Journal of Health Care Finance · Frederick, MD 21704, or call 1-800-234-1660. To place an order, call 1-800-638-8437. Subscribers in all countries other than the United States, Canada,

1 2 3 4 5

For more information resources, directories, articles, and a searchable version of Aspen Publishers’ full catalog,visit us at: www.aspenpublishers.com

Journal of Health Care Finance (USPS 048-090) (ISSN 1078-6767) is published quarterly by Aspen Publishers, 76 Ninth Avenue, New York, NY 10011. Copyright © 2011 CCH Incorporated. All rights reserved. www.aspenpublishers.com. Reproduction in whole or in part without permission is strictly prohibited. Postmaster: Send address changes to Subscription Dept. IP. P.O. Box 3000, Denville, NJ 07834.

Subscription rate is $345 (plus postage and handling)per year in the United States and Canada (four issues),payable in advance. The two-year subscription rate is$587, the three-year subscription rate is $828, and thecost of one issue is $104. Subscribers may specify anyissue to begin the subscription. Subscribers in theUnited States and Canada: Address inquiries toFulfi llment, Aspen Publishers, 7201 McKinney Circle,Frederick, MD 21704, or call 1-800-234-1660. To placean order, call 1-800-638-8437. Subscribers in allcountries other than the United States, Canada, andJapan: Address inquiries to Aspen Publishers, c/o Swets& Zeitlinger, P.O. Box 825, 2160 SZ Lisse, TheNetherlands, telephone 31 252 435111, fax 31 252415888.

Permission requests: For information on how toobtain permission to reproduce content, please go tothe Aspen Publishers Web site at www.aspenpublishers.com/permissions.

Purchasing reprints: For customized article reprints,please contact Wright’s Media at 1-877-652-5295 orgo to the Wright’s Media Web site at www.wrightsmedia.com.

Indexing: JHCF is indexed in AcademicSearch/CD-ROM, the Business Periodicals Index, theCumulative Index to Nursing & Allied HealthLiterature (CINAHL), EMBASE, Health Source, IndexMedicus, MEDLINE, MEDLARS, the UP-TO-DATELibrary/Health Services Management and WilsonBusiness Abstracts®, Wilson Business Abstracts FullText® by The H.W. Wilson Company. Currentlyavailable on CD-ROM and online via the WilsonWeb.For more information, please visit http://www.hwwilson.com.

“This publication is designed to provide accurate andauthoritative information in regard to the Subject Mattercovered. It is sold with the understanding that thepublisher is not engaged in rendering legal, accounting,or other professional service. If legal advice or otherexpert assistance is required, the services of acompetent professional person should be sought.”(From a Declaration of Principles jointly adopted by aCommittee of the American Bar Association and aCommittee of Publishers and Associations.)

Issue: Vol. 37, No. 3, 9900610016ISSN: 1078-6767Printed in the United States of America

The paper used in this publication meets therequirements of the American National Standard forInformation Sciences—Permanence of Paper for PrintedLibrary Materials, ANSI Z39.48-1992, effective withVolume 13, Issue 3.

Journal of Health Care Finance

Editors Editors Emeritus

James J. Unland, MBA Judith J. Baker, PhD, CPAPresident PartnerThe Health Capital Group Resource Group, Ltd.Chicago, IL Dallas, TX

Paul Gibson, Publisher William O. Cleverley, PhDJoanne Mitchell-George, Senior Managing Editor ProfessorElizabeth Venturo, Managing Editor Ohio State UniversityDom Cervi, Marketing Director Columbus, OH

Editorial Board

Editorial Board

Dana A. Forgione, PhD, CPA, CMA, CFE, Janey S. Briscoe Endowed Chair in the Business of Health, and Professor of Accounting, College of Business, University of Texas at San Antonio,TX.

Ellen F. Hoye, MS, Principal, Hoye Consulting Services, Elmhurst, IL

Daniel R. Longo, ScD, Professor and Director of Research, ACORN Network Co-Director, Department of Family Medicine, Virginia Commonwealth University, Richmond, VA

Kevin T. Ponton, President, SprainBrook Group, Hawthorne, NY

Elizabeth Simpkin, President, The Lowell Group, Inc., Chicago, IL

Elaine Scheye, President, The Scheye Group, Ltd., Chicago, IL

Pamela C. Smith, PhD, Associate Professor, Department of Accounting, The University of Texas at San Antonio, San Antonio, TX

Jonathan P. Tomes, JD, Partner, Tomes & Dvorak, Overland Park, KS

Mustafa Z. Younis, Professor of Health Economics & Finance, Jackson State University, School of Health Sciences, Department of Health Policy & Management, Jackson, MS

1 An Examination of Contemporary Financing Practices and the Global Financial Crisis on Nonprofi t Multi-Hospital Health Systems

Louis J. Stewart and Pamela C. Smith

25 Interactive Financial Decision Support for Clinical Research TrialsBenjamin Holler, Dana A. Forgione, Clinton E. Baisden, David A. Abramson,and John H. Calhoon

38 The Role of Financial Market Performance in Hospital Capital InvestmentKristin L. Reiter and Paula H. Song

51 Nursing Home Safety: Does Financial Performance Matter?Reid M. Oetjen, Mei Zhao, Darren Liu, and Henry J. Carretta

62 Cigarette Taxes and Respiratory Cancers: New Evidence from Panel Co-Integration Analysis

Echu Liu, Wei-Choun Yu, and Hsin-Ling Hsieh

72 Revisiting the Cost of Medical Student Education: A Measure of the Experience of UT Medical School–Houston

Elizabeth Gammon and Luisa Franzini

87 Cost-Volume-Profi t Analysis and Expected Benefi t of Health Services: A Study of Cardiac Catheterization Services

Mustafa Z. Younis, Samer Jabr, Pamela C. Smith, Maha Al-Hajeri, and Michael Hartmann

ContentsJHCF 37:3, Spring 2011

From the Editor—About This Issue

Once again, this issue of the Journal of Health Care Finance is illustrative of the breadth of topics we cover. We are always interested in new article ideas that directly or indirectly relate to health care fi nance. To submit ideas or articles, please send an email to: [email protected].

—James J. Unland The Health Capital Group 244 South Randall Road, Ste 124 Elgin, IL 60123 (800) 423-5157 healthfi [email protected]

iv Copyright © 2011 CCH Incorporated

1

An Examination of Contemporary Financing Practices and the Global

Financial Crisis on Nonprofi t Multi-Hospital Health Systems

Louis J. Stewart and Pamela C. Smith

This study examines the impact of the 2008 global fi nancial crisis on large US nonprofi t health systems. We proceed from an analysis of the contemporary capital fi nancing practices of 25 of the nation’s largest nonprofi t hospitals and health systems. We fi nd that these institutions relied on operating cash fl ows, public issues of insured variable rate debt, and accumulated investment to meet their capital fi nancing needs. The combined use of these three fi nancial instruments provided these organizations with $22.4 billion of long-term capital at favorable terms and the lowest interest rates. Our analysis further indicates that the extensive utilization of bond insurance, auction rate debt, and interest rate derivatives created signifi cant risk exposures for these health systems. These risks were realized by the broader global fi nancial crisis of 2008. Findings indicate these health systems incurred large losses from the early retirement of their variable rate debt. In addition, many organizations were forced to post nearly $1 billion of liquid collateral due to the falling values of their interest rate derivatives. Finally, the investment portfolios of these large nonprofi t health systems suffered millions of dollars of unreal-ized capital losses, which may minimize their ability to fi nance future capital investment requirements. Key words: derivatives, auction rate securities, variable rate demand obligations, nonprofi t, fi nancing .

This article examines the impact of the 2008 global fi nancial markets crisis on 25 of the largest nonprofi t

health systems in the United States. The delivery of 21st century health care serv-ices in the United States is an increasingly capital intensive enterprise. As a result, the nation’s nonprofi t hospitals and health sys-tems have a large and growing need for low cost capital fi nancing to fund their grow-ing capital budgets. 1 The need to replace and renovate aging facilities and a growing demand for new diagnostic and treatment equipment driven by a rapidly advancing medical technology increases demand for capital fi nancing. Furthermore, the escalat-ing need for large investments in informa-tion technology required to implement fully integrated electronic health records (EHRs) and computerized provider order entry sys-tems (CPOEs) also create a demand for signifi cant capital fi nancing. Few, if any, nonprofi t hospitals and health systems can

fund these capital needs from their operat-ing cash fl ows or capital fund-raising cam-paigns. 2 Moreover, all nonprofi t entities are constrained from seeking equity fi nancing due to their tax-exempt status. 3

The purpose of this article is to investi-gate the extent to which the nation’s larg-est nonprofi t health systems have utilized contemporary capital fi nancing instruments, including variable rate long-term debt and interest rate derivatives.

J Health Care Finance 2011; 37(3):1–24Copyright © 2011 CCH Incorporated

Louis J. Stewart, PhD, CPA, is an Associate Profes-sor of Accounting in the School of Business at Howard University in Washington, DC. He can be reached at [email protected].

Pamela C. Smith, PhD, is an Associate Professor of Accounting in College of Business at University of Texas at San Antonio in San Antonio, Texas. She can be reached at [email protected].

2 JOURNAL OF HEALTH CARE FINANCE/Spring 2011

Our exploratory analysis of the audited fi nancial statements and other public dis-closures proceed from an assessment of the pre-crisis ( i.e ., fi scal year (FY) 2007) fi nan-cial condition. Findings indicate that all of these nonprofi t organizations made sig-nifi cant use of variable rate debt and inter-est rate swaps. These health systems also invested signifi cant resources in various debt and equity securities as a means of meeting anticipated debt service requirements, fund-ing capital investment needs, and provid-ing a liquid reserve for normal operating contingencies. The primary effects of the 2008 global fi nancial markets crisis arose from the realization of many potential risks associated with the use of these innovative capital fi nancing instruments and their posi-tion as major investors in the global securi-ties market. As a result, many of the nation’s large nonprofi t health systems faced higher debt service payments, incurred large non- recurring losses, experienced sizeable unre-alized losses in their investment portfolios, as well as diminished liquidity. Analysis further indicates that these outcomes will diminish the access to long-term capital at the low costs that were previously available just a few years ago.

To our knowledge, we are the fi rst to document the combined use of variable rate long-term debt and derivative fi nancial instruments in large nonprofi t multi-hospital health systems. This analysis contributes to the discussion of the use of these long-term debt and risk management instruments in the municipal markets. The use of variable rate long-term debt and derivative fi nancial instruments by these organizations, com-bined with the volatile markets, raises con-cerns over the long-term fi nancial health of nonprofi t health providers. Policymakers

may begin to question whether regulation of the derivative markets is warranted for this sector. Future research should investigate the long-term effects of the use of variable rate long-term debt and derivative fi nancial instruments in the municipal markets. Addi-tional analysis of potential long-term use of derivative instruments may assist in creating sound policy to protect the fi nancial health of nonprofi t health systems.

The remainder of the article is organized as follows: the section entitled “Financial Management on the Eve of the Crisis” pro-vides an overview of the fi nancial condition and fi nancing behavior of our sample non-profi t health systems, including the use of derivative fi nancing instruments. The section entitled “The Financial Crisis” describes the aspects of the fi nancial crisis and auction failures that signifi cantly impacted the inter-est costs our sample health systems, while the last section offers concluding remarks and directions for future research.

Financial Management on the Eve of the Crisis

Extensive Municipal Debt Market Participation

The last two decades of the 20th century marked a fundamental change in the organi-zation of the nation’s health service deliv-ery system. Many community hospitals and academic medical centers sought the simul-taneous pursuit of vertical and horizontal integration. The objectives of these strate-gies were three-fold:

1. To achieve economies of scale through acquiring other hospitals;

2. To achieve economies of scope by acquiring physician practice groups,

The Global Financial Crisis on Nonprofi t Multi-Hospital Health Systems 3

nursing homes, and other dissimilar health care providers; and

3. To accumulate market power to facili-tate aggressive negotiations with third-party payers and managed care organi-zations on pricing and other terms of trade. 4

The growing capital budgets of contempo-rary nonprofi t health care enterprises relied on various forms of long-term debt instruments as their predominant mode of capital fi nanc-ing. Moreover, competitive pressures among institutional health care providers and rate containment pressure from aggressive pay-ers increases the demand for low cost capital fi nancing alternatives. The nation’s munici-pal securities markets have historically played a central role in providing long-term debt fi nancing for health care facilities at the lowest possible cost. Public issues of long-term tax-exempt debt have been and remain the largest source of low cost capital fi nanc-ing for most nonprofi t hospitals and health systems. 5 Research suggests that large non-profi t hospitals and integrated delivery sys-tems relied extensively on tax-exempt debt to meet their large and growing long-term capi-tal needs over the past two decades. 6 These needs included the fi nancing of extensive mergers and acquisition activities—which created large, multiple-hospital integrated service delivery systems—while allowing the replacement and expansion of aging facilities, and advancement in their infor-mation technology infrastructure. Many of the merger and acquisition activities that resulted in the development of large multi-hospital integrated delivery systems were fi nanced through the issuance of tax-exempt debt. Wedig and colleagues fi nd that the availability of tax-exempt fi nancing provides

nonprofi t organizations with their own tax-based incentives. 7 Their empirical fi ndings indicate that nonprofi t hospitals and health systems gain an indirect arbitrage from tax-exempt debt issuance in lieu of spending its own cash balances. Against this incentive to issue tax-exempt debt lies a collection of factors that cause the marginal cost of bor-rowing to increase as leverage increases. In addition, nonprofi t tax-exempt debt issu-ers are also subject to the Internal Revenue Service (IRS) regulatory project fi nancing constraint. 8 This constraint requires that the cash fl ows from the tax-exempt debt fi nanc-ing must be less than or equal to the invest-ment cash fl ows to capital projects that are consistent with the charitable purposes of the debt-issuing nonprofi t organization.

Figure 1 presents an overall fi nancial picture of our sample of 25 of the largest nonprofi t health systems. 9 We relied upon Modern Healthcare’s 2008 list of the ten largest health systems—in particular, the ten largest Catholic, non-Catholic religious, and secular not-for-profi t health care systems. 10 The data for this study were drawn from publicly available sources—including the audited fi nancial statements for fi scal years 2007 through 2009, as well as public bond offering statements. 11

On average, these health systems had 4,395 acute care beds for FY 2007. More-over, these 25 institutions refl ect a signifi cant percentage of the nation’s health care service delivery capacity. Collectively, their 109,865 acute care beds represent nearly 12 percent of the total staffed beds reported for the nation’s 5,815 hospitals. 12 Average operating revenue was $4.42 billion, and average total assets were $6.21 billion. The average ratio of long-term debt to total assets is 25.46 per-cent. The distribution of sample long-term


Figure 1. Financial Overview—Fiscal Year 2007 ($ in Millions)

No. of

Acute Care

Beds

Operating

Revenue

($)

Total

Assets

($)

Total Long-

Term Debt

($)

Long-Term

Debt to

Total Assets

Ascension Health 11,157 12,304 17,009 4,304 25.3%

Partners Health Care System 8,493 6,462 10,277 2,147 20.9%

Catholic Health Iniatives 7,787 7,731 11,387 3,403 29.9%

Catholic Health East 7,554 4,325 5,924 1,405 23.7%

Catholic Healthcare West 7,148 7,477 10,520 4,075 38.7%

Christus Health 5,580 2,762 4,297 938 21.8%

Trinity Health 5,193 6,110 8,933 2,117 23.7%

Sutter Health 5,173 7,651 9,048 2,366 26.1%

North Shore Long Island

Jewish Health System

5,164 4,172 3,551 904 25.5%

Providence Health & Services 4,622 6,348 7,773 1,518 19.5%

Bon Secours Health System 4,397 2,437 2,838 1,133 39.9%

Catholic Healthcare Partners 3,791 3,629 5,173 1,645 31.8%

SSM Health Care 3,478 2,521 3,795 1,141 30.1%

University of Pittsburg Medical

Center

3,334 6,277 7,354 2,193 29.8%

BJC Health Care 3,233 3,060 5,623 750 13.3%

Jefferson Health System 2,998 2,864 4,522 612 13.5%

Banner Health 2,978 3,355 5,188 1,833 35.3%

Texas Health Resources 2,946 2,429 4,050 1,133 28.0%

Cleveland Clinic Health System 2,837 4,733 6,884 1,460 21.2%

Advocate Health Care 2,813 3,457 5,264 787 15.0%

Baylor Health Care System 2,240 2,849 3,822 769 20.1%

Clarian Health 2,201 2,814 3,988 1,466 36.8%

Iowa Health System 2,092 1,874 2,329 509 21.9%

Ohio Health 1,783 1,765 2,326 631 27.1%

Northwestern Memorial

Healthcare

873 1,299 3,545 621 17.5%

Mean 4,395 4,428 6,217 1,594 25.46%

Median 3,478 3,457 5,188 1,405 25.30%

Maximum 11,157 12,304 17,009 4,304 39.92%

Minimum 873 1,299 2,326 509 13.34%


debt to total asset ratio ranges from a low of 13.3 percent to a high of 39.9 percent, with a standard deviation of 7.4 percent. Based on these results, these institutions are large-scale business enterprises whether meas-ured in terms of their economic resources (total assets) or volume of economic activity (operating revenues).

A review of the sample entity fi nancial statement footnotes indicates that tax-exempt debt generally exceeds 95 percent of the total long-term debt reported on each system’s FY 2007 balance sheet. These fi nancial data provide additional empirical support to prior empirical fi ndings that large multi- hospital chains use tax-exempt debt up to a tar-geted level. 13 This level represents a balance between the rising marginal cost of debt and the income tax–related incentives associ-ated with greater use of tax-exempt debt. In particular, our sample debt issuers utilize tax-exempt debt with a very narrow range. The widespread use of tax-exempt long-term debt by hospitals and health systems was an integral part of an extended period of US municipal securities market growth. The US municipal security market has undergone a period of remarkable growth over the past two decades. 14 Outstanding municipal debt securities have grown in volume from less than $400 billion in 1980 to over $2.4 trillion at the end of 2007, while more than $487 bil-lion of new bonds and notes were issued in 2007. 15

Variable Rate Debt

Nonprofi t health systems of all sizes were very often major issuers of variable debt. Fitch Rating reports that more than 63 per-cent of tax-exempt health care debt issues in 2004 were variable rate bonds. 16 D’Silva, Gregg, and Marshall 17 observed that the

primary appeal of tax-exempt variable rate debt to nonprofi t health care enterprises was a debt fi nancing instrument that had long maturity (20 years or more) with short-term interest rates. The variable debt issue’s long maturity allowed a debt issuer to build a debt amortization schedule that matched the peri-odic cash fl ows generated by the related cap-ital investment. Paying short-term interest rates enables variable rate debt issuers to pay lower interest rates by allowing the issuer to take advantage of the short end of the yield curve. McCue and Kim 18 observed that aver-age short-term interest rates of the fi rst fi ve years of the 21st century of 2.01 percent were less than half of long-term fi xed rate of 5.36 percent. Lower interest rates and peri-odic interest expense translated directly into improved profi t margins and enhanced cash fl ows that are essential to maintaining access to the public debt markets. 19

McCue and Kim 20 examined a variety of operational, environmental, and fi nan-cial factors that infl uenced an individual health enterprise’s decision to use variable rate debt. They found that facilities located in certifi cate of need states that possessed high case mix acuity, earned higher profi t margins, generated higher debt service cov-erage, and held less long-term debt, were more likely to issue variable rate debt. Clev-erly and Baserman 21 present a study of the nation’s fi ve largest (by revenue) investor-owned health systems and fi ve largest non-profi t health systems. They found that about 31 percent of long-term debt was variable rate fi nancing among investor-owned health systems, while more than 54 percent of long-term debt was variable rate fi nancing among nonprofi t health systems.

Figure 2 indicates that the mean ratio of variable rate debt to total long-term debt


among our sample of nonprofi t health sys-tems is 56.3 percent. This fi nding is consist-ent with the Cleverly and Baserman 22 average of 54 percent for the sample nonprofi t health

systems in their study. Our results, how-ever, exhibit considerable variance ranging from a low of 17.6 percent to a maximum of 94.3 percent with a standard deviation

Figure 2. Variable Rate Debt—Fiscal Year 2007 ($ in Millions)

Variable Rate

Long-Term

Debt ($)

Fixed Rate

Long-Term

Debt ($)

Total Long-

Term Debt

($)

Variable Rate

Debt / Long-

Term Debt %

Ascension Health 2,475 1,829 4,304 57.5%

Partners Health Care System 479 1,668 2,147 22.3%

Catholic Health Iniatives 1,775 1,628 3,403 52.2%

Catholic Health East 1,165 240 1,405 82.9%

Catholic Healthcare West 2,447 1,628 4,075 60.0%

Christus Health 796 142 938 84.9%

Trinity Health 1,315 863 2,117 62.1%

Sutter Health 417 1,949 2,366 17.6%

North Shore Long Island Jewish Health System 318 586 904 35.2%

Providence Health & Services 735 783 1,518 48.4%

Bon Secours Health System 402 731 1,133 35.5%

Catholic Healthcare Partners 924 721 1,645 56.2%

SSM Health Care 938 203 1,141 82.2%

University of Pittsburg Medical Center 913 1,280 2,193 41.6%

BJC Health Care 342 408 750 45.6%

Jefferson Health System 119 493 612 19.4%

Banner Health 1,512 321 1,833 82.5%

Texas Health Resources 429 704 1,133 37.9%

Cleveland Clinic Health System 633 827 1,460 43.4%

Advocate Health Care 692 95 787 87.9%

Baylor Health Care System 475 391 769 61.8%

Clarian Health 720 746 1,466 49.1%

Iowa Health System 480 29 509 94.3%

Ohio Health 500 131 631 79.2%

Northwestern Memorial Healthcare 427 194 621 68.8%

Mean 56.3%

Median 56.2%

Maximum 94.3%

Minimum 17.6%


of 21.4 percent. This large variation in the extent of variable rate debt utilization sug-gests strongly that there is considerable vari-ation among fi nancial management practice among similar sized fi rms facing presumably similar conditions.

Variable rate debt exposes debt issuers to the risk of rising interest rates. 23 To limit the potential negative cash fl ow impact from ris-ing interest rates, variable debt issuers can enter into a variety of interest rate hedges including swaps, caps, or collars. Stewart and Trussel 24 found that the extent of vari-able rate debt in a nonprofi t health system’s capital structure along with its size offered considerable explanatory power regarding its interest rate derivative use. Their study found that the most common approach is to fi x all or part of their variable-rate exposure through a ‘‘fl oating to fi xed’’ interest rate swap. Under a fl oating to fi xed swap, a varia-ble-rate borrower will pay a fi xed amount to the fi nancial institution, which in turn pays a fl oating amount to the borrower to settle the underlying variable-rate loan obligations. The payment amounts are computed as the product of the appropriate interest rate (fi xed or fl oating) and an agreed upon or “notional” amount. If the notional amount of the inter-est rate swap matches the principle value of the related debt and the interest rate reset mechanism of the debt is compatible with the swap’s fl oating rate index, then the borrower can convert their variable-rate cash fl ows into fi xed-rate cash fl ows without changing the structure of the underlying bond issue. An advantage of fi xing all of the exposure is that it results in a net fi xed-rate payment obligation by the issuer, which makes debt service planning easier. Moreover, this syn-thetic fi xed interest payment can afford a nonprofi t organization issuing variable rate

bonds with considerable interest cost sav-ings when compared to the interest cost associated with issuing traditional fi xed rate bonds. 25 The interest rate of any bond will always include a principal repayment default risk premium while the fi xed rate of a swap does not. Swap counter parties promise only to make periodic interest payments.

Figure 3 presents the total swap notional value, as well as the ratio of swap notional value to total variable rate debt. The notional principal amount of an interest rate swap is the specifi ed amounts on which the inter-est payments are based that are exchanged by the two counter parties. Hence, a swap’s notional value is a measure of its underlying hedging transaction.

The variability of interest rate derivative utilization in the sample is even more pro-nounced than the distribution of variable rate debt utilization among our sample health systems. The range is zero percent to 373 percent for the ratio of swap notional value to total variable rate debt range, with a mean value of 98 percent. A review of the foot-notes to the published fi nancial statements indicates that several organizations reported other reasons than hedging the interest risk exposure from their variable rate debt port-folio. Two health systems (Catholic Health East and Christus Health System) used fl oat-ing rate to fl oating rate swaps to hedge the basis risk associated with their fl oating to fi xed interest rate swaps. Catholic Health East also used fl oating to fi xed swaps to increase their exposure to the prevailing low market interest rates of the pre-2007 period and synthetically refi nance a portion of their fi xed rate debt portfolio as a broader inter-est cost minimization strategy. Neverthe-less, the mean ratio of swap notional value to total variable rate debt of 98 percent does


Figure 3. Interest Rate Swaps—FY 2007 ($ in Millions)

Notional Value

Swap Notional Value

to Total Variable

Rate Debt %

Ascension Health 1,229.8 49.7%

Partners Health Care System 619.6 67.1%

Catholic Health Iniatives 973.2 36.1%

Catholic Health East 1,215.0 233.7%

Catholic Healthcare West 1,547.1 65.8%

Christus Health 944.4 118.6%

Trinity Health 669.5 50.3%

Sutter Health 62.0 14.9%

North Shore Long Island Jewish Health System 223.3 70.2%

Providence Health & Services - -

Bon Secours Health System 467.8 111.6%

Catholic Healthcare Partners 1,266.0 137.0%

SSM Health Care 605.1 64.6%

University of Pittsburg Medical Center 661.3 67.9%

BJC Health Care 368.6 108.1%

Jefferson Health System 443.6 372.8%

Banner Health 1,426.2 94.3%

Texas Health Resources 60.0 14.0%

Cleveland Clinic Health System 560.3 63.7%

Advocate Health Care 623.0 90.0%

Baylor Health Care System 534.5 99.9%

Clarian Health 529.9 73.6%

Iowa Health System 624.5 129.8%

Ohio Health 184.7 36.8%

Northwestern Memorial Healthcare 558.5 130.8%

Total 16,397.9

Mean (3.27) 92.1%

Median (2.90) 70.2%

Maximum 33.90 372.8%

Minimum (49.90) 0.0%


reinforce the Stewart and Trussel 26 fi ndings that the proportion of variable rate debt in a nonprofi t health system’s long-term debt portfolio is positively related to the extent of its interest rate derivative utilization. In other words, interest rate derivatives are used pri-marily as hedging instruments for the inter-est rate volatility of variable rate debt.

Auction Rate Debt

Within this environment of overall munic-ipal bond market growth, the past 20 years were also marked by a period of fi nancial engineering of new fi nancing and risk man-agement instruments. In particular, there was a growing market in municipal secu-rities with long-term maturity dates and short-term (nine months or under) interest rate reset periods in recent years. Auction Rate Securities (ARS) and Variable Rate Demand Obligations (VRDO) currently comprise more than 25 percent of outstand-ing municipal debt. 27 The number of trans-actions in this sector reported to the MSRB Transaction Reporting Program increased from approximately 32,000 transactions per month in 2000 (approximately 6 percent of all transactions), to over 190,000 trades per month in 2007 (approximately 25 percent of all transactions). D’Silva et al . 28 argue the theory behind ARS and VRDO was a debt-fi nancing instrument that behaved like a long-term bond for an issuer, but resem-bles a short-term security for the investor. ARS and VRDO are similar in that they are long-term securities with short-term inter-est rates. In both types of securities, interest rates are reset periodically through programs operated by dealers on behalf of the issuers of the securities. Issuers are always able to issue these long-term debt obligations at par because they carry current market interest

rates. Nonprofi t health systems of all sizes were very often major issuers of fl oating rate municipal debt. Fitch Rating Agency reports that more than 60 percent of tax-exempt health care issues over the past fi ve years were long-term fl oating rate debt. 29

Figure 4 presents the ARS and VRDO debt for the sample. Our sample averages a fl oating rate debt mix of 57.5 percent, with the Iowa Health System reporting a high of 94.5 percent. A review of Figure 4 clearly indicates that these fl oating rate long-term debt instruments had become very popular sources of long-term debt fi nancing by 2007. Our sample health systems used signifi cant amounts of fl oating rate long-term debt in their capital structure.

One example from our sample of an entity using contemporary capital fi nancing is Northwestern Memorial Healthcare (NMH). NMH utilizes ARS and VRDO variable rate debt, four fl oating to fi xed interest rate swaps, letters of credit, and bond insurance. NMH issued $207.9 million in auction rate bonds in May 2004. The related public bond offering statement reports that a portion of the bond issue’s proceeds were combined with other funds to construct a $500 million state-of-the-art facility to house their expand-ing obstetrics and gynecology program. The bonds were marketed by J.P. Morgan Securi-ties and UBS Financial Services as the initial brokers/dealers for their respective portion of the total $207.9 million bond issue. NMH’s ARS bonds were insured by Financial Secu-rity Assurance, Inc. (FSA) and carried an AAA bond rating (S&P: AAA, and Moody: Aaa) as a result.

In December 2007, NMH issued $364.5 million of VRDO bonds, of which a portion of the bond issue’s proceeds were used to retire its outstanding Series 2004A fi xed rate


bonds. The Series 2004A fi xed rate bonds (which had a principle value of 194 mil-lion) carried an interest rate of 5.5 percent. J.P. Morgan Securities and UBS Financial Services marketed the VRDO bonds and

also served as the initial remarketing agents. NMH’s VRDO bonds were not insured and carried an AA bond rating (S&P: AA+, and Moody: Aa2). The Bank of Nova Scotia, J.P. Morgan Securities, and UBS Financial

Figure 4. Floating Rate Debt—FY 2007 ($ in Millions)

ARS Debt

($)

VRDO Debt

($)

Floating Rate

Debt / LT Debt

Ascension Health 1,504 917 56.3%

Partners Health Care System 450 473 43.0%

Catholic Health Iniatives 1,002 1,696 79.3%

Catholic Health East 520 37.0%

Catholic Healthcare West 1,716 635 57.7%

Christus Health 379 417 84.9%

Trinity Health 607 725 62.9%

Sutter Health 417 17.6%

North Shore Long Island Jewish Health System 318 35.2%

Providence Health & Services 732 3 48.4%

Bon Secours Health System 314 105 37.0%

Catholic Healthcare Partners 924 56.2%

SSM Health Care 493 444 82.1%

University of Pittsburg Medical Center 156 818 44.4%

BJC Health Care 244 97 45.5%

Jefferson Health System 119 19.4%

Banner Health 228 1,284 82.5%

Texas Health Resources 300 129 37.9%

Cleveland Clinic Health System 622 257 60.2%

Advocate Health Care 623 69 87.9%

Baylor Health Care System 242 293 69.6%

Clarian Health 720 49.1%

Iowa Health System 481 94.5%

Ohio Health 187 315 79.6%

Northwestern Memorial Healthcare 208 219 68.8%

ARS = Auction Rate Securities

VRDO = Variable Rate Demand Obligations


Services also served as the initial liquidity facility providers. As such, the three ini-tial liquidity facility providers collectively offered NMH the capacity to repurchase all of its VRDO bonds in response to potential investor demand.

