tomas philipson
TRANSCRIPT
Endogenous CEA in Health Care Technology
Adoption(NBER WP #15032)
Anupam Jena Harvard University
Tomas J. Philipson University of Chicago
Leonard Davis InstituteDecember 4, 2009
Motivation New technology is a driving force behind
growth in health care spending How do we value new technologies?
“Cost-Effectiveness” (CE): “Bang-for-the-Buck” CE Analysis largest subfield of health economics?
Research Question: Efficiency implications of adopting new technologies based on CE?
Cost-Effectiveness in Practice European Union
“Fourth hurdle” Prior to 1993, few countries had agencies responsible for economic assessments of new medical products
Now, majority do (Drummond, 1991; OECD, 2001; Cookson et al., 2003)
United Kingdom Threshold for adopting new technologies by NICE appears to be ~
$60,000 per QALY (Raftery et al., 2001) Australia
First country to require pharmacoeconomic assessments of all new drugs submitted for national coverage
By 2001, only 2 of 26 new submissions were accepted whose cost per QALY exceeded $57,000 (Bethan et al., 2001)
Cost-effectiveness and the probability of treatment adoption, NICE 1999-2005
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
< 10,000 £ 10,000 - 20,000 £ 20,000 - 30,000 £ 30,000 - 40,000 £ 40,000 - 50,000 £ > 50,000 £
Cost-effectiveness (£ per QALY/LYG)
Pro
babi
lity
of
acce
ptan
ce
Preview of Punch Lines Exogenous CE uses resource COSTS
Determines economic efficiency or gains from trade Endogenous CE uses PRICES
Mark-ups above costs affected by: Patient & Doctor Demand + Adoption Rules (!)
Bang for the Buck? “Buck” depends on Demand Endogenous CE reverses exogenous CE
When mark-up differences reverse cost differences (Devices vs Drugs?)
How to Test for Reversals Data from NICE 1999 - 2005
Exogenous and Endogenous CE p = price of medical product (drug, device, service) q = quality or “effectiveness” of product (QALY) c = cost of producing product Exogenous CE : c/q Endogenous CE : p/q However: Endogenous prices are affected by the
reimbursement rule used! Example: Fixed thresholds cause firms to price up to
threshold regardless of costs
Mark-ups and Reversals Prices marked up above costs
p = m*c For two technologies, reversals occur whenever
Treatment 1 is more cost effective exogenously: c1/q1 < c2/q2
Treatment 2 is more cost effective endogenously: p1/q1 > p2/q2
Mark-ups offset exogenous cost-effectiveness m1/m2 > [c2/q2]/[c1/q1]
Example: NICE pricing p/q=T m=T/[c/q]
Profits and Technology Adoption Demand: y(p,q ) Profits conditional on approval
π(p) = [p-c(q)]y(p,q) A(p) = Probability of technology approval falls in
price Example: CE ratios lowers adoption A(p/q)
Expected Profits=Probability of Approval*Profits A(p)*π(p)
Mark-up Determination Mark-ups depend on demand In standard monopoly pricing models, markups falls
with the elasticity of demand E Lerner condition p = m*c where m = 1/[1+E]
Here, markups depend on two demand sides Price sensitivity of adoption rule: A(p) Price sensitivity of ex-post demand: y(p,q)
Both demand sides affect mark-up P = m(Demand,Approval)*c If CEA is used by governments for adoption, then this
determines endogenous CE!
Optimal Pricing- Nonzero rejection- Reduced price due to technology adoption
0
1A(p)
p
A(p)π(p)
π(p)
Adoption Probability Profits
Optimal price balances gains in profits with increased rejection:
A’π + A π’=0
π’/π h π’/π h
Price, p
Class Dummies and Reversals Cannot directly identify reversals without
information on prices, costs, and quality Test for reversals of a “Procedure”
Adoption not solely driven by endogenous CE Low Goodness of Fit consistent with political factors
affecting adoption Class heterogeneity induces reversals Class Dummies to test for reversals
Reversals in cost-effectiveness & Class heterogeneity in adoption
Price
Costs and Exogenous CE
Low Adoption Class p(c)
High Adoption Class p(c)
cL cH
pL
pH
pM
Empirical Analysis – Data from NICE
Since 1999, NICE issued 141 guidances Our data includes 86 guidances involving 145
treatments 30 percent recommended unconditionally 32 percent w/ minor restrictions 22 percent w/ major restrictions
76 of these treatments have explicit CE data 12/76 of these treatments flat out rejected
Estimated unconditional acceptance (A) and hazard (h) as a function of CE levels (p/q), NICE 1999 - 2005
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 10 20 30 40 50 60 70
Endogenous cost-effectiveness, p/q (£ per QALY)
Acc
epta
nce
prob
abil
ity,
Haz
ard
Acceptance probability
Hazard
Number of treatments submitted and accepted by disease class (k) and endogenous cost-effectiveness
(p/q), NICE 1999-2005
Endogenous Cost-effectiveness (1,000£/QALY)
Disease Class < 10 10 - 20 20 - 30 30 - 40 40 - 50 > 50
Arthritis 0/0 5/5 0/0 2/2 0/0 0/1
Cancer 6/6 8/8 3/4 5/5 2/3 0/0
Heart 6/6 1/1 4/4 0/0 0/0 0/0
Infectious 2/2 0/0 2/2 0/3 1/1 ¼
Mental 0/1 4/4 0/0 1/2 0/0 0/1
Prevention 1/1 1/1 2/2 0/0 0/0 0/0
Other 2/2 1/1 1/1 1/1 1/1 1/1
Source: NICE published treatment guidances, 1999 – 2005. Each cell reports the number of accepted treatments/submitted treatments for a given disease class and endogenous cost-effectiveness range.
Impact of endogenous cost-effectiveness and disease class on probability of treatment acceptance Variable
Mean cost-effectiveness (1,000£/QALY) -0.009*
(0.002)
Cancer -0.034
(0.098)
Heart -0.031
(0.122)
Infectious -0.322*
(0.120)
Mental health -0.310*
(0.132)
Prevention -0.008
(0.171)
Constant 1.154
(0.096)
R2 0.38
F-test of equality of disease indicators p = 0.03
Source: NICE published treatment guidances, 1999 – 2005. Table presents coefficients of a linear probability model of the impact of cost-effectiveness and disease class (excluded class: diabetes) on the probability of treatment adoption by NICE. Standard errors are in parentheses. * Significant at p < 0.05.
Limitations & Future Issues Sample Reversals vs Procedure Reversals
Difficult as markups unobservable Endogenous Effectiveness as opposed to Costs
Learning by doing rises with lower price (devices) Transparency
Measured by goodness if fit of criteria explaining adoption
Endogenous Comparative Effectiveness