Cost-Effectiveness Analysis Econ 737.01 April 14, 2011
Outline • I. Introduction • II. Example • III. Costs • IV. Benefits
I. Introduction • How do you determine the appropriate approaches to testing, treatments, etc.? • For HIV (Elgar Ch. 44): • Which antiretroviral drug regimen should be offered? • What criterion should be used for starting antiretroviral therapy? • What tests should be used to monitor patients before and after antiretroviral treatment? • Should a second line of treatment be offered if necessary?
I. Introduction • Cost-benefit analysis is typically used to guide decisions in economics, but applying it to health care would put us in the morally sticky situation of explicitly placing a dollar value on a year of life. • Cost-effectiveness is the somewhat more politically correct alternative. • 1) Compute dollars required to save each year (or quality-adjusted year (QALY)) of life • 2) Accept treatment if $/year is below a given threshold. • Determining the threshold still involves implicitly placing a $ value on a year of life, but it’s somewhat more palatable.
I. Introduction • We can’t just “let the market decide” because of all the market failures involved with health care (imperfect information, moral hazard, etc.). • Shouldn’t we just use the approaches that work the best? Not necessarily – as economists, we care about utility, not just health. • Who would care about whether a treatment is “cost-effective”? • Patients (if they’re paying) • Doctors (are they health maximizers, patient utility maximizers, or social welfare maximizers?) • Insurance companies (they can structure coverage to incentivize cost-effective treatments) • Government (same; they have budgets to worry about)
I. Introduction • Incremental cost-effectiveness ratio: ratio of changes in costs to changes in benefits if we switch from a “default” treatment to a new treatment • C1: net present value of total lifetime costs of new treatment • C0: net present value of total lifetime costs of default treatment • E1: effectiveness of new treatment, measured in (discounted) life years or QALYs gained • E0: effectiveness of default treatment • Average cost-effectiveness ratio: special case where C0 and E0 are assumed to be zero • Incremental cost-effectiveness ratio is what’s typically used, since for most conditions there is already some available treatment.
II. Example (from Elgar) • We are in charge of a country’s health system. We have $6 million to allocate, and 11 programs to choose from. (See Table 44.2.) • “Shopping spree” approach: Order them by average cost-effectiveness, then pick them until budget is exhausted (Table 44.3). • We end up funding programs E, A, B, I, F, C, D, and K.
II. Example • Now four new (mutually exclusive) programs are developed for a particular condition (Table 44.4). • Should we divert resources from one of the currently funded programs to make room for one of the new ones? • Start with the cheapest option M1. It’s C/E ratio $1250 is lower than the C/E ratio for the worst program that made the cut ($5500 for K). So, M1 would indeed be funded.
II. Example • Now, compute incremental C/E ratios for each of the remaining options and see if any are better than the cheapest option. • M2: Incremental C/E ratio for M2 relative to M1 is $5000, which is less than the threshold of $5500 so switch to M2. • M3: Incremental C/E ratio is $15000>$5500, so no. • M4: Drop M3 from the table and compute incremental C/E ratio for M4 relative to M2 (Table 44.5). C/E ratio is $6000>$5500 so no.
II. Example • Weak dominance: M3 is “weakly dominated” because the incremental C/E ratio for M4 relative to M3 is smaller than the incremental C/E ratio for M3 relative to M2. This means there is no threshold for which we would ever pick M3. • Strict dominance: Suppose we had another option that cost $200,000 and produced 25 QALYs. This would be “strictly dominated” because M2 costs less and produces more QALYs. • We can express each incremental move among the Ms as part of the larger table (Table 44.6).
II. Example • Key assumptions • Fixed budget determines threshold. How would you determine threshold otherwise? (Economists have started using utility maximization to do this.) • Budgets are often period-specific, not based on net present values over a long time horizon • Predictable utilization • Independence of programs • Divisibility (can spend whatever’s left over on the marginal option) • Costs and benefits can be accurately measured
III. Costs • Costs are generally considered more straightforward to measure than benefits. • Do you use marginal costs, average costs, or prices? • These can be very different … for instance, with pharmaceuticals fixed costs are huge (R&D), marginal costs are small, and price is well above marginal cost. • Often price is used in this case, but this ignores producer surplus.
IV. Benefits • Issues: • Defining the health states • Assigning probabilities to the health states • Assigning utilities to the health states • Fi=probability person is still alive at age i • δ=time discount factor • qi:=Expected value of preference weight at age i; 1 indicates perfect health, 0 death, between 0 and 1 alive but imperfect health
IV. Benefits • Survival probabilities • Experimental data: usually limited to short (<5 years) follow-up periods, extrapolation necessary • Observational data: endogeneity concerns • Preference weights • Source of considerable uncertainty • How do you deal with heterogeneity in preferences? • Given difficulties estimating key parameters, sensitivity analysis is crucial