In September 2003, NMH entered into two forward interest rate swaps with J.P. Morgan Securities and USB Financial Serv-ices respectively in order to hedge the vari-able rate debt exposure and interest rate risk of its August 2004 ARS debt issue. The two swap agreements provided that, on August 15, 2004, NMH would pay a fi xed rate of 3.64 percent and receive a fl oating rate based on one-month London Inter Bank Offered Rate (LIBOR) plus a fi xed constant based on a notional amount of $148.6 million. Furthermore, in July 2007, NMH entered into two forward interest rate swaps with J.P. Morgan Securities and USB Financial Services respectively in order to hedge the variable rate debt exposure and interest rate risk of its December 2007 VRDO debt issue. The two swap agreements provided that, on December 19, 2007, NMH would pay a fi xed rate of 3.89 percent and receive a fl oat-ing rate based on one-month LIBOR plus a fi xed constant based on a notional amount of $214.5 million.

NMH’s 2007 VRDO bond issue in combi-nation with the two fi xed to fl oating interest rate swaps exemplifi es the benefi ts associ-ated with a synthetic fi xed rate obligation. NMH was effectively able to refund a $194 million debt carrying a 5.5 percent interest rate with a comparable debt carrying a syn-thetically fi xed rate of 3.89 percent. Overall, NMH had a 68.8 percent ratio of variable rate debt to total long-term debt and 130.8 percent ratio of swap notional value to total variable rate debt.

ARS and VRDO have several signifi cant structural differences. 30 VRDO bonds are debt securities for which the interest rate is reset periodically, typically through a remar-keting process, or according to a specifi ed index. The bond’s demand feature permits the bondholder to require the purchase of the bonds by the issuer or by a specifi ed third party, either periodically at a certain time prior to maturity, or upon the occurrence of specifi ed events or conditions. This process is often referred to as “putting” a bond or exercising a “tender option.” 31 Interest rates are generally based on market conditions and the length of time until the bondholder can exercise the put option. Financially strong VRDO issuers are able to maintain suffi cient liquid assets to meet this potential current obligation. Alternatively, less liquid VRDO issuers will rely upon the fi nancial support of an external “liquidity facility” to meet these put options. An external liquidity facility may take the form of a letter of credit (LOC) or a stand-by purchase agreement (SPA) and are typically issued by a major commercial bank. Thus, the VRDO fi nancing instrument eliminates the possibility that the investor will be required to hold this security when that investor desires to liquidate its invest-ment for whatever reason. However, the cost of obtaining a LOC or SPA in an issuance of VRDO, along with risks associated with the elimination and/or renewals of the LOC or SPA, can make the cost of funds for an issu-ance of VRDO even more expensive than that of an issuance of long-term debt through an ARS mode.

Auction Rate Securities were fi rst devel-oped in 1984, and the ARS market has grown to $325 to $360 billion of securities, with state and local governments account-ing for about $166 billion of the outstanding


auction-rate debt. 32 ARS have a long-term (20 to 30 years) nominal maturity with interest rates reset through a modifi ed Dutch auction, at predetermined short-term intervals—usually 7, 28, or 35 days. Existing holders and potential investors enter a competitive bidding process through broker/dealer(s). The lowest bid rate at which all the shares can be sold at par establishes the interest rate, also known as the “clearing rate.” This rate is paid on the entire issue for the upcoming period. As a result, ARS trade at par and are callable at par on any interest payment date at the option of the issuer. Interest is paid in the current period based on the interest rate determined in the prior auction period. A “failed auction” can occur due to a lack of demand and no clearing bid received. In the event of a failed auction, existing bond-holders must hold their positions and receive the maximum rate as set in the offi cial state-ment until suffi cient bids are entered to set a clearing bid at the next auction. Although the underwriting broker/dealers are not required

to do so, they can provide a “clearing bid” to ensure the success of each auction and provide liquidity to investors who wish to sell. Until recently, failed auctions were very rare and were associated with down-grades in credit quality of either the indi-vidual issuer or the individual insurer of the issue. McCon nell and Sarretto 33 found that nearly 75 percent of these municipal ARS issues have received the highest credit rat-ing available from the major credit agencies, generally because of bond insurance. Bond insurance essentially enables debt issuers of varying investment quality to purchase the AAA bond rating of the bond insurer.

Figure 5 presents an overview of key fea-tures of ARS and VRDO fi nancing instru-ments. Auction rate debt became a very popular investment and fi nancing option during the fi rst decade of the 21st century. Over $200 billion of these fi nancing instru-ments were issued between 2000 and 2007. 34 2004 refl ected the peak of the municipal auc-tion rate bond market, with 755 new auction

Figure 5. Key Features—Variable Rate Demand Obligations and Auction Rate Securities

Feature VRDO ARS

Interest Reset Mechanism Remarketing Agent Dutch Auction

Interest Reset Period Daily or Weekly 7 to 35 days

Bondholders can redeem (put)

their bond at par on demand

Yes No

Imposition of Penalty Rate if

interest reset auction fails

No Yes

Penalty Rate Determination No Fixed Rate (up to 15%)

or Floating Rate

(% of LIBOR)

Need for Bank Letter of Credit

or Standby Purchase Agreement

Yes No




rate debt issues exceeding $39.6 billion. 35 This amount represented nearly 15 percent of the total outstanding auction rate debt out-standing at the 2007 year end. The tremen-dous popularity of tax-exempt auction rate debt during this period refl ects the fact that these debt instruments clearly met the needs of issuers, investors, and investment bank-ers in the stable low interest environment of the period. Auction rate debt offered various public utilities, student loan fi nancing agen-cies, large nonprofi t organizations, and other municipal debt issuers with the lowest cost long-term debt fi nancing available. Moreover, the use of auction rate long-term debt fi nancing also offered several other advan-tages in comparison to variable rate demand obligations. ARS have no “put” or tender feature, no letter-of-credit requirement, and no need for an annual short-term bond rat-ing, all of which increase the cost of issuance and maintenance of VRDO. Tax-exempt auction rate debt offered various nonprofi t hospitals and health systems with the low-est cost long-term debt fi nancing available. 36 Corporate and individual investors were attracted to ARS as short-term instruments because of the liquidity provided through the interest rate reset auction. This liquidity in conjunction with their exemption from income taxes and the highest possible credit rating help explain the popularity of ARS as a short-term investment for corporate treas-urers, high net-worth individuals, and mutual funds. D’Silva et al . 37 observed that one of the keys to the growth of the ARS market was the belief on the part of many investors that these instruments were equivalent to money market funds. In fact, many corporate investors classifi ed their ARS investments as cash equivalents. Finally, investment bankers received signifi cant compensation for their

role in the ARS market. An ARS broker/dealer will earn an ongoing fee as an auction agent as well as one-time fees for the origi-nal debt offering and subsequent secondary bond market sales.

Investor demand for ARS debt, therefore, rests solidly on investor perceptions that their ARS investments are readily market-able on demand via the auction mechanism. However, when a failed auction occurs, its reset rate reverts to a contractually specifi ed “default,” “penalty,” or “maximum” rate, which is established in the bonds’ offering statement. McConnell and Saretto 38 observed that these maximum rates function as an embedded interest rate cap within the ARS debt structure. There are two generic types of interest rate caps—fi xed and fl oating. As implied by their name, the computation of a penalty interest rate under a fi xed rate caps is straightforward. For example, the public bond offering statement for University of Pittsburgh Medical Center series 2003A $63 million bond issue specifi es an ARS maxi-mum rate of 15 percent. Floating rate caps, on the other hand, are calculated by means of a complex formula and reference rate. Northwestern Memorial Healthcare’s ARS fl oating rate cap used the one-month LIBOR as a reference rate. Its maximum rate was 125 percent of LIBOR as long as it maintains an AAA bond rating. The applicable percent-age rises to 150 percent or 200 percent if Northwestern’s bond rating falls to AA or A respectively. In the event of a failed auction, existing bondholders must hold their posi-tions and receive the maximum rate as set in the offi cial statement until suffi cient bids are entered to set a clearing bid at the next auc-tion. ARS debt issuers are obligated to pay interest to its bondholders based upon this maximum rate. Moreover, these increased


interest costs are, at most, only partially offset by the fl oating to fi xed interest rate swaps that have been widely entered into by many nonprofi t variable rate debt issuers as a hedge against rising market interest rates. Although the underwriting broker/dealers are not required to do so, they can provide a “clearing bid” to ensure the success of each auction and provide liquidity to inves-tors who wish to sell. From the inception of auction rate debt instruments in 1984 until fall 2007, failed auctions were very rare and were associated with downgrades in credit quality of fewer than 20 individual issuers. 39

The Financial Crisis

The 2008 global fi nancial crisis served to transform many of the latent risks associated with bond insurance, variable rate debt, and interest rate derivatives that were used by our sample health systems. Our analysis of entity fi scal year 2007 fi nancial statements indicates that 22 of the sample health sys-tems reported aggregate ARS debt totaling $12.2 billion. The February 2008 collapse of the municipal ARS debt market was the central event that negatively impacted the current fi nancial condition and future bond market prospects of all of our sample enti-ties. This event actually refl ected the culmi-nation of many factors over the previous two years. The two largest ARS brokers/deal-ers were Citigroup and UBS. These fi rms became a primary focus of the investigations being conducted by Security and Exchange Commission (SEC) Enforcement staff, as well as the state attorneys general in New York, Massachusetts, and Texas. The SEC Enforcement team investigation found these two fi rms misrepresented to their customers that ARS were safe, highly liquid investments

that were equivalent to cash or money mar-ket funds through their sales forces, market-ing materials, and account statements. Late in 2006, the SEC’s Division of Corporate Finance reaffi rmed its judgment that ARS were properly disclosed in SEC registrants’ fi nancial statements as long-term invest-ments and not cash equivalents. This reaf-fi rmation had the effect of decreasing the attractiveness of ARS to corporate treasurers seeking to “clean up” their balance sheets while earning a high interest rate on tempo-rary investments. When many corporations reclassifi ed the ARS investments from cash equivalents to investments on their 2007 fi nancial statements, many of these corpora-tions also sought to limit or reduce their ARS investments.

The marketability of ARS bonds was fur-ther impaired by the credit problems of the bond insurance industry. All 22 health sys-tems in this study that issued ARS bonds also purchased bond insurance for these bonds. Historically, MBIA, AMBAC Financial, and the other so-called monoline insurance com-panies earned a modest yet consistent profi t by insuring municipal bonds, which rarely default, and were characterized by relatively predictable risks. However, these monoline insurance companies sought to improve their profi tability by expanding their product lines into other areas of the world’s fi nancial markets. In 1998, they succeeded in obtain-ing New York state regulatory permission to insure the complex securities and deriva-tives that Wall Street created by packaging mortgages, including subprime ones, for investors. 40 As the nation’s real estate and mortgage markets boomed in the early years of the 21st century, so too did the revenues, profi ts, and market capitalization of the bond insurance companies. However, the 2007


subprime mortgage market decline created billions of dollars of potential claims that had been inadequately reserved for the bond insurance industry. As a result, AMBAC Financial was downgraded from AAA to AA, Financial Guaranty Insurance Co. was downgraded to A, and ACA Financial’s bond rating has fallen all of the way down to CCC. 41

The credit rating downgrades of many of these bond insurers had an immediate and negative impact on the municipal debt mar-kets. Institutional and corporate investors began to doubt the credibility of the mono-liners’ default guarantees on the bonds that they insured. Moreover, the general uncer-tainty and volatility that characterized the world’s fi nancial markets led many investors to forsake municipal bonds of perceived ris-ing risks for the safety of risk free US gov-ernment securities. 42 These events appeared to combine greatly to diminish investor demand for ARS bonds. Beginning on Thursday, February 7, 2008, ARS auctions began to fail en masse. Historically, the ARS brokers/dealers, such as Citicorp, USB, and others, had stepped into a potential failed auction and purchased any excess supply in order to clear such ARS interest reset auc-tions. The broader capital market meltdown put these fi nancial institutions and their capital under extreme stress. As a result, the auction-running banks refused to step in to bid on the excess ARS bond supply. 43 Eighty percent of these auctions failed on February 13, 2008. On February 20th, 62 percent of these auctioned failed (395 out of 641 auc-tions). As a point of comparison, there were a total of 44 failed auctions from 1984 until the end of 2007.

The impact of widespread auction failures had varying impacts upon the interest costs

of our sample nonprofi t health systems and hospitals. Both McConnell and Saretto 44 and Han and Li found that failed auctions were much more likely to occur for ARS bonds with fl oating rate caps than those with fi xed rate caps. Christus Health issued $379.3 million in ARS bonds in Novem-ber 2005. According to the related public debt offering statement, the interest rate on Christus Health’s ARS bonds were subject to a fi xed maximum rate cap of 12 percent. While Christus Health’s ARS bonds did not experience auction failures, the fi scal year weighted average interest rate on these bonds did increase from its fi scal year 2007 level of 3.53 percent to the fi scal 2008 level of 4.20 percent. On the other hand, both Northwestern Memorial Healthcare (a fl oat-ing rate cap) and University of Pittsburgh Medical Center (a fi xed rate cap) reported ARS auction failures. University of Pitts-burgh Medical Center saw the interest rate on its outstanding ARS bonds more than tri-ple from an average rate in 2007 of 3.72 per-cent to the contractually specifi ed auction failure fi xed cap rate of 15 percent. Accord-ing to a March 2008 public bond offering statement, University of Pittsburgh Medical Center had approximately $464 million of ARS bonds outstanding out of its total long-term debt of $2.43 billion as of December 31, 2007. Northwestern Memorial Health-care reported that a weighted average inter-est rate of 3.55 percent for its ARS debt for its fi scal year ending August 31, 2008. Its fi scal year 2008 average interest rate for its ARS debt was actually marginally lower than its fi scal 2007 average interest rate of 3.61 percent due to its fl oating rate cap.

ARS debt issuers sought other options in the face of failed auctions and dramati-cally increased interest costs associated


with their outstanding ARS debt. As shown in Figure 6, 21 systems in this study retired some or all of their outstanding ARS debt during 2008. The range of available debt retirement options depended on the conver-sion provisions of the underlying bond cov-enants and the issuer’s overall credit quality. In some cases, the bond covenants of certain ARS debt issues permitted its conversion to VRDO or fi xed rate interest modes. The conversion to VRDO mode will typically involve incurring the costs of securing a line of credit, standby purchase facility or other appropriate liquidity facilities necessary to meet potential bondholder redemption requests. Eleven health systems in this sam-ple exercised the conversion option in their ARS debt covenants to retire, at least, their outstanding ARS bonds in favor of VRDO and fi xed rate bonds.

If conversion to alternative debt instru-ments is not permissible, then the ARS debt issuer may consider an early retirement of its ARS debt. Such an early debt retirement can be fi nanced out of existing operating cash fl ows, if suffi cient; a bank line of credit; or a refunding bond issue. Evans 45 reports that cities, hospitals, and other municipal borrow-ers have now refi nanced approximately $96.2 billion of their $166 billion in auction rate bonds, which amount to about 58 percent of all auction-rate bonds. The current $3 billion a week average rate of municipal bond refi -nancing has slowed a bit from the $5.5 billion a week average in April and May. The abil-ity to retire outstanding municipal ARS debt is contingent upon the issuer’s credit quality and will involve considerable transaction costs. These transaction costs include pro-fessional fees for lawyers, accountants, and fi nancial advisers; loan origination fees; and other related costs. These debt conversion,

redemption, and refunding costs must be con-sidered and weighed against the high interest rates associated with a failed auction. Figure 6 results also indicate that 16 health systems in this sample paid off more than $5.7 billion in outstanding ARS bonds out of operating cash fl ows, short-term lines of credit, or the proceeds of refunding bonds.

The University of Pittsburgh Medical Center has an excellent credit rating (AA+). It launched a program to repurchase and retire all of its outstanding ARS bonds on Febru-ary 18, 2008. The fi rst step was to increase a previously existing line of credit from $300 million to $650 million in order to provide suffi cient funds to redeem the outstanding ARS bonds as soon as possible. Permanent fi nancing for the refunding of University of Pittsburgh Medical Center’s $464 million ARS bonds was affected through the issue of $491 million of fi xed rate bonds in March 2008. Northwestern Memorial Healthcare also has an AA+ credit rating. It used a line of credit draw to prepay $109.7 million of its outstanding ARS bonds in November 2008 during its fi scal 2009 reporting period. On January 13, 2009, it issued $207.4 million in VRDO bonds to repay its line of credit draw as well as prepay the remaining $109.7 mil-lion in outstanding ARS debt.

Prevailing fi nancial reporting standards for bond issuers classifi es the conversion as well as the refi nancing and retirement of outstanding bonds as a debt extinguish-ment . Under SFAS 140 (FASB, 2000) a debt extinguishment occurs when a bond issu-ing entity either pays off its bondholders in cash or converts its fi nancial obligation to its bondholders from one form ( i.e ., ARS debt) to another form (fi xed rate or VRDO bond). A loss on debt extinguishment is rec-ognized when the cash paid or fair value of


Figure 6. Retired ARS Bonds and Losses from Early Extinguishment of Debt Transactions ($ in Millions)

Retired

ARS Debt

($)

Paid

During

2008 ($)

Converted

to Fixed

Rate

Mode ($)

Converted

to VRDO

Mode ($)

2008 Debt

Extinguishment

Loss ($)

Ascension Health 1,453 1,118 335 23.3

Partners Health Care System 300 150 150 0.2

Catholic Health Iniatives 897 300 597 33.4

Catholic Health East 68 68 2.7

Catholic Healthcare West 1,387 27 615 745 68.2

Christus Health n/a

Trinity Health 607 607 15.9

Sutter Health 417 329 88 8.4



n/a

Providence Health & Services 721 335 165 221 11.2

Bon Secours Health System 314 314 19.9

Catholic Healthcare Partners 648 291 357 11.4

SSM Health Care 344 344 7.8

University of Pittsburg

Medical Center

484 484 6.2

BJC Health Care 244 244 5.0

Jefferson Health System n/a

Banner Health 228 228 21.6

Texas Health Resources 300 300 11.0

Cleveland Clinic Health System 622 622 16.5

Advocate Health Care 623 623 9.6

Baylor Health Care System 212 212 4.7

Clarian Health 720 720 3.8

Iowa Health System n/a

Ohio Health 187 187 5.8


Healthcare

208 208 4.4

Total 10,984 5,702 1,860 3,422 291.0

Mean 13.9

Median 9.6

Maximum 68.2

Minimum 0.2




debt securities transferred to the bondhold-ers is greater than carrying value of the extinguished bond. Figure 6 illustrates that 21 systems in this study recorded losses on debt extinguishment on their 2008 fi nancial statements. These losses range in amount from $216,000 to $68.2 million to with a median value of $9.6 million. The mate-riality of these 2008 debt extinguishment losses to the 2008 operating income of these health systems ranged from 56.2 percent to 1.8 percent. Finally, while many ARS debt issuers used interest rate swaps to create a synthetic fi xed rate debt from an ARS debt issue, the swap contract remains a separate contract with its own rights and obligations. Consequently, the retirement of an outstand-ing ARS debt issue whether by conversion, redemption, or an advanced bond refunding may also require the unwinding of a debt issuer’s interest rate swap position. Several health systems in this study included these costs in their computation of their loss on debt extinguishment.

The impact of 2008 global fi nancial cri-sis upon the VRDO bond issuers among our sample nonprofi t health systems was much more modest than that of ARS bonds. Many health systems suffered from remarketing failures during this time period. A remarket-ing failure occurs when VRDO bondholders tender their investments to a remarketing agent and the agent is unable to fi nd a buyer. Under these conditions, the issuing health system purchases the tendered bonds with a draw from its letter of credit. For example, during its fi scal year ending June 30, 2008, Christus Health had $514 million in VRDO bonds that failed to remarket. These bonds were purchased by various banks under standby purchase agreements. The agree-ments, which carry an interest rate ranging

between LIBOR and prime, provide for amortization periods of three to fi ve years with full payment due in 2012 through 2013. While such transactions did not generate the debt extinguishment losses or the multiplica-tive infl ation of interest related costs, VRDO remarketing failures could negatively impact a health system’s medium-term liquidity and solvency.

Many variable rate debt issuers found that the basis risk from their fl oating to fi xed swap could be a very substantial exposure that can undo its hedging charac-teristics. For issuers of variable rate bonds, the variable rates they received from their swap counter party, based on a percentage of the LIBOR, roughly matched the cost of variable rate bonds for more than fi ve years. This relationship diverged sharply in 2008 as the rates on variable rate munici-pal bonds increased and LIBOR fell. Spe-cifi cally, the ratio of short-term tax-exempt to taxable rates (Securities Industry and Financial Markets Association (SIFMA) tax-exempt index vs. one-month LIBOR) had averaged 108 percent over the last four months of 2008, signifi cantly higher than the 67 percent which is imbedded in many swap agreements. 46 This divergence ended the effectiveness of many fi xed payer inter-est rate swaps as an interest rate hedge. In addition, the drastic fall in the LIBOR received by nonprofi t health system vari-able rate debt issuers drastically reduced the swap’s valuation. Under these circum-stances, SFAS 133 mandates that the loss in the swap fair value must be recognized currently as a non-operating item. 47 Moreo-ver, standard interest rate swap contract lan-guage mandates that a swap counter party (variable rate debt issuer) whose swap valuation becomes negative (a liability) and


falls below a contractually specifi ed level must post liquid collateral to secure its posi-tion with its counter party (the investment banker). A posting of a material portion of a health system’s liquid assets can result in a reduction of its cash resources or lead to a violation of its bond covenant provisions. 48 Figure 7 indicates that 2008 swap losses for our sample health systems averaged over $74 million, ranging from $2 million to $311 million. Eight health systems were required to post collateral with the amount of posted liquid collateral ranging in value from $36.9 million to $273 million.

Figure 8 indicates that 2008 was a very diffi cult year for the investment portfo-lios of our sample health systems. Median and mean percentage investment losses exceeded 15 percent. All but one health sys-tem reported negative returns ranging from 6 percent to 28 percent of pre-2008 invest-ments level. The role of alternative invest-ments in explaining the investment results is not clear. These losses in the investment portfolios have the direct effect of materially diminishing their available reserve funds and potential liquid collateral. Lower invest-ment balances indirectly inversely impacts the liquidity, solvency, and capital ratios that are used by investors and rating agencies to evaluate the risk and credit worthiness of health system debt issues. It is not yet clear how these losses will ultimately impact the ability of these 25 health systems to attract the long-term debt fi nancing needed to sup-port their large and growing future capital expenditure needs.

Concluding Remarks

Recent trends in facility construction across the nation appear to lag the reported

increases in inpatient admissions and outpa-tient services that have been driven primarily by population growth on the West Coast and in the South. Moreover, the nation’s aging population will create an increasing demand for health care resources beyond the impact of this population increase. Additionally, advances in pharmaceuticals and medical technology will continue to increase the type and number of services available to patients. As a result, the nation’s nonprofi t hospitals and health systems have a demand for low cost capital fi nancing to fund these projects. 49 Billions of dollars for capital spending for facilities and equipment are needed to sup-port this technologically driven growth in service utilization. Both internal and exter-nal pressures will spur signifi cant invest-ments in information technology to support the clinical and administrative infrastructure of the nation’s health care providers. Many nonprofi t health systems are using creative fi nancing to meet such demands.

The nonprofi t health systems in this study made extensive use of credit enhancements, variable rate debt, and interest rate deriva-tives. By 2007, these entities obtained $22.4 billion of variable rate long-term debt at favorable terms and the lowest interest rates. Recently, most of these health systems found their use of bond insurance, variable rate debt, and interest rate derivatives exposed them to a variety of entity-specifi c and systemic risks. The use of bond insurance exposes credit enhanced long-term debt issuers to the bond insurers’ credit problems. Beginning in the winter of 2008, dramatically diminished investor demand led to signifi cantly increased interest costs for many health systems, costly debt refunding, unscheduled swap termina-tion payments, and large collateral postings. Eighty-eight percent of our sample health


Figure 7. FY 2008 Interest Rate Swaps Losses and Related Collateral Postings ($ in Millions)

FY 2008 Loss in

Swap Fair Value

Posted Swap

Collateral

Ascension Health 39.3

Partners Health Care System 73.0

Catholic Health Iniatives 64.7

Catholic Health East 113.9 98.0

Catholic Healthcare West 17.9

Christus Health 53.8

Trinity Health 9.5

Sutter Health 61.0



31.2

Providence Health & Services -

Bon Secours Health System 32.5

Catholic Healthcare Partners 83.1 83.0

SSM Health Care 94.4

University of Pittsburg Medical

Center

31.8 55.5

BJC Health Care 81.2 36.9

Jefferson Health System 18.6

Banner Health 310.5 273.0

Texas Health Resources 2.1

Cleveland Clinic Health System 116.0 105.0

Advocate Health Care 93.5 69.0

Baylor Health Care System 39.0

Clarian Health 249.4 45.0

Iowa Health System 101.2

Ohio Health 15.3


Healthcare

31.4

Total 1,764.3 765.4

Mean 73.5 95.68

Median 57.4 76.00

Maximum 310.5 273.00

Minimum 2.1 36.90


Figure 8. FY 2008 Investment Losses ($ in Millions)

Pre-September

2008 Investment $

Investment

Gains/(Losses) $

Investment

Gains/(Losses) %

Ascension Health 7,314 (1,063) -14.5%

Partners Health Care System 4,996 (429) -8.6%

Catholic Health Iniatives 5,078 (623) -12.3%

Catholic Health East 1,557 (320) -20.6%

Catholic Healthcare West 4,476 (387) -8.6%

Christus Health 1,550 (225) -14.5%

Trinity Health 4,181 (645) -15.4%

Sutter Health 3,025 8 -0.3%



1,268 (148) -11.7%

Providence Health & Services 2,607 (413) -15.8%

Bon Secours Health System 949 (109) -11.5%

Catholic Healthcare Partners 2,042 (350) -17.1%

SSM Health Care 1,894 (356) -18.8%

University of Pittsburg

Medical Center

3,240 (597) -18.4%

BJC Health Care 3,106 (790) -25.4%

Jefferson Health System 1,511 (92) -6.1%

Banner Health 2,238 (536) -23.9%

Texas Health Resources 1,527 (91) -6.0%

Cleveland Clinic Health

System

2,914 (509) -17.5%

Advocate Health Care 2,780 (530) -19.1%

Baylor Health Care System 1,489 (188) -12.6%

Clarian Health 1,445 (384) -26.6%

Iowa Health System 743 (211) -28.4%

Ohio Health 1,171 (236) -20.2%


Healthcare

1,845 (171) -9.3%

Mean (376) -15.3%

Median (356) -15.4%

Maximum 8 0.3%

Minimum (1,063) -28.4%


systems incurred more than $291 million of losses from the early retirement of their variable rate debt. In addition, seven sample organizations were forced to post liquid col-lateral in an aggregate amount of $709 mil-lion as a result of the falling values of their interest rate derivatives. The median value of the posted collateral was $83 million.

It may be that the contemporary ARS and VRDO fi nancing instruments are inherently unstable and structurally unsound. It may not be possible for the same debt security to be a long-term obligation for an issuer and a short-term investment for an investor. 50 How-ever, if these debt instruments are to be used successfully, health care system treasurers must learn to fully evaluate the potential costs and other exposures associated with variable rate debt issues and interest rate derivatives. Debt issuing and carrying costs for ARS and VRDO debt instruments may include broker-age fees, remarketing fees, auction fees, and liquidity provider fees, as well as periodic scheduled principle and interest payments. Basis risk, unscheduled termination pay-ment risk, and other risks associated with the use of interest rate swap should be identifi ed explicitly. Where possible, provisions of the underlying swap contract such as the selec-tion of reset mechanisms and market interest indices should be negotiated to minimize if not eliminate these risks.

In any event, health system treasurers and fi nancial executives must acquire the capac-ity to independently model future debt serv-ice requirements and derivative valuations

under a variety of conceivable scenarios. Health care managers cannot delegate their due diligence duties to the same investment bankers who stand to benefi t fi nancially from the use of variable rate debt and interest rate derivative instruments. Those charged with governance responsibility over these organizations must insist upon the develop-ment, implementation, and, maintenance of policies regarding the use of variable rate debt and derivative fi nancial instruments. These policies must be fi rmly grounded in an overall policy governing the use of debt in fi nancing the organization’s capital needs. A corporate debt management policy should include a defi nition of the organization’s overall debt capacity, a description of the amount and type of permissible debt instru-ments, and its strategy for the use of interest rate derivatives in the context of a corporate risk management program. This corporate debt management program must be fully integrated into an enterprise-wide strategic plan to insure that priority capital needs are funded. Analysis of company treasurers’ selection of fi nancing and risk management mechanisms used to meet capital fi nanc-ing needs should be investigated in future research. Finally, appropriate controls and periodic interim reporting must be imple-mented to insure that corporate policies concerning the use of variable rate debt and interest rate derivatives are followed. This may ultimately lead to the establishment of industry policies within the municipal debt markets.

REFERENCES

1. Allen, M, “FMA Testimony to House Financial Services Committee—Municipal Bond Issues,” May 21, 2009, available at: http://www.hfma.

org/Knowledge-Center/Finance-and-Business-Strategy/HFMA-Testimony-to-House-Financial-Services- Committee—Municipal-Bond-Issues/.


2. Carlson, D, “Access to Capital—A Growing Concern / Response from a Feature Author,” Frontiers of Health Services Management , 21(2): 153 (2004).

3. Robinson, J, “Capital Finance and Ownership Conversions in Health Care,” Health Affairs , 19(1): 56–71 (2000).

4. Conrad, DA, Dowling, WL, “Vertical Integra-tion in Health Services: Theory and Manage-rial Implications,” Health Care Management Review , 15(4): 9–22 (1990).

5. Carpenter, C, McCue, M, Hossack, J, “Asso-ciation of Bond Market, Operational, and Financial Factors with Multi Hospital System Bond Issues,” Journal of Health Care Finance , 28(2): 29–37 (2001).

6. Id .; Cleverly, W, Baserman, S, “Patterns of Financing for the Largest Hospital Systems in the United States,” Journal of Healthcare Management , 50(6): 361–365 (2005).

7. Wedig, GJ, Hassan, M, Morrisey, MA, “Tax-Exempt Debt and the Capital Structure of Nonprofi t Organizations: An Application to Hospitals,” Journal of Finance , 51(4): 1247–1283 (1996).

8. Id. 9. Our list of 25 of the largest health care sys-

tems includes those for which we were able to obtain system-wide fi nancial statement data.

10. Evans, M, Galloro, V, “Clouds Gathering?” Modern Healthcare , 38(24): 40–46 (2008); American Hospital Association (AHA), Fast Facts for US Hospitals , Chicago, IL, Health Forum LLC (2009).

11. Public bond offering statements are avail-able from several Web sites, including www.dacbond.com , www.munios.com , and www.msrb.org.

12. American Hospital Association (AHA), Fast Facts for US Hospitals , Chicago, IL, Health Forum LLC (2009).

13. Supra , n.7. 14. Sirri, ER, Testimony Concerning Municipal

Bond Turmoil: Impact on Cities, Towns and States, Testimony Before the committee on Financial Services, US House of Represen-tatives (Mar. 12, 2008), available at http://www.sec.gov/news/testimony/2008/ts031208ers.htm.

