Medical Research and Health Care Financing: Academic Medical Centers Following the 1997 Medicare Cuts

Pierre Azoulaypa2009@columbia.eduColumbia UniversityGraduate School of Business

Wharton School, University of PennsylvaniaLeonard Davis Institute of Health Economics

February 6th, 2004

Abigail Tayat436@columbia.eduColumbia UniversityDepartment of Economics

Research Agenda

Academic Medical Centers (AMCs) play a crucial role in the American system of biomedical innovation

Research within AMCs comes from 4 sources:– Public sources — mostly NIH, about 65%– Foundations — ignored in this paper, about 10% – Industry — mostly clinical trials, about 15%– “Institutional Funds” — X-subsidies from patient-care activities, about 10%

Cross-subsidies have been a traditional source of seed-research funds, especially for physician-scientists

What have been the effects of changes in health care financing on the level and composition of research in AMCs?

Two views of X-subsidies

Old Boys’ Network– Substitute with other sources of funding

Essential Lubricant– Complement other sources of funding

FinancialSlack

X-Subsidies Research

+ +- or ?

+- or ?

Evolution of Extramural NIH Funds, by Degree of Investigator, 1970-2002

Research Strategy

Pure time series analysis will be contaminated by secular trends such as the massive expansion of the NIH budget during the 1990s

Cross-sectional comparisons across disease areas or research institutions will suffer from omitted variable bias (scientific opportunities, etc.)

We focus on the impact of a discrete shock to hospital finances:Cuts in the Medicare Indirect Medical Education (IME) subsidy following the Balanced Budget Act of 1997

Compare grant awards, before and after 1997, between hospitals that faced a potential large decrease in the level of Medicare reimbursements with those that faced merely a modest decrease

Preview of Results Elasticity of NIH grant awards with respect to health care reimbursements: About .15

– Endowments cushion the effect of the reform– Only very weak evidence that effect is driven by substitution of residents by full-time faculty– Results consistent with the view that hospital X-subsidies complement external sources of funding

Effect shows up “too soon”– One would have expected a 1-2 year lag– Suggests that the reform was anticipated

No response of industry-funded research activity– But more affected hospitals see a rebalancing of their research portfolio towards

clinical trials away from NIH-funded research

Important differences in the magnitude of the response across types of investigators– MDs & MD/PhDs are more affected than PhDs– Human-subjects research is more affected than lab-based research– Young investigators are more affected than experienced faculty members

(the result is not consistent across measures of experience)– No difference between competing and noncompeting grants; only noncompeting funds appear to be affected

(very counterintuitive)

A Primer on the Medicare PPS System

Since 1984, Medicare reimburses inpatient care prospectively, based on the following formula:

$ Reimbursed = Std. Amount × DRG weight × (1+ Teaching Adjustment + Medicaid Adjustment)

%IME %DSH

%IME = × [(1+ #Residents/#Beds).405– 1]<1997 = 1.89

1998 = 1.72

>1999 = 1.60

Data

NIH Consolidated Grant Applicant File

Clinical trial grant data from FastTrack Systems, Inc.

American Hospital Association Survey

AAMC Faculty Roster

HCFA/CMS Cost Reports and IMPACT Files

Area Resource File

HMO penetration variable

Data Issues: JHU and Affiliated Hospitals

Johns Hopkins Hospital

Franklin Square Hospital

Howard County General Hospital

Johns Hopkins Bayview Med. Center

Good Samaritan Hospital

Greater Baltimore Med. Center

Sinai Hospital of Baltimore

Johns Hopkins School of Medicine

Basic Science Departments(Anatomy, Microbiology…)

Clinical Departments (Medicine, Surgery…)

Johns Hopkins Bayview Med. Center

Good Samaritan Hospital

Greater Baltimore Med. Center

Sinai Hospital of Baltimore

Kennedy Krieger Children's Hospital

1

2

3

5

4

Descriptive Statistics:163 Hospitals/“Hospital Aggregates”

  # Obs. Mean Std. Dev. Min. Max.

NIH Grant Awards, Total 1,301 $20,221,863 $30,200,380 $0 $186,525,440

NIH Grant Awards, MDs only 1,301 $11,275,818 $17,139,616 $0 $105,685,576

NIH Grant Awards, PhDs only 1,301 $5,906,290 $8,673,354 $0 $53,164,632

NIH Grant Awards, MD/PhDs only 1,301 $2,990,594 $5,713,233 $0 $43,639,076

NIH Grant Awards, clinical only 1,301 $10,058,137 $15,589,789 $0 $100,212,728

NIH Grant Awards, nonclinical only 1,301 $10,114,564 $15,157,144 $0 $94,685,632

NIH Grant Awards, Career Age < 5 1,301 $2,238,714 $5,008,235 $0 $64,818,416

NIH Grant Awards, Career Age > 5 1,301 $14,237,951 $22,118,375 $0 $156,667,904

NIH Grant Awards, Compet. Funds 1,301 $ 5,297,328 $ 8,167,470 $0 $ 55,544,696

NIH Grant Awards, Noncompet. Funds 1,301 $ 13,935,713 $ 21,261,604 $0 $ 135,425,456

