Do physicians cost shift?

At the Intersection of Health, Health Care and Policy Cite this article as: T Rice, S Stearns, S DesHarnais, D Pathman, M Tai-Seale and M Brasure Do physicians cost shift? Health Affairs, 15, no.3 (1996):215-225 doi: 10.1377/hlthaff.15.3.215

The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/15/3/215

For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml

Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 1996 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.

Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

DO

PHY SICIANS

CO ST SH I FT?

Do Physicians Cost Shift? Hospitals shift costs, charging privately insured patients more to cover losses from other payers. To date, no research has examined whether physicians do the same. by Thomas Rice, Sally Stearns, Susan DesHarnais, Donald Pathman, Ming Tai-Seale, and Michelle Brasure A B S T R A C T : T his study analy z es w hether phy sicians charg e their privately insure d patients m o re - a pra ctice kno w n as co st shifting - in re spo nse to Me dicare pay me nt re ductio ns. A s part o f co ng re ssional leg islatio n in 1989 and 1990, Medicare reduced its pay ment rates for selected procedures by as much as 30 percent. H ere we ex amine whether reductions in Medicare rates increase how much phy sicians charg e privately insured patients. O ur data provide no evidence that phy sicians respond to Medicare pay ment reductions by shifting costs to their privately insured patients.

F

have been accused of engaging in a practice known as cost shifting. This term is, unfortunately, a bit of a misnomer in that it is really charges, not costs, that are being shifted. Cost shifting refers to the practice of charging privately insured patients more to make up for losses from patients who do not pay full freight-the uninsured, Medicaid patients, and, more recently, Medicare beneficiaries. Cost shifting was roundly criticized during the recent health care reform debates as saddling the working population with an undue economic burden. There is ample evidence that cost shifting occurs in the hospital sector. The Prospective Payment Assessment Commission (ProPAC) has estimated that privately insured patients are charged, on average, 28 percent more than costs.’ To our knowledge, however, no one has examined specifically whether physicians engage in the OR MANY YEARS HOSPITALS

DA T AWA T CH

Thomas Rice is a professor and chair of the Department of Health Services at the University of California, Los Angeles (UCLA), School of Public Health. Sally Stearns is an assistant professor in the Department of Health Policy and Administration at the University of North Carolina (UNC) School of Public Health and a research fellow at UNC’s Cecil G. Sheps Center for Health Services Research. Susan DesHarnais is an associate professor in that department, and Michelle Brasure is a doctoral candidate there. Donald Pathman is research director and assistant professor of family medicine at UNC and a research fellow at the Sheps Center. Ming Tai-Seale is an assistant professor at the Indiana University School of Public Health and Environmental Affairs. H E AL T H AF F AI R S - F a l l 1 9 9 6 © 1996 The People-to-People Health Foundation, Inc content.healthaffairs.org Downloaded from

by Health Affairs on January 21, 2016 by guest

215

D

ATA

W

AT C H 2

practice as well. In this study we analyze whether physicians engage in cost shifting in response to recent substantial reductions in Medicare payment rates. As part of the Omnibus Budget Reconciliation Acts of 1989 and 1990 (OBRA 1989 and 1990) Medicare’s prevailing charges for thirty-six groups of procedures whose payment rates were deemed “overvalued” (a total of 245 procedure codes) were reduced by as much as 30 percent. The fact that different geographic areas experienced different magnitudes of payment reduction makes it easier to isolate the effects of interest in the statistical analysis. Using data on physician charges and payment rates from both public and private sources, we construct and estimate multivariate models that are designed to detect whether physicians increased their charges to privately insured patients in response to these Medicare payment reductions,