15. Hildreth, WB, Zorn, CK, “Evolution of the State and Local Government Municipal Debt Market,” Public Budgeting and Finance , 25(4): 127–153 (2005).

16. Fitch Ratings, Investment and Debt Portfo-lio Trends of Hospitals and Health Systems 1993–2003, New York, NY: Fitch Rating Agency (2005).

17. D’Silva, A, Gregg, H, Marshall, D, “Explain-ing the Decline in the Auction Rate Securities Market,” Chicago Fed Letter Essays on Issues , no. 256, Nov. 2008.

18. McCue, M, Kim, T, “Evaluating the Factors Underlying Variable Rate Debt,” Health Care Management Review , 32(4): 300 (2007).

19. Martin, L, “Annual Sector Outlook for Not-for-Profi t Healthcare for 2010,” Moody’s Investors Services US Public Finance Outlook , 1–31 (Jan. 2010).

20. Supra , n.18. 21. Cleverly, W, Baserman, S, “Patterns of Financ-

ing for the Largest Hospital Systems in the United States,” Journal of Healthcare Man-agement , 50(6): 361–365 (2005).

22. Id. 23. Id. 24. Stewart, LJ, Trussel, J, “The Use of Interest

Rate Swaps by Nonprofi t Organizations: Evi-dence from Nonprofi t Health Care Providers,” Journal of Health Care Finance , 33(2): 6–22 (2006).

25. Venkataramani, P, Johnson, T, O’Neal, P, Poindexter, V, Rooney, J, “The Effect of Inter-est Rate Derivative Transactions on Debt Savings for Not-for-Profi t Health Systems,” Journal of Health Care Finance , 33(2): 23–38 (2006).

26. Supra , n.24. 27. Supra , n.15. 28. Supra , n.17. 29. Supra , n.16. 30. California Debt and Investment Advisory

Commission, Auction Rate Securities: An Issue Brief (Aug. 2004).

31. Sultzberger, F, Flynn, A, “Lessons from Tough Times: Understanding VRDO Failures,” The Bond Buyer (July 21, 2008).

32. Supra , n.17. 33. McConnell, J, Sarretto, A, “Auction Failures

and the Market for Auction Rate Securities,”


Journal of Financial Economics , 97(3): 451–469 (2010).

34. Id. 35. Id. 36. McCue, M, Peterman, J, “What Is the Hospital

Industry’s Exposure from the ARS Collapse?” Healthcare Financial Management , 63(10): 80–87 (2009).

37. Supra , n.17. 38. Supra , n.33. 39. Supra , n.17. 40. Satow, J, “Next Big Crisis Could Involve Bond

Insurers Finance Costs Could Rise for City, State,” NY Sun (Jan. 31, 2008) available at http://www.nysun.com/business/next-big-crisis-could-involve-bond-insurers/70488/ .

41. Howard, J, “Municipal Bond Insurance—Recent Rating and Insurer Actions,” WM Financial Strategies (2009).

42. Han, S, Li, J, “Liquidity, Runs, and Secu-rity Design: Lessons from the Collapse of

the Auction Rate Municipal Bond Market,” unpublished Working Paper (2008).

43. Supra , n.17. 44. Supra , ns.33, 42. 45. Evans, M, “Auction-Rate Endgame: Restruc-

turing Debt Has Cost Hospitals Millions,” Modern Healthcare , 38 (48): 42 (2008).

46. Steingard, D, “Interest Rate Swaps Cause New Liquidity Stresses for Some Healthcare, Higher Education, and Other Nonprofi t Bor-rowers, Feb. 2009 Comment Letter New York NY: Moody’s Investors Service (2009).

47. Financial Accounting Standards Board (FASB) “Statement of Financial Accounting Standards No. 133: Accounting for Deriva-tive Instruments and Hedging Activities,” Stanford, CT, Financial Accounting Founda-tion (1998).

48. Supra , n.46. 49. Supra , n.1. 50. Supra , n.17.

25

Interactive Financial Decision Support for Clinical Research Trials

Benjamin Holler, Dana A. Forgione, Clinton E. Baisden, David A. Abramson, and John H. Calhoon

The purpose of this article is to describe a decision support approach useful for evaluating proposals to conduct clinical research trials. Physicians often do not have the time or background to account for all the expenses of a clinical trial. Their evaluation process may be limited and driven by factors that do not indicate the potential for fi nancial losses that a trial may impose. We analyzed clinical trial budget templates used by hospitals, health science centers, research universities, departments of medicine, and medical schools. We compiled a databank of costs and reviewed recent research trials conducted by the Department of Cardiothoracic Surgery in a major academic health science center. We then developed an interactive spreadsheet-based budgetary decision support approach that accounts for clinical trial income and costs. It can be tailored to provide quick and understandable data entry, accurate cost rates per subject, and clear go/no-go signals for the physician. Key words: clinical research trials, decision support, cardiothoracic surgery, pharmaceuticals, biotechnology, fi nancial management, cost.

Clinical Trials are the key to the development of new drugs. Over $24 billion in 2005 was spent on

clinical research trials, with a single drug costing approximately $800 million. The physicians, through their time and fi nancial resources consumed, incur 30 percent of that cost, which is typically funded under con-tract with a sponsor such as a governmental agency, pharmaceutical corporation, or bio-technology fi rm. Fewer than 90,000 physi-cians participate in clinical research trials. This equates to $7.2 billion dollars being distributed to 90,000 physicians who must properly manage their fi nances in order to conduct the clinical trials on a cost-effi cient basis. Current estimates project research costs rising at 4.6 percent a year, reaching $32.1 billion by 2011. BCC Research, a lead-ing information resource producing high-quality market research reports, estimates a growth rate of 5.8 percent. 1 This means that by 2011, more than 13,000 research trials will take place within the United States. 2 Although research companies are expect-ing signifi cant growth in this market, it may only be feasible in today’s economy if phy-sicians can fi nd ways to lower the cost of

Benjamin Holler, BBA , is an Honors Graduate of the Col lege of Business at the University of Texas at San Antonio.

Dana A. Forgione, PhD, CPA, CMA, CFE , is the Janey S. Briscoe Endowed Chair in the Business of Health and a Professor of Accounting in the College of Business at the University of Texas at San Antonio.

Clinton E. Baisden, MD, FACS , is a Clinical Profes-sor and Director of Clinical Research & Faculty Sup-port for the Department of Cardiothoracic Surgery in the School of Medicine at the University of Texas Health Science Center at San Antonio.

David A. Abramson, MBA, is the Administrator for the Department of Cardiothoracic Surgery in the School of Medicine at the University of Texas Health Science Center at San Antonio.

John H. Calhoon, MD , is the Rowena C. Gorman Distinguished Chair in Pediatric Cardiac Surgery, President’s Council Chair in Surgery and Professor and Head of the Department of Cardiothoracic Sur-gery in the School of Medicine at the University of Texas Health Science Center at San Antonio.

Acknowledgement: The authors thank everyone in the Department of Cardiothoracic Surgery of the School of Medicine at the University of Texas Health Science Center at San Antonio for their valuable time, assist-ance, and support throughout this project, as well as the participants in the Cardiothoracic Research Meet-ings for their valuable comments and feedback.



conducting these new trials. The need for effective fi nancial management tools in the health care professions is greater than ever. With the proper use and application of such tools, and special consideration of their lim-itations, interactive fi nancial decision sup-port systems for clinical research trials can give a major advantage to physicians and assist in improving the allocation of medi-cal services within the economy.

Clinical Research Trials

Clinical research trials are conducted to evaluate the safety and effectiveness of new medical products and procedures. Research physicians or investigators seek patient vol-unteers to register for a study. The size of the trial depends on the particular testing phase of the product. Often, when initial product safety and effi cacy results are found to be favorable, the number of participating vol-unteers for subsequent trials will increase. Clinical trials can be conducted at a variety of centers and their scope may range from a single center to multiple centers worldwide.

Types

The US National Institutes of Health (NIH) has organized clinical research trials into six different types: 3

• Prevention Trials: designed to look for better ways to prevent a disease or its recurrence. This may include testing medicines, vitamins, vaccines, miner-als, or lifestyle changes; • Screening Trials: designed to identify the best way to detect certain diseases or health conditions; • Diagnostic Trials: conducted to develop better tests or procedures to

accurately diagnose a particular disease or condition; • Treatment Trials: designed to test experimental treatments, new combi-nations of drugs, or new approaches to surgery or radiation therapy; • “Quality of Life” Trials: designed to explore ways to improve comfort and the quality of life for individuals with a chronic illness; and • Compassionate Use Trials: designed to provide experimental therapeutics, prior to fi nal US Food and Drug Administra-tion (FDA) approval, to patients whose options with other remedies have been unsuccessful. Usually, case by case approval must be granted by the FDA for such exceptions.

Three Phases

New drugs must pass through three key steps, or phases, and many years of testing before they reach the market. Once they have passed through Phases I, II, and III, the national regulatory authority usually approves them for general public use. The three phases involved in a pharmaceutical drug test, as described by the US Depart-ment of Health and Human Services, 4 are as follows.

• Phase I: researchers determine dosing, document how a drug is metabolized and excreted, and identify side effects; • Phase II: researchers gather further safety data and preliminary evidence of the drug’s effi cacy. During this phase, the drug is studied in a larger number of people with the disease. This phase further tests the product’s effectiveness, monitors side effects, and, in some cases, compares the product’s effects to

Interactive Financial Decision Support for Clinical Research Trials 27

a standard treatment, if one is already available; and • Phase III: groups of physicians recruited by the sponsoring drug com-pany conduct research. It is the most expensive phase for the physicians. The sponsoring company pays the physi-cians a set amount of money per partici-pating subject in the trial.

The amount of sponsored funding, how-ever, may not be enough to cover the actual costs incurred by the physicians in conduct-ing the clinical trial. Thus, it is in the physi-cian’s interest to have effective tools that will help to calculate the expected work effort and costs required per subject. These costs can then be compared to the payment rates offered by the sponsor and can facilitate con-tract agreement.

History

In the year A.D. 1025, Avicenna published a book entitled, “The Cannon of Medicine.” It provides the fi rst known mention of the concept of clinical trials. This book sets the foundation of rules that defi ne clinical research trials to this day. Avicenna’s rules included the use of controlled experiments for testing drugs. He wrote an amazingly unambiguous guide for discovering the effectiveness of medical drugs through sys-tematic experimentation. His rules are listed below. 5

1. The drug must be free from any extra-neous, accidental quality;

2. The drug must be used on a simple, not a composite, disease;

3. The drug must be tested with two con-trary types of diseases, because some-times a drug cures one disease by its

essential qualities, and another by its accidental ones;

4. The quality of the drug must corre-spond to the strength of the disease. For example, there are some drugs whose “heat” is less than the “cold-ness” of certain diseases, so that they would have no effect on them;

5. The time of action must be observed, so that essence and accident are not confused;

6. The effect of the drug must be seen to occur constantly or in many cases, for if this did not happen, it was an acci-dental effect; and

7. The experimentation must be done with the human body, for testing a drug on a lion or a horse might not prove anything about its effect on man.

One of the fi rst and most famous clinical trials was conducted in 1747 by James Lind, who discovered that the cure for scurvy was citrus fruits. 6 While working as a surgeon on a British ship, he identifi ed 12 men with scurvy, and divided them into six pairs. Some were given cider, others seawater, others a mixture of garlic, mustard, and horserad-ish, others spoonfuls of vinegar, and others oranges and lemons. The sailors who were fed the citrus fruits experienced a remark-able recovery. While the benefi ts of lime juice had been known for centuries, Lind defi nitively demonstrated the superiority of citrus fruits over all other proposed cures for scurvy. 7

Challenges

Economic incentives are playing an ever-increasing role in the environment of health care. While physicians are trained to not


allow economics to affect their clinical deci-sion making, they are nonetheless respon-sible for running a business. That includes oversight of the managerial tasks of human resources, legal and regulatory compliance, marketing, fi nance, accounting, budgets, coding and billing, collections, and informa-tion systems needed to operate a successful medical practice. They also make the clini-cal decisions that drive the majority of costs incurred by hospitals. Yet, many, if not most, medical schools do not equip their students with the knowledge or skills needed to run a business organization.

During clinical pretrial discussions with a sponsoring agency, the physician’s com-pensation is addressed. The amount a phar-maceutical company is willing to pay is typically expressed as a fi xed amount per study subject enrolled. This fi gure then needs to be compared to the physicians known or estimated expenses for conducting the trial. Since the physician frequently does not have a systematic way to accumulate and summa-rize the projected expenses for a clinical trial, the result is often a premature acceptance, or denial, of a sponsor’s proposal. In essence, the current process is often little more than an educated guesstimate. The consequences can be a fi nancial loss for the physician or the physician’s group practice. This may impinge on other important clinical activi-ties, resulting in the closing of a program, needless uncertainty, worry, and stress, and limitations to future research activity that may be important to society.

Interactive Spreadsheet for Decision Support

To address these issues, we designed an interactive spreadsheet model that readily

accounts for all of a physician’s projected clinical trial costs. It provides for quick and straightforward data entry, complete and accurate measures of the costs per subject, comparison to the costs of standard-of-care (SOC) protocols, and affords the physician straightforward go/no-go decision support for a clinical trial. This allows physicians to better manage their resources in a more effi -cient and effective manner. While there are many fi nancial decision support tools used by hospitals and other health care providers, we found in general that they were not easily adaptable to a specifi c fi eld. They tended to use broad headings, such as “services,” “sub-ject tests,” and “subject visit expenses,” and had few descriptions of the costs under such headings. 8 We concluded that a more useful tool was needed.

Model Development

We relied on several sources of information to develop our interactive spreadsheet, includ-ing previously conducted research trials, ongo-ing trials, and fi nancial tools used by other organizations. We worked with a new clini-cal trial sponsored by a large pharmaceutical company at a major academic medical center to test and refi ne our model. While prepara-tions for that new trial were being made, the study coordinator and the physicians involved provided us with prior estimates of how long the trial was expected to last and their expected time involved for each specifi c task. During the course of the succeeding nine months of the trial, the study coordinator and physicians randomly monitored their actual time usage for completing the tasks in various parts of the trial. This information gave us both prior esti-mates and actual results to use for comparison in our model.


Key Features

A detailed budget that accounts for and justifi es each cost item is an essential fea-ture of a well-designed system, and is criti-cal for the acceptance decision regarding clinical trials. 9 Our interactive spreadsheet is divided into ten main sections. Each section is dedicated to a particular task with lists of its related costs. The sections include:

1. Protocol; 2. Pre-screening visit; 3. Baseline visit; 4. Diagnostics; 5. Subject visits; 6. Unscheduled visits; 7. Closing visit; 8. Consultant fees; 9. Close-out costs; and 10. Miscellaneous.

An optional procedure section can also be used to refl ect the costs of required surgical or other procedures for a specifi c depart-ment. One fi nal section provides a conven-ient summary of the information from the previous ten sections, along with any insti-tutional overhead charge rates, the sponsor’s payment rates, and a clear go/no-go indica-tor, for quick reference.

Each of the ten main sections has similar, important features. First, is the ability to tag any cost item as supporting the SOC pro-tocol. The SOC tag signifi es that physician compensation will be made by the patient’s third-party payer and is thus deducted from the total research cost. The second common feature is the ability to tag any expense as a fi xed or variable cost and include a corre-sponding total or per-unit cost rate, as appro-priate for each department. This allows for accurate cost accumulation and automatic

adjustment for changes in the number of par-ticipating patients.

Each cost item is initially recorded as an estimate, and then actual costs are included as the trial progresses. The estimated cost is a forecast derived from the researcher’s previous experience with research trials. By comparing the initial estimate to the fi nal actual costs (on both an aggregate and detailed level), the researcher can quickly evaluate their estimation skills and gain insights for making better cost estimates in the future.

The Ten Main Sections

The fi rst step in any research trial is to write and submit a protocol for approval. Fol-lowing this is the process of writing a review paper on the fi nal results obtained. Our pro-tocol section includes these costs as well as any institutional review board fees, pharmacy set-up costs, and the personnel time used in this process—including costs of the primary investigator, co-primary investigator, author, co-author, and research coordinator, among others. The costs are determined using each individual’s salary rate and time spent on each task.

Before a patient can participate in a clini-cal research trial, the patient must fi rst qual-ify via a pre-screening visit and then follow up with the introductory baseline visit. The tests and tasks conducted during these two visits, similar to a physical examination ( see Figure 1), are recorded in the corresponding sections—pre-screening visit and baseline visit.

Diagnostic costs, such as MRI and EKG tests, are also important. The personnel involved in this part of the research proc-ess include the principal investigator, study coordinator, nurse, and technician. Their


costs are recorded as in the previous section, with both time and salary rates.

The next three sections, subject visits, unscheduled visits, and closing visit, all use the same cost categories. However, each one serves a different type of patient visit. These costs include the time used by the person-nel involved: the principal investigator, study coordinator, nurse, and technician. The sub-ject visits section accounts for all costs of the visits that a subject must make as prescribed by the research protocol. The unscheduled visits section refl ects the costs arising from

complications or when the subject needs an unforeseen appointment with the physician conducting the research trial. The closing visit is the last visit that the subject needs to make for the trial. Each of these three sec-tions is divided into two parts. The fi rst part includes the costs associated with forms and charts. This is the preliminary paperwork and related tasks completed prior to the examina-tion ( see Figure 2). The second part is for the examination itself. This includes costs such as lab tests, vital signs, the physical exam, and others ( see Figure 3).

During the course of a research trial, the expertise of a physician in a different spe-cialty may be sought. This is refl ected in our consultant fees section. Each clinical depart-ment, from Anesthesiology to Urology, is listed and the default personnel included in this section are: physician, nurse, technician, and assistant. These can be easily altered as needed. The time spent per visit is recorded along with their salary rate to derive the total consultant costs.

An optional “procedure” section, between sections 8 and 9, can be used for a specifi c department. For example, if a patient needs to have a surgical operation performed as

Figure 1. Common Physical Examination Tasks

Physical Exam Tasks

Vital signs

Height

Weight

Physical exam

Review adverse reactions

Lab testing

Lab handling

Treatment plan

Figure 2. Pre-Visit Time Costs Commonly Done Before

an Examination

Tasks Done Before Examination

Prepare for visit

Chart review

Demographics

Medical history

Medication history

Case report form

Figure 3. Tasks Done During an Examination

Common Exam Tasks

EKGs

Pharmacy dispensing fee

Screening failures

Monitoring visit

FDA audit fee

Adverse Event (AE)

Serious Adverse Event (SAE)


part of the research, the procedure section is designed to include these costs.

The last two sections, close out and mis-cellaneous, contain short lists of tasks. The close out costs refl ect the time the research coordinator, author, principle investiga-tor, and others spend on the events needed to bring a clinical research trial to closure ( see Figure 4). Miscellaneous costs include such expenses as printing, postage, and other minor administrative costs.

A total costs summary presents a concise display from each of the previous sections. It also accounts for indirect overhead charge rates that may be assessed by the institution hosting the trial, as a percentage of the direct costs. This summary refl ects the total costs of the trial, including both the SOC costs and the total research costs. For studies that take more than one year to complete, an optional infl ation rate adjustment can also be applied. The sponsoring agency’s payment rates are included in the summary. If the offered pay-ment is greater, or equal to the estimated cost, then the clinical trial can be accepted. If it is lower, the physician will lose money on

the trial. We include a clear go/no-go indica-tor to readily signal these outcomes for fi nal decision making.

Results

We tested our system with actual data from a new clinical research trial at a major academic medical center. We relied on four main assumptions:

1. The number of patients enrolled was 25;

2. There were no changes in salaries of the medical faculty involved during the trial;

3. There were no changes in the assessed indirect cost rate during the trial; and

4. The time incurred by the health pro-fessionals involved in the trial was recorded accurately.

A unique feature of the interactive spread-sheet is its ability to show how much each major section contributes to the total cost. When examining the change in composition of total costs from the time of estimation, to the fi nal actual amounts, we discov-ered two major changes ( see Figure 5). The pre-screening visit and baseline visit costs were estimated to constitute 17.5 per-cent of the total cost, but were actually 48.6 percent of the total. That is 31.1 percent more than estimated. Subject visits costs were estimated to constitute 33.9 percent of the total cost, but were actually only 10.1 percent. That is 23.8 percent less than estimated.

However, the actual “profi t” was not dra-matically changed when compared with the estimated “profi t.” This is because it was originally estimated that all patients who

Figure 4. Common Tasks Done to Close out

a Clinical Trial

Common Tasks

Close out visit

Closure letter

Data queries

Archiving time

Audits

Institutional Review

Board (IRB) report

Follow-up review


were screened would enroll in the trial. In fact, only one third of the screened patients actually enrolled. Since additional patients had to be screened to meet the 25 patient minimum quota, the variable costs incurred for pre-screening visits and baseline visits increased in total. The fi xed costs associ-ated with subject visits then remained essen-tially fi xed, with only a slight decrease. The combined result was a higher than estimated total cost per patient. To help compensate, the sponsoring pharmaceutical company provided a 41 percent increase in their pay-ment rates. The net effect was a bottom-line “profi t margin” that was 7.7 percent less than the original estimate.

This cost-volume-profi t relationship is illustrated in Figures 6 and 7 for a range of zero to 25 patients. With no patients, a loss of $4,300 occurs. With one patient enrolled, a “profi t” of $3,300 occurs ( see Figure 6). The break-even point is between zero and

one patient ( see Figure 7). Thus, from the physician’s perspective, just one patient pro-duces a fi nancially positive result for this clinical research trial.

Figure 8 illustrates that the slope of the profi t margin line increases steeply as the number of patients starts to increase from zero. However, once enrollment approaches ten patients, the profi t margin line levels off and the profi t margin continues at a constant rate per patient. While the margin no longer gets bigger per patient, the trial nonetheless does continue to make a positive profi t per patient after ten. So not only can we draw conclusions about the accuracy of estimates and the cost proportions of a research trial, we can also assess the relationship between costs, patient volume, and “profi tability.” These factors can then be reliably con-sidered, along with the clinical value of a proposed trial, in the physician’s decision-making process.

Figure 5. Change in Cost Composition—Estimate to Actual

Section No. Cost Component Estimate Actual Change

1 Protocol 2.7% 1.7% –1.0%

2 & 3 Pre-screening visit

and Baseline visit

17.5 48.6 +31.1

4 Diagnostics 11.9 8.0 –3.9

5 Subject visits 33.9 10.1 –23.8

6 Unscheduled visits 1.3 0.9 –0.4

7 Closing visit 0.0 0.0 +0.0

8 Consultant fees 10.5 10.5 +0.0

Optional Procedure 21.3 19.6 –1.7

9 Close-out costs 0.2 0.1 –0.1

10 Miscellaneous 0.9 0.6 –0.3

Total Cost 100.0% 100.0% 0.0%

Note: Columns may not sum to exactly 100.0% due to rounding.


Figure 6. Cost-Volume-Profi t Relationships

Figure 7. Break-Even Point


Caveats

There are several potential limitations in the use of our approach to decision support. These include the experience of the physi-cians and study members in clinical research trials in estimating their time, the reliability of the person inputting the information, and the time required to input the information. These may affect the integrity of the data and conse-quently the predicted cost. Each of these fac-tors is interrelated and has a cumulative effect on the quality of the results ( see Figure 9).

The time invested by personnel in acquir-ing experience in clinical research trials has an impact on the accuracy and hence the reliability of the information input. If a fourth-year surgical resident and a 20-year veteran in cardiothoracic surgery were each asked how long it should take to perform a

bronchoscopy, two different answers would likely be obtained. The more experienced surgeon would likely provide the more accu-rate estimate.

Reliability relates to the amount of care-ful attention given to easily overlooked par-ticulars when entering information. When

Figure 8. Slope of the Profi t Margin Line

Figure 9. Relation of Limitations

Experience

Reliability

Time

Integrity


recording the time to visit a patient, care must be taken to include major tasks, such as time to prepare for the visit, seeing the patient, and discussing the procedures, and answering questions with the patient. In addition, minor tasks, such as time to travel to and from the meeting site, subsequent discussion with the nurse, and time to update the charts, must also be included. We found that these minor tasks can account for as much as 25 percent of the incurred time, and nearly 20 percent of the cost of a clinical research trial. If overlooked, these can lead to a signifi cant amount of unac-counted cost. When both major and minor tasks are included in a reliable manner, the results will be considerably more accurate.

The time taken to input the information in the spreadsheet is a direct refl ection of the level of detail chosen. We found that it takes about three to fi ve hours to input a suffi ciently detailed set of data for accurate cost estima-tion. Generally, the most familiar task is the clinical protocol, and the more experience the person has, the closer he or she will be to the three-hour benchmark for data entry.

The integrity of the system is a culmi-nation of the preceding three factors. The

adage, garbage-in, garbage-out, applies to any data assimilation and entry process. The quality of the results can be no better than the care and quality invested in the inputs.

Conclusions

The advantages of using an interactive, spreadsheet-based fi nancial decision sup-port tool are signifi cant. First, all estimated costs are accounted for and appropriately attested as they are entered. Second, it is readily customizable for as much or as little detail as desired. Third, it prompts the phy-sician to think within specifi c timeframes when estimating the duration and costs of a proposed clinical trial. Since labor costs must be included, the physician must project a reasonable timeframe in which to complete the research trial. Most impor-tantly, an interactive clinical trial decision support system provides clarity and an increased level of certainty to the deci-sion-making process. Investing the time to assimilate and input the data will pay off in terms of better resource allocation and bet-ter decisions.

REFERENCES

1. “The Clinical Trials Business,” available at www.bccresearch.com/report/PHM027B.html , accessed July 15, 2010.

2. “Clinical Research Management in Pakistan,” available at www.epb.gov.pk/v1/exporters/docs/report_clinical_research_management_pakistan.pdf , accessed July 15, 2010.

3. US Institutes of Health, “Clinical Trials Terms,” available at http://clinicaltrials.gov/ct2/info/glossary , accessed Nov. 30, 2009.

4. US Food and Drug Administration, “Inside Clinical Trials: Testing Medical Products in People,” available at www.fda.gov/Drugs/ResourcesForYou/Consumers/ucm143531.htm , accessed July 11, 2009.

5. Tschanz, DW, “Arab Roots of European Medi-cine,” Heart Views , 69–80 (2003); Brater, CD, Walter, DJ, “Clinical Pharmacology in the Middle Ages: Principles that Presage the 21st Century,” Clinical Pharmacology & Therapeu-tics , 447–450 (2000).

6. Lind, JA, “Treatise of the Scurvy,” available at www.bruzelius.info/Nautica/Medicine/Lind(1753).html , accessed Nov. 30. 2009.

7. “Historic Figures: James Lind (1716–1794),” British Broadcasting Company, available at www.bbc.co.uk/history/historic_fi gures/lind_james.shtml , acccessed Sept. 25, 2010.

8. Eastern Virginia Medical School, “List of Government and Private Funding Sources,”


available at www.evms.edu/offi ce-of-research/evms-offi ce-of-research-extramural-funding-opportunities.html , accessed July 15, 2010; University of Texas Health Science Center at Houston, “Clinical Budget Development Resources,” available at www.uth.tmc.edu/ctrc/budgetdevelopment.html , accessed July 15, 2010; University of Texas Health Science Center at Houston, “Compliance Program for Clinical Research Budgeting/Billing,” avail-able at www.uth.tmc.edu/research/training/nidp/docs/carter.ppt , accessed July 15, 2010; University of California San Francisco, School of Medicine, “Clinical Research Tools,” avail-able at http://medschool.ucsf.edu/clinical_research/tools/index.aspx , accessed July 15, 2010; “The Northwestern University Clinical and Translational Sciences Institute,” avail-able at www.nucats.northwestern.edu/centers/CCR_CRU/CCR_CRU%20Web%20Pages/CCR%20Main/index.html , accessed July 15, 2010; Dartmouth-Hitchcock Medical Center, “Clinical Trials Handbook,” available at www.dartmouth.edu/~osp/resources/ct-handbook/clinicaltrialprocess.html , accessed Apr. 4, 2009; “Advocate Health Care,” available at http://www.advocatehealth.com/default.cfm , accessed July 15, 2010; University of Iowa, “F&A Costs and Budget Templates for Indus-try-Sponsored Research,” available at http://research.uiowa.edu/dsp/main/?get=budget-templates&q=&action , accessed July 15, 2010; “Melbourne Health Guidelines for Clinical Trial Costing Financial Process,” available at www.mh.org.au/secure/downloadfi le.asp?fi leid=1012854 , accessed July 15, 2010; KUMC Research Institute, “How to Create a Clinical Trial Budget,” available at www2.kumc.edu/PDFATraining/Admin/documents/DANIELS%2010.15.07%20CRA%20Budget%20Presentation.ppt , accessed July 15, 2010; The University of Tennessee, “Health Science Center Sponsor Study Budget Tem-plate,” available at www.uthsc.edu/research/research_administration/clinical_trials/forms.php , accessed July 15, 2010; “Managing Clin-ical Trials: Contracts, Budgets, and Finances of Industry-Sponsored Clinical Trials,” avai-lable at www6.miami.edu/research/SeminarHandouts/Handout072105.pdf , accessed

July 15, 2010; The Northern Alberta Clini-cal Trials and Research Centre, “Clinical Trial Budgets,” available at www.clinicaltrials.ualberta.ca/resources_budgeting.php , accessed July 15, 2010; The University of Utah School of Medicine, “Clinical Research Compli-ance and Education,” available at http://medicine.utah.edu/crce/ , accessed July 15, 2010; “Industry Sponsored Clinical Study Budget Checklist,” available at www.ohsu.edu/research/crp/docs/budgetchecklist.doc , accessed July 15, 2010; Child & Family Research Institute, “Budget Template,” availa-ble at www.cfri.ca/research_support/contract/template_checklist.asp , accessed July 15, 2010; University of Minnesota, “Sponsored Projects Symposium II,” available at www.ospa.umn.edu/announcements/documents/session12.ppt#310,39,References , accessed July 15, 2010; “Beth Israel Deaconess Medi-cal Center Clinical Trials Offi ce,” available at www.bidmc.org/Research/ResearchandAcademicAffairs/ClinicalTrialsOffice.aspx , accessed July 15, 2010; Oregon Health & Sci-ence University, “Industry Sponsored Clinical Trial Preaward Process,” available at www.ohsu.edu/research/rda/rgc/docs/ctpreaward.doc?fi x , accessed July 15, 2010; “St Vincent’s Hospital Research Offi ce Study Budget Tem-plate,” available at www.stvincents.com.au/assets/files/pdf/R/study%20budget_template.xls , accessed July 15, 2010; Auckland District Health Board, “Budget Template,” available at www.adhb.govt.nz/ResearchOffi ce/ Budget/develop_a_budget.htm , accessed July 15, 2010; “OCR Guidance for Manage-ment of UTHSCSA Clinical Trials,” available at http://research.uthscsa.edu/ocr/OCR%20Guidance%20for%20Mgt%20of%20Clinical%20Trials.pdf , accessed July 15, 2010; “Implementation and Utilization of a Clini-cal Studies Management System in Support of Clinical Research Billing Compliance,” available at www.hcca-research-conference.org/pastconf/2008/pre/P5/Kukuljan-Moran_P5.pdf , accessed July 15, 2010; “University of Colorado at Boulder Budget Training,” available at www.colorado.edu/pba/budget/training/ , accessed July 15, 2010; “University of California, Santa Barbara Budget Offi ce,”


available at http://bap.ucsb.edu/budget.html , accessed July 15, 2010; “Thoracic Surgery Foundation for Research and Edu-cation,” available at www.scangrants.com/grant/2009/10/17/research-grantscardiothoracic-surgery.aspx , accessed Apr. 3, 2009; “Indirect Costs Associated with Clinical Tri-als,” available at www.hamiltonhealthsciences.ca/body.cfm?id=1069 , accessed July 15, 2010; “Guide to completing the Research Budget Template,” available at www.adhb.govt.nz/researchoffice/Budget/Guide%20to%20completing%20the%20Budget%20Template%20V5.1%20July%202008%20.doc , accessed July 15, 2010; “The Cost of

Clinical Trials Drug Discovery & Develop-ment,” available at www.dddmag.com/the-cost-of-clinical-trials.aspx , accessed July 15, 2010; Ferris, LE, Naylor, CD, “Physician Remuneration in Industry-sponsored Clinical Trials: The Case for Standardized Clinical Trial Budgets,” Canadian Medical Association , 171: 883–886 (2004); Pitler, LR, Bonomi, PD, “Developing an Effective and Compliant Plan for Billing Clinical Trials,” Journal of Oncol-ogy Practice , 2: 265–267 (2006).