Industry Grant Awards 1,301 $1,580,419 $1,591,144 $0 $10,156,377

Counterfactual Medicare Payments 1,301 $73,951,984 $56,429,525 $2,223,885 $362,655,768

Hospital Employment 1,301 4,626 3,384 135 28,643

Hospital Employment in the HSA 1,301 36,550 31,228 1,831 104,031

Population in HSA 1,301 2,502,211 2,461,715 166,977 12,527,938

Fraction of the pop. 65+ in HSA 1,301 12.30% 2.00% 7.04% 21.00%

Per-capita income in HSA 1,301 $27,223 $6,725 $15,567 $53,340

HMO penetration in HSA 1,301 26.60% 12.70% 1.00% 77.80%

Distribution of Average Yearly NIH Awards,1994-2001

0

10

20

30

40

50

Nb.

of A

MC

s

0 50 100 150Amounts ($ Millions)

Note: Amounts are Def lated by the Biomedical R&D Price Index

Distribution of Average Yearly Industry Awards, 1994-2001

0

5

10

15

Nb.

of A

MC

s

0 2 4 6 8Amounts ($ Millions)

Note: Amounts are Def lated by the Biomedical R&D Price Index

Evolution of Industry Expenditures on Clinical Trials by Type of Site, 1991-2000

0

200

400

600

800

1,000

$ M

illio

ns

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

AMCs For-Profit Centers

“Parameterizing” the Reform

Regressing research outputs on actual Medicare reimbursements is problematic, because hospitals can change behavior in response to the reform

We create a measure of “Counterfactual” Medicare Payments

– Before the Act, Counterfactual = Actual

– After the Act, CMP corresponds to the payments that would have accrued to the hospital based on the new formula if the underlying determinants of reimbursement levels (patient mix, #residents, #beds, #Medicare discharges) had remained at their average pre-reform level

The CMP variable is defined entirely as of the before period: nothing that the hospital does after the passage of the act (e.g., close beds, DRG “upcoding”...) will affect it

Mean Counterfactual Medicare Payments(Balanced PPS Sample, 1994 Real $ ×106)

65

68

70

73

75

78

80

$ M

illio

ns

1994 1995 1996 1997 1998 1999 2000 2001

Impact of BBA Reform [1]

0

20

40

60

80

Num

ber

of O

bser

vatio

ns (

NT

=1,

301)

0 100 200 300 400

Counterfactual Medicare Payments, 1994-2001 ($ Millions)

Impact of BBA Reform [2]

0

5

10

15

Num

ber

of A

MC

s (N

=16

3)

-60 -40 -20 0 20

Change in Counterfactual Medicare Payments Between 1996 and 2000 ($ Millions)

Impact of BBA Reform [3]

0

5

10

15

20

Num

ber

of O

bser

vatio

ns (

N=

163)

-0.10 -0.05 0.00 0.05

Change in Counterfactual Medicare Payments Between 1996 and 2000 as a Fraction of Average Yearly Total Inpatient Revenues

Regression Analyses

2 < 0 or > 0 ?

– Regression weighted by average grant amounts in the pre-period (unweighted residuals exhibit extreme form of heteroskedasticity)

– Standard errors clustered by medical schools– Equations estimated jointly by SUR to account for

contemporaneous correlations of the residuals

Scatterplot of Unweighted Residuals Against Chosen Weights

-15

-10

-5

0

5

10

Unw

eigh

ted

Res

idua

l

0 50 100 150

Average NIH Grant Amounts, 1994-1997 ($ Millions)

Table 3: “After” Dummy Summarizes the Passage of the Reform

NIH Grants

Industry Grants

Industry Grants/

Total Grants to MDs

MDsOnly

PhDsOnly

MD/PhDs Only

Clinical Research

Non-clinical

Research

Ln(CMP)-0.077 -0.295 -0.164 0.068 -0.592 0.348 0.049 -0.142

[0.219] [0.386] [0.327] [0.277] [0.394] [0.643] [0.369] [0.317]

Ln(CMP)×After0.147** 0.076 -0.109† 0.224** 0.125* 0.322** 0.247** 0.097†

[0.035] [0.062] [0.056] [0.045] [0.063] [0.103] [0.059] [0.051]

Ln(#Employees)0.220* 0.090 -0.050 0.194 0.284 0.098 0.210 0.167

[0.100] [0.176] [0.172] [0.127] [0.180] [0.293] [0.169] [0.145]

Observations 1,301 1,301 1,158 1,301 1,301 1,301 1,301 1,301

R2 0.89 0.79 0.82 0.92 0.84 0.84 0.88 0.87

Table 4: Robustness Checks

Basic Specification

State-specific

Time Trends

Hospital Employment/

Year Interactions

Total Inpatient Revenue Control

Ln(CMP)-0.077 -0.065 -0.051 -0.115

[0.219] [0.111] [0.089] [0.094]

Ln(CMP)×After0.147** 0.210** 0.093† 0.151**

[0.035] [0.080] [0.049] [0.042]