Incentives And Barriers To Physician Cost Shifting 216

DATAWATCH

Physicians have an economic incentive to engage in cost shifting: Medicare and Medicaid payments are low compared with reimbursement for privately insured patients, and both Medicare and Medicaid restrict how much physicians can charge program beneficiaries beyond what the public insurers reimburse. The Physician Payment Review Commission (PPRC) estimates that Medicare pays only 68 percent as much as private insurers do, and that Medicaid 3 pays an average of around 50 percent. A large proportion of physicians do not treat Medicaid patients because of low payment rates. Anecdotal evidence suggests that the same thing may have begun to happen with Medicare, as its payment rates gravitate toward those 4 of Medicaid. However, this has yet to be confirmed by data. One possible physician response to public insurers’ low payment rates would be to charge publicly insured patients amounts in excess of each program’s payment rates. We define the amount charged over and above what an insurer deems reasonable as “excess charges.” However, such charges are not permitted under Medicaid, and generating additional income through excess charges has become increasingly difficult (and unprofitable) under Medicare. Since the mid-1980s Medicare has instituted a number of incentives for physicians to accept Medicare patients on assignment (that is, to accept Medicare’s payment rate or “allowed charge” as payment in full.) Furthermore, as part of the Medicare payment reform legislation of 1989, there are strict limits on physician excess charges to Medicare patients– a maximum of 15 percent per service. These limits, combined with assignment rates that now exceed 90 percent,

H E AL T H AF F AI R S - V o l u m e 1 5 , N u m b e r 3 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

DO

PHY SICIANS

CO ST SH I FT?

have effectively prevented physicians from recouping income by charging their Medicare patients more than the amount of the 5 Medicare payment rate. Such restrictions and limits, however, typically do not exist for charges to privately insured patients. Thus, physicians might have an incentive to increase their charges to privately insured patients. Since-it is difficult to generate more revenue by increasing excess charges to Medicare patients, we might expect this type of behavior to become more pronounced in the wake of the Medicare payment reductions brought about by OBRA 1989 and 1990. However, several forces combine to make this type of cost shifting either difficult or unattractive: (1) Physicians already may be setting their private charges at a profit-maximizing level. In fact, rather than raising charges in response to Medicare payment reductions, they might wish to lower them to attract more privately in6 sured patients. (2) Physicians-particularly surgeons, who are the focu s of our study-may experience competitive pressures (especially from managed care plans) that induce them to keep their fees down. (3) Physicians may perceive that insurers are taking a closer look at their billed charges, especially in the wake of payment reductions by Medicare and even by some private insurance companies, (4) Physicians may be using other methods of increasing their revenues. For example, in earlier research we found some evidence to indicate that physicians have increased the volume of services they 7 provide to privately insured patients. Physicians might find it easier to explain higher volume to insurers than to explain higher fees.

DATAWATCH

Data And Methods Procedure groups selected. In this study we used pooled, cross8 sectional, time series data for sixteen procedure groups. Individual procedures in each group were subjected to substantial Medicare payment reductions as part of OBRA 1989 and 1990. The procedure groups (by specialty) included the following: thoracic surgery (coronary artery bypass graft [CABG]); ophthalmology (cataract extraction); cardiology (right heart catheterization and pulmonary artery wedge monitoring); gastroenterology (colonoscopy); gynecological surgery (hysterectom y and dilatation and curettage); general surgery (extended simple mastectomy, mediastinoscopy, partial or total colectomy, and hernia repair); orthopedic surgery (carpal tunnel release, total hip replacement, and knee arthroscopy); and urology (transurethral resection of the prostate). Data manipulation. Private billed charges and Medicare payment data (the primary dependent and independent variables) were obtained from a convenience sample of seventy-six geographic areas H E AL T H AF F AI R S - F a l l 1 9 9 6 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

217

D

218

ATA

W

DATAWATCH

AT C H

around the country for each procedure group from 1988 through 1991. We obtained data on private billed charges from The MEDSTAT Group, and we obtained Medicare payment rates from the 9 Medicare programs statistical files. The geographic areas represented Medicare charge localities in 10 thirty-two states. The Medicare charge localities were used to define the sample areas because Medicare prevailing charges historically were set according to these areas. Price data were not available for all procedure groups in all locations because of the criteria used for selecting hospitals for the other portion of the study. A total of 2,915 annual observations for procedure group prices in various areas were used in this analysis. In constructing annual measures, we aggregated those data according to the Medicare fee schedule year because changes in the Medicare payment rate did not occur on a regular calendar year 11 basis during 1988-1990. Because different procedures were aggregated into procedure groups for the analysis, it was necessary to create price and payment variables as weighted averages of the allowed charges for individual procedures, using baseline-year procedure volumes as weights. Dependent variables. To capture physician cost shifting, we tested two specifications of the dependent variable. The first was excess private charges-that is, the difference between average billed charges and average payment rates by private insurers in a charge locality area during each of the study years. The second was average billed charges to privately insured patients. Independent variables. The independent variable of primary interest was the Medicare payment rate or allowed charge. This was defined as the average Medicare payment rate in the Medicare 12 charge locality for each group of procedures. The OBRA reductions in Medicare payment rates for “overvalued” Medicare procedures are useful from a research standpoint because the size of the reductions varied geographically. When different physicians are subjected to different changes in payment rates, it is possible to disentangle the impact of payment rate changes on physicians’ pricing behavior from that of other factors that change over tune, such as technology. Two other categories of time-varying independent variables were included in the model: measures of nursing wages and county characteristics. The nursing wage variables, constructed for each year from the Current Population Survey (CPS), were designed to act as proxies for private insurance payment rates or, more precisely, as measures of physician cost. We chose not to use actual payment rates because of the endogenous nature of private payment-that is,