9. Hatfi eld, E, Dicks, E, Parfrey, E, “Budgeting, Funding, and Managing Clinical Research Projects,” Methods of Molecular Biology , 299–311 (2009).

38

The Role of Financial Market Performance in Hospital

Capital Investment Kristin L. Reiter and Paula H. Song

Many not-for-profi t hospitals hold large portfolios of fi nancial investments, making them vulnerable to fl uctuations in market performance. This article examines the association of bond and equity market performance with investment in property, plant, and equipment by 194 not-for-profi t general hospi-tals in California over the period 1997 to 2006. The study combines retrospective panel data from the California Offi ce of Statewide Health Planning and Development with year-end returns on the S&P 500 and ten-year US Treasury bonds. Using fi xed-effects regression, we fi nd a signifi cant positive association between S&P 500 performance and hospitals’ capital investment; investment is not corre-lated with ten-year Treasury bond performance. Key words: not-for-profi t hospitals, capital investment, fi nancial markets.

Among health delivery organizations, not-for-profi t hospitals are unique in that they hold signifi cant portfo-

lios of long-term fi nancial investments. New fi nancial investment securities have created opportunities for hospitals to achieve sub-stantial returns on their fi nancial asset portfo-lios; however, trends in investment allocation suggest that not-for-profi t hospitals may also be increasingly vulnerable to fl uctuations in fi nancial market performance. A recent sur-vey of 143 not- for-profi t hospitals and health systems revealed that while hospitals typi-cally invest more than half of their endow-ment assets in cash and other traditionally low-risk, liquid securities, the remaining endowment funds are invested in stock and, increasingly, in alternative investments including venture capital, real estate, hedge funds, and private equity. 1 In good times, investments in fi nancial securities can boost returns for hospitals; however, in bad times, certain securities introduce risk into invest-ment portfolios and can result in substantial losses. 2 Recent years have been marked by tremendous fl uctuation in fi nancial market performance, a trend that is likely to con-tinue as global economies emerge from the fi nancial crisis that began in 2007.

Given the size of fi nancial investment portfolios in not-for-profi t hospitals, gains and losses from fi nancial investments have the potential to infl uence not-for-profi t hos-pitals’ capital spending in at least two ways. First, market performance may induce man-agers to trade off fi xed asset and fi nancial investment opportunities based on return potential. Second, and more importantly, market performance can directly affect the

Kristin L. Reiter, PhD, is an Assistant Professor in the Department of Health Policy and Management at The University of North Carolina at Chapel Hill in Chapel Hill, North Carolina. She can be reached at [email protected].

Paula H. Song, PhD, is an Assistant Professor in the Division of Health Services Management and Policy at The Ohio State University in Columbus, Ohio. She can be reached at [email protected].

Acknowledgement: Dr. Reiter received support from Award Number KL2RR025746 from the National Center for Research Resources. The content of this article is solely the responsibility of the authors and does not necessarily represent the offi cial views of the National Center for Research Resources or the National Institutes of Health.


The Role of Financial Market Performance in Hospital Capital Investment 39

resources available to support capital expen-ditures. In the stock market downturn of 2000, for example, the Cleveland Clinic Foundation was reported to have lost over $500 million of the value of its investment portfolio causing downgrades in the hospi-tal’s bond rating, delays in routine expen-ditures on computer systems and nursing stations, and postponement of a new heart center. 3

The implications of this are clear. Hospitals that fail to invest in capacity, new treatment technologies, and information technology may not be prepared to meet the needs of a growing and aging population, or to address challenges related to health reform, quality, and patient safety. 4 Thus, it is critical that managers and policymakers understand the extent to which hospital capital expenditures are infl uenced by fi nancial market returns in order to anticipate the effects of economic fl uctuations on health care access and quality. This article provides a fi rst look at this ques-tion by examining the relationship of bond and equity market performance with invest-ment in property, plant, and equipment in a sample of not-for-profi t hospitals over the period 1997 to 2006.

Background and Literature Review

Not-for-profi t hospitals hold, on average, 21 percent of their total assets in fi nancial investments versus only 4 percent held by investor-owned hospitals. 5 The dispropor-tionate reliance on fi nancial investments among not-for-profi t hospitals results prima-rily from not-for-profi t hospitals’ inability to raise funds by selling stock, their use of debt over internal equity reserves, and deliberate strategies to hold cash to protect or enhance bond ratings. 6 In the absence of funds from

stock sales, fi nancial investments help support not-for-profi t hospitals’ operations and capital investments. Specifi cally, fi nan-cial investments provide precautionary sav-ings and a source of untaxed interest income to supplement earnings from patient care. 7 Financial investments also strengthen not-for-profi t hospitals’ balance sheets and allow them to achieve favorable bond ratings that reduce the cost of borrowing. 8 Since not-for-profi t hospitals have access to tax-exempt bonds, relatively inexpensive debt is often substituted over internal reserves to support capital projects. 9

Although we are not aware of any empir-ical studies that address the effect of fi nancial market performance on hospital capital investment, ample evidence suggests an association. In a series of interviews, chief fi nancial offi cers (CFOs) of 12 lead-ing not-for-profi t health care systems con-fi rmed the importance of cash and debt in supporting capital investment. 10 These same CFOs cited investment and bond ratings as two primary reasons for holding large bal-ances of cash and marketable securities. 11 Another study found that hospital fi nancial performance was positively related to capital investment, suggesting that fi nancial market performance may affect investment if a hos-pital is particularly reliant on non-operating income. 12

Debt capacity has also been shown to positively infl uence capital expenditures. Excess tax-exempt debt capacity stimulates hospital investment. 13 In contrast, risky debt ( i.e ., debt that substantially raises the like-lihood of bankruptcy) has been shown to raise the required rate of return on capital projects resulting in lower levels of invest-ment. 14 Finally, the availability of internally generated cash fl ow has been found to be


positively associated with capital spending, but only when capital market imperfections such as information asymmetry raise the cost of borrowing. 15

Together, fi ndings from previous studies suggest a role for fi nancial market perform-ance in determining hospital capital expen-ditures through its effects on non-operating income and debt capacity, but empirical evidence is needed to begin to understand the extent of the association. To address this gap in the literature, we examine the

association between debt and equity market performance and capital investment in a sample of not-for-profi t hospitals, control-ling for other factors known to infl uence capital spending.

Conceptual Framework

The conceptual framework for the study is presented in Figure 1. Not-for-profi t hospitals are assumed to have an objective to maximize the quantity and quality of services, subject

Figure 1. Conceptual Framewrk

Note: Constructs in boxes with dashed lines are not measured in the empirical model.

Financial Market Performance

(Bond and Equity)

Fair value of financial assets

Non-operating (investment)

income

Bond rating

Cost of debt

financing

HospitalInvestment

Hospital operating characteristics

Local marketcharacteristics

Hospital size System affiliation Lagged investment

CONTROL VARIABLES

Existing debt Operating

profitability Financial assets


to a target profi t or breakeven constraint. 16 As a result, hospitals’ capital spending will depend on the fi nancial condition of the hos-pital. 17 Bond and equity market performance are expected to affect hospital capital invest-ment through their effects on non-operating income, the fair value of fi nancial assets, and the cost of borrowing.

Higher returns on both bond and equity investments would be expected to increase non-operating income, while lower returns would be expected to decrease non- operating income. Bond investments, for example, provide earned interest and capital gains or losses on sales of seasoned (outstanding) bonds. Equity investments provide divi-dend payments or realized capital gains and losses on sales of securities. For many hos-pitals, non-operating income is an important source of overall profi tability, so we would expect to see a positive correlation between fi nancial market performance and capital expenditures. 18

In addition to affecting non-operating income, changes in fi nancial market per-formance may affect the resources available to support capital expenditures by changing the strength of the hospital’s balance sheet. In particular, fi nancial market performance may affect the hospital’s liquidity ( i.e ., its ability to pay debts as they come due). Accounting principles require that most investments in debt or equity securities be reported at fair value (versus capital investments, such as buildings and equipment, which are held at historical cost); thus, unrealized gains or losses on fi nancial investments resulting from changes in market value will be refl ected in the hospital’s balance sheet. 19 The strength of the balance sheet, in turn, affects the cost of borrowing and the hospital’s ability to fi nance capital projects. 20

We expect that the fair value of most hos-pitals’ fi nancial assets will be correlated with average market returns, so that increases in market performance will reduce the cost of borrowing while declines in market perform-ance will increase the cost of borrowing. As with non-operating income, this relation-ship suggests a positive correlation between fi nancial market performance and capital expenditures.

To examine the association, we model hos-pital capital investment as a function of debt and equity market performance, controlling for hospital and county-level market charac-teristics expected to infl uence capital spend-ing. We include lagged capital expenditures to control for the fact that capital spending is “lumpy” rather than constant over time. In our model, hospital size and system affi lia-tion affect the resources available to support investment. Similarly, existing levels of debt and fi nancial assets affect debt capacity—the hospital’s ability to borrow and the price it will pay for borrowing—while hospital operating profi t determines the internally generated funds available to pay for future capital expenditures.

We capture demand for capital spend-ing by including a series of hospital and market characteristics expected to affect the need for and profi tability of fi xed asset investments. Return on past capital invest-ment, patient volumes, and patient revenue signal future demand for hospital services and are likely positively related to capital spending. 21 Labor intensity of the hospital is included as a measure of the marginal productivity of capital. 22 County-level market characteristics are included to control for the effect of competition and socio-demographic factors on hospital investment. 23


Methods

Data and Study Sample

This study combines retrospective panel data from several sources over the period 1997 to 2006. Hospital fi nancial data come from the Annual Hospital Disclosure Reports collected and maintained by the California Offi ce of Statewide Health Planning and Development (OSHPD). California, which is the largest state requiring annual reporting, provides the most comprehensive publicly available source of hospital fi nancial data. Although Medicare Cost Reports are avail-able for most hospitals nationally, fi nancial data are more limited and less detailed than OSHPD data. 24 Financial data are merged with county-level market characteristics from the 2006 Area Resource File. Hospital system affi liation is obtained from the 2006 American Hospital Association’s (AHA) Annual Hospital Survey. Data on equity mar-ket performance are obtained from historic data compiled from S&P 500 market index values at the end of each calendar year. Ten-year Treasury bond returns including coupon interest payments and price appreciation/depreciation at the end of each calendar year are from the Federal Reserve of St. Louis. 25

The study sample includes all nonfederal, not-for-profi t, general acute care hospitals in California except those owned by Kaiser Permanente (the state of California exempts the 29 Kaiser-owned hospitals from report-ing fi nancial data at the individual hospital level). We exclude investor-owned hospitals since they have different motivations for holding fi nancial investments, and may be affected by fi nancial market performance not only through fi nancial asset holdings, but also through any effects on the market value of their own stock. Government hospitals are

excluded because they operate under differ-ent fi nancing mechanisms and regulations. For example, budgeting processes differ among government hospitals, and unused surpluses must often be returned.

We use the report with the greatest number of reporting days for two not-for-profi t hos-pitals that submitted duplicate reports in the same reporting period. We exclude 120 hos-pital reports (approximately 7 percent) with missing data on any of our included variables; 84 percent are missing system affi liation, 12 percent are missing capital expenditure data, and the remaining 4 percent are miss-ing fi nancial or market data. Hospitals that merged during the study period are refl ected as individual observations in the pre-merger period. In the post-merger period, fi nancial data from these hospitals are refl ected in the reports of the acquiring hospitals. The fi nal study sample includes 1,329 pooled hospital observations representing 194 nonfederal, not-for-profi t, general acute care hospitals in California from 1997 to 2006.

Empirical Model and Specifi cation

We estimate a model of hospital capital investment as presented in Equation 1:

INV i,t = α

0 + β

1 Rbond

t-1 + β

2 Requity

t-1

+ β 3 INV

i,t-1 + β

4 Size

i,t + β

5 System

i,t

+ β 6 Financial

i,t-1 + β

7 Operating

i,t-1

+ β 8 Market

i,t-1 + Time + Time 2

+ µ i + ε

i,t (1)

Hospital capital investment, INV i,t , which

is measured by total capital purchases as reported on the hospital’s annual disclosure report, refl ects true capital expenditures excluding the effects of asset sales or retire-ments. Annual capital expenditure values are adjusted for infl ation to 2006 dollars using the consumer price index.


Our key variables of interest, bond mar-ket performance ( Rbond

t-1 ), and equity mar-

ket performance ( Requity t-1

) are measured based on index values reported by the Fed-eral Reserve of St. Louis and Standard and Poor’s, respectively. To measure bond market performance, we include the actual annual returns, including coupon interest payments and price appreciation/depreciation on ten-year Treasury bonds lagged one period. To measure equity market performance, we include actual annual returns on the S&P 500, an index constructed to broadly refl ect the stock market, lagged one period.

To control for the periodic nature of capi-tal spending, we included the lagged value of hospital capital investment ( INV

i,t-1 ). The

possibility that large hospitals or those affi li-ated with a health system may have greater resources is controlled for by the number of licensed beds ( Size

i,t ), and a dummy vari-

able set equal to one if the hospital reported system affi liation in the AHA annual survey ( System

i,t ).

Since fi nancial statement data elements are determined contemporaneously by defi -nition, and capital planning occurs prior to capital expenditures, we measure all other explanatory variables in the year preced-ing the dependent variable. Income state-ment and cash fl ow measures are adjusted for infl ation to 2006 dollars using the con-sumer price index. Balance sheet measures are more problematic since they refl ect the cumulative effect of asset acquisition and disposition over time. For example, a hos-pital’s fi xed assets balance may include a building purchased 20 years ago and equip-ment purchased last year, each of which is valued at its historical cost at the time of pur-chase. While we do not adjust balance sheet measures for infl ation, most balance sheet

measures are expressed as ratios. The one balance sheet measure that is not expressed as a ratio, fi nancial assets, is likely to refl ect current market values in the year it is meas-ured since most fi nancial assets are held at fair value. 26

Financial i,t-1

refl ects a vector of variables that capture existing fi nancial resources of the hospital available to support investment. Debt capacity is refl ected by the existing level of debt in the hospital, measured as the ratio of total liabilities to total assets; and by the value of fi nancial investment assets, measured as the sum of unrestricted cash and marketable securities, total assets lim-ited as to use, and total restricted fi nancial investment assets. Cash fl ow from opera-tions, a measure of hospital operating prof-itability, as reported on the hospital’s cash fl ow statement refl ects the hospital’s ability to generate cash to support future capital expenditures.

Operating i,t-1

and Market i,t-1

are vectors of hospital operating and market characteristics that refl ect the need for and potential profi t-ability of fi xed capital investments. Hospital operating characteristics include measures of hospital productive output, return on pre-vious capital investment, and labor intensity. Hospital output measures, intended to refl ect the demand for a hospital’s services and therefore the need for future capital invest-ment, is measured by the total number of discharges, the total number of outpatient visits, net patient service revenue, and the hospital’s occupancy rate as reported on the hospital’s annual disclosure report. Return on previous capital investment, intended to refl ect the information value of past invest-ment performance, is measured as the return on assets, or net income divided by total assets. Finally, labor intensity, intended to


refl ect the marginal productivity of capital investment, is measured by total salary expense including fringe benefi ts.

We include a vector of county-level meas-ures to control for time-varying market con-ditions that may affect a hospital’s demand for capital investment. The number of hospi-tal beds in the county controls for the effect of competition. 27 To control for the effects of socio-demographic characteristics of the local population on demand and reimbursement

for hospital services, we include per capita income, the unemployment rate, and percent of the population over 65.

Time fi xed effects are not included in our model because they are exactly collinear with the fi nancial market performance measures; however, linear and quadratic time trends are included to control for the upward trend in real capital investment over the study period. Because county-level data may not precisely refl ect hospital markets, and because of the

Figure 2. Descriptive Statistics 1997–2006 (n=1,329)

Variable Mean/Percent Standard Deviation

Capital investment (in millions) $14.47 $23.00

Financial market returns

Return on ten-year T-Bonds 7.07% 7.50%

Return on S&P 500 8.61% 18.86%

Control Variables

Number of licensed beds 265 188

Percent system affiliated 80%

Existing fi nancial resources

Debt-to-assets 61.62% 35.94%

Financial assets (in millions) $41.62 $82.62

Operating cash fl ow (in millions) $13.49 $30.34


Occupancy rate 67.57% 19.06%

Return on assets 2.71% 11.09%

Net patient revenue (in millions) $153.77 $166.96

Salary expense (in millions) $77.58 $83.85

Number of discharges 10,707 7,712

Number of outpatient visits 157,127 158,789

County-level market characteristics

Number of beds 6,756 9,054

Per capita income $31,312 $9,258

Unemployment rate 6.19% 2.87%

Percent of population over 65 11.33% 2.27%


potential for hospital-level variation in aver-age capital spending for reasons related to management preferences or other time-in-variant characteristics, we include hospital fi xed effects ( µ

i ). Hausman tests indicated

that fi xed-effects regression was preferred over random effects. Robust standard errors are calculated to address heteroskedasticity, and standard errors are adjusted to account for clustering of observations within hos-pitals. All analysis is performed using STATA 10.0. 28

Results

Descriptive data on the study sample are presented in Figure 2. Over the period 1997 to 2006, mean annual capital expenditures are approximately $14 million. The average return on ten-year US Treasury Bonds over the study period is 7.07 percent with a range from negative 8.25 percent in 2000 to over

16 percent in 2001. The average return on the S&P 500 over the study period is approx-imately 8.7 percent, although returns vary widely from year to year, ranging from a low of negative 22 percent in 2002 to a high of over 31 percent in 1997.

Figure 3 shows unadjusted trends in bond and equity market performance and average annual capital expenditures over the study period. There does not appear to be any relationship between hospital capi-tal investment and the performance of ten-year US Treasury bonds; however, trends suggest a possible correlation between equity market performance and hospital investment. Specifi cally, as returns on the S&P 500 decline over the period 1997 to 2002, hospital capital investment remains relatively fl at (all values adjusted to 2006 dollars). In contrast, following large gains in the S&P 500 from 2002 to 2003, annual hospital capital expenditures grew from

Figure 3. Financial Market Returns and Capital Investment 1997–2006


just under $14 million in 2003 to over $20 million by 2006.

The fi xed effects regression, presented in Figure 4, estimates a coeffi cient on the lagged treasury bond return that is not

statistically signifi cantly different from zero (p=0.525). In contrast, the coeffi cient on the lagged S&P 500 return is large and signifi cant (β=5.560, p=0.023), indicat-ing that a one percentage point increase in

Figure 4. Fixed Effects Regression of Hospital Capital Investment

Variable Coefficient Standard Error p-value

Financial market returns

Return on 10-year T-Bonds 3.508 5.503 0.525

Return on S&P 500 5.560** 2.432 0.023

Control variables

Number of licensed beds 0.013 0.018 0.467

System affiliation -1.061 2.076 0.61

Lagged capital investment 0.249*** 0.070 0.000

Existing fi nancial resources

Debt-to-assets -0.030 0.021 0.153

Financial assets 0.046 0.029 0.120

Operating cash fl ow -0.014 0.013 0.540


Occupancy rate 0.017 0.035 0.642

Return on assets 0.015 0.033 0.664

Net patient revenue 0.087** 0.038 0.022

Salary expense 0.015 0.051 0.774

Number of discharges -0.002** 0.001 0.013

Number of outpatient visits 6.788 7.735 0.381

County-level market characteristics

Number of beds (in thousands) -0.001 0.000 0.21

Per capita income (in thousands) 0.000 0.000 0.313

Unemployment rate -0.214 0.579 0.712

Percent of population over 65 0.749 0.851 0.38

Time Trend

Time 1.407 0.907 0.122

Time-squared -0.124* 0.064 0.054

R-squared 0.452

Sample Size 1,329

Unique Hospitals 194

*p<0.10, **p<0.05, ***p<0.01

Note: Model controls for hospital fi xed effects, uses robust standard errors and adjusts for

clustering of observations by hospital.


the S&P 500 return results in an additional $5.6 million of capital investment.

Statistically signifi cant control variables include lagged capital expenditures, net patient revenue, and total discharges. As expected, lagged capital investment is posi-tively associated with capital investment in the current period (β=0.25, p=0.000). Con-sistent with previous research, net patient revenue is positively associated with capital investment, with an additional $1 million of revenue predicting $87,000 in capital expenditures (p=0.022). The number of dis-charges, surprisingly, was negatively asso-ciated with capital expenditures (β=-0.002, p=0.013) although the effect size was small. None of the market characteristics is signifi -cantly related to hospital capital spending; however, the negative and signifi cant coef-fi cient on the squared time variable suggests that hospital capital spending was increasing at a decreasing rate over time (β=-0.12, p=0.054).

Conclusion and Discussion

Although the effect size is surprisingly large, fi ndings from this study suggest that capital investment by not-for-profi t hospitals in California is positively asso-ciated with equity market performance as measured by annual returns on the S&P 500. In contrast, we fi nd no evidence that capital investment is related to bond market performance.

Not-for-profi t hospitals, unlike their investor-owned counterparts, hold substan-tial portfolios of fi nancial assets. These port-folios expose not-for-profi t hospitals to risk associated with market fl uctuations. Since fi nancial asset holdings play a central role in maintaining favorable bond ratings to support

borrowing for capital investment, and in some cases, provide a non-trivial source of non-operating income, our fi ndings suggest that downturns in the market may constrain the resources available to support investment in property, plant, and equipment, particu-larly when the downturns are substantial and persistent.

The lack of a statistically signifi cant relationship between bond market returns and investment in this study may, in part, refl ect imprecision in our bond market return measure for any given hospital. Lack of data availability prevented us from precisely measuring the effects of fi nancial market performance on hospi-tal capital investment. There are no pub-lic data currently available to understand asset allocation or investment strategy at the individual hospital level. As a result, our measures of fi nancial market perform-ance are limited to returns on the S&P 500 and ten-year US Treasury bonds, and the effects of each market performance meas-ure are constrained to be the same for all hospitals in the study sample. Actual effects will depend on the extent to which hospitals hold specifi c types of securities in their fi nancial asset portfolios; thus, it is likely that our results are overstated for some hospitals and understated for others. Still, this study is the fi rst that we know of to examine the association between fi nan-cial market returns and hospital capital spending, and fi ndings suggest that future research is warranted.

Other limitations of the study are noted below. First, the study is limited to hospitals in California; therefore, fi ndings may not generalize to not-for-profi t hospitals nation-ally. The competitive landscape for hospi-tals in California may differ substantially


from other areas of the United States. To the extent that differences in competition gen-erate differences in fi nancial asset holdings and allocation, our fi ndings may not refl ect the population of not-for-profi t hospitals.

Second, our study is not couched in a fully developed, theoretical model of hos-pital capital investment. We have attempted to control for demand for capital spending using hospital operating and market charac-teristics. We include hospital fi xed effects to capture the infl uence of unobserved, time-invariant factors on hospital investment. It is possible, however, that we have omitted important time-varying covariates. Existing theories of investment have been developed primarily for investor-owned fi rms, where demand for capital spending is measured pri-marily through the market value of a compa-ny’s stock (which should refl ect the value of future investment opportunity if capital mar-kets are effi cient). Such a measure does not currently exist for not-for-profi t hospitals. While development of a complete model of hospital investment was beyond the scope of this article, our fi ndings suggest that future models of hospital capital investment should consider equity market performance as a covariate.

Finally, 80 percent of the hospitals in our sample are members of hospital systems, but we are only able to observe capital invest-ment for individual hospitals located in Cali-fornia. It is possible that multi-state hospital systems pool resources at the corporate level and allocate capital investment to individual hospitals across the system. In this case we may not observe the full effect of fi nancial market performance on hospital capital spending.

Despite these limitations, our fi ndings are suggestive of a relationship between

equity market performance and hospital capital investment. The possibility of such a relationship raises several concerns in light of the recession in the early 2000s and the more recent fi nancial crisis that began in 2007. Previous research has shown that when access to debt fi nancing is constrained, hospitals are reliant on internally generated cash fl ow to support investment. 29 The most recent credit crisis has resulted in limitations on access to debt capital for hospitals with even the strongest balance sheets. 30 At the same time, hospitals have watched the value of their investment portfolios plummet. Findings from a survey of 143 not-for-profi t hospitals revealed that the average return on endowment funds fell by 21.2 percent in 2008 threatening liquidity. 31 The combi-nation of lower investment earnings, losses in fair value of investment assets, and the credit crunch may bring capital investment to a halt.

Hospitals that are not able to make invest-ments in infrastructure and innovation may suffer losses in quality, physicians, patients, or capacity. Recent fi ndings by Bazzoli et al . support this concern, showing that hospitals with low operating margins not offset by alternative sources of revenue ( e.g ., investment income) have excess incidents in nursing or surgical-related adverse events and higher excess deaths in low- mortality DRGs. 32 Given the increasing focus on hospital quality and patient safety and the need for ongoing investment in information technology, not-for-profi t hospital manag-ers should assess and actively manage the risk in their fi nancial investment portfolios to avoid long-term constraints on resources to support capital investment. Additionally, long-term capital plans should allow for variation in fi nancial investment returns,


and hospitals with greater near-term capi-tal needs should carefully weigh the risk of including equity and alternative investments in their fi nancial investment portfolios. In the case of extended downturns in fi nancial

markets such as the current recession, policymakers may need to consider short-term bridge fi nancing options to support any capital spending that cannot be safely delayed.

REFERENCES

1. Evans, M, “Dippin’ Endowments,” Modern Healthcare, >http://www.modernhealthcare.com/article/20091026/REG/910229985 >, accessed May 12, 2010 (Oct. 26, 2009).

2. Song, PH, Smith, DG, and Wheeler, JRC, Jr, “It Was the Best of Times, It Was the Worst of Times: A Tale of Two Years in Not-for-Profi t Hospital Financial Investment,” Health Care Management Review , 33: 234–242 (2008).

3. Cowan, AL, “Investment Losses Hurt Major Hospital,” The New York Times , New York (2003).

4. Bazzoli, GJ, Clement, JP, Lindrooth, RC et al ., “Hospital Financial Condition and Opera-tional Decisions Related to the Quality of Hospital Care,” Medical Care Research and Review , 64: 148–168 (2007); Byrd, CWJ, McCue, MJ, “Can Your Hospital Remain Inde-pendent?” Healthcare Financial Management , 57: 40–44 (2003); Campbell, C, “Hospital Plant and Equipment Replacement Decisions: A Survey of Hospital Financial Managers,” Hospital and Health Services Administration , 39: 538–547 (1994); Levitt, SW, “Quality of Care and Investment in Property, Plant and Equipment in Hospitals,” Health Services Research , 28: 713–728 (1994); Unland, JJ, “Can Community Hospitals Survive With-out Large Scale Health Reform?” Journal of Health Care Finance , 30: 49–58 (2004).

5. Song, PH and Reiter, KL, “Trends in Invest-ment Portfolios Between Not-for-Profi t and Investor-Owned Hospitals,” Medical Care Research and Review , E-pub ahead of print (2010).

6. Robinson, JC, “Bond-Market Skepticism and Stock-Market Exuberance in the Hos-pital Industry,” Health Affairs , 21: 101–117 (2002).

7. Rivenson, HL, Wheeler, JR, Smith, DG, Reiter, KL, “Cash Management in Health Care

Systems,” Journal of Health Care Finance , 26: 59–69 (2000).

8. McCue, MJ, “A Trend Analysis of Hospitals with High Cash and Security Investments,” Hospital Topics , 79: 23–27 (2001); McCue, MJ, Thompson, JM, Dodd-McCue, D, “Asso-ciation of Market, Mission, Operational, and Financial Factors with Hospitals’ Level of Cash and Security Investments,” Inquiry , 37: 411–422 (2000/2001); Wheeler, JR, Smith, DG, Rivenson, HL, Reiter, KL, “Capital Struc-ture Strategy in Health Care Systems,” Journal of Health Care Finance , 26: 42–52 (2000).

9. Gentry, WM, “Debt, Investment and Endow-ment Accumulation: The Case of Not-for-Profi t Hospitals,” Journal of Health Economic , 21: 845–872 (2002).

10. Smith, DG, Wheeler, JR, Rivenson, HL, Reiter, KL, “Sources of Project Financing in Health Care Systems,” Journal of Health Care Finance , 26: 53–58 (2000).

11. Supra , n.7. 12. Bazzoli, GJ, Clement, JP, Lindrooth, RC et al .,

“Hospital Financial Condition and Opera-tional Decisions Related to the Quality of Hospital Care,” Medical Care Research and Review , 64: 148–168 (2007).

13. Gentry, WM, “Debt, Investment and Endow-ment Accumulation: The Case of Not-for-Profi t Hospitals,” Journal of Health Economic , 21: 845–872 (2002); Wedig, GJ, Hassan, M, Mor-risey, MA, “Tax-Exempt Debt and the Capi-tal Structure of Nonprofi t Organizations: An Application to Hospitals,” Journal of Finance , 51: 1247–1283 (1996).

14. Wedig, GJ, “Risk, Leverage, Donations and Dividends-in-Kind: A Theory of Nonprofi t Financial Behavior,” International Review of Economics and Finance , 3: 257–278 (1994).

15. Calem, PS, Rizzo, JA, “Financing Constraints and Investment: New Evidence from Hospital


Industry Data,” Journal of Money, Credit and Banking , 27: 1002–1014 (1995); Reiter, KL, Wheeler, JR, Smith, DG, “Liquidity Con-straints on Hospital Investment When Credit Markets Are Tight,” Journal of Health Care Finance , 35: 24–33 (2008).

