Week 9: (4/1)

University of Minnesota
Medical Technology Evaluation and
Market Research
Course: MILI/PUBH 6589
Spring Semester, 2013
Stephen T. Parente, Ph.D.
Carlson School of Management, Department of Finance
=
Lecture Overview
•
•
•
•
•
Cost / Benefit Review
Discounting
Uncertainty & Risk
Monte Carlo Analysis
SQL Server
Review of Cost/Benefit Analysis
• CE Ratio = Change in Costs / Change in
QALYs.
• To calculate changes in costs measure the
differences in accumulated costs associated with
treatment
• To calculate QALYs—compare patient
outcomes.
• Importance of Time and Dynamics
Discounting
• The impact of most medical technologies
stretches over time.
• The future is “worth less” than the present.
• The key question is how much less?
t
 1 
NPV   
 Cost ( j ) t  Cost (0) t 
t 1  1  r 
T
Whose discount rate?
•
•
•
•
Individuals?
Corporations?
Governments?
Social Planner?
How much does it matter?
• Sometimes it is important, sometimes not. Yes for
disease screenings and many preventions, no for
influenza vaccine.
• Important if there are long time frames and if the
intervention is marginally CE.
• Suppose an intervention increases HRQL by .2 for 10
years. Assume QALYs = $100,000.
• Annual discount rates and PDV of QALYs:
•
•
•
•
0% — $200,000
3% — $175,722
5% — $162,156
10% — $135,180
Discounting Costs
• Two strategies for picking discount rates
– Use the market (e.g. long-term bonds)
– Use a political process.
• Advantage of using the market is that there is a
theoretical justification for doing so.
• However, market based discount rates may not
reflect the type of transactions that are relevant
for health care.
Discounting QALYs
• Some controversy about how (and if) to discount QALYs
• Is a future life year worth less than a contemporaneous one?
• The prevailing practice is to use the same discount rate for costs
as one uses for health consequences.
• If discounts rates between costs and health benefits are different
then either you want to delay interventions or you want to ‘over
implement.’
• Want ‘Horizontal Equity’: (i.e., assuming equal healthcare access
to those who are the same in a relevant respect (such as having
the same 'need').
Issues?
• What about inflation?
– Analysis should be conducted in ‘real’ terms.
• What about uncertainty?
– Use ‘certainty equivalence’ in the analysis: (e.g., The amount of payoff
(e.g. money or utility) that an agent would have to receive to be indifferent between that
payoff and a given gamble is called that gamble's 'certainty equivalent'. For a risk averse
agent (as most are assumed to be) the certainty equivalent is less than the expected value of
the gamble because the agent prefers to reduce uncertainty.)
– Assume risk neutrality to address certainty equivalence
• Should we use a constant discount rate?
– Although, individuals display ‘time-inconsistency’, it is
probably better to use a constant ‘rational’ discount rate.
• Keep valuations and discounting separate!
Which interest rate?
• There are many interest rates out there, which
one, if any, is the right one?
• Captures the rate at which society is willing to
trade off current for future consumption.
• Many interest rates contain a riskpremium/inflation premium.
Which Interest Rate?
• Most CE analysis use 5%.
• Empirical literature suggests a somewhat lower rate.
–
–
–
–
–
Lind (1982) – 2%;
Lesser and Zerbe (1994) 2.5-5%
UK NHS 6%
World Bank 3%
CDC 5%
• Bottom line: Use 3% to 5%
• Sensitivity Analysis: 0% to 7%
Uncertainty and Risk
• There are several sources of uncertainty in CE
analysis.
• Statistical Uncertainty
• Randomness in health & cost outcomes
• Analyzing the CE values for alternative
parameter values is called Sensitivity Analysis.
Statistical Uncertainty
• Parameter Uncertainty
– Intervention reduces chance of heart attack by 12%
with 5% error.
– Cost of intervention is $200 with a standard error of
$40.
• One-way sensitivity analysis changes one
variable at a time.
• Multi-way changes many at a time—Likely need
to do Monte Carlo analysis.
Monte Carlo Analysis
• Computer simulation methodology to examine the impact of
uncertainty.
• Use a random number generator.
• The parameters of interest in CE analysis will have a given
distribution, usually Normal, with a mean and variance.
• Each iteration, take a draw from the distribution of parameters.
• Calculate the CE ratio given the draw and keep track of it.
• The distribution of the CE ratios from all of the iterations is the
distribution of the CE ratio