3,4,5,6,7,8,9. Most studies have focused on admission rates or rates of use for discrete procedures and have examined small geographic areas. We examined all Medicare payments for physicians' services in providing inpatient and outpatient care and considered every metropolitan area in the nation. Despite the high degree of aggregation in terms of the combination of services and the relatively large geographic areas (both factors that tend to dampen variation), the magnitude of variation was substantial.
Although physicians' services are generally reported as a single item in presentations of national health care expenditures,16 our analyses used an accounting structure that reflects more accurately the clinical decisions that generate services. The first is the decision to admit a patient to the hospital. The second is the choice of professional services once a patient is hospitalized. The third is the choice of the professional services given to outpatients.
This construct allows us to reject two hypotheses that attempt to explain the marked geographic variation in expenditures for physicians' services. This variation cannot be attributed to the substitution of inpatient for outpatient care; indeed, metropolitan areas with high levels of payments to physicians for inpatient care tend to be those that have high levels of payments for outpatient care. In addition, the variation cannot be attributed to the disproportionate admission of mildly sick patients in high-admission areas; indeed, in areas with high admission rates physicians tend to deliver extensive services to patients once they are hospitalized. Instead, unless a hypothesis of marked variation in the unmeasured burden of disease is entertained, the variation in payments to physicians must be attributed to broad-based differences in practice style among communities that cut across the three major clinical decisions generating the payments.
Consider the differences between Miami and San Francisco. After adjustment for differences in age and sex, beneficiaries in Miami were 1.15 times more likely to be admitted to the hospital than their counterparts in San Francisco. Once admitted, they received physicians' services that were 1.7 times more costly. The relatively low level of payments to physicians for inpatient care in San Francisco did not, however, translate into a corresponding increase in payments for outpatient care. In fact, payments per beneficiary for outpatient care in Miami were 2.2 times those in San Francisco.
Our analysis of the supply of physicians points up the important effect of the proportion of physicians engaged in primary care on overall payments to physicians. Higher proportions of primary care physicians in an MSA were associated with a less expensive practice of medicine (i.e., lower payments for both in-hospital and out-of-hospital care). When cross-sectional differences are studied, a community's practice style appears to be determined more by physicians' specialties than by their aggregate numbers. The somewhat surprising finding that the number of physicians per capita was unrelated to use of services is most likely the effect of three limitations of this study. First, the data studied represented only Medicare beneficiaries. Second, although the crossings of MSA borders had no effect on our population-based measures of use, they may have affected the validity of our measure of supply. Third, the data were cross-sectional; whereas our results suggest that the supply of physicians does not explain cross-sectional differences in use, it may explain growth. A nationwide increase in the supply of physicians, therefore, would presumably result in increased expenditures for their services.
Considerable effort was devoted to removing variation due to mobility of patients, prices, demographic variables in the population, and admission diagnoses. First, to account for patients who traveled outside their MSA to obtain care, the analysis assigned a beneficiary's use of services to the metropolitan area where he or she lived. Second, to ensure that the variations were not caused by variations in the prices paid for services, each physician's bill was deflated according to an index of Medicare's prevailing charges. Finally, to account for differences among populations, rates of use were adjusted for age and sex, and payments to physicians for inpatient care per admission were adjusted for case mix.
Nevertheless, three potential criticisms of the analysis merit discussion. One is that these results may be peculiar to 1989. To investigate this possibility, we used aggregate data from the Health Care Financing Administration to calculate Medicare Part B payments (mainly to physicians) for the same MSAs in 1984, 1985, 1986, and 1987. After adjustment for age, sex, and price, the pattern was remarkably stable over time, with correlation coefficients of approximately 0.85 between consecutive years. Given the magnitude of the variations and their stability over time, we are confident that our results are not statistical artifacts.
A second potential criticism is that unmeasured differences in disease burden might explain the variation among MSAs. Previous work has found that differences in use are not explained by differences in disease burden9,17,18,19,20. However, it is possible that physicians in Miami, for example, serve a much more severely ill population than do their colleagues in San Francisco. To explain the differences, the disparity in disease burden would have to have three characteristics. First, it could not be reflected in the age and sex distributions of the respective populations. Second, differences among inpatients with respect to disease could not be reflected in the case mix. Third, the difference in disease burden would have to be substantial. In fact, to explain differences in use, one must posit that on average beneficiaries in Miami are twice as ill as their counterparts in San Francisco.