16. Newhouse, JP, “Toward a Theory of Nonprofi t Institutions: An Economic Model of a Hospi-tal,” American Economic Review , 60: 64–74 (1970); Hoerger, TJ, “’Profi t’ Variability in For-Profi t and Not-for-Profi t Hospitals,” Journal of Health Economics , 10: 259–289 (1991).

17. Supra , n.12. 18. McKay, NL, Gapenski, LC, “Nonpatient Reve-

nues in Hospitals,” Health Care Management Review , 34: 234–241 (2009).

19. AICPA, AICPA Audit and Accounting Guide: Health Care Organizations, New York, Ameri-can Institute of Certifi ed Public Accountants, Inc. (2006).

20. Moody’s, Moody’s Public Finance Healthcare Ratings Not-for-Profi t Healthcare: Fiscal Year 2005 Medians, Moody’s Investors Services (2006).

21. Supra , n.12; Campbell, C, “Hospital Plant and Equipment Replacement Decisions: A Survey of Hospital Financial Managers,” Hospital and Health Services Administration , 39: 538–547 (1994); Wedig, GJ, Hassan, M, Sloan, FA, “Hospital Investment Decisions and the Cost of Capital,” Journal of Business , 62: 517–537 (1989).

22. Calem, PS, Rizzo, JA, “Financing Constraints and Investment: New Evidence from Hospital Industry Data,” Journal of Money, Credit and Banking , 27: 1002–1014 (1995).

23. Supra , n.12; Wheeler, JR, Burkhardt, J, Alexander, JA, Magnus, SA, “Financial and

Organizational Determinants of Hospital Diversifi cation into Subacute Care,” Health Services Research , 34: 61–81 (1999).

24. Magnus, SA, Smith, DG, “Better Medicare Cost Report Data Are Needed to Help Hospi-tals Benchmark Costs,” Health Care Manage-ment Review , 25: 65–76 (2000).

25. Damodaran, A, “Damodaran Online,” http://pages.stern.nyu.edu/~adamodar/ , accessed May 10, 2010.

26. Supra , n.19. 27. We explored the possibility of including the

number of HMOs and the Medicare HMO penetration rate, but could not include them due to substantial missing data. We estimated our model including and excluding these variables and found insignifi cant coeffi cients on the managed care variables. Although the standard errors were larger due to the reduced sample size, there was no meaning-ful difference in the size of the coeffi cients on the fi nancial market performance variables. Therefore, to preserve observations we meas-ure competition solely with the number of hospital beds.

28. StataCorp., Stata Statistical Software: Release 10, College Station, Texas, StataCorp LP (2007).

29. Supra , n.15. 30. Pollard, L, “Short-Lived Financing Options

Can Make Capital Projects Feasible in Tough Economy,” Healthcare Financial Management Association Tarheel News , 44: 10–11 (2010).

31. Supra , n.1. 32. Bazzoli, GJ, Chen, H, Zhao, M, Lindrooth,

RC, “Hospital Financial Condition and the Quality of Patient Care,” Health Economics , 17: 977–995 (2008).

51

Nursing Home Safety: Does Financial Performance Matter?

Reid M. Oetjen, Mei Zhao, Darren Liu, and Henry J. Carretta

Objectives: This study examines the relationship between fi nancial performance and selected safety measures of nursing homes in the State of Florida.

Methods: We used descriptive analysis on a total sample of 1,197. Safety information was from the Online Survey, Certifi cation and Reporting (OSCAR) data of 2003 to 2005, while the fi nancial perform-ance measures were from the Medicare cost reports of 2002 to 2004. Finally, we examined the most frequently cited defi ciencies as well as the relationship between fi nancial performance and quality indicators.

Results: Nursing homes in the bottom quartile of fi nancial performance perform poorly on most resident-safety measures of care; however, nursing homes in the top two fi nancial categories also experienced a higher number of defi ciencies. Nursing homes in the next to lowest quartile of fi nancial performance category best perform on most of these safety measures.

Conclusions: The results reinforce the need to monitor nursing home quality and resident safety in US nursing homes, especially among facilities with poor overall fi nancial performance.

Key words: nursing homes, safety, quality, fi nancial performance.

Florida’s nursing home industry has experienced close scrutiny since the mid-1990s when it had a signifi cant

increase in litigation for poor quality. 1 In 2004, Florida’s state legislature enacted Senate Bill 1202, which raised nursing home staffi ng standards in an attempt to improve the quality of care provided in nursing homes. Its nursing homes have achieved a reduction in total defi ciencies, in both frequency and severity, since the enactment of this law. The quality of care in Florida nursing homes, however, still cannot compete with that of the majority of states, with Florida ranking 35th in total number of nursing home defi ciencies in 2006. 2 These issues warrant a closer look at nursing home safety.

Around the same time of the passage of this bill, Florida’s nursing home industry experienced signifi cant fi nancial diffi culties. In 2003, its average shortfall in Medicaid reimbursement equaled $11.76 per Medicaid patient-day. This number increased to $14.38 in 2004 and again to $14.58 in 2006. 3 Medi-care has historically cross-subsidized these nursing homes, but this source of revenue

has declined with refi nements to the Medi-care payment system. 4

The multidimensional concept of qual-ity includes many dimensions, especially as it pertains to nursing homes. 5 The Joint Commission 6 and the Institute of Medicine 7 include safety as one of their dimensions of quality; the Joint Commission defi ning it as “the degree to which the healthcare


Reid M. Oetjen, PhD, MSHSA, is an Assistant Professor, Director Health Services Administration Undergraduate Program, University of Central Flor-ida, Orlando, Florida.

Mei Zhao, PhD, is an Associate Professor, University of North Florida, Jacksonville, Florida.

Darren Liu, DrPH, MS, MHA, is a Visiting Assistant Professor, University of Central Florida, Orlando, Florida.

Henry J. Carretta, PhD, MPH, is an Assistant Profes-sor, Florida State University, Tallahassee, Florida.

Acknowledgments: This research was supported in part by a grant from the University of North Flor-ida Foundation/Brooks Health Foundation Dean’s Professorship.


intervention minimizes risks of adverse out-comes for both patient and provider and the degree to which the risk of an intervention and risk in the care environment are reduced for the patient and others, including the healthcare provider.” 8

Despite the fact that many continue to examine nursing home industry quality, only a few pieces of literature address nursing home safety as compared with care delivered in other settings. Previous research on nurs-ing home quality focused on organizational issues in nursing homes, such as quality improvement initiatives, teamwork, commu-nication, and leadership. Other studies inves-tigated such specifi c patient-safety issues as falls, pressure sores, restraint usage, medi-cation administration, and infection rates. 9 To date, only O’Neill, Harrington, Kitch-ener, and Saliba 10 have linked nursing home fi nancial performance to quality of nursing home care. Although many of these studies examined safety issues in nursing homes, no research to date has investigated the rela-tionship between fi nancial performance and nursing home safety.

Despite the lack of research in this area, a link regarding the fi nancial performance and safety can be hypothesized based upon the fact that one can argue that investments that affect staffi ng levels, training budgets, and the provision of additional direct care services ultimately impact nursing home quality and safety. Given that many of these activities that enhance resident care qual-ity involve considerable costs, the fi nancial performance of nursing facilities may prove to be a valuable predicator of safety, one of the key dimensions of quality performance. Therefore, nursing homes experiencing poor fi nancial performance may eliminate some of these activities. In fact, existing

quality studies in the hospital industry sup-port this assertion. 11 This study is explora-tory in nature and examines the relationship between fi nancial performance and select safety measures in nursing homes in Florida from 2003 to 2005.

Financial viability and quality of nurs-ing home care are of signifi cant interest to Florida’s taxpayers and government agen-cies, as nearly 17 percent of the population has reached the age of 65 or older—the highest percentage of any state. 12 Efforts in monitoring fi nancial stability and safety of the care in Florida’s nursing homes prove critical to ensuring care for the state’s elder population.

Methods

Sample and Data

Data were collected from Florida’s fed-erally certifi ed nursing homes for the years 2003 to 2005. Only freestanding nursing homes were included in the analysis. Nurs-ing homes that were part of another facility, such as an acute-care or rehabilitation hos-pital, were excluded because these facilities generally have higher reimbursement and staffi ng levels.

The data sources for this study included the Online Survey, Certifi cation and Report-ing (OSCAR) data reports for 2003 to 2005, and the Centers for Medicare & Medicaid Services (CMS) Medicare cost reports for 2002 to 2004.

A national database, OSCAR comprises self-reported nursing home information pro-vided to state surveyor agencies during annual inspections. This database included data on quality and resident safety indicators—the total number of defi ciencies cited during each

Nursing Home Safety: Does Financial Performance Matter? 53

standard recertifi cation survey. The Medicare cost-reports data for nursing homes provided data on fi nancial performance.

To eliminate extreme outliers, nursing homes with a total margin beyond three standard deviation points from the mean in either direction were excluded. Approxi-mately 400 nursing homes remained in each year, for a total sample of 1,197; the study used descriptive analysis.

Safety Measures

This study utilized nine safety measures from OSCAR data—specifi cally, the total number of defi ciencies reported for each nursing home by surveyors as a result of the annual federal survey mandated by CMS for all facilities participating in the Medic-aid or Medicare programs. Although federal law requires nursing homes to be surveyed each fi scal year, investigators can conduct surveys up to 15 months from the previ-ous survey. 13 CMS defi nes the standards the nursing home industry must meet to receive

reimbursement from Medicare and Med-icaid. Violations of these standards result in defi ciencies, reported through OSCAR databases. 14

Figure 1 shows the most frequently cited defi ciencies in Florida related to safety for 2006. 15 Among the top-20 cited defi ciencies, this study focused on the following nine to represent nursing home safety:

1. Food sanitation (F-371); 2. Records complete (F-514); 3. Accuracy of assessments (F-278); 4. Assessment by qualifi ed staff (F-282); 5. Drug storage (F-432); 6. Pharmacy procedures (F-426); 7. Infection control (F-441); 8. Medication errors greater than 5 per-

cent (F-332); and 9. Unnecessary drugs (F-329).

These nine measures represent both organizational issues and specifi c clinical processes related to resident safety.

Figure 1. Most Frequently Cited Nursing Home Defi ciencies in Florida, 2006

Rank Tag Requirement Facilities Cited (%)

1 F-371 Food sanitation 50.73%

5 F-514 Records complete 28.01%

7 F-278 Accuracy of assessments 20.09%

8 F-282 Assessment by qualifi ed staff 18.18%

10 F-432 Drug storage 17.45%

12 F-426 Pharmacy procedures 16.86%

13 F-441 Infection control 16.42%

18 F-332 Medication errors > 5% 14.22%

20 F-329 Unnecessary drugs 11.44%

Adapted from Cowles CM: Nursing Home Statistical Yearbook.

McMinnville, OR, Cowles Research Group, 2007.


To better understand what each meas-ure entails, a brief description of the intent and safety implications of each regulation follows together with the validity of these measures.

Food Sanitation

The food sanitation regulation aims to prevent the spread of food-borne illness by ensuring that nursing homes store, prepare, distribute, and serve food under sanitary conditions. Compliance with this regulation proves especially important because food-borne illness often becomes fatal for nursing home residents. 16

Records Complete

The records complete regulation endeav-ors to ensure each nursing home maintains accurate, organized, accessible, and com-plete clinical records for each resident. The presence of a complete record that provides an accurate functional representation of the actual experience of the resident signals an indication that the facility knows the status of its patients and has adequately planned for each resident’s care. 17

Previous research has shown that the com-munication of inappropriate or inaccurate information in nursing homes were barri-ers to timely care, which ultimately has the potential of affecting the safety of nursing home residents. Based upon this reasoning, the completeness and accuracy of a resi-dent’s record could have a signifi cant impact on the resident’s safety. 18

Accuracy of Assessments

The accuracy of assessments regulation attempts to guarantee each resident receives an accurate assessment by appropriate

personnel to ascertain each resident’s medi-cal, functional, and psychosocial problems. This assessment provides a baseline for ongoing assessment and assures residents have safe-care plans. 19

Assessment by Qualifi ed Staff

The purpose of the assessment by quali-fi ed staff regulation is to ensure qualifi ed staff members care for residents. Direct car-egivers must be knowledgeable about the care, services, and expected outcomes of the care they provide; otherwise, the safety of this care can become suspect. 20 This is an important indicator of safety because research has shown that improved out-comes are associated with use of properly trained nurses. 21 Thus, residents who are assessed by unqualifi ed, or poorly trained staff, may be at risk of receiving improper assessments which could negatively impact patient safety.

Drug Storage

The drug storage regulation maintains only authorized personnel should have access to drugs. Nursing homes must limit access to drugs to authorized personnel and store all drugs in locked compartments in accordance with all state and federal laws to prevent residents from becoming exposed to potentially dangerous situations. 22

Pharmacy Procedures

The pharmacy procedures regulation attempts to safeguard the drug needs of each resident. To maintain the optimal health and functional status of each resident, each facil-ity “must provide pharmaceutical services (including procedures that assure the accu-rate acquiring, receiving, dispensing, and


administering of all drugs and biologicals) to meet the needs of each resident.” 23 Previ-ous research regarding adverse drug events (ADEs) in nursing homes found that most errors occurred most often at the stage of prescribing and monitoring. 24 Thus, there is a clear link between the proper distribution and storage of drugs (drug storage, phar-macy procedures) and patient safety.

Infection Control

The infection control regulation requires the facility to have an effective infection control program for investigating, control-ling, and preventing infection; it ensures each facility protects its residents from the transmission of disease and infection and provides a safe, sanitary, and comfortable environment. 16 Better nursing home out-comes have been linked to the presence of an effective infection control program; thus, an effective infection control program is critical in maintaining patient safety. 25

Medication Errors Greater Than 5 Percent

This regulation aims to safeguard residents from medication errors in excess of 5 per-cent. If facilities have error rates in excess of 5 percent, this indicates systemic problems exist within their drug-distribution systems. 26 Previous study regarding medication errors in nursing homes has shown that medication errors have the potential to impact patient safety. In fact, researchers have shown that 7 percent of medication errors have the poten-tial to harm patients. 27

Unnecessary Drugs

The unnecessary drugs regulation ad -dresses each resident’s drug regimen to en -sure they are not prescribed any unnecessary

drugs. In the interest of safety, residents should only receive prescriptions for psy-chopharmacological drugs when they suffer from mental illness and not from underlying environmental or psychosocial stressors. 28

Florida’s Most Frequently Cited Defi ciencies

These nine defi ciencies help paint a pic-ture of resident safety. Unnecessary drugs, drug storage, pharmacy procedures, and medication errors represent critical meas-ures necessary to assuring residents receive proper medications and dosages. Complete records, accurate assessments conducted by qualifi ed staff members, and infec-tion control serves as a proxy underlying management practices. The presence of defi ciencies in these areas may indicate managerial oversight lacks something criti-cal to providing the best quality of care. Among these measures, food sanitation represents the number 1 defi ciency in Flor-ida’s nursing homes. 29

Financial Performance Measures

Total margin, the excess of revenue over expenses divided by total revenues, repre-sented fi nancial performance in this analy-sis. It refl ects profi ts from both nursing home operations and nonoperational sources (typi-cally investment income). Owing to typical lag in time that occurs between a facility’s fi nancial problems and any subsequent qual-ity and safety changes, nursing home fi nan-cial measures from the previous year were collected.

This study classifi ed the fi nancial per-formance of each nursing home relative to all nursing homes based on their rank in the percentile distribution of total margin. In particular, it distinguished facilities based on whether they fell into the fi rst, second, third,


or fourth quartiles. Previous studies have used the percentile distribution of fi nancial indicators to distinguish nursing facilities’ fi nancial performance. 10 The advantage of classifying nursing homes in this way emerges from the ability to assess whether relatively broad fi nancial performance cat-egories, rather than small incremental dif-ferences, are associated with resident safety problems for study subjects.

Results

The median total margin for this analysis was 0.8 percent (ranging from -93.2% to 29.3%); the fi rst quartile was -4.2 percent, and the third quartile was 4.2 percent. Figure 2 presents results for the four drug-related quality indicators. Drug storage, prescrip-tion administration, and distribution play

important roles in meeting the needs of residents and ensuring nursing home safety. Figure 3 presents the results for the infection, staffi ng, and management quality indicators. The food sanitation indicator stands alone, because it represents the most commonly occurring defi ciency ( see Figure 3).

Drug-related defi ciencies were found in 5.9 percent to 24.1 percent of the facilities, depending on their fi nancial performance ( see Figure 2), and from 8.7 percent to 23.1 percent for the indicators in Figure 3. The proportion of nursing homes with defi cien-cies was higher in the lowest fi nancial per-formance group (quartile 1), compared with facilities in the second lowest fi nancial per-forming category (quartile 2). For example, 21.4 percent of nursing homes in the low-est fi nancial performance category received citations for defi ciencies in accuracy of

Figure 2. % of Facilities Cited for Defi ciencies over Four Financial Categories: Drugs and Management

0<25%

Total Margin25%–50%

Total Margin50%–75%

Total Margin>75%

Total Margin

5

10

15

20

25

30

% o

f F

acili

ties

Cit

ed

Unnecessary DrugsDrug StoragePharmacy ProceduresMedication Errors > 5%


assessments; this number was only 13.9 percent in the second lowest fi nancial per-formance category. Similarly, drug storage citation was 24.1 percent and 16.4 percent for the poorest fi nancial performance cat-egory and the second poorest category, respectively.

Interestingly, six of these eight safety indicators occurred less often among nurs-ing homes in the second lowest perform-ance category, rather than those in the highest fi nancial performers in the third and fourth quartiles. For example, the pro-portion of nursing homes cited for infec-tion control was 12.3 percent in the second lowest fi nancial performance quartile, compared with 17.1 percent in the second highest fi nancial performing category ( see Figure 3).

Similarly, unnecessary drug citations occurred in only 8.9 percent of the facilities in the second quartile compared with 12.7

percent in the third quartile ( see Figure 2). This fi nding implies nursing homes experi-encing very poor fi nancial performance may not have suffi cient resources to ensure resi-dent safety and the proper level of quality to meet the requirements for federal and state survey agencies. Therefore, they are more likely to receive citations. On the other hand, nursing homes with the highest margins may have achieved these results by sacrifi cing the quality of care provided to and safety of its residents.

Figure 4 illustrates the relationship be -tween fi nancial performance and food- sanitation citations. Nursing homes in the poorest fi nancial group had the highest per-centage of food sanitation citations, whereas those in the third fi nancial group had the lowest proportion of citations for food defi -ciencies. This contradicts the fi ndings for most of the other indicators, which occurred least in the second fi nancial group.

0<25%

Total Margin<25%–50%


Total Margin>75%

Total Margin

5

10

15

20

25

30

% o

f F

acili

ties

Cit

ed

Accuracy of AssessmentsAssessment by Qualified StaffInfection ControlRecords Complete

Figure 3. % of Facilities Cited for Defi ciencies over Four Financial Categories: Staffi ng and Management


Discussion

The nursing home industry in Florida has undergone an intense period of public scru-tiny regarding quality and resident safety while simultaneously experiencing wide-spread fi nancial diffi culties. The fi ndings from this study suggest these two phenom-ena may have some relation to each other.

Results indicate nursing homes in the bot-tom quartile of fi nancial performance per-form poorly on most of the resident safety measures of care. However, nursing homes in the next to lowest quartile of fi nancial per-formance category perform the best on most of these safety measures. Nursing homes in the higher margin categories (quartiles 3 and 4) tend to have more defi ciency citations, as compared with those in the second quartile of fi nancial performance. These fi ndings are consistent with previous study fi ndings that found increased margin is more likely

to affect quality adversely in proprietary facilities. 30

Our study had several limitations. First, although Florida has more than 600 nursing homes, this study represents only 4 percent of US nursing facilities. Therefore, the results of this analysis cannot be general-ized to other states. Future studies should expand the sample size to the national nursing homes and further examine the relationship between fi nancial perform-ance and specifi c safety indicators. Sec-ond, because this study only includes nine measures of nursing home safety, these measures cannot fully refl ect the dimension of nursing home quality. Future research should examine other indicators that relate to safety.

Third, data in the OSCAR and CMS cost report fi les have limitations associated with the study of nursing home safety. That is, the surveyors who collect OSCAR exercise

Figure 4. % of Facilities Cited for Defi ciencies over Four Financial Categories: Food Sanitation

43<25%



Total Margin>75%

Total Margin

44

45

46

47

48

49

50

51

52

% o

f F

acili

ties

Cit

ed

Food Sanitation


judgment could only base on their education and experience. This may lead to a low inter-rater reliability among surveyors. This may understate or overstate the frequency and severity of quality issues in the annual sur-vey process. 31

Fourth, data from OSCAR surveyors summarize facility-level data, rather than resident-level data, which preclude a more refi ned approach than one might use with patient-level assessment of safety measures. Additionally, OSCAR data only provide a snapshot of safety at a point in time; they may not accurately represent some facilities over longer periods. Nursing home fi nancial data from CMS may have shortcomings in relation to their accuracy and completeness because many fi nancial statements were not audited.

Finally, the analysis only comprised one measure of fi nancial performance; future research should include many other factors that could infl uence nursing home resident safety, such as size, ownership, chain affi lia-tion, and staffi ng, so that the potentially con-founding effects could be controlled for.

Despite these limitations, this study does suggest the importance of certain issues for policy makers and providers. In particular, the results reinforce the need to monitor nursing home quality and resident safety in US nursing homes, especially among facili-ties with poor overall fi nancial performance. Nursing homes with the weakest fi nancial performance appear to have more defi -ciency citations relevant to patient-safety measures. This suggests that such homes may not have fi nancial resources to pro-vide residents with a safe environment. Pol-icymakers may want to consider increased surveillance of these poor performers or changes to reimbursement formulas to

increase the likelihood that residents in these facilities remain safe.

Nursing homes in the top two fi nancial categories also experienced a higher number of defi ciencies. This suggests that the top performers may overemphasize fi nancial performance at the expense of resident safety. On the other hand, nursing homes in the sec-ond performance category outperformed both lower performing and higher perform-ing facilities in all but two safety measures. This implies they may have arrived at a more optimal level of resource allocation for patient safety measures than other facilities in this study; this might prove useful in fur-ther assessment of the relationship between fi nancial performance and nursing home patient safety indicators.

Conclusion

The nursing home industry plays a critical role in providing long-term care to elders. This study makes several contributions to the literature on nursing home safety. First, it examines the overall fi nancial health of nursing homes, rather than the effects of a specifi c policy change. Second, this study has clearly shown the poorest performing nursing homes produce the least favorable safety results. In order for this segment of the industry to improve quality, it must either improve effi ciency or rely on policy makers to increase reimbursement. Lastly, it provides evidence that nursing homes in the top two fi nancial performance categories have more defi ciency citations compared with those in the second quartile of fi nancial performance. This fi nding warrants closer examination and further study to ensure Florida nursing homes provide their residents a safe, quality environment.


REFERENCES

1. Tanner, R, Nursing Home Liability Insurance (issue brief), National Conference of State Legislatures Health Policy Tracking Service (July 2004).

2. Schaefer, M, Burwell, B, Medstat, T, “The Nursing Home Liability Insurance Market: A Case Study of Florida,” US Department of Health and Human Services ( aspe.hhs.gov/daltcp/reports/2006/nhliab-fl .pdf ) (June 1, 2006); Cowles, CM, Nursing Home Statis-tical Yearbook , McMinnville, OR, Cowles Research Group (2007).

3. A Report on Shortfalls in Medicaid Funding for Nursing Home Care, BDO Seidman, LLP ( ahca.org/brief/seidmanstudy0606.pdf ) (June 2006).

4. Konetzka, RT, Norton, EC, Stearns, SC, “Medi-care Payment Changes and Nursing Home Quality: Effects on Long-Stay Residents,” Intl J Health Care Finance Econ , 2006; 6:173–189.

5. Zimmerman, DR, Karon, SL, Arling, G, et al ., “Development and Testing of Nursing Home Quality Indicators,” Health Care Finance Rev , 16:107–127 (1995).

6. 2006–2007 Comprehensive Accreditation Manual for Behavioral Health Care, Oakbrook Terrace, Illinois, Joint Commission (2006).

7. Crossing the Quality Chasm: A New Health System for the 21st Century, Washington, DC, National Academies Press (2001).

8. Supra , n.6. 9. Scott-Cawiezell, J, Vogelsmeier, A, “Nursing

Home Safety: A Review of the Literature,” Annu Rev Nurs Res , 24:179–215 (2006).

10. O’Neill, C, Harrington, C, Kitchener, M, Saliba, D, “Quality of Care in Nursing Homes: An Analysis of Relationships Aamong Profi t, Quality, and Ownership,” Med Care , 41:1318–1330 (2003).

11. Bazzoli, G, Clement, J, Lindrooth, R, Chen, H, Aydede, S, et al ., “Hospital Financial Con-dition and Operational Decisions Related to the Quality of Hospital Care,” Med Care , 64 (2):148–168 (2007); Ciliberto, F, Lindrooth, R, “Exit from the Hospital Industry,” Economic Inquiry , 45(1):71–82 (2007).

12. Bernstein, R, “Census Bureau Estimates Number of Children and Adults in the States

and Puerto Rico (press release), US Cen-sus Bureau ( www.census.gov/Press-Release/www/releases/archives/population/004083.html )(Mar. 10, 2005).

13. Annual Quality Improvement Report on the Nursing Home Survey Process and Progress Reports on Other Legislatively Directed Activities, Minnesota Department of Health (health.state.mn.us/divs/fpc/AQIRnhs2004.pdf) (2004).

14. Harrington, C, Carrillo, H, Thollang, SC, et al ., Nursing Facilities, Staffi ng, Residents and Facility Defi ciencies, 1993–1999 , San Fran-cisco, University of California Department of Social and Behavioral Sciences (2000).

15. Cowles, supra , n.2. 16. Long-Term Care Survey , Washington, DC,

American Health Care Association (Dec. 2006).

17. Id. 18. Longo, D, Youngh, J, Mehr, D, Lindbloom, E,

Salerno, L, “Barriers to Timely Care of Acute Infections in Nursing Homes: A Preliminary Qualitative Study,” Journal of the American Medical Directors Association , 5:s5–s10 (2004).

19. Bernstein, supra , n.12. 20. Supra , n.16. 21. Mor, V, “Hospitalization of Nursing Home

Residents: A Review of Clinical, Organiza-tion and Policy Determinants,” unpublished manuscript, Brown University Gerontology Center, Providence, Rhode Island (1999).

22. Supra , n.16. 23. Id. 24. Gurwitz, J, Field, T, Avorn, J, McCormick, D,

Jain S, Eckler, M, “Incidence and Prevent-ability of Adverse Drug Events in Nursing Homes, The American Journal of Medicine 109:87–94 (2000); Gurwitz, J, Ochon, P, Har-rold, L Cadoret, C, “The Incidence of Adverse Drug Events in Two Large Academic Long-Term Care Facilities,” The American Journal of Medicine , 118(3):251–258 (2005).

25. Supra , n.18. 26. Supra , n.16. 27. Barker, K, Flynn, E, Pepper, G, Bates, D,

Mikeal, R, “Medication Errors Observed in 36


Health Care Facilities, “ Archives of Internal Medicine , 162(16):1897–1903 (2002).

28. Supra , n.16. 29. Cowles, supra , n.2. 30. Supra , n.10.

31. Nursing Homes: Despite Increased Oversight, Challenges Remain in Ensuring High-Quality Care and Resident Safety. Washington, DC, Government Accountability Offi ce publica-tion No. (GAO) 06-117, 2005.

62

Cigarette Taxes and Respiratory Cancers: New Evidence from Panel

Co-Integration Analysis Echu Liu, Wei-Choun Yu, and Hsin-Ling Hsieh

Using a set of state-level longitudinal data from 1954 through 2005, this study investigates the “long-run equilibrium” relationship between cigarette excise taxes and the mortality rates of respiratory cancers in the United States. Statistical tests show that both cigarette excise taxes in real terms and mortality rates from respiratory cancers contain unit roots and are co-integrated. Estimates of co-integrating vectors indicated that a 10 percent increase in real cigarette excise tax rate leads to a 2.5 percent reduction in respiratory cancer mortality rate, implying a decline of 3,922 deaths per year, on a national level in the long run. These effects are statistically signifi cant at the one percent level. Moreover, estimates of co-integrating vectors show that higher cigarette excise tax rates lead to lower mortality rates in most states; however, this relationship does not hold for Alaska, Florida, Hawaii, and Texas. Key words: cigarette tax, panel co-integration, panel unit roots.

Since the fi rst Surgeon General’s report on the health hazards of smoking was issued in 1964, various levels of gov-

ernment have consistently implemented tobacco control policies in order to promote the reduction of tobacco use. According to the US Department of Health and Human Services, 1 the most important economic policy for promoting reduction of tobacco use is increased taxation on tobacco prod-ucts. Governments (especially state gov-ernments) have vigorously used cigarette excise taxes as a tool to discourage smoking and to increase state revenue.

The majority of economics literature ana-lyzing the policy effect of cigarette excise taxes focuses on the estimation of the effect of increased taxation on cigarette consumption, and shows that increased cigarette taxation signifi cantly reduces cigarette smoking by discouraging teenagers from initiating smok-ing, increasing smoking cessation among adults, and decreasing cigarette consump-tion among ongoing smokers. 2 As pointed out by Warner, 3 the ultimate importance of tax-induced changes in cigarette consump-tion is improved health consequences. How-ever, studies that quantify the health benefi ts

of increases in cigarette taxes are rare. Our article sheds light on this line of research.

Echu Liu, PhD, is an Assistant Professor of Health Care Management in the School of Allied Health, Southern Illinois University at Carbondale, Carbon-dale, Illinois. His main research interests are in health economics, applied microeconomics, and applied econometrics. He can be reached at [email protected].

Wei-Choun Yu, PhD, is an Associate Professor of Economics at Winona State University, Winona, Min-nesota. His main research interests are in forecasting methods, time series econometrics, applied economet-rics, empirical fi nance, empirical macroeconomics, and empirical microeconomics. He can be reached at [email protected].

Hsin-Ling Hsieh, PhD, is an Assistant Professor of Economics, Northern Michigan University, Mar-quette, Michigan. Her main research interests are in labor economics and applied microeconomics. She can be reached at [email protected].

Acknowledgements: The authors are very grateful to Peter Pedroni for providing us the computer codes. We also thank participants at the 79th Annual Conference of the Southern Economic Association (San Antonio, 2009) and the 47th Annual Meeting of the Missouri Valley Economic Association (St. Louis, 2010) for helpful comments.


Cigarette Taxes and Respiratory Cancers: New Evidence from Panel Co-Integration Analysis 63

To the best of our knowledge, only two previous studies have estimated the effect of increased taxation on the general popula-tion’s health, and they provided mixed evi-dence. Liu et al . 4 used a set of state-level panel data from 1970 through 2000 in order to estimate the effect of cigarette excise taxes on changing the rates of morbidity caused by strokes and heart attacks, which are among the top diseases caused by smoking. They found that tax increases led to small and statistically insignifi cant decreases in inci-dences of stroke and heart attacks. Moore 5 used a set of state-level panel data from 1954 to 1988 in order to analyze the effect of ciga-rette excise tax changes on smoking-related mortality rates. The evidence shows that a 10 percent tax increase leads to a 1.5 percent decrease in respiratory cancer mortality rates and a 0.5 percent decrease in cardiovascular disease mortality rates. Both effects are sta-tistically signifi cant at the one percent level.