Ln(#Employees)0.220** 0.212* 0.157† 0.203*

[0.100] [0.103] [0.080] [0.079]

Ln(Total Inpatient Revenue)

0.094

[1.059]

R2 0.89 0.90 0.89 0.89

Table 5: Year-specific Slopes for the Impact of the Reform

NIH Grants

Industry Grants

MDsOnly

PhDsOnly

MD/PhDs Only

Clinical Research

Non-clinical Research

Ln(CMP)×19950.007 0.125 0.018 -0.195 0.216 0.048 -0.090

[0.067] [0.118] [0.085] [0.120] [0.196] [0.113] [0.097]

Ln(CMP)×19960.030 0.131 0.077 -0.237* 0.327† 0.078 -0.075

[0.067] [0.119] [0.085] [0.121] [0.197] [0.113] [0.097]

Ln(CMP)×19970.046 0.103 0.097 -0.174 0.466* 0.035 -0.054

[0.068] [0.120] [0.086] [0.122] [0.199] [0.115] [0.099]

Ln(CMP)×19980.114† 0.118 0.261** -0.094 0.523** 0.222† -0.004

[0.068] [0.121] [0.087] [0.123] [0.200] [0.116] [0.099]

Ln(CMP)×19990.138* 0.250* 0.289** -0.022 0.764** 0.278* 0.049

[0.068] [0.120] [0.086] [0.123] [0.200] [0.115] [0.099]

Ln(CMP)×20000.153* 0.202† 0.301** -0.043 0.676** 0.288* 0.056

[0.069] [0.122] [0.087] [0.124] [0.202] [0.116] [0.100]

Ln(CMP)×20010.280** 0.093 0.245** 0.047 0.349† 0.371** 0.069

[0.070] [0.123] [0.088] [0.125] [0.203] [0.117] [0.101]

Ln(#Employees)0.229* 0.084 0.180 0.303† 0.035 0.222 0.173

[0.101] [0.178] [0.128] [0.181] [0.294] [0.170] [0.146]

R2 0.89 0.79 0.92 0.84 0.84 0.88 0.87

Are X-subsidies Driving the Effect? [1]

Endowment Below Median

Endowment Above Median

Ln(CMP)-0.120 -0.061

[0.197] [0.103]

Ln(CMP)×After0.182* 0.096**

[0.072] [0.031]

Ln(#Employees)0.420 0.122*

[0.266] [0.056]

Observations 656 645

R2 0.83 0.99

Note: Endowment Measure is the sum of investment income and contributions, bequests and gifts during the 4 years before the reform.

Are X-subsidies Driving the Effect? [2]

All Hospitals

Hospitals Above the BBA Cap in the Pre-Reform

Period

Hospitals Below the BBA Cap inthe Pre-Reform

Period

Ln(CMP)0.250† 0.158 0.381

[0.134] [0.136] [0.237]

Ln(CMP)×After-0.005 -0.038† 0.022

[0.018] [0.023] [0.022]

Ln(Beds)0.559** 0.574** 0.427**

[0.158] [0.189] [0.117]

Observations 1301 640 661

R2 0.98 0.99 0.98

Dep. Variable: Log of number of FTE Residents

Concluding Thoughts

Our results do not suggest that cutting the IME subsidy was a bad idea; rather, we highlight unintended consequences of the reform

Health economists have examined how insurance type, for profit/not-for-profit care, etc. influence current health outcomes

– Meta-analysis of this literature: health care financing does not seem to explain much once selection issues are dealt with adequately

– Our results suggest that financing may affect future health outcomes through its effect on the pace of medical progress

Congress, NIH, academics often focus on the efficiency of the horizontal allocation of public research funds across diseases (Lichtenberg, 2001)

– But the imbalances along the vertical chain of biomedical innovation may ultimately be of greater importance

Youngsters vs. Old-timers:Contradictory Results

First Grantees

Repeat Grantees

Career Age < 5 years

Career Age > 5 years

NoR01Yet

At least One R01

< $500K cum.

funding

> $500k cum.

funding

Ln(CMP)-0.968 -0.086 -0.843 0.865* -0.179 -0.057 -0.285 -0.339

[0.858] [0.239] [0.777] [0.398] [0.394] [0.325] [0.404] [0.367]

Ln(CMP)×After0.016 0.158** 0.446** 0.034 0.188** 0.206** 0.182** 0.413**

[0.138] [0.038] [0.125] [0.064] [0.063] [0.052] [0.065] [0.059]

Ln(#Employees)0.062 0.226* -0.488 0.274 0.102 0.307* 0.149 0.422*

[0.392] [0.109] [0.355] [0.182] [0.180] [0.149] [0.184] [0.168]

R2 0.66 0.88 0.80 0.90 0.77 0.87 0.83 0.99

Competing vs. Noncompeting Funds:Counterintuitive Results

Competing Funds Noncompeting Funds

Ln(CMP)-0.137 0.012

[0.436] [0.259]

Ln(CMP)×After0.088 0.125**

[0.070] [0.042]

Ln(#Employees)0.066 0.290*

[0.199] [0.118]

R2 0.83 0.92