H E AL T H AF F AI R S - V o l u m e 1 5 , N u m b e r 3 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

D O PHY SI CI ANS COST SHI FT?

physicians’ past billed charges are a major determinant of subsequent payment rates. Therefore, the regression represents a 13 reduced-form model for private excess charges. Four measures of wages for registered nurses (RNs) and licensed practical nurses (LPNs) in hospital and other (nonhospital) settings were con14 structed. The other set of independent variables included timevarying area (county) characteristics in each of the four study years: physicians per thousand population, hospital beds per thousand population, health maintenance organization (HMO) membership per thousand population, and per capita income. We used three sets of fixed-effect “dummy” variables in the regressions. First, we used three fee-year dummies to control for overall trends in billed charges and excess charges. Second, we included seventy-five dummy variables to control for the seventy-six geographic areas represented. Finally, we used fifteen dummy variables to control for the sixteen procedure groups, Statistical techniques. We used fixed-effects regression models to estimate the relationship between OBRA 1989 and 1990 Medicare payment reductions and the two specifications of the dependent variable: changes in private excess charges, and changes in private billed charges. The fixed-effects approach provides a strong control for groupwise (area or procedure) heteroskedasticity. In the main set of results, we regress the two dependent variables on the independent variables described earlier. In a second set of regressions (which are summarized but not presented in the exhibits) we allow the coefficient on the Medicare payment rate to differ across procedure groups. This was done by interacting the payment rate with each of the procedure-specific dummy variables. This specification makes it possible to determine whether cost shifting occurs for a subset of procedures. Study limitations. This study has several limitations, most of which concern the data sources that have been compiled and combined for conducting the analysis. First, the data used for the study are based on a convenience sample of geographic areas that included hospitals that agreed to allow researchers to use their discharge abstracts and that had sufficient volume to meet inclusion criteria 12 for the other portion of the study. Although the data sample includes seventy-six geographic areas representing thirty-two states, it should not be considered a representative national sample. Second, the procedures chosen for the study were restricted to only sixteen of the thirty-six groups that experienced substantial payment reductions in response to OBRA 1989 and 1990. Perhaps the most important restriction was that most radiology and pathology services were not included in the study, nor were services pro-

DATAWATCH

H E AL T H AF F AI R S - F a l l 1 9 9 6 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

219

D

ATA

W

AT C H

vided in physicians’ offices, because they were not part of the data 16 set collected for the study. Third, it is possible that excess charges were not collected by some physicians in a geographic area; the dependent variable indicates liability for these charges, but not the actual payment. Physicians might forgive these charges selectively or not pursue payment vigorously. Furthermore, physicians in preferred provider organizations (PPOs) probably would not be allowed to collect such charges because of their contracts. Fourth, the estimates presented here may represent short-run effects and may not necessarily indicate how physicians might alter their pricing behavior in the longer run. Finally, the data sources conclude in December 1991. Changes in pricing behavior since that time obviously are not reflected in this investigation.