Our research extends the discussion from these two previous studies and estimates the effectiveness of cigarette taxes in reducing the mortality rates of respiratory cancers, which are responsible for many deaths and some of the most devastating disabilities in the United States. We expect that the estimate will be negative and statistically signifi cant, because many medical studies have provided solid evidence regarding the relationship between cigarette smoking and various negative health outcomes. 6 In addi-tion, the literature has shown that increased cigarette taxation signifi cantly reduces the prevalence of cigarette smoking, as men-tioned previously.

Our study has three contributions. First, compared with Moore, 7 which used data from 48 states (Alaska, Hawaii, and the District of Columbia excluded) over 1954 to 1988 and

Liu et al. , 8 our data cover a longer and more contemporary time span (from 1954 through 2005). Second, to avoid the famous “spuri-ous regression” problem, 9 which leads to problematic inference and spurious results of estimation, non-stationarity of the data was tested before proceeding with our estima-tion. Third, our study estimates a “long-run equilibrium” relationship between cigarette taxation and the death rates of a leading smoking-attributable disease by conducting co-integration analysis. Estimation of this “long-run equilibrium” relationship would provide a better idea of the health benefi ts induced by increasing taxes on cigarette because respiratory cancers can develop many years after smoking begins, and the risk of cancer decreases slowly after smok-ing stops. 10

The rest of this article is organized as fol-lows: a description of the sources of the data, an outline of the estimation methods, results, conclusions, along with directions for future research.

Data

Our data cover the years 1954 to 1959, 1961 to 1964, and 1966 to 2005 (data for 1960 and 1965 are unavailable) and the 50 states and the District of Columbia. The sources of mortality rates for respiratory cancers (ICD8 160-164, ICD9 160-165, and ICD10 C30-C39) are Vital Statistics of the United States, Annual Summary and the WONDER Database of The Centers for Disease Control and Prevention (CDC). The mortality rates are age-adjusted to the 2000 US standard population. Because the comparability ratio of ICD10 to ICD9 for cancers is 1.01, no signifi cant problems should occur by using mortality rate data


across the boundary of ICD versions for our analysis.

The nominal state and federal ciga-rette excise tax rates were obtained from Orzechowski and Walker. 11 The real cigarette excise tax rate is defi ned as the sum of nominal state and federal cigarette excise tax rates (in cents) per pack defl ated by the consumer price index (CPI) with a value of 100 in 1983.

The panel data of mortality rates and real cigarette tax rates are shown in Figures 1 and 2. As Figure 1 demonstrates, the real cigarette tax rates fl uctuated before 1970,

declined between 1970 and 1980, and increased gradually after 1980. Also, the overall increases in the real cigarette tax rates after 1980 are notable in 46 states and the District of Columbia. Moreover, the series of mortality rates in Figure 2 do not fl uctuate around a stable mean, and show signifi cantly positive trends. Together, these suggest the need to check the non-stationarity, or the existence of unit roots, of our data before applying any standard approaches, such as the fi xed effects model, in order to analyze the linkages between cigarette taxation and

Figure 1. The Real Cigarette Excise Tax Rate (per Pack, Defl ated by CPI in 1983)—50 States and the District of Columbia, 1954 to 2005


mortality rates from respiratory cancers. Otherwise, a “spurious regression” problem, including invalid t -statistics, high R 2 , and a low value of the Durbin-Watson statistic, could occur when both series of mortality and real cigarette tax rates contain unit roots ( i.e ., they are non-stationary), and any stand-ard estimation approach is applied. 12 There-fore, we will test the existence of unit roots in our data fi rst, and afterwards apply an appropriate approach to estimate the effect of raising cigarette taxes on the mortality rates from respiratory cancers.

Methods and Results

Panel Unit Root Tests

In order to test for the existence of unit roots in the context of panel data, we fi rst use the tests proposed by Levin, Lin, and Chu (LLC), 13 and Im, Pesaran, and Shin (IPS). 14 The ideas of LLC and IPS tests in our study are carried out using Equation 1, where i = 1, 2, ..., 51, and t = 1, 2, ..., 50.

Δy it = α

i + γ

it + δ

i y

it – 1 + Σ

L=t

Pi

ϕiΔ y

it – L + ε

it (1)

Figure 2. The Mortality Rate of Respiratory Cancers (Cases per 100,000 People)—50 States and the District of Columbia, 1954 to 2005


In Equation 1, y it is the logarithm of real

cigarette excise tax rate or the logarithm of mortality rate of respiratory cancers for state i in year t ; α

i denotes the state-

specifi c fi xed effects, t denotes determin-istic trends, and δ

i = ρ

i − 1, where ρ

i is the

autoregressive coeffi cient. Δ is the lag operator, and ε

it is the error term. Under

the null hypothesis of δi = 0, y

it contains

unit roots ( i.e ., non-stationary). Under the alternative hypothesis, LLC assumes that y

it is stationary whereas IPS assumes that a

fraction of y it is stationary. In our analysis,

the number of lags (Pi ) in (1) is selected

based on Schwarz Information Criterion (SIC), a Bartlett kernel is used for spectral estimation, and Newey-West 15 is used to choose bandwidth.

Several studies have noted that the panel unit root tests of LLC and IPS have low power and have suggested testing the null hypothesis of stationarity against the alter-native of non-stationarity instead. 16 Hadri’s 17 residual-based Lagrange multiplier test is one example of those suggested tests, and we will present the results of Hadri’s test with those of LLC and IPS for the purpose of checking the robustness of our results.

The results of our panel unit root tests are presented in Figure 3. For both LLC and IPS tests, we do not reject the null hypothesis of unit roots for both data series (the loga-rithm of real cigarette excise tax rates and the logarithm of mortality rates of respira-tory cancers). Using Hadri’s test, we rejected the null hypothesis of stationarity for both data series. Based on these test results, we conclude that the two data series—the loga-rithm of real cigarette excise tax rates and the logarithm of mortality rates of respira-tory cancers—contain unit roots.

Panel Co-Integration Tests

It has been shown that the logarithms of real cigarette excise tax rates and mortality rates of respiratory cancers contain unit roots; therefore, it seems necessary to eliminate their non-stationarity by differencing and to estimate the relationship between these two data series using only differenced vari-ables. However, as suggested by Kennedy, 18 we should test for co-integration before dif-ferencing if the data are found to have unit roots. If any co-integration relationship can be found, then we should estimate the co-integrating vector. Therefore, our next

Figure 3. Tests of Panel Unit Roots for the Logarithms of Mortality and Real Cigarette Tax, 1954 to 2005

Variables Levin, Lin, and Chu Im, Pesaran, and Shin Hadri

Statistic p-value Statistic p-value Statistic p-value

Log mortality of

respiratory cancers

4.179 1.000 10.642 1.000 28.769 0.000

Log cigarette tax

per pack

5.221 1.000 4.785 1.000 10.781 0.000

Notes: 1. The null hypothesis in Levin, Lin, and Chu (2002) and in Im, Pesaran, and Shin (2003) is

that all individual series have unit roots. 2. The null hypothesis in Hadri (2000) is that all individual

series are stationary.


step is to test whether the logarithms of real cigarette excise tax rates and mortality rates of respiratory cancers are co-integrated. If so, we could suggest that a “long-term equi-librium” relationship exists between these two variables and estimate the strength of this relationship.

In this study, we apply the panel co- integration tests developed by Pedroni 19 and Kao. 20 Pedroni’s idea is carried out by esti-mating the following equation:

y it = α

i + γ

it + β

i x

it + ε

it (2)

where i =1, 2, …, 51 and t =1, 2,…, 50. yit is

the logarithm of mortality rate of respiratory cancers for state i in year t , and x

it is the loga-

rithm of real cigarette excise tax rate for state i in year t . The individual state fi xed effect (intercept, α

i) would capture the effects of

characteristics of states that are constant over time. γ

it represents deterministic time trends

that are specifi c to state i , and this term cap-tures the effects of those independent vari-ables other than x

it , such as state income

and demographic variables, on yit. ε

it is the

error term. Pedroni 21 suggests using seven residual-

based statistics to test co-integration for panel data. Four of these statistics are based on estimators that pool the autoregressive coeffi cient across different states for the unit root tests on the estimated residuals (called “panel co-integration statistics”), whereas three of them are based on estimators that average the individual estimated autore-gressive coeffi cients for each state i (called “group mean co-integration statistics”). The null hypothesis of seven tests proposed by Pedroni 22 assumes that there is no co- integration between x

it and y

it.

The results of Pedroni’s 23 panel co- integration tests are reported in Figure 4.

Figure 4. Panel Co-Integration Tests, 1954 to 2005

Pedroni Tests Kao Test

Panel ν-statistic 1.468 (0.071) ADF-Test 19.963 (0.000)

Panel ρ-statistic -7.271 (0.000)

Panel t-statistic

(non-parametric)

-16.317 (0.000)

Panel t-statistic

(parametric)

-16.309 (0.000)

Group ρ-statistic -2.061 (0.020)

Group t-statistic

(non-parametric)

-7.828 (0.000)

Group t-statistic

(parametric)

-7.401 (0.000)

Notes: 1. The null hypothesis in Pedroni (1999, 2004) and Kao (1999) is

that two panels are not co-integrated. 2. p-values are in parentheses. 3.

To choose the appropriate lags, we use the automatic selection method

to minimize Schwarz information criteria (SIC): 2(L/T)+klog(T)/T, where

L is the selected Newey-West (1994) bandwidth parameter, T is the sample

size, and k is the number of degrees of freedom used in model fi tting.


All seven statistics show evidence that the null hypothesis of no co-integration should be rejected. For the purpose of checking the robustness of Pedroni’s 24 test results, we also performed Kao’s 25 augmented Dickey-Fuller test for co-integration of panel data. The test statistic of Kao 26 reported in Figure 4 also rejects the null hypothesis that no co- integration exists between x

it and y

it .

FMOLS Results

Now that we have established that the logarithms of real cigarette excise tax rates

Figure 5. FMOLS Results by State

Individual Coefficient

State Coefficient t-statistic

Alabama -0.440 -2.880***

Alaska 0.290 3.910***

Arizona -0.110 -1.830*

Arkansas -0.490 -2.320**

California -0.680 -5.940***

Colorado -0.410 -2.880***

Connecticut -0.560 -5.240***

Delaware -0.060 -0.610

District of

Columbia

-0.340 -4.690***

Florida 0.340 6.010***

Georgia -0.040 -0.450

Hawaii 0.120 2.880***

Idaho -0.230 -2.610**

Illinois -0.490 -4.570***

Indiana -0.050 -0.720

Iowa 0.020 0.310

Kansas -0.050 -0.480

Kentucky -0.740 -17.390***

Louisiana 0.010 0.300

Maine 0.080 1.360

Maryland -0.290 -4.850***

Massachusetts -0.290 -2.260**

Michigan -0.010 -0.420

Minnesota -0.080 -2.800***

Mississippi -0.600 -4.370***

Missouri 0.040 0.990

Montana -0.180 -1.920*

Nebraska 0.010 0.130

Nevada -0.020 -0.140

New Hampshire -1.190 -3.820***

New Jersey -0.520 -4.870***

New Mexico -0.340 -3.360***

New York -0.730 -6.570***

North Carolina -1.500 -5.440***

State Coefficient t-statistic

North Dakota 0.040 0.170

Ohio -0.050 -1.290

Oklahoma -0.090 -2.210**

Oregon -0.040 -1.110

Pennsylvania -0.010 -0.240

Rhode Island -0.230 -1.080

South Carolina -0.740 -10.980***

South Dakota -0.760 -3.320***

Tennessee -0.380 -4.590***

Texas 0.430 4.160***

Utah -0.270 -0.910

Vermont 0.250 1.260

Virginia -0.170 -2.150**

Washington -0.340 -7.680***

West Virginia -0.600 -5.190***

Wisconsin 0.000 0.000

Wyoming -0.010 -0.070

Panel Coefficient

-0.250 -15.790***

Notes: 1. * indicates signifi cance at the

10 percent level. 2. ** indicates signifi cance

at the 5 percent level. 3. *** indicates

signifi cance at the one percent level.


and mortality rates of respiratory cancers are co-integrated, we use the Fully Modifi ed Ordinary Least Square (FMOLS) technique proposed by Pedroni 27 in order to estimate the co-integrating vector between these two variables. The FMOLS was designed for non-stationary panel data and is able to esti-mate the “long-run equilibrium” relation-ship while allowing the short-term dynamics and fi xed effects to be heterogeneous among members in the panel. FMOLS also accounts for the endogeneity of the independent vari-ables and the correlation and heteroscedas-ticity of the error terms, which are possibly present in time-series or panel data with long time-spans.

The results of FMOLS are displayed in Figure 5. Individual FMOLS estimates and t -statistics for each state are reported in the fi rst 51 entries, and the panel estimate, which is the average of individual FMOLS esti-mates, is reported at the bottom. The major-ity of individual coeffi cients on the logarithm of the tax rates have the expected signs, and the corresponding t -statistics are statisti-cally signifi cant. For example, a 10 percent increase in real cigarette excise tax rate will lead to a 15 percent reduction in respiratory cancer mortality rates in North Carolina, an 11.9 percent reduction in New Hampshire, a 7.6 percent reduction in South Dakota, a 7.4 percent reduction in Kentucky, a 7.4 percent reduction in South Carolina, and a 7.3 per-cent reduction in New York. Also, it seems that cigarette taxation could lead to greater health improvements in areas with rela-tively more tobacco production, e.g ., North Carolina, Kentucky, and South Carolina. 28 However, based on the individual FMOLS estimates reported in Figure 5, greater cig-arette taxation leads to higher respiratory cancer mortality rates in Alaska, Florida,

Hawaii, and Texas. The non-intuitive results deserve future research attention.

Most importantly, based on the panel FMOLS results at the bottom of Figure 5, a 10 percent increase in real cigarette excise tax rate leads to a 2.5 percent reduction in the respiratory cancer mortality rate on a national level in the long run, and this effect is statistically signifi cant at the one percent level (with a t -statistic -15.79). Given the total population of size of 299,398,000 in 2006, this estimate implies that a 10 per-cent increase in taxation would avert 3,922 deaths per year in the United States, because according to the CDC data, there were on average 52.4 respiratory cancer deaths per 100,000 people between 2001 and 2006. This estimate of the number of deaths averted by increasing cigarette taxation is much larger than the estimate calculated by Moore’s 29 fi nding: 2,353 deaths averted per year. We believe that this discrepancy is due to the difference between our statistical modeling and that of Moore.

Conclusions

Many empirical studies have shown that raising taxes on cigarettes is an effective tool for reducing the prevalence of smok-ing in the United States. Yet, few studies have addressed the health benefi ts of rais-ing taxes on cigarettes. Using state-level data for the years 1954 to 1959, 1961 to 1964, and 1966 to 2005, this study attempts to fi ll this gap by estimating the “long-run equilibrium” relationship between respi-ratory cancer mortality rates and cigarette taxes. We fi nd that both data series of respi-ratory cancer mortality rates and cigarette taxes contain unit roots, meaning that they are not stationary. Moreover, we fi nd that


these two data series are co-integrated. Our estimated co-integrated vector shows that in the long run, a 10 percent increase in real cigarette excise tax rate leads to a 2.5 percent reduction in the respiratory cancer mortality rate nationally, which means that 3,922 deaths are averted per year based on the population size of the United States in 2006. Compared with the fi ndings from previous literature, our results suggest a greater magnitude of health benefi ts from cigarette excise taxes. Therefore, increas-ing excise taxes on cigarettes seems to be a valid tool for improving people’s health in the United States. However, the health-improving effects of cigarette taxes differ across various states in our data. The pub-lic health benefi t appears to be larger in tobacco-producing states, such as North Carolina, South Carolina, and Kentucky.

In contrast with the prediction, higher ciga-rette taxation leads to higher mortality rates in Alaska, Florida, Hawaii, and Texas. This result deserves future research attention.

Some limitations of the study should be noted. First, in order to make sure that the co-integrating relationship is uniquely identifi ed, 30 some important factors, such as per capita income, population distribution, environmental legislation ( e.g ., Clean Air Act), are not explicitly controlled in our sta-tistical modeling. Second, our analysis only covers the data up to 2005 because the WON-DER Database of CDC does not provide estimates of mortality rates for respiratory cancers after that year. It would be interest-ing to see if the “long-term equilibrium” relationship would change if newer data are available and more covariates are included in the estimation in a future research.

REFERENCES

1. US Department of Health and Human Serv-ices, “Surgeon General’s Report: Reducing Tobacco Use,” Atlanta (2000).

2. Baltagi, BH, Goel, RK, “Quasi-Experimental Price Elasticities of Cigarette Demand and the Bootlegging Effect,” American Journal of Agricultural Economics , 69(4), 750–754 (1987); Baltagi, BH, Levin, D, “Cigarette Tax-ation: Raising Revenues and Reducing Con-sumption,” Structural Change and Economic Dynamics , 3(2), 321–335 (1992); Chaloupka, F, “Rational Addictive Behavior and Ciga-rette Smoking,” Journal of Political Econ-omy , 99(4), 722–742 (1991); Chaloupka, FJ, Wechsler, H, “Price, Tobacco Control Policies and Smoking Among Young Adults,” Journal of Health Economics , 16(3), 359–373 (1997); Hu, TW, Mao, Z, “Effects of Cigarette Tax on Cigarette Consumption and the Chinese Economy,” Tobacco Control , 11, 105–108 (2002); Lewit, EM, Coate, D, Grossman, M, “The Effects of Government Regulation on Teenage Smoking,” Journal of Law and

Economics, 24(3), 545–569 (1981); Lewit, EM, Coate, D, “The Potential for Using Excise Taxes to Reduce Smoking,” Journal of Health Economics ,1(2), 121–145 (1982); Lewit, EM, Hyland, A, Kerrebrock, N, Cummings, KM, “Price, Public Policy, and Smoking in Young People,” Tobacco Control , 6 (supplement 2), s17–s24 (1997); Meier, KJ, Licari, MJ, “The Effect of Cigarette Taxes on Cigarette Con-sumption, 1955 Through 1994,” American Journal of Public Health , 87(7), 1126–1130 (1997); Peterson, DE, Zeger, SL, Remington, PL, Anderson, HA, “The Effect of State Ciga-rette Tax Increases on Cigarette Sales, 1955 to 1988,” American Journal of Public Health , 82(1), 94–96 (1992); Walbeek, CV, “A Simu-lation Model to Predict the Fiscal and Public Health Impact of a Change in Cigarette Excise Taxes,” Tobacco Control , 19, 31–36 (2010); Wasserman, J, Manning, WG, Newhouse, JP, Winkler, JD, “The Effects of Excise Taxes and Regulations on Cigarette Smoking,” Journal of Health Economics , 10(1), 43–64 (1991).


3. Warner, KE, “Smoking and Health Implica-tions of a Change in the Federal Cigarette Excise Tax,” Journal of American Medical Association , 255(8), 1028–1032 (1986).

4. Liu, E, Rivers, PA, Sarvela, PD, “Does Increas-ing Cigarette Excise Tax Improve People’s Health? The Cases of Heart Attacks and Stroke,” Journal of Health Care Finance , 34(3), 91–109 (2008).

5. Moore, MJ, “Death and Tobacco Taxes,” RAND Journal of Economics , 27(2), 415–428 (1996).

6. Bartecchi, CE, MacKenzie, TD, Schrier, RW, “The Human Costs of Tobacco Use: First of Two Parts,” New England Journal of Medi-cine , 330(13), 907–912 (1994).

7. Supra , n.5. 8. Supra , n.4. 9. Granger, CWJ, Newbold, P, “Spurious Regres-

sions in Econometrics,” Journal of Economet-rics , 2(2), 111–120 (1974).

10. Lubin, JH, Caporaso, NE, “Cigarette Smok-ing and Lung Cancer: Modeling Total Expo-sure and Intensity,” Cancer Epidemiology, Biomarkers and Prevention , 15(3), 517–523 (2006).

11. Orzechowski, W, Walker, R, The Tax Burden on Tobacco: Historical Compilation (2007).

12. Entorf, H, “Random Walks with Drifts: Non-sense Regression and Spurious Fixed-Effect Estimation,” Journal of Econometrics , 80(2), 287–296 (1997).

13. Levin, A, Lin, C-F, Chu, C-S, “Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties,” Journal of Econometrics , 108, 1–24 (2002).

14. Im, KS, Pesaran, MH, Shin, Y, “Testing for Unit Roots in Heterogeneous Panels,” Journal of Econometrics , 115(1) 53–74 (2003).

15. Newey, WK, West, KD, “Automatic Lag Selection in Covariance Matrix Estimation,” Review of Economics Studies, 61(4), 631–653 (1994).

16. DeJong, DN, Whiteman, CH, “Reconsidering ‘Trends and Random Walks in Macroeconomic Time Series’,” Journal of Monetary Econom-ics , 28(2), 221–254 (1991); Kwiatkowski, D, Philips, PC, Schmidt, P, Shin, Y, “Testing the Null Hypothesis of Stationarity Against the Alternative of a Unit Root,” Journal of Econo-metrics , 54, 159–178 (1992).

17. Hadri, K, “Testing for Stationarity in Hetero-geneous Panel Data,” Econometrics Journal , 3, 148–161 (2000).

18. Kennedy, P, A Guide to Econometrics , Black-well Publishing: Malden, MA (2008).

19. Pedroni, P, “Critical Values for Cointegration Tests in Heterogeneous Panels with Multiple Regressors,” Oxford Bulletin of Economics and Statistics, 61(S1), 653–670 (1999); Pedroni, P, “Panel Cointegration: Asymptotic and Finite Sample Properties of Pooled Time Series Tests with an Application to the Purchasing Power Parity Hypothesis,” Econometric Theory , 20, 597–625 (2004).

20. Kao, C, “Spurious Regression and Residual-Based Tests for Cointegration in Panel Data,” Journal of Econometrics , 90, 1–44 (1999).

21. Supra , n.19. 22. Id. 23. Id. 24. Id. 25. Supra , n.20. 26. Id. 27. Pedroni, P, “Fully Modifi ed OLS for Hetero-

geneous Cointegrated Panels,” Advances in Econometrics, 15, 93–130 (2000).

28. Based on the data from 2002 National Agri-cultural Statistics Service, USDA, North Carolina (167,677 acreage) and Kentucky (110,734 acreage) have the largest tobacco farm acreages. South Carolina (30,241 acre-age) has the highest average tobacco acreage per farm (34.6 acreage).

29. Supra , n.5. 30. Supra , n.18.

72

Revisiting the Cost of Medical Student Education: A Measure

of the Experience of UT Medical School–Houston

Elizabeth Gammon and Luisa Franzini

This study uses a cost construction model to estimate the cost of a four-year undergraduate medical education at the University of Texas–Houston Medical School (UT–Houston) in 2006–2007 compared to 1994–1995. The model computes the cost by measuring increasingly inclusive defi nitions of the edu-cational mission: instructional (direct-contact teaching), educational (instructional plus general supervi-sion), and milieu (educational plus research costs). Using the model and adjusting for infl ation, annual cost per student enrolled decreased by 16 percent in 2006–2007 compared to 1994–1995 and total cost decreased by 9 percent. Additionally, the model predicted 190 full-time equivalent (FTE) faculty and 187 FTE residents for 2006–2007 compared to 201 FTE faculty and 258 FTE residents for 1994–1995. Decreases in the cost of educating medical students were driven by (1) the reduction in the number of educator contact hours required for curriculum delivery; (2) change in the mix of educators; and (3) an increase in medical school class size. Key words: costs, cost analysis, curriculum, medical education, medical schools, economic model.

The sea change experienced by aca-demic health centers (AHC) in providing and fi nancing the under-

graduate medical education in Texas and the United States over nearly two decades is well known. 1 AHC faculty and administra-tors alike have responded to the pressures of changing reimbursement structures, ris-ing health care costs, and evolving delivery systems by employing a variety of manage-ment solutions. These efforts have been dis-cussed in department meetings, deliberated on in institutional planning events, debated in state legislatures and in the US Congress, and reported on in the literature. With the recent national debate on health insurance reform including predictions for shortages of physicians, pressures to educate more health care providers will exacerbate the issue of the high cost of medical education. However, the debate regarding the cost of undergraduate medical education has been neglected after the fl urry of interest prompted

by the managed care revolution. This study re-examines the cost of undergraduate med-ical education in anticipation of increased enrollment and cost management.

Introduction

In the late 20th century, there was a strong interest in computing the cost of medi-cal student education at AHC, 2 concerns about tuition costs limiting access to the profession were delineated, 3 and reports on “mission-based management” have became

Elizabeth Gammon, PhD, CPA, is an Assistant Pro-fessor, Management, Policy and Community Health, at the University of Texas School of Public Health.

Luisa Franzini, PhD, is an Associate Professor, Management, Policy and Community Health, Univer-sity of Texas School of Public Health.


Revisiting the Cost of Medical Student Education 73

de rigueur . 4 Measures of outcome of these efforts have been limited to the relatively short time horizon of one year. 5 The present study reports on the cost of medical stu-dent education at the University of Texas–Houston Medical School (UT–Houston) in 2006–2007 compared to 1994–1995 as estimated by a cost-construction model. To determine the cost of medical student edu-cation, we used the cost construction model which was employed in the 1994–1995 study and reported on previously. 6

Measuring the costs at UT–Houston using the same model at two points in time over ten years apart allowed us to identify changes in the cost of undergraduate medical education in a post–managed care environment.

Methods

The cost construction model computes the cost of providing “an adequate level of resources for an educational program in an acceptable educational environment” for the four-year undergraduate medical educa-tion program. 7 The acceptable educational environment is one in which there is ongo-ing research and patient services. Because of the diffi culties of reaching a consensus on the level of research and patient services considered acceptable in a medical school and the diffi culty of allocating costs associ-ated with joint products (for example as dur-ing general supervision of students, when education and patient care occur simultane-ously), the cost construction model consid-ers increasingly inclusive defi nitions of the educational mission and computes three types of costs:

1. Instructional costs (costs for direct-contact teaching);

2. Educational costs (instructional costs plus general supervision); and

3. Milieu costs (educational costs plus research costs).

Variables used in the model include stu-dent contact hours, student enrollment, full-time equivalent (FTE) faculty and resi-dents, activity profi les by educator type, educator salaries, and supporting resource costs. The cost construction model does not include costs of training residents, fellows, nor biomedical graduate students. While the model includes all costs, regardless of fund-ing source, it does not consider offsetting revenues. Measures of teaching effective-ness are not included in the model. The cost construction model offers one approach to isolate AHC costs of undergraduate medical student education.

In order to provide for comparability of costs, the method for the current cost study relied on the model and the method of the original study as previously reported on in the literature. This methodology is described in general below and the reader is encour-aged to review the detailed steps described by the previous study.

Data Sources

Actual expenses for all funds sources for UT–Houston were taken from the Finan-cial Statements Fiscal Year Ended August 31, 2007, which summarized the fi scal year September 1, 2006, to August, 31, 2007. Costs from the previous study were adjusted to 2006–2007 price levels assum-ing US infl ation rates (using the consumer price index (CPI)). 8 Class size statistics were drawn from the LCME Annual Medi-cal School Questionnaire Part II 2006–2007.


Educator involvement for years one and two was determined using the computer-ized course schedule for fi rst and second year courses, which detailed encounter type, number of educators, encounter time length, and number of students present. Educator involvement for the third and fourth years was determined by surveying clerkship directors to estimate encounter type, number and type of educators (faculty, resident, or volunteer physician), encounter time length, and educator-to-student ratio. Faculty salary, including all fund sources as well as merit and supplemental pay, was calculated on actual faculty salary expense for individual faculty member for the 2006–2007 fi scal year. Resident salary expense was calculated on the weighted average for PGY1-5 levels for 2006–6007.

Model Formulation

The cost construction model predicts medical student education cost using the variables of student contact hours, student enrollment, FTE faculty and residents, pro-fessional-activity profi les, salaries, and sup-porting resources costs. The model employs fi ve steps to compute costs.

Step 1: Demand for Educator Contact Hours

Educator contact hours (ECH), direct-contact teaching hours provided by fac-ulty or residents to medical students, were categorized as one of two encounter types: didactic encounters and direct supervision. Didactic encounters included lectures, small-group activities, laboratory sessions, confer-ences, grand rounds, teaching rounds, and examinations. Direct supervision included time where the educator was physically present and directly supervised the medical

student’s performance. General supervision, time where the educator was not physically present but immediately available to the medical student if needed, was not included in the ECH. Supervision by faculty of resi-dents was not included in this study. Demand for ECH was determined as follows:

1. Computing ECH per encounter per student (ECH per encounter multi-plied by number of educators present and divided by number of students present);

2. ECH per encounter per student was then summed for all encounters in a department and multiplied by the med-ical school class size to obtain the total ECH by department for the medical school program.

ECH were computed separately for fac-ulty and residents.

Calculations for elective clerkships were based on average demand for ECH in the required clerkships, length of the elective, and number of students taking the elective. ECH provided by volunteers (community physicians and faculty from other schools) were allocated to the department that offered the clerkship.

Step 2: Supply of ECH

Supply of ECH was assessed using the activity profi le assumptions validated and applied in the 1994–1995 study ( see Figure 1). Profi le assumptions allot educator time to one of four activity categories: education, research and scholarship without students or residents, patient service without students or residents, and, for residents, study time. Education time is further categorized into one of the four types: direct-contact teaching


(ECH), teaching preparation, administration related to education, and general supervi-sion of research/service. Profi le assump-tions differentiate by faculty or resident status and appointment type (basic or clini-cal science).

Step 3: Calculation of Educator Cost

Educator cost was assessed for faculty, residents, and volunteer faculty. Educator cost included all fund source faculty sala-ries and the cost of supporting resources, including allocated school and institutional support. To calculate cost of supporting resources, we included faculty benefi t costs and allocated department, medical school, and institutional expenditures (except fac-ulty salary) against faculty salary by depart-ment of primary appointment. Educator cost

for residents was based on weighted-average PGY1-5 resident salaries and related sup-porting resources. Educator cost for volun-teer faculty was imputed as replacement cost of medical school faculty, either as basic or clinical science as appropriate using actual average faculty salary.

Step 4: Calculation of Educator Requirements

Educator requirements by department, measured the FTE educators needed to teach the 2006–2007 curriculum to medical stu-dents enrolled for 2006–2007, were com-puted by dividing the total demand for ECH in a department (from Step 1) by the supply of ECH provided by the educator (from Step 2). Separate calculations were made for faculty and residents. In these computations, we

Figure 1. Activity Profi le by Type of Medical School Educator

Activity

Basic Science Faculty Clinical Science Faculty Residents

Hours/

Week (%)

Hours/

Week (%)

Hours/

Week (%)

Education

Direct-contact teaching 8 (14.5) 20 (33.3) 12 (16.7)

Teaching preparation 12 (21.8) 4 (6.7) 3 (4.2)

Administration related

to education

1.5 (2.7) 2 (3.3) -

General supervision of

research/service

6 (11) 4 (6.7) 30 (41.7)

Subtotal 27.5 (50) 30 (50) 45 (62.5)

Research and scholarship

(without students or

residents present)

27.5 (50) 20 (33.3) 6 (8.3)

Patient service (without

students or residents

present)

- 10 (16.7) 12 (16.7)

Residents’ study time - - 9 (12.5)

Total 55 60 72


assumed that faculty spends all their teach-ing time educating medical students, allow-ing us to differentiate the cost of educating medical students from the cost of educating residents, fellows, and graduate students.