Finally, our study could be criticized for measuring expenditures without linking them to outcomes. The only outcome measure available for the study populations was mortality. We studied 1989 age-adjusted mortality for Medicare beneficiaries in the 317 MSAs and found no correlation between overall payments to physicians and mortality. The only variables significantly correlated with mortality were the rate of admissions and consequently payments to physicians for inpatient care. These correlations were positive (r = 0.40 and 0.34, respectively), suggesting that areas with higher rates of hospital admission also had higher death rates in the population. The direction of causation, however, is ambiguous. Either higher mortality rates could reflect higher levels of illness in the population, leading to higher rates of hospitalization, or higher rates of hospitalization could result in higher mortality.
Regardless of their cause, the magnitude of the variations in physicians' services raises difficult questions for Medicare's policies toward fee-for-service payment. The variations underscore the lack of a firm basis on which to judge the appropriate rate for most procedures,21 let alone the rate for the aggregate differences in per capita health care expenditures. No evidence is available to support the contention that the residents of Miami benefit from their high treatment rates; they may just as plausibly be harmed. The high cost differential underscores the need to assess their effect on health.
Changing such patterns will be challenging. The geographic index in the Medicare fee schedule, which was implemented in January 1992, will dampen the geographic variation in fees. Because we adjusted for charges prevailing locally, however, this index will have little effect on the variations in adjusted payments reported here. The relative-value scale of the fee schedule will increase payments to primary care physicians in relation to payments to specialists. This change could either mitigate the variations reported here or, if lower fees increase volume, it could exacerbate them. Attempts to control variation by micromanagement of the clinical encounter require more knowledge than is now available about the usefulness of specific services in specific clinical settings. Furthermore, physicians understandably have a negative view of such micromanagement, because it is enforced by people with no direct contact with or responsibility for the patient.
Global limits avoid micromanagement. They achieve cost containment by specifying an overall budget and allowing physicians the autonomy to allocate resources as they think best. As such, the recently enacted volume-performance standard is a move in the right direction. Because the entire nation is the risk pool, however, this may not have much effect on geographic variations as demonstrated here.
To address geographic variation, the volume-performance standard needs to consider smaller populations. The difficulty lies in linking beneficiaries to some organization of physicians. A reasonable place to start might be with the inpatients of each hospital, because the episode of care including their hospitalization has been identified and these patients are served by a well-defined group of physicians22. Strategies that deal with payments to physicians for inpatient care per admission must ultimately be complemented by strategies that account for the variation in admission rates and outpatient services. As initial steps, however, the number of hospital beds might be regulated23 and a volume-performance standard for outpatient services defined for each state24.
It is clear from our study and others that there is substantial variation in the use of health care resources among distinct populations. Historically, policy makers could not respond to these variations, simply because they could not measure them systematically. With the availability of large population-based data bases, they now can. As policy makers consider either true global limits or volume-performance standards, it is important that they consider the existing distribution of health resources and make an effort not to lock in current inequities.
Supported in part by the Health Care Financing Administration under a cooperative agreement (17-C-99489/3-01) with the Urban Institute. Dr. H.G. Welch is the recipient of a Veterans Affairs Career Development Award in health services research and development. Any opinions expressed herein are those of the authors and do not represent the opinions or policies of the Health Care Financing Administration, the Department of Veterans Affairs, the Urban Institute, Dartmouth Medical School, or their sponsors.
We are indebted to Mr. Steve Norton and Ms. Juhi Chawla for research assistance; to Ms. Paula Beasley and her staff at Social and Scientific Systems for programming; and to Ms. Sherry Terrell, the project officer of the Health Care Financing Administration, for her support.
Source Information
From the Urban Institute, Washington, D.
(W.P.W., M.E.M.); the Department of Veterans Affairs Medical Center, White River Junction, Vt. (H.G.W., E.S.F.); and the Center for the Evaluative Clinical Sciences, Dartmouth Medical School, Hanover, N.H. (H.G.W., E.S.F., J.E.W.).
Address reprint requests to Dr. W.P. Welch at the Urban Institute, 2100 M St., NW, Washington, DC 20037.
References
| |||||||||||||||||||||||||||||||||||||||
Related Letters:
Geographic Variations in Payments to Physicians
Huff E., Brownsberger W. N., McMillan D. E., Brunhild G., Peven D. R., Welch H. G., Miller M. E., Welch W. P.
Extract |
Full Text
N Engl J Med 1993;
329:666-667, Aug 26, 1993.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