Results

220

DATAWATCH

To illustrate the patterns found in the data, we chose one procedure for each of the eight specialties (Exhibit 1). These procedures are represented by four bars on the horizontal axis (one bar for each of the study years 1988-1991). These data therefore show changes in the two dependent variables over time. Deflated (real) private charges stayed roughly constant on average for most of the procedure groups, although real charges declined somewhat for some 17 procedure groups. Not surprisingly, excess private charges are greatest for CABG, which has the highest real private charges. Excess private charges show increases over time for some procedure groups but are fairly constant for several other groups. Exhibit 2 presents descriptive statistics for all of the independent variables except for the time, procedure, and geographic dummies. The average Medicare payment rate was $931. The primary reason that its range is so great (from $119 to $6,146) is that some procedures are far more complicated and time-consuming (and therefore more expensive) than others. The main set of regression results is presented in Exhibit 3. The first column shows the coefficients and p-values when excess charges are the dependent variable, and the second column, when billed charges are the dependent variable. We are primarily interested here in the sign and significance of the first independent variable, the Medicare payment rate; a negative sign provides evidence of cost shifting to privately insured patients. The results show no evidence of cost shifting. Although the sign of the Medicare payment coefficient is negative when the dependent variable is excess charges, the coefficient is not statistically significant (p-value of .567). In the second regression, with billed charges

H E AL T H AF F AI R S - V o l u m e 1 5 , N u m b e r 3 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

DO

PHY SICIANS

COST

SHIFT?

E XH IB IT 1 Private Charges And Excess Private Charges, 1988-1991

SOURCE: Authors’ calculations from MEDSTAT data. NOTE: Sets of four bars per procedure represent four years: 1988, 1989, 1990, and 1991.

as the dependent variable, the Medicare payment coefficient is statistically significant, but the positive sign of the coefficient is not consistent with the cost-shifting hypothesis. The coefficient of .121 can be interpreted as an elasticity, which means that as Medicare payment rates decline by 10 percent, billed charges to privately insured patients fall by about 1.2 percent. Such a result is not implausible-as discussed earlier, physicians may reduce their billed charges to attract more privately insured patients in the wake of Medicare payment red uctions-especially in light of Exhibit 1, H E AL T H AF F AI R S - F a l l 1 9 9 6 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

D

ATA

W

ATC H

E XH IBIT 2 Characteristics Of The Study Sample, Medicare Charge Localities, 1988-1991 S tandard D ev iation

Minim um

$930.5 $14.0 $12.6

$726.9 $1.4 $1.5

$118.9 $10.0 $8.6

$9.3 $8.8 2.0

$1.4 $1.7 1 . 2

$6.0 $5.6 0.32

H ospital beds per 1,000 4.1 H MO membership per 1,000 145.0 Per capita income $16,349.2

1.7 197.1 $3,890.5

1.2 0.15 $7,427.0

Mean

V ariable Medicare payment a Hospital RN wages O ther RN wagesa a

Hospital LPN wages O ther LPN wagesa P hysicians per 1,000

Maxim u m $6,145.7 $16.8 $17.6 $13.0 21.8 8.3 9.5 1, 000. 0 b $33,733.9

SOURCE: Authors’ calculations. NOTES: RN is registered nurse; LPN is licensed practical nurse. N = 2,915. Annual observations for procedure group prices in various areas. a Dollars Per hour. b One county In Minnesota had an estimated 100 percent health maintenance organization (HMO) penetration rate.

E XH IBIT 3 Results For Regression Model Of Excess Private Charge And Billed Charge Amounts Independent v ariable

E xcess charges a

Medicare payment rate Hospital RN wages O ther RN wages

- .109 (.567) 1.492 (.024) ** - .157 (.614)

Hospital LPN wages O ther LPN wages Physician density

- .275 (.242) - .172 (.394) - .717 (. 021) **

.009 (.704) - .002 (.941) (.003) (.925)

H ospital bed density HMO density P er capita income

.528 (.060) * .056 (.054) * - .521 (.517)

.134 ( .000) ** (.012) (.000) ** .172 (.039) **

1989 dummy 1990 dummy 1991 dummy

.262 (.000) ** .231 (.004) ** .392 (.002) **

Billed charges b .121 (.000) ** .079 (.246) .089 (.005) **

- .003 (.656) - .016 ( .062) * - .002 (.887)

SOURCE: Authors’ calculations. NO TE : T he first number for each variable is its coefficient. T he p value of this estimate is shown in parentheses. P rocedure group and geographic area dummy variable coefficients are not include d in the exhibit. All continuous variables are natural log. arithms. N = 2,915. RN is registered nurse; LPN Is licensed practical nurse. HMO Is health maintenance organization. a R 2 = .437 b R 2 = .970 * Statistically significant at the 10 percent level. ** Statistically significant at the 5 percent level.