Step 5: Calculation of Education Costs

The total education cost for the medical school program was calculated by:

1. Multiplying the number of educators required in each department (from Step 4) by the related educator cost (from Step 3). Separate calculations were made for faculty and residents;

2. Allocating to education cost only the cost of time spent by educators as education time, which included direct-contact teaching, teaching preparation, administration related to education, and general supervision of research/service (from Step 2);

3. Adding education cost for faculty and residents over all departments. This total education cost represents the cost of the four-year medical school pro-gram. Next, the education cost per stu-dent per year was obtained by dividing the total education cost for the medical school program by the number of stu-dents enrolled, 868 in 2006–2007.

Joint Products

Because the AHC environment is compli-cated by the interaction of the patient care, research, and education missions, an analy-sis was performed to allow for varying cost estimates. Both the original and the present study examined the effect of joint products, the simultaneous multiple demands on educa-tor time in the clinical and research settings,

in measuring program cost. Joint products are produced when faculty and residents engage in general supervision of medical students while conducting patient care or research. Joint products may also occur during direct supervision of students but we did not assess their effects in this study. Relying on the edu-cator profi le (from Step 2), we computed the instructional cost by eliminating costs for general supervision from the education costs.

Next, we computed the milieu cost, which is an inclusive cost because it encompasses the cost of research which contributes to pro-viding an adequate educational environment for a quality medical school program. The milieu cost was computed by adding to the educational costs the costs associated with research and scholarship without students present ( see Figure 2).

Sensitivity Analyses

A sensitivity analysis was performed to estimate costs under varying scenarios ( see Figure 3). Scenario 1 refl ected the base assumptions of the model described in Steps 1–5. Scenarios 2–4 excluded selected costs: Scenario 2 excluded institutional overhead; Scenario 3 excluded volunteer faculty costs that are not paid by the medical school and are imputed in the base model; and Scenario 3 excluded resident costs which are not paid by the medical school. Scenario 5 estimated costs if faculty devoted 25 percent less time in dir-ect-contact teaching than in the base profi le.

Results

Demand for ECH

Figure 4 displays estimated demand for ECH for the undergraduate medical


curriculum at UT–Houston for 2006–2007 and 1994–1995 and the percent change com-pared to 1994–1995. ECH for the fi rst two years included direct contact with educators through lectures, laboratory sessions, small group meetings, and examinations. ECH for the third and fourth years, the clerkship

years, included direct supervision by edu-cators in patient service /research settings and didactic experiences in the clinical set-ting. ECH are reported by educator type and department affi liation.

Changes in the estimated demand for ECH in the fi rst two years of the undergraduate

Instructional

Cost

Educational

Cost

Milieu

Cost

Direct contact teaching X X X

Teaching preparation X X X

Administration related to education X X X

General supervision X X

Research and scholarship X

Figure 2. Measuring Program Cost

Figure 3. Five Scenarios of the Estimated Cost per Student per Year, 2006–2007

Scenario

Instructional

Cost ($)*

Educational

Cost ($)†

Milieu

Cost ($)‡

1. Base assumptions 47,438 61,862 97,759

2. Base assumptions

excluding institutional

overhead

40,065 52,583 83,095

3. Base assumptions

excluding volunteer

faculty costs

47,188 61,607 97,788

4. Base assumptions

excluding resident

costs

45,243 52,818 88,441

5. Base assumption

with 25% less direct-

contact teaching

by faculty

41,245 65,961 104,237

* Includes direct-contact teaching, teaching preparation, and education-

related administration costs.

† Includes instructional costs and the cost of general supervision.

‡ Includes educational costs and the cost of research/scholarship by

faculty responsible for teaching medical students.

78 JOURNAL OF HEALTH CARE FINANCE/Spring 2011Fi

gure

4. E

duca

tor

Con

tact

Hou

rs (

ECH

), F

our-

Year

Und

ergr

adua

te M

edic

al P

rogr

am, 2

006–

2007

Com

pare

d to

199

4–19

95

Dep

art

men

t

20

06–20

07

20

06–20

07

1994–1995

% C

han

ge

Co

mp

are

d

to 1

994–

1995

20

06–20

07

20

06–20

07

1994–1995

% C

han

ge

Co

mp

are

d

to 1

994–

1995

1st

an

d 2

nd

Year

MS

1 &

2M

S 1

& 2

3rd

an

d 4

th Y

ear

MS

3 &

4M

S 3

& 4

Facu

lty

Ho

urs

Resid

en

t

Ho

urs

To

tal

Ho

urs

To

tal

Ho

urs

Facu

lty

Ho

urs

Resid

en

t

Ho

urs

To

tal

Ho

urs

To

tal

Ho

urs

Basic

Scie

nces

Bio

chem

istr

y378

-378

471

-20%

--

--

-

Inte

gra

tive

Bio

logy

823

823

724

14%

--

Mic

robio

logy

309

-309

240

29%

--

--

-

Neuro

bio

logy

1,9

41

1,9

41

1,3

82

40%

1,0

01-

1,0

01535

87%

Volu

nte

ers

176

-176

643

-73%

--

--

-

Subto

tal

3,6

27

3,4

60

3,4

60

5%

1,0

01-

1,0

01535

87%

Clin

ical S

cie

nces

Anesth

esio

logy

--

%2

-100%

2,9

79

2,1

15

5,0

33

5,0

33

1%

Derm

ato

logy

6-

66

0%

1,0

47

1,0

47

2,5

21

3,7

61

-33%

Em

erg

ency

Medic

ine

154

72

226

297

-24%

2,0

50

3,4

60

3,4

60

2,8

76

20%

Fam

ily M

edic

ine

334

-334

281

19%

40,2

47

7,150

42,9

03

42,9

03

10%

Inte

rnal

Medic

ine

1,4

33

72

1,4

33

18,8

10

-15%

17,

302

17,

302

55,1

81

55,1

81

-35%

Neuro

logy

42

-2

299

-86%

4,3

56

2,4

00

6,5

66

6,5

66

3%

Obste

tric

s-

Gynecolo

gy

182

-182

444

-59%

14,2

87

14,2

87

53,6

63

53,6

63

-45%

Revisiting the Cost of Medical Student Education 79O

phth

alm

olo

gy

122

122

43

430%

553

393

94

1,4

94

-37%

Oto

lary

ngolo

gy

5-

557

-91%

391

278

669

1,4

94

-55%

Path

olo

gy

929

558

1,4

87

2,5

92

-43%

1,0

66

1,0

66

2,5

67

1,6

04

60%

Pedia

tric

s368

-368

444

-17%

21,8

78

5,8

50

23,2

04

23,2

04

19%

Psy

chia

try

729

729

1,3

41

-46%

9,1

35

6,8

54

10,1

87

10,1

87

57%

Radio

logy

22

204

226

135

67%

7,01

14,9

89

11,9

89

3,2

08

274

%

Phys

ical M

ed

38

-38

100%

579

411

990

--

Surg

ery

68

-68

371

-82%

35,4

54

39,2

08

74,6

62

103,5

64

-28%

Volu

nte

ers

364

-364

1,9

29

-81%

--

--

Subto

tal

4,7

23

906

5,6

29

9,9

75

-44%

104,7

49

104,7

49

266,4

96

314,7

38

-15%

Tota

l8,3

50

906

9,2

56

162,7

48

-31%

104,7

49

104,7

49

267,

497

315,2

73

-15%

* E

CH

s a

re the h

ours

of direct-

conta

ct te

achin

g s

tudents

receiv

ed fro

m e

ducato

rs (

regula

r and v

olu

nte

er

faculty,

and r

esid

ents

). E

CH

were

calc

u-

late

d o

n a

depart

ment basis

and w

ere

the s

um

of E

CH

s for

all

educational encounte

rs p

rovid

ed b

y a

depart

ment.

Subto

tals

and tota

ls a

re n

ot direct

sum

s d

ue to r

oundin

g.


curriculum occurred in both the basic and clinical science departments. Estimated demand for basic science ECH was 3,627 hours, an increase of 5 percent, or 167 hours, for 2006–2007 compared to 1994–1995. Estimated demand for clinical science ECH was 5,629 hours, a decrease of 44 percent, or 4,346 hours for 2006–2007 compared to 1994–1995. Clinical science estimated demand in years one and two included 906 ECH provided by residents for which there was no ECH demand in the previous study. Total volunteer faculty demand for ECH decreased 2,032 hours, or 467 hours for basic science and 1,565 for clinical sciences faculty, and constituted the largest single component of reduced ECH demand. Over-all, demand for ECH in the fi rst two years was 9,256 hours, a 31 percent decrease, or 4,179 hours, for the current study compared to the previous study.

The trend of overall decrease in the esti-mated demand for ECH continued in the third and fourth years. Estimated demand for basic science ECH was 1,001 hours, an increase of 87 percent, or 466 hours, for 2006–2007 compared to 1994–1995. Esti-mated demand for clinical science ECH was 266,496 hours, a decrease of 15 percent, or 48,242 hours for 2006–2007 compared to 1994–1995. Overall, demand for ECH in the third and fourth years was 267,497 hours compared to 315,273, a 15 percent decrease, or 47,776 hours, for the current study compared to the previous study. Of the 315,273 ECH refl ected in the original study for the third and fourth years, 145,764 were attributable to residents. Thus, resi-dent-provided ECH decreased by 41,015 hours. The reduction in resident ECH rep-resents 86 percent of the reduced ECH from the previous study.

Supply of ECH

The cost construction model relies on an activity profi le by type of educator to calcu-late ECH provided by faculty and residents. To insure study comparability, we used the same assumptions as the previous study: edu-cators work full time (basic science faculty 55 hours per week, clinical science faculty 60 hours per week, and residents 72 hours per week) for 47 weeks a year. The profi les were further validated in the current study by sam-pling Annual Faculty Reviews for basic and clinical sciences faculty, and by reviewing assumptions with institutional administrators familiar with faculty time and effort. Figure 1 displays these profi les which allocated edu-cator time to one of three categories: educa-tion, research, or patient service. Resident time allocation included a fourth category, study time. Supply of ECH was measured by the hours of “direct-contact teaching” of the respective educator type, basic science/clini-cal science faculty, or resident.

Educator Cost

Annual costs per FTE faculty member included average salary plus the cost of sup-porting resources. Compared to the previ-ous study, the average FTE basic sciences educator costs have risen from $416,328 to $495,609 or 19 percent, and average FTE clinical sciences educator costs have risen from $471,796 to $523,021 or 11 percent ( see Figure 5). The average cost per resi-dent was $67,223, or $40,908 for salary and $26,315 for supporting resources.

Number of Educators Required

The cost construction model estimates the number and the mix of FTE educa-tors required to teach the 2006–2007 total


enrollment in the undergraduate medical education ( see Figure 6). It estimates 13 basic science faculty, 179 clinical science faculty, and 188 residents for the current study period medical student enrollment of 868. This compares to 11 basic science fac-ulty, 190 clinical science faculty, and 258

residents for the 800 student enrollment in 1994–1995. The model estimated a decrease in the number of educators required, 380 edu-cators for 2006–2007 compared to 459 edu-cators for 1994–1995 against an increased enrollment of 868 medical students in the current study.

Figure 5. Annual Costs per FTE Faculty Member, by Department, 2006–2007 Compared to 1994–1995

Department

2006–2007

Average

Salary ($) †

2006–2007 Cost

of Supporting

Resources ($) ‡

2006–2007

Total Cost

($)

1994–1995

Total Cost

($)*

Basic sciences

Biochemistry 128,157 441,047 569,204 389,956

Integrative biology and

pharmacology

120,379 307,319 427,698 814,644

Microbiology 85,742 323,842 409,584 549,861

Neurobiology and anatomy 111,187 430,564 541,751 353,994

Average § 114,542 381,067 495,609 416,328

Clinical sciences

Anesthesiology 248,912 759,837 1,008,749 424,248

Dermatology 183,233 350,622 533,855 748,536

Emergency medicine 174,572 291,207 465,779 398,703

Family practice 133,011 106,516 239,527 329,079

Internal medicine 153,677 242,382 396,059 507,507

Neurology 181,334 380,724 562,058 436,003

Obstetrics-gynecology 185,371 246,309 431,680 492,244

Opthalmology 102,430 204,670 307,100 334,240

Otolaryngology 330,625 343,508 674,133 461,568

Pathology and laboratory medicine 147,251 280,672 427,923 520,683

Pediatrics 146,242 767,562 913,804 342,711

Physical medicine and rehabilitation 120,955 129,361 250,316

Psychiatry 132,765 181,976 314,741 497,310

Radiology 267,336 297,058 564,394 563,510

Surgery 203,451 337,665 541,116 639,006

Average § 175,702 347,319 523,021 471,796

*1994–1995 costs adjustedto 2006–2007 dollars using the CPI infl ation rate

†Salary data from fi scal year 2006–2007, September 2006–August 2007.

‡Costs are based on FY07 all funds expenditures.


Figure 6. Estimated Full-Time Equivalent (FTE) Educators Required and Education Costs Four-Year Undergraduate Medical Program

University of Texas Health Science Center at Houston—Medical School 2006–2007

Compared to 1994–1995

Department

2006–2007 1994–1995

No. FTE

Faculty*

No. FTE

Residents

Education

Costs ($)

No. FTE

Faculty*

No. FTE

Residents

Education

Costs ($)†

Basic sciences

Biochemistry 1.01 286,116 1.25 244,334

Integrative biology and

pharmacology

2.19 402,109 1.92 402,109

Microbiology 0.82 168,300 0.64 175,542

Neurobiology and

anatomy

7.82 2,119,457 5.10 902,218

Volunteers 0.47 1355,983 1.71 355,983

Subtotal 12.31 3,157,945 10.62 2,080,185

Clinical sciences

Anesthesiology 3.17 3.75 1,778,074 2.95 4.13 788,693

Dermatology 1.57 1.86 494,680 2.15 3.09 927,718

Emergency medicine 2.42 2.63 665,496 1.96 2.36 483,503

Family practice 43.17 12.68 5,757,463 36.23 16.19 6,598,213

Internal medicine 21.54 30.80 5,361,525 28.53 53.28 9,337,539

Neurology 4.68 4.26 1,500,528 2.75 7.58 898,974

Obstetrics-gynecology 16.50 25.33 4,477,193 24.07 55.81 8,123,196

Opthalmology 0.72 0.70 136,214 0.88 1.23 194,977

Otolaryngology 0.42 0.49 365,749 0.91 1.23 259,065

Pathology and laboratory

medicine

3.18 2.88 795,681 3.67 8,327 1,008,327

Pediatrics 23.67 10.37 11,558,317 18.83 10.54 3,641,978

Physical medicine and

rehabilitation

0.66 0.73 108,154 - - -

Psychiatry 10.49 12.15 2,091,397 8.67 5.99 2,391,456

Radiology 7.70 9.19 2,548,929 1.98 2.63 660,246

Surgery 37.79 69.52 12,793,177 54.73 93.06 21,151,808

Volunteers 0.39 105,316 2.05 484,093

Subtotal 178.06 187.33 50,537894 190.37 258.44 56,949,787

Total 190.37 187.33 53,695,839 200.97 258.44 59,029,972

* The FTE numbers shown have been rounded; therefore, they cannot be used to produce the educator

costs shown.

† 1994–1995 costs adjusted to 2006–2007 dollars using the CPI infl ation rate.


Education Cost

The cost construction model estimates educational costs based on the assumptions for educator activity profi le displayed in Figure 1. Educational costs by department in 2006–2007 and 1994–1995 are reported in Figure 6. Education costs in the basic science departments increased to $3,157,945 from $2,080,185 or 52 percent. But in the clinical departments, education costs decreased to $50,537,894 from $56,949,787 or 11 percent. Overall, educational costs for the four-year medical program decreased to $53,695,839 in 2006–2007 from $59,029,972 in 1994–1995 or 9 percent. The cost per student per year in 2006–2007 when 868 students were enrolled was $61,862. The corresponding cost in 1994–1995 when 800 were enrolled was $73,787, or $11,925 higher or 16 per-cent than in 2006–2007.

Joint Products

In order to address the impact of joint products on the cost of educating medical students, we computed the different costs that excluded and included specifi c joint products ( see Figure 3). Instructional costs, which include direct-contact teaching, teach-ing preparation, and administration related to education, were $47,438 per student per year in 2006–2007. The milieu cost per stu-dent per year was estimated at $97,759 in 2006–2007. The corresponding costs for 1994–1995 were $56,582 and $116,603.

Sensitivity Analysis

The instructional, education, and milieu costs per student per year under varying sce-narios are reported in Figure 3. In Scenario 2, which excluded institutional overhead, education costs were 15 percent lower. In

Scenario 3, which excluded volunteer fac-ulty costs that are not paid by the medical school and are imputed in the base model, costs were almost unchanged. In Scenario 3, that excluded resident costs which are not paid by the medical school, educational costs were 15 percent lower. Educational costs were 7 percent higher in the Scenario 5, which estimated costs if faculty devoted 25 percent less time in direct-contact teach-ing than in the educator activity profi le in Figure 1.

Discussion

Estimates of the cost of medical student education in the literature vary by method-ology and refl ect a point in time measure-ment. These factors preclude analysis of the impact of curriculum changes, variations of teaching methodology, or class size on edu-cation cost. This study assesses the cost of educating undergraduate medical students at UT–Houston in 2006–2007 and com-pares it to costs assessed for 1994–1995. The cost assessments for both years were estimated using identical methodology, a cost construction model, to insure compa-rability. After adjusting for infl ation, the annual cost per student enrolled decreased by 16 percent and the total cost of pro-viding the four-year medical school pro-gram decreased by 9 percent. Decreases in cost of education medical students can be traced to:

1. A reduction in the number of ECH required for curriculum delivery;

2. A change in the mix of educators pro-viding the ECH; and

3. An increase in medical school class size.


Overall, fewer ECH were required to deliver the curriculum in 2006–2007 than in 1994–1995. Three factors accounted for the reduction on ECH:

1. Reduction of faculty intensive class-room instruction;

2. Reduction of resident ECH in the third and fourth years; and

3. Changes in two clerkships in the third and fourth years.

The curriculum delivery in the fi rst two years experienced an overall decrease in ECH: ECH for basic sciences faculty grew by 167 hours, while the ECH for clinical sci-ences faculty decreased by 4,346 hours. The same direction of change occurred in the third and fourth years: ECH for basic sci-ences faculty increased in the clerkship years by 466 hours, while the clinical sciences fac-ulty ECH decreased by 7,225 hours. These changes most likely refl ect adjustments in teaching methodologies instituted during the 12-year interval between the fi rst and second study. In response to fi nancial pressures, fac-ulty intensive classroom instructional meth-ods, such as clinical sciences’ team teaching or panel discussions, have been reduced in order to maximize billable clinic hours. Res-ident ECH experienced the largest decrease, 41,015 hours. Change in resident ECH would appear to be driven by the Accredi-tation Council for Graduate Medical Educa-tion’s limits on residents’ work hours, which took effect in July, 2003. Additionally, two of the required clerkships restructured the clerkship experience, contributing to reduc-ing the intensity of the ECH required.

From 1994–1995 to 2006–2007, there has been a shift away from using more expensive clinical faculty and toward more teaching by

less expensive basic science faculty in teach-ing medical students. While the number of basic science faculty required to deliver the curriculum increased slightly (by two FTEs), the number of FTE clinical science faulty and FTE residents required decreased by 11 and 70 FTEs, respectively.

Changes in medical school enrollment at UT–Houston have occurred from 1994–1995 to 2006–2007. In the 1994–1995 origi-nal study total enrollment was 800 and in the present study it was 868. This 8.5 per-cent increase in enrollment contributes to the fi ndings of decreased annual cost per student enrolled.

Cost, or expense, management has been a trademark of the last decade in health care delivery and in the AHC environ-ment. Despite medical school management actions to improve effi ciency and reduce costs through typical expense management techniques such as annual budget reduc-tions, FTE reductions, hiring freezes or slow downs, and delayed salary increases, costs for FTE educators increased both in the basic and clinical sciences. However, the increase in educator costs was more than offset by the reduction in ECH and the change in the mix of educators, which drove down the cost of educating medical students.

The cost construction model provides one approach to isolating costs of undergraduate medical student education in the AHC envi-ronment. Limitations of the model described in the original study apply to this study as well. The costs reported in this study do not rep-resent the costs of running a medical school, but the societal costs of undergraduate medi-cal education, given that all real and imputed costs were included. The reported costs do not include the cost of training residents, fel-lows, basic science graduate students, or the


cost of research and patient service, which are part of a medical school mission. Neither the original nor the present study examined program effectiveness measures, such as the quality of the education. Using the strength of the cost construction model, which care-fully builds cost from the bottom up, this study focused on addressing the magnitude and direction of any change in the cost of medical student education at one publicly funded AHC. The next steps would be to determine whether costs have changed per equivalent educational experience, assuming equal cost per student, or whether the root cause of change is due to the distribution of the educational experiences, assuming vary-ing cost per student. The complex nature

of the AHC setting, demand and supply of educator time, assumptions about time allo-cation, and lack of consensus regarding opti-mal mix of research and service in providing undergraduate medical education are vari-ables whose impact are no better measured in 2007 than they were in 1995. The present study, however, demonstrates the changes in program requirements and costs for under-graduate medical education in one AHC, UT–Houston, in the last 12 years, pre- and post-managed care. It confi rms that delivery of undergraduate medical education at this AHC has responded to fi nancial pressures and has become more effi cient as measured by reduction in estimated ECH for program delivery and decreased cost per student.

REFERENCES

1. Kirch, DG, “Financial and Organizational Turmoil in the Academic Health Center: Is It a Crisis or an Opportunity for Medi-cal Education?” Acad. Psychiatry , Jan–Feb, 30(1): 5–8 (2006); Magill, MK, Catinella, AP, Haas, L, Hughes, CC, “Cultures in Confl ict: A Challenge to Faculty of Academic Health Centers,” Acad. Med ., 73: 871–875 (1998); Barzansky, B, “United States Medical School Financing: Beyond the Black Box,” Journal of the International Association of Medical Sci-ence Educators , (12): 5–8 (2002); Watson, RT, “Rediscovering the Medical School,” Acad. Med ., 78(7) :659–665 (July 2003); Mul-hausen, R, Kaemmerer, C, Foley, J, Schultz, A, “Education Costs in Two Public Teaching Hospitals, Acad. Med ., 64(6): 314–9 (June 1989); Langabeer, J, “Competitive Strategy in Turbulent Healthcare Markets: An Analysis of Financially Effective Teaching Hospitals,” Journal of Healthcare Management , 43(6): 512–26 (Nov.–Dec. 1998); Levey, S, Ander-son, L, “Painful Medicine: Managed Care and the Fate of America’s Major Teaching Hospitals,” Journal of Healthcare Manage-ment , 44(4): 231–249, discussion 249–251

(Jul.–Aug. 1999); Kuttner, R, “Managed Care and Medical Education,” New England Jour-nal of Medicine , 341(14): 1092–1096 (Sept. 30, 1999); Shine, KI, “Challenges Facing Academic Health Centers and Major Teach-ing Hospitals,” Journal of Nursing Adminis-tration , 27(4): 21–6 (Apr. 1997); Pardes, H, “The Future of Medical Schools and Teaching Hospitals in the Era of Managed Care,” Acad. Med ., 72(2): 97–102 (Feb. 1997).

2. Rein, MF, Randolph, WJ, Short , JG, Coolidge, KG, Coates, ML, Carey, RM, “Defi ning the Cost of Educating Undergraduate Medical Students at the University of Virginia,” Acad. Med ., 72(3): 218–227 (Mar. 1997); Goodwin, MC, Gleason, WM, Kontos, HA, “A Pilot Study of the Cost of Educating Undergraduate Medical Students at Virginia Commonwealth University,” Acad. Med ., 72(3): 211–217 (Mar. 1997); Jones, RF, Korn, D, “On the cost of educating a medical student. Acad. Med ., 72(3):200–10 (Mar. 1997); Franzini, L, Low, MD, Proll, MA, “Using a Cost-Construction Model to Assess the Cost of Edu-cating Undergraduate Medical Students at the University of Texas–Houston Medical School,” Acad. Med ., 72(3): 228–237 (Mar. 1997).


3. Morrison, G, “Mortgaging Our Future—The Cost of Medical Education,” New England Journal of Medicine , 352(2): 117–119 (Jan. 13, 2005); Anonymous, “Panel: Racism, Cost Keep Minorities Out of Health Professions, Medicine & Health , 58(35): 4–5, (Sept. 27, 2004); Dunn, MR, Miller, RS, “The Shifting Sands of Graduate Medical Education, JAMA , 276: 710–731 (1996); American Association of Medical Colleges, “Medical Education Costs and Student Debt: A Working Group Report to the AAMC Governance,” Mar. 2005, https://services.aamc.org/Publications/index.cfm?fuseaction=Product.displayForm&prd_id=121&showReadingList=true.

4. Stites, S, Vansaghi, L, Pingleton, S, Cox, G, Paolo, A, “Aligning Compensation with Edu-cation: Design and Implementation of the Educational Value Unit (EVU) System in an Academic Internal Medicine Department, Acad. Med. , 80(12): 1100–1106 (Dec. 2005); Sloan, TB, Kaye, CI, Allen, WR, Magness, BE, Wartman, SA, “Implementing a Simpler Approach to Mission-Based Planning in a Medical School,” Acad. Med ., 80(11): 994–1004 (Nov. 2005); Howell, LP, Hogarth, MA, Anders, TF, “Implementing a Mission-based Reporting System at an Academic Health Center: A Method for Mission Enhance-ment. Acad. Med ., 78(6): 645–651 (June 2003); Ridley, GT, Skochelak, SE, Farrell, PM, “Mission Aligned Management and Alloca-tion: A Successfully Implemented Model

of Mission-Based Budgeting,” Acad. Med ., 77(2): 124–129 (Feb. 2002); Sloan, TB, Kaye, CI, Allen, WR, Magness, BE, Wartman, SA, “Implementing a Simpler Approach to Mis-sion-Based Planning in a Medical School,” Acad. Med ., 80(11): 994–1004 (Nov. 2005).

5. Rein, MF, Randolph, WJ, Short , JG, Coolidge, KG, Coates, ML, Carey, RM, “Defi ning the Cost of Education Undergraduate Medical Students at the University of Virginia,” Acad. Med ., 72(3): 218–227 (Mar. 1997); Good-win, MC, Gleason, WM, Kontos, HA, “A Pilot Study of the Cost of Educating Undergraduate Medical Students at Virginia Commonwealth University,” Acad. Med ., 72(3): 211–217 (Mar. 1997); Franzini, L, Low, MD, Proll, MA, “Using a Cost-Construction Model to Assess the Cost of Educating Undergraduate Medical Students at the University of Texas–Houston Medical School,” Acad. Med ., 72(3): 228–37 (Mar. 1997).

6. Franzini, L, Low, MD, Proll, MA, “Using a Cost-Construction Model to Assess the Cost of Educating Undergraduate Medical Students at the University of Texas–Houston Medical School,” Acad. Med ., 72(3): 228–237 (Mar. 1997).

7. Hanft, RS, “Cost of Education in the Health Professions (Parts 1–3), PB, 238–329, Wash-ington, DC: National Academy of Sciences (1974).

8. Halfhill, T, “Tom’s Infl ation Calculator,” 2008, http://www.halfhill.com/infl ation.html.

87


Cost-Volume-Profi t Analysis and Expected Benefi t of Health

Services: A Study of Cardiac Catheterization Services

Mustafa Z. Younis, Samer Jabr, Pamela C. Smith, Maha Al-Hajeri, and Michael Hartmann

Aim: Academic research investigating health care costs in the Palestinian region is limited. Therefore, this study examines the costs of the cardiac catheterization unit of one of the largest hospitals in Pales-tine. We focus on costs of a cardiac catheterization unit and the increasing number of deaths over the past decade in the region due to cardiovascular diseases (CVDs).

Methods: We employ cost-volume-profi t (CVP) analysis to determine the unit’s break-even point (BEP), and investigate expected benefi ts (EBs) of Palestinian government subsidies to the unit.

Results: Findings indicate variable costs represent 56 percent of the hospital’s total costs. Based on the three functions of the cardiac catheterization unit, results also indicate that the number of patients receiving services exceed the break-even point in each function, despite the unit receiving a government subsidy.

Conclusions: Our fi ndings, although based on one hospital, will permit hospital management to realize the importance of unit costs in order to make informed fi nancial decisions. The use of break-even analysis will allow area managers to plan minimum production capacity for the organization The economic benefi ts for patients and the government from the unit may encourage government offi cials to focus efforts on increasing future subsidies to the hospital.

Key words: cost analysis, cost-volume-profi t (CVP) analysis, break-even point (BEP), health care costs.

Academic research investigating health care costs in the Palestinian region is limited. This study seeks

to add to the literature concerning hospi-tal costs in the Palestinian territories. The objective of this study is to examine the

Mustafa Z. Younis, PhD, MA, MBA, is Tenured Professor at Jackson State University, Jackson, Mis-sissippi, and an international consultant in health economics, fi nance, and policy. He can be reached at [email protected].

Samer Jabr is Director of the Health Economics Department, Health Planning & Policy Unit Ministry of Health, in Nablus, Palestine, and a lecturer at the Institute of Community & Public Health, Birzeit Uni-versity, Department of Economics, An-Najah National University. He can be reached at [email protected].

Pamela C. Smith is an Associate Professor in the Department of Accounting, College of Business, Uni-versity of Texas at San Antonio. She can be reached at [email protected].

Maha Al-Hajeri, PhD, is an Assistant Professor of Health Information Administration, Faculty of Allied Health at Kuwait University, Kuwait, and can be reached at [email protected].

Michael Hartmann is Director of Pharmacy, School of Medicine, Friedrich Schiller University of Jena, Germany. His area of expertise is Health Econom-ics. He can be reached at [email protected].

Financial and Competing Interests Disclosure: At the time this research was conducted, Drs. Younis, Smith, Al-Hajeri, and Hartmann were college pro-fessors and Mr. Jabr was employed at the Palestin-ian Ministry of Health. The article was derived from their academic research. The authors have no other relevant affi liations or fi nancial involvement with any organization or entity with a fi nancial interest in or fi nancial confl ict with the subject matter or materi-als discussed in the manuscript apart from those disclosed. No writing assistance was utilized in the production of this manuscript.


costs of the cardiac catheterization unit of one of the largest hospitals in Palestine. We employ cost-volume-profi t (CVP) analy-sis, also known as break-even analysis, to determine the fi xed and variable costs of the hospital’s cardiac catheterization unit. We also estimate whether the Palestinian gov-ernment subsidizes the costs of the cardiac catheterization unit.

Palestine allocates a signifi cant part of its resources to the health sector. The escala-tion of health expenditure has been a univer-sal phenomenon over the past decades, and while its extent has been signifi cant in high-income countries, low- and middle-income countries have similarly experienced much the same. Many factors may have affected the escalation of health expenditure, such as:

Evolving high-cost medical technologies;• Changes in disease patters; or• Increasing demand for health care • services.

An analysis of health care costs in the Pal-estinian region is timely and necessary due to the region’s economic impact. 2008 data indicated that the gross domestic product (GDP) of Palestine was estimated to be US $6,108.2 million (current price) or about US $1,697 per capita. The health expenditure in 2008 is estimated to be 15 percent of the GDP. 1 These economic indicators, coupled with rising global health care costs, support further analysis of hospital operations. Kiy-maz et al . 2 analyze the relationship between health care expenditures and GDP, and their fi ndings support prior fi ndings that a posi-tive correlation exists between a country’s GDP and its health expenditures. This article seeks to contribute to the discussion of glo-bal health care expenditures.

We focus on costs of a cardiac catheteri-zation unit and to the increasing number of deaths over the past decade in the region due to cardiovascular disease (CVD). Data indicate that in 2003, 3,894 people died from various CVDs, with the distribution of deaths due to CVD approximately the same between males and females (52.4 percent for males, 47.6 percent for females). 3 Pal-estine and its neighboring countries are facing increasing mortality rates due to dis-eases of the circulatory system. For exam-ple, in Jordan, CVD was the leading cause of death. 4

The incidence of coronary heart disease is growing signifi cantly in the region, includ-ing Bahrain, Egypt, Iraq, Kuwait, and Qatar. Data from these countries indicate deaths attributable to diseases of the circulatory system range from 25 percent to 40 percent of all deaths. 5 Mataria et al . 6 fi nd that the overall quality of life of Palestinians is very poor, yet “the Palestinian people provide a valuable public health lesson with respect to the broader understanding of health and the problem of ‘medicalization’ of health.” Giacaman et al . 7 point out the importance of examining both fatal and non-fatal health outcomes in order to guide health policy in the region.

This article focuses on a treatment of CVD, where cardiac catheterization services are the most expensive, most accurate, and most advanced as a diagnostic and therapeu-tic tool and thereby the most cost effective when considering the signifi cant reduction in mortality and morbidity with its socio-economic consequences. 8

We employ a case-study analysis of the costs of a hospital’s cardiac catheterization unit. We concentrate on the level of patient activity needed for the facility to break even.

Cost-Volume-Profi t Analysis and Expected Benefi t of Health Services 89

The use of cost analysis by administrators and regulators may improve the quality of fi nancial information, as well as enhance the effi cient use of scarce resources. Our fi nd-ings, though based solely on one hospital, permit hospital management to realize the importance of unit costs in order to make informed fi nancial decisions within the unit. Furthermore, the economic benefi ts for patients and the government from the cardiac catheterization unit may encourage govern-ment offi cials to focus efforts on the unit and increase future subsidies to this unit.

Prior Literature

Prior literature examines health care break-even analysis in numerous settings—from the expansion of dental practices to long-term care facilities. 9 Stritecky and Pubal 10 note that break-even analysis is one of many managerial tools that help a hospital survive under market competition and forces management to reply to fl exible market cir-cumstances. Chotiwan et al . 11 probe the lab-oratory costs of a Thai hospital, and fi nd the material costs of the unit comprise the high-est proportion of direct and total costs.

Despite this variety of break-even analysis in prior literature, research investigating cost allocation in Middle Eastern hospitals exists on a limited basis. 12 Younis et al . 13 employ CVP analysis in a Palestinian multi-service hospital for the period from 2005 to 2007, and Chodick et al . 14 assess the direct medical cost of CVD in an Israeli health maintenance organization (HMO).

The impact of the Palestinian region’s economic resources, combined with the need for fi scal responsibility, supports the need for further research on health care costs. We seek to expand upon the use of CVP analysis

through a case-study format in order to add to the literature concerning hospital costs in the Palestinian region.

Methods

Overview

This study is a case study of the cardiac catheterization unit of the Ramallah Hos-pital. The hospital, originally built in 1961, is located in Ramallah City, West Bank. The hospital employs 62 physicians, 129 nurses, 46 paramedics, and 87 administra-tive employees. The hospital is owned and operated by the Ministry of Health (MoH), has a capacity of 150 beds, and its primary objective is to provide general and surgi-cal services to the community. As a MoH hospital, Ramallah hospital operated under MoH regulations and guidelines. It is a gen-eral hospital providing numerous units of specialty, including orthopedics, internal medicine, surgical operations, cardiac cath-eterization, burns, neonatology, general sur-gery, intensive care unit, intermediate care unit, delivery, and specialized surgery. The occupancy rate was estimated at about 90 percent in 2009.

In 2010, the hospital became part of the Palestine Medical Complex (PMC). The PMC consists of two newly established hospitals and two existing facilities with the addition of the central blood bank. The consolidation of the fi ve facilities into one entity serves as a national center of excellence and as a pilot approach to decentralized hospital management. The cardiac catheterization unit of the hospital performed over 2,000 procedures in 2004. Without this unit, these procedures would have been transferred to Jordan or Israel.


The objectives of the cardiac catheteriza-tion unit are to:

1. Provide a local, essential, and life-saving service;

2. Alleviate the burden of travel to the heart disease patient;

3. Save money for the facility; 4. Provide further training in invasive

cardiology for Palestinian doctors and nurses; and

5. Provide necessary support for local cardiovascular surgeons. 15

The cardiac catheterization unit has two main purposes:

1. The diagnosis of various CVDs; and 2. Therapeutic treatment of CVDs.

Cost-Volume-Profi t Analysis

Cost-volume-profi t (CVP) analysis allows hospital management to discern the proba-ble effects of changes in sales price, product mix, or sales volume. 16 We focus on CVP analysis of the cardiac catheterization unit of the Ramallah hospital during 2003. Cost allocation involves the transfer or allocation of costs from one department to another. These cost allocations assist in determin-ing unit price, which can be used for price setting. Cost allocations also help deter-mine the relationship of total revenue and total cost for a department or service, and to determine profi tability on a product line or departmental basis. In health care, most allocations involve cost transfers from over-head centers (non-revenue department) to revenue centers (revenue department). 17 The costs of any overhead departments are dis-tributed to the intermediate and fi nal service

departments through a step-down method, according to allocation criteria devised to resemble as closely as possible the actual use of resources by each of the departments. The step-down method is a more advanced cost-fi ndings technique than the direct dis-tribution method because it involves the distribution of costs from overhead depart-ments and fi nally to intermediate and fi nal service departments. 18

The step-down cost accounting (SDCA) approach identifi es the range of resources needed to run a facility, and then assigns these resources to chosen “cost centers” on an allocation basis. Those costs in turn fi lter down until the fi nal cost centers of interest are left. The following are necessary consid-erations to compute unit costs:

1. Rank the support cost centers in order; the one that is consumed the most by others would be ranked the fi rst, and next would be the second;

2. Allocate direct costs from the fi rst rank down unit, the last support cost center;

3. Allocate denominators—the sum of allocation criteria of all cost centers that are not yet allocated; and

4. Never allocate back to the previous support cost centers. 19

Figure 1 contains the allocation criteria for the various cost centers of the Ramallah Hospital.

The costs of the cardiac catheterization unit can be classifi ed into two elements: fi xed cost and variable cost. Fixed costs are those costs that tend to remain constant over the course of many accounting periods, and are not infl uenced by changes in volume or intensity of service. Thus, fi xed costs are incurred regardless of how much service is


provided. Four main cost items are taken into consideration to calculate fi xed costs:

1. Cost of medical equipment; 2. Cost of furniture and equipment; 3. Salary of doctors, nurses, and all staff

in the cardiac catheterization unit; and 4. Overhead allocation from overhead cen-

ters to the cardiac catheterization unit.

Variable costs are those that relate to the direct cost of providing care, and are expressed as costs per unit of service deliv-ered. 20 Variable costs thus rise and fall in relation to changes in the level of activity. Examples include disposable supplies; for example, intravenous tubing, contrast dye, and stents. Variable costs may be paid from the cardiac catheterization unit’s earnings or from government subsidies.

We use the following formula to calculate unit variable costs for each function inside the cardiac catheterization unit:

TC = γ 1 Diag + γ

2 Ball + γ

3 Pace + FC (1)

Where TC = total cost. γ = gamma (unit variable cost). Diag = number of Diagnosis patients.

Ball = number of Balloon patients. Pace = number of Pacemaker patients. FC = fi xed cost.

Ordinary least square (OLS) or linear least square is a method for estimating the unknown parameters in a linear regression model. It has been used to analyze the results that determine the unit cost per each function by entering the data on a monthly basis. We convert Equation 1 to fi nd the equivalent number of patients (total output) for all unit functions (diagnosis, balloon (angioplasty), and pacemaker), which will be used to estimate the break-even point for the cardiac catheterization unit.

γ 2 /γ

1 Ball = Diag

γ 3 /γ

1 Pace = Diag

γ 1 * Q = γ

1 Diag + γ

2 Ball + γ

3 Pace

Q = Diag + γ 2 /γ

1 Ball + γ

3 /γ

1 Pace (2)

Where Q means equivalent diagnosis (output) patients

We assume that γ 1 is the unit variable cost.

Equivalent diagnosis has been used because the majority of patients are diagnosis patients; this also should minimize statistic error. However, using γ

2 or γ

3 as unit variable

cost will give the same result. The cardiac catheterization unit earns

revenue from patient and government pay-ments (out-of-pocket, copayments, and glo-bal budget allocation). Therefore, the total revenue (TR) for the unit is as follows:

TR = ƒ (patients charges, government)

Thus the following formula is used to set the price for services:

Equivalent Diag = Diag + γ 2 /γ

1 Ball +

γ 3 /γ

1 Pace (3)

Figure 1. Allocation Bases Criteria

Number Cost Center Allocation Criteria

1 Building Square meter

2 Administration Salary

3 Cleaning Square meter

4 Guard Salary

5 Utility Square meter

6 Maintenance-

building

Square meter

7 Rent Square meter


Average price = Σ Total revenue/ Equivalent Diag

The Break-Even Point with Constant Price and Linear Costs

Our analysis is based on data obtained from the following sources within the Ram-allah hospital—the cardiac catheterization unit, medical storage, fi nance department, central storage, and maintenance depart-ment. Other data were also obtained from the Palestinian Health Information Sys-tem. It holds that, with constant price, the revenue develops according to this relation:

R = p*q

Where R = revenue. q = number of patients. p = price.

And the expenses with linear (propor-tional) develop according to this relation:

TC = FC + b*q

Where TC = total cost. FC = fi xed cost. b = unit variable cost.

Then, if that is true, then profi t is the dif-ference between revenues and expenses

P = R – TC

It is obvious that we get profi t when R > TC; in the opposite situation there is loss. When R= TC, there is neither profi t nor loss. We can analytically derive it from the equations above:

R = TC; (4)

p*q = FC + b*q

FCq = p – b

The break-even point (BEP) is an analytic technique that helps determine the level of vol-ume needed to reach the fi nancial break-even point, the point at which net revenue exactly equals cost. At this point there is neither a loss nor a profi t. 21 After calculating the BEP, the number of patients where total costs = total revenue is allocated on the functions of the cardiac catheterization unit to estimate the number of patients for each function by using Equation 3 above.

Break-even analysis assumes that the fi rm’s average variable costs are constant in the rel-evant output range; hence, the fi rm’s total cost function is assumed to be a straight line. Since average variable cost is constant, the extra cost of an additional unit of output—marginal cost—must be constant too, and equal to aver-age variable cost.

Government Expected Benefi t (GEB)

In this article, we also calculate how the government subsidizes the cardiac catheteri-zation unit if the patient BEP is more than the real number of patients who visited the cardiac catheterization unit. Thus, we ana-lyze whether or not the government covers any difference between costs and total reve-nue. We base our analysis on the following:

Operation loss = Total cost – Total revenue + (any cost subsidy that is included in fi xed cost and variable cost).

We analyze the economic value of the car-diac catheterization unit over several periods. The difference between the price and the unit


cost is estimated. Depending upon the dif-ference obtained, the government may need to cover the cost if the new price is less than unit cost, but the government may benefi t if the new price is more than unit cost. The expected benefi ts (EB) for the government, in terms of GDP, from extending patient life, is estimated using sensitivity analysis over a fi ve-, ten-, 15-, and 20-year period. We use the following for EBs:

(a) Economic value for patient (EVP) = number of years saved * GDP per capita

(b) Economic value for government (EVG) = price – cost for treatment = difference subsidy from government compare with EB

(c) EB

EB = (1 – x)* – $23 +

years saved * GDP per capita – $23x (5) employee / population

x = proportion of population with CVD -$23 = Transportation ($6.4) and Time

Loss ($16.6)

Equation 5 shows for the average patient who visits the cardiac catheterization unit. The probability of this patient without CVD is (1-x), and the probability with CVD is x; (1-x) is multiplied by -$23 because some patients without CVD disease may use diag-nostic tests, which means the cardiac cath-eterization unit doesn’t save the life of these patients. But the patients incur transportation costs and the loss of time. The individual employee is divided by the total population because it is assumed that not all patients work; for example, some patients are less than 15 years of age. (The working age in

Palestine starts above 15 years of age.) By excluding patients under 15 years of age, GDP per worker can be obtained by divid-ing GDP per capita upon working age (more than 15 years of age). Overall, the third for-mula estimates the government benefi t using sensitivity analysis for fi ve, ten, 15, and 20 years.

Results

Total Costs

In the analysis of fi xed costs and variable costs of the cardiac catheterization unit, this study reveals that variable costs represent the greatest portion of total costs (56 percent), and fi xed costs are 44 percent. Total costs for the cardiac catheterization unit during 2003 are $ 613,544 ( see Figure 2).

Expanding Equation 1 above, we derive total variable costs (VC) as follows:

VC = γ 1 Diagnosis + γ

2 Balloon +

γ 3 Pacemaker

Using the unit’s data, it was found that unit variable cost is

$140.5139 for diagnosis; • $532.3362 for balloon; and • $1,689.898 for pacemaker. •

Regression analysis using the OLS method has been used in estimating the result for

Figure 2. Fixed Cost and Variable Cost of Cardiac Catheterization Unit

in Ramallah Hospital, 2003

Cost Categories Amount in US $ %

Fixed cost $270,903.42 44%

Variable cost $342,641.21 56%

Total cost $613,544.63 100%


unit variable cost of diagnosis, balloon, and pacemaker ( see Figure 3). However, in test-ing the signifi cance of individual independ-ent variables, t-tests indicate that none of the independent variables is statistically sig-nifi cant at the 5 percent level. Further, high standard errors revealed that the independent variables are not good estimators in explain-ing variable costs. When t-tests indicate that all independent variables are statically insig-nifi cant, it’s important to check the overall signifi cance of the model using the F-test. From the F-test, it can be seen that the overall result is statistically insignifi cant. Therefore, this model is not statistically signifi cant in explaining the variable cost of patients. The coeffi cient of determination, r 2 , is 0.185624, which means that only 18.56 percent of the variation in the total variable cost can be explained by the explanatory variables (diagnosis, balloon, and pacemaker).

In examining the results in Figure 3, the model is not statistically signifi cant; this is because the number of observations (n=12 based on a monthly analysis) is not adequate to run a regression model. Despite the lack

of statistical signifi cance, this model has practical fi ndings, since these three variables (diagnosis, balloon, and pacemaker) play a signifi cant role in determining total variable cost. Therefore, in our break-even analysis, the above estimated unit variable cost will be used. From this result, the variable cost formula can be written as follows:

VC = γ 1 Diagnosis + γ

2 Balloon +

γ 3 Pacemaker (6)

$342,641.21 = $140.5139*Diagnosis + $532.3362*Balloon + $1,689.898*Pacemaker

Break-Even Analysis

According to the previous formula, equiva-lent diagnosis (output) can be estimated by converting each balloon and pacemaker to diagnosis. Equivalent diagnosis (output) will be used to calculate variable cost, total rev-enue, and BEP for all functions together in the cardiac catheterization unit. Cost ratio has been used to calculate equivalent diag-nosis (output) by dividing unit variable cost for each balloon and pacemaker by the unit

Figure 3. Regression Analysis of Variable Costs

VC= γ1Diagnosis+γ

2Balloon+γ

3Pacemaker

VC = 140.5139 Diagnosis + 532.3362 Balloon +

1,689.898 Pacemaker

Diagnosis Balloon Pacemaker

se (121.3900) (870.3585) (3939.648)

t (1.15754) (0.611629) (0.428947)

Prob (0.2768) (0.5559) (0.68780)

r2 = 0.185624

Adj r2 = 0.004651

n = 12


variable cost for diagnosis. The result will then be multiplied with the number of patients (the number of balloon patients for the year is 163, and the number of pacemaker patients is 23).

532.3362 = = 3.7885 140.5139

γ 2 /γ

1 * Balloon = 3.7885*163 = 617.5255

1,689.898 γ

3 /γ

1 = = 12.0266

140.5139

γ 3 /γ

1 * Pacemaker = 12.0266*23 = 276.612

Equivalent Diagnosis (output) = Diag + γ

2 /γ

1 Ball + γ

3 /γ

1 Pace

= 1557 + 617.5255 + 276.612 = 2451.1375

This result means that there are 2451.1375 equivalent diagnosis (output) patients visited the cardiac catheterization unit in 2003.

We then calculate the average price for equivalent diagnosis by using the following formula:

Σ Total Revenue Average price =

Equivalent Diagnosis

$987,500.00 = = $402.874 2451.1375

The amount $402.874 represents the aver-age price per patient for equivalent diagnosis (output).

We calculate the break-even point by using the following formula:

Fixed Cost (7) Price/unit – VC/Unit

$270.903.00 = 402.874 – 140.5139

$270.903.00 = = 1032.562 $262.36

From this result, the BEP where total cost = total revenue were received with 1,032.562 equivalent number of patients. Figure 4 ex plains BEP for the cardiac catheteriza tion unit.

We next calculate the break-even point for each function (Diagnosis, Balloon, and Pacemaker). To calculate break-even point for each function, the following steps are used.

Step 1: Calculate the equivalent number of patients for each function by dividing the equivalent diagnosis (output) that represents for BEP by the equivalent number of patients requested. The result will be multiplied by the equivalent number of patients requested for each function (number of patients requested for diagnosis is 1557, number of

Figure 4. Break-Even Point (BEP) for Cardiac Catheterization Unit, 2003

Total Revenue Total Cost

Fixed cost $270,903.00

Variable cost = 1032.562*140.5139 $145,089.32

Total revenue = 1032.562*402.874 $415,992.32

Total $415,992.32 $415,992.32


patients requested for balloon is 617.5255, and number of patients requested for pace-maker is 276.612).

Equivalent number of diagnostic patients 1032,562 = *1557 = 655.899 2451.1375

Equivalent number of balloon patients 1032,562

= *617.5255 = 260.138 2451.1375

Equivalent number of pacemaker patients 1032,562 = *276.612 = 116.525 2451.1375

Step 2: Convert equivalent patients to real number of patients for each function in the cardiac catheterization unit to estimate the BEP for each function by using cost ratio. To receive at BEP for each function, unit variable cost for diagnosis (γ

1 ) is divided

upon unit variable cost for balloon and pace-maker (γ

2 , γ 3 ), the result will be multiplied

by the equivalent number of patients for bal-loon and pacemaker (equivalent number of patients for balloon is 260.138, and equiva-lent number of patients for pacemaker is 116.525).

140.5139 γ

1 /γ

2 = = 0.26396

532.3363

γ 1 /γ

2 * Equivalent Balloon = 0.26396

*260.138 = 68.665

140.5139 γ

1 /γ

3 = = 0.08315

1689.898

γ 1 /γ

3 * Equivalent Pacemaker = 0.08315*

116.525 = 9.689

From these results, this study reveals that the BEP for patients in the cardiac catheteri-zation unit in 2003 actualize with approxi-mately 656 diagnostic patients, 69 balloon patients, and ten pacemaker patients. Figure 5 demonstrates the BEP for each function and illustrates the number of patients served by each function was higher than the BEP.

Expected Benefi t

The EB from the cardiac catheterization unit includes the economic value for both the patients and the government. By using sensi-tivity analysis for fi ve, ten, 15, and 20 years, EB can be calculated for each CVD patient and for the government. Calculating the EB of a patient can be obtained by multiplying the years saved by GDP per capita in 2003

Figure 5. The Number of Patients Requested and the Break-Even Point

Cardiac

Catheterization

Unit Function

Break-Even

Point (Number

of Patients)

Number of

Patients

Served

Diagnosis 656 1,557

Balloon 69 163

Pacemaker 10 23

Total 735 1,743


(which is $896). If the treatment in the car-diac catheterization unit results in preserving the patient’s life (years saved) for fi ve years, then the EB is $4,480.00; for ten years it is $8,960.00, and for 15 years it is $13,440.00. Finally, if the treatment results in preserv-ing the patient’s life for 20 years, the EB is $17,920.00.

The EB for the government can be calcu-lated by using Equation 5 above:

EB = (1 – 20%)* – $23 + 20%

5 years * $896 – $23 = $1,636.25 (5) 54%

This shows that the government achieves an EB in addition to the benefi t obtained from the difference between the revenues and the costs. Everyone is expected to visit the cardiac catheterization unit, patients hav-ing CVD make 20 percent and the EB will be $1,636.25 if the patient’s life is preserved for fi ve years. For ten years the benefi t is $3,295.50, for 15 years it is $4,954.75, and for 20 years it is $6,614.

Discussion

From the study of the cardiac catheteriza-tion unit in fi scal year 2003, we fi nd that vari-able costs are the highest cost. By analysis of the variable cost, it was shown that materials from the hospital’s central storage unit con-tributed the highest portion (51 percent). We also fi nd that the number of patients who vis-ited the cardiac catheterization unit in 2003 was more than the required number so as to reach the BEP. The cardiac catheterization unit works more effi ciently than the required BEP to cover the expenses. These results are similar to fi ndings by Souzdalnitski et al . 22

and Chotiwan et al. 23 —where the number of patients served is higher than the hosp-ital’s BEP.

In terms of governmental subsidies (which consists of salaries, and other oper-ating costs for the unit), the Palestinian gov-ernment offered a subsidy of $480,548.50 to the cardiac catheterization unit in 2003, taking into consideration that the economic revenue of the cardiac catheterization unit was $987,500.00. Considering this revenue level for the unit, the government’s subsidy to the cardiac catheterization unit is covered by the revenue of the cardiac catheterization unit, leading to a difference between the revenue and the expenses of $373,955.37. This also shows that the government can support the cardiac catheterization unit rev-enues before spending on this unit from the government’s revenues as the facts manifest in 2003.

Conclusion

From this study, results reveal that vari-able costs represent 56 percent of the total costs, while fi xed costs are 44 percent of the total costs. Further, we fi nd that the number of patients who visited the cardiac catheteri-zation unit in the year 2003 exceeded the BEP. Despite the subsidy offered by the gov-ernment of $480,548.50 to the cardiac cath-eterization unit for 2003, earnings from the three functions of the unit were suffi cient for the unit to exceed their BEP.

This study’s use of CVP analysis of the cardiac catheterization unit within the Ram-allah hospital can be applied to other units and other hospitals in the region. The use of break-even analysis will allow area manag-ers to plan minimum production capacity for the organization. However, our focus on the


cardiac catheterization unit may not allow our results to be generalizable to other Mid-dle Eastern health care facilities.

Our results are benefi cial to hospital administrators and government regulators. Policy makers may fi nd these results ben-efi cial in future planning and potential revi-sion of the government subsidies offered to health care facilities.

Expert Commentary

CVP analysis is an often overlooked fi nan-cial measurement tool. Health care adminis-trators and managers must know all aspects of providing care to patients, not simply measured by health outcomes. Cardiac cath-eterization services continue to be one of the most expensive, yet accurate, treatments of CVD. Middle Eastern hospitals must use these advanced diagnostic treatments to save lives while simultaneously providing treat-ment that is cost effective for the unit and hospital as a whole. We support the use of CVP analysis for hospital units, not merely the cardiac catheterization unit. All units of a health care facility will benefi t from such a useful cost analysis tool.

The overall health system in Palestine is predominantly provided by the Ministry of Health (MOH), with a centralized govern-ance structure. Recent broad-scale public fi nancial management reforms were initi-ated by the Palestinian National Authority (PNA). The goal of these reform measures is to improve accountability and fi nancial control of operations. Through these ini-tiatives, the MOH will disburse funding to area health care facilities in the hope of improving accountability and fi nan-cial performance. This focus by the PNA on accountability supports the need for

further research of health care costs and cost allocation.

CVP analysis is one of the tools neces-sary to allow hospital managers to discern any probable effects of changes in price, product mix, or sales volume. The use of cost analysis by regulators and administra-tors should improve a facility’s use of scarce resources.

Five-Year View

The current economic condition of Pal-estine, including other Middle Eastern countries, is in a constant state of change. Therefore, a fi ve-year project is complicated; however, one crucial element needed for change is the stabilization of global health care costs.

Due to current economic conditions, health care facilities must strive to oper-ate at their most effi cient level. Combined with the increasing incidence of coronary heart disease, hospital administrators must become more involved in determining their costs of facility operations and effi ciency levels. Assuming some degree of economic stability in the region, health care managers, in conjunction with the MOH will be able to allocate resources to increase the level of hospital effi ciency.

It is projected that government subsidies to hospital units may increase and health care fi nancing should be reformed in three ways—revenue collection, risk pooling, and purchasing—in order to sustain effi cient use of resources over the next fi ve years. This projection is supported by our fi ndings of exceeding necessary BEP despite govern-ment subsidies. An increase in hospital effi -ciency will optimistically lead to improved cardiac care in the region.


REFERENCES

1. Palestinian Central Bureau of Statistics, Pales-tinian National Health Accounts (2010).

2. Kiymaz, H, Akbulut, Y, Demir, A, “Tests of Stationarity and Cointegration of Health Care Expenditure and Gross Domestic Product—An Application to Turkey,” European Journal of Health Economics , 7: 285–289 (2006).

3. Health Management Information System, “The Status of Health in Palestine,” Palestin-ian Ministry of Health (2002).

4. Eastern Mediterranean Regional Offi ce (EMRO), “Cardiovascular Diseases (CVD),” available at http://www.emro.who.int/ncd/cardiovascular-PriorityProg.htm.

5. World Health Organization, Eastern Mediter-ranean Regional Offi ce (EMRO), Introduction, (2000), accessed Feb. 20, 2005 at www.emro.who.int/publications/EMRO%20PUBTPS-GEN-PRT2-CHPT4.

6. Mataria, A, Giacaman, R, Stefanini, A, Naidoo, N, Kowal, P, Chatterji, S, “The Quality of Life of Palestinians Living in Chronic Confl ict: Assess-ment and Determinants,” European Journal of Health Economics , 10: 93–101, p. 100 (2009).

7. Giacaman, R, Mataria, A, Nguyen-Gillham, V, Abu Safi eh, R, Stefanini, A, Chatterji, C, “Quality of Life in the Palestinian Context: An Inquiry in War-Like Conditions,” Health Policy , 81: 68–84 (2007).

8. Merrill, CT, Elixhauser, A, Procedures in U.S. Hospitals, 2003 , Rockville, MD: Agency for Healthcare Research and Quality (2006).

9. Austin, CJ, Boxerman, SB, Quantitative Analy-sis for Health Services Administration , AUPHA Press/ Health Administration Press, pp. 30–32 (1995).

10. Stritecky, R, Pubal, K, “An Analysis of the Ward Medical Apparatus Effi ciency by Means of the Break-Even Point,” Economic Univer-sity of Prague, 2000, available at http://www.urpana.un.org/intradoc/groubs/public.

11. Chotiwan, P, Hempisut, P, Kamolratanakul, P, Dhanamun, B, Tungcharoensathien, V, Hiran-suthikul, N, “Unit Costs of Laboratory Tests at the Outpatient Department of Chulalongkorn Hospital,” Chula Medical Journal , 801–813 (1996).

12. As-Sayaideh, A, Shafei, A, Banks, D, Muhtaseb, A “Implementing Hospital Autonomy in Jor-dan: An Economic Cost Analysis of Al-Karak Hospital,” Partners of Health Reform (PHR), 17–55 (2002); Department of Planning, Ministry of Health and Population, Data for Decision Making, Harvard School of Public Health, and University of California, Berke-ley, “Cost Analysis and Effi ciency Indicators for Health Care: Report Number 3 Summary Output for El Gamhuria General Hospital, 1993–1994,” Boston, Harvard University Press (1997); Zaman, S, “Cost Analysis for Hospital Care, The Case of Embaba Hospital, Cairo, Egypt, Health Financing and Sustain-ability Project,” Technical Note No. 32. Abt Associates: Bethesda, Maryland (1993).

13. Younis, MZ, Jaber, S, Smith, PC, Hartmann, M, Bongyu, M, “The Determinants of Hospital Cost: A Cost-Volume-Profi t Analysis of Health Services in the Occupies Territories: Palestine,” International Journal of Pharmacy Practice , 18 (3): 167–173 (2010).

14. Chodick, G, Porath, A, Alapi, H, Sella, T, Flash, S, Wood, F, Shalev, V, “The Direct Med-ical Cost of Cardiovascular Diseases, Hyper-tension, Diabetes, Cancer, Pregnancy and Female Infertility in a Large HMO in Israel,” Health Policy , 95: 271–276 (2010).

15. Palestinian Health Information Centre, Health Status in Palestine, Palestinian Ministry of Health (2003).

16. Gleim, I, Flesher, D, CMA Review , Gleim Publication, Inc. (2006).

17. Ward, WJ, “An Introduction to Health Care Financial Management,” WTA Communica-tions (1992).

18. Zaman, S, “Cost Analysis for Hospital Care: The Case of Embaba Hospital, Cairo, Egypt, Health Financing and Sustainability Project,” Technical Note No. 32. Abt Associates: Bethesda, Maryland (1993).

19. Shepard, D, Hodgkin, D, Anthony, Y, “Analysis of Hospital Costs: A Manual for Managers,” Institute for Health Policy, Heller School–Brandeis University, 18–23 (1998); Hanson, K, and Gilson, L, “Cost, Resource Use and


Financing Methodology: A Practical Manual,” Bamako Initiative Technical Report Series, No. 16. New York: UNICEF (1993); Conteh, L, Walker, D, “Cost and Unit Cost Calculations Using Step-Down Accounting,” Health Policy and Planning , 19(2): 127–135 (2004).

20. Supra , n.9.

21. Supra , n.17. 22. Supra , n.11; Souzdalnitski, D, Sourovtseva,

E, Kay, J, “Cost Description and Break-Even Analysis of an Acute Pain Service,” Working Paper, American Society of Anesthesiologists (2004).

23. Supra , n.11.

Documents

Journal of Health Care Finance · Frederick, MD 21704, or call 1-800-234-1660. To place an order, call 1-800-638-8437. Subscribers in all countries other than the United States, Canada,