H E AL T H AF F AI R S - V o l u m e 1 5 , N u m b e r 3 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

DO

PHY SICIANS

CO ST SH I FT?

which shows declining, inflation-adjusted billed charges for some procedures over the study period. Although we have concerns about the validity of this result, it is noteworthy that neither specification of the dependent variable shows any evidence of cost shifting’s We also conducted two more sets of regressions (not shown) in which the Medicare payment rate is interacted with a dummy variable representing each procedure group. This specification allows the coefficient on the Medicare payment rate to differ across procedure groups. These results showed almost no evidence of cost shifting. Of the thirty-two coefficients representing Medicare payment rates in the two regressions, only two coefficients (both in the regression with excess charges as the dependent variable) were negative and statistically significant at the 10 percent level.

Conclusions Our results show no evidence that physicians respond to Medicare payment reductions by engaging in cost shifting-that is, by raising their charges to their privately insured patients. Although the data set used here does not allow us to investigate why cost shifting did not occur, there are several possible explanations. First, physicians may have found it increasingly difficult to charge more as the health care market became more competitive. Second, physicians may have perceived that insurers have been monitoring their billing patterns more closely than before. Third, physicians may have set their billed charges at profit-maximizing levels already; raising them could result in a reduction in privately insured patients. And fourth, perhaps physicians found it easier to increase the volume of services, rather than their billed charges. In a companion study we found evidence that in response to Medicare payment reductions, physicians did increase the volume of some of the serv19 ices they provided to privately insured patients. However, we found no evidence in the study presented here that this same behavior occurs with respect to the fees physicians charge for services. We find no indication that the physician services market has experienced the same type of cost shifting that has been found for hospital services in response to the OBRA 1989 and 1990 payment reductions. It would seem most worthwhile, therefore, for administrators and public policymakers to focus their efforts instead on continuing to monitor the volume and appropriateness of services provided to privately insured patients. These factors seem the most likely routes for cost shifting in the physician services sector.

DATAWATCH

H E AL T H AF F AI R S - F a l l 1 9 9 6 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

223

D

AT A

W

A T C H

This project was supported by The Robert Wood Johnson Foundation (Grant no. 20038). The authors thank Paul Ginsburg of the Center for Studying Health System Change and Christopher Hogan of the Physician Payment Review Commission for providing data as well as for suggestions on conceptual issues. Karen Kmetik of the American Medical Association provided critical assistance in the early phases of the study, and Thomas Walke of the University of North Carolina provided helpful research assistance. Anne Gauthier and Deborah Regal of the Alpha Centerprovided support and encouragement throughout the study. All methodological decisions, conclusions, and errors are solely the responsibility of the authors. NO TES

224

DATAWATCH

1. Prosp ective Payment Assessment Commis sion, Optional Hospital Payment Rates, Congression al Report C-92-03 (Washington: ProPAC, March 1992). 2. Some of the more acad emic r esearch on hosp ital cost shifting occu rr ed some time ago, before the implementation of diagnosis-related groups (DRGs). See, for example, F.A. Sloan and E .R. Becker, “Cro ss-Subsidies and Payment for Hospital Care,” Journal of Health Politics, Policy and Law (Winter 1984): 660-685. Using older studies like these for inform ing cu rr ent policy su ffers fr om one major drawback: Because they pr ed ated implementation of the DRG system, th eir f in d in gs m ay n o t b e ap p licab le to th e cu r r en t h os p ital m ar ket. Two r ecen t go ver n m en t s tu d ies h ave attem p ted to calcu late th e d egr ee of co s t shifting that now exists. See ProPAC. Optional Hospital Payment Rates; and Congressional Budget Office, Responses to Uncompensated Care and Public-Program Controls on Spending: Do Hospitals ‘Cost Shift’? (Washington: CBO, May 1993). 3. Physician Payment Review Commission, Annual Report to Congress, 1995 (Washington: PPRC, 1995). 4. PPRC, Monitoring Access of Medicare Beneficiaries (Washington: PPRC, May 1995). 5. PPRC, Monitoring the Financial Liabili ty of Medicare Benefici aries (Was hington: PPRC, May 1995). 6. We thank Paul Ginsburg for pointing this out to us. 7. T. Rice et al., “Physician Response to Medicare Payment Reductions: Impacts on the Public and Pr ivate Sectors,” Final Report (Submitted to The Robert Wood Jo hns on Fou nd ation, September 1994). 8. Ibid. The procedure groups were chosen primarily to facilitate analysis of how Med icare payment redu ctions affected the volume of services. The criter ia u s ed fo r s electin g p r o ced u r es in clu d ed the f o llo win g: (1) They h ad to be su rgical pro cedur es s o that they were in clud ed in the dis char ge abstract data that were obtained for a different component of the study; (2) they had to have a clean crosswalk between International Classification of Diseases, Ninth Revision, Cli ni ca l Modi fi ca ti on (ICD -9-CM) an d Current Procedura l Termi nol ogy (CP T) co d es , b ecau s e th e d is char ge ab s tr acts u s ed in th at s tu d y d id n ot con tain infor mation on CPT codes to which the Medicar e payment r ed uction s were applied over the study period; and (3) they had to be defined as “overvalued b y Med icar e. B ecau s e th e pr oced u r es s elected ar e a cr o s s-s ection of o ver valu ed pro cedur es repr esenting eight s pecialties, we believe that they r epresent an adequate sample for the analysis conducted in this DataWatch. 9 . The ME D STAT d ata wer e aggr egated to th e m etr op o litan s tati s ti cal ar ea (MSA) or non-MSA state level. Thes e d ata wer e lin ked to the Medicare data (which wer e pro vided at the Med icar e char ge locality level) by using inform a-

H E AL T H AF F AI R S - V o l u m e 1 5 , N u m b e r 3 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

DO

10.

11.

12. 13.

14.

15. 16. 17.

18.

19.

PHY SICIANS

CO ST SH I FT?

tion on the cou nties comprising the charge localities. To help ensur e cleanliness of the d ata, we gave MED STAT r anges for “valid” r eimbur sement rates and billed physician charges for each procedure code, by state, by year. During file con stru ction it was determined that the frequency of claims us ed in constr ucting the ME DSTAT variables was extremely limited for s ome areas (that is, only one or two claims are used to construct the “averages”) and that there were no private insurance price data available in some areas in which there are Commission on Profess ional Hosp ital Activities (CPHA) hospitals . Addition al data obtain ed to check whether the ran ges caused excessive elimination of data did not indicate a pr oblem, and the overall aver ages for the private-pay char ges and r eimbur sements seem to be in line with what would be expected based on the level of Medicare fees. Therefore, missing data were filled in by creating a second set of average private charges and reimbursements based on region al MSA and non-MSA area aggr egations. Selected areas contained one or mor e hospitals that agr eed to allow u se of their dischar ge abstr acts thr ou gh agr eements with the CPHA. Although the dischar ge abstr acts are not u sed in this investigation. charge and payment data were obtained only for the areas defined by this set of hospitals. Fr om 1988 thr ou gh 1990 Medicare fee screens wer e chan ged in Apr il r ather than on a calendar year basis. Medicare fees were set according to the calendar year beginning in Janu ary 1991. All charge and payment variables were deflated to 1988 levels by the phys ician services component of the Consu mer Pr ice Index (CPI). This has no effect on the estimated regression coefficient for the primary ind epend ent variable, the Medicare payment r ate, since the CPI provides a national rather than a regional adjustment. If the number of available claims was too small to construct a reliable measure of payment for the charge locality, the carrier payment level was used instead. Although we exper imented with s tatistical techniques that are designed to deal explicitly with endogenou s independent var iables, the data so ur ces available were not sufficient for producing stable results. Because of the limited number of su r vey r espondents who wer e nur ses, we estimated these four categories of wages for urban and rural areas in the nine regions of the country. See Note 8. Ibid . These declines may represent a shift in the mix of procedures within the group (that is, an increase in the relative proportion of less expensive procedures in the group), rather than a decline in real price for any individual procedure. One concern is that when the dependent variable is specified as billed charges, the R-sq u are valu e is near unity, equ aling .970 (compared with .437 for the other specification). This could be caused by the omission of some important independent var iables that are corr elated with both the pr ivate billed charge and the Medicare payment rate; the most likely are those representing the relative expense of differen t geographic areas. To investigate this fur ther , we used a Ramsey RESET test to look for evidence of omitted variables. Although the test found su ch evidence for both specifications, it was particularly strong when the dependent variable was defined as billed charges. Rice et al., “Physician Response to Medicare Payment Reductions.”

DATAWATCH

H E AL T H AF F AI R S - F a l l 1 9 9 6 Downloaded from content.healthaffairs.org by Health Affairs on January 21, 2016 by guest

225