Cost-Effectiveness Analysis

Transcription

1 Cost-Effectiveness Analysis Henry A. Glick, Ph.D. Pharmacoeconomics April 9, 0 Outline Introduction to cost-effectiveness analysis (CEA) Choice criteria for CEA The cost-effectiveness frontier Net benefits (a transformation of CEA) and choice criteria Additional topics Cost-Effectiveness Analysis (I) Estimates costs and outcomes of intervention Costs and outcomes are expressed in different units If outcomes are aggregated using measures of preference (e.g., quality-adjusted life years saved), referred to as cost utility analysis

2 Cost-Effectiveness Analysis (II) Results meaningful if: They are compared with other accepted and rejected interventions (e.g., against league tables), or There exists a predefined standard (i.e., a maximum acceptable cost-effectiveness ratio or an acceptability criterion) against which they can be compared (e.g., $0,000 per year of life saved might be considered the maximum acceptable ratio), or We can define utility curves that trade off health and cost (not discussed further) Cost-Effectiveness History $/Life saved $/Year of life saved (YOL) $/Quality adjusted life year saved (QALY) Why CEA Rather Than CBA? Not precisely clear Potential difficulties in measurement Discomfort with placing a dollar value directly on a particular person's life (rather than years of life in general) Potential ethical issues

3 Potential Ethical Issues / life years more equally distributed than wealth Gini Coefficients for life expectancy and wealth (measure of equality between 0 and., with larger values representing greater inequality) Birth cohort: 0. Current population:. Wealth: 0. Health more a right than a commodity, thus person vote may be more appropriate than dollar vote Cost-effectiveness analysis uses QALY/year vote Cost-Effectiveness Ratios Cost-effectiveness ratio Costs - Costs Effects - Effects A ratio exists for every pair of options option (case series), no ratios calculated options, ratio options, ratios (option versus option, option versus option, and option versus option ) In the efficient selection algorithm, we don t necessarily calculate all the possible ratios Average Vs. Incremental C-E Ratios Some dispute about definitions e.g., Some use average cost-effectiveness ratio to refer to the practice of dividing a therapy s total cost by its total effect (including Treeage, a fairly ubuiqitious piece of decision analysis software)

4 Dividing a Therapy s Costs by Its Effects is Generally Uninformative Cost Effect Example Rx 00.0 Rx (780-00) / (.06-.0) = 80,000 Example Rx 00.0 Rx 00.0 (00-00) / (.0-.0) = 6,667 Ratio 0,000 0,000 0,000 0,000 Average Vs. Incremental C-E Ratios We don t define the average CER by dividing a therapy s total cost by its total effect Treeage, a fairly ubuiqitious piece of decision analysis software, does We recommend against calculation of these ratios They provide little to no information We instead define the average cost-effectiveness ratio as the comparison of costs and effects of each intervention with a single option, often the "do nothing" or usual care option Example: Average Ratios and the Sixth Stool Guaiac # Guaiac Tests Cost * (C i C ) / (E i E ) Cases Detected Avg Cost/ Case Detected * 9 88,78,79 6,80 Neuhauser and Lewicki, NEJM, 97;9:6-8.

5 Incremental Cost-Effectiveness Ratios Comparison of costs and effects among the alternative options (i.e., excluding the comparator used for the average cost-effectiveness ratios) When there are only options being evaluated, the average and incremental cost-effectiveness ratios are the same Guaiac Average and Incremental Ratios # Guaiac Cases tests Cost Detected * (C i C ) / (E i E ) ** (C i C i- ) / (E i E i- ) Neuhauser and Lewicki, NEJM, 97;9:6-8. Average CER * 9 88,78,79 6,80 Increm CER ** ,86,687,00,000,000 Cost-Effectiveness Plane -oo oo (-) Difference in Cost (+) Alternative therapy dominates Alternative therapy more effective but more costly New therapy more effective but more costly New therapy dominates Axes Origin Average ratios Incremental ratios oo -oo (-) Difference in Effect (+)

6 Choice Criteria For Cost-Effectiveness Ratios Choose options with acceptable average and incremental cost-effectiveness ratios (i.e., whose ratios with all other options are acceptable) Subject to: Budget Constraint? Acceptable Ratio? Not accounting for uncertainty around the ratios Consider mutually exclusive options Choice Criteria, Example Option Option Option Expected Costs 0,000,000 70,000 Expected 0 0 Ratios Option Option Option,000 6,000 Option 7,000 Adopt? Choice Criteria, Example Option Option Option Expected Costs 0,000,000,000 Expected 0 6 Ratios Option Option Option,000 7,00 Option 00,0000 Adopt? 6

7 Choice Criteria, Example Option Option Option Expected Costs 0,000 0,000 0,000 Expected 0. Ratios Option Option Option 00,000 6,667 Option 0,000 Adopt? Multitherapy Example Suppose 6 screening strategies have the following discounted costs and life expectancies: Treatment Cost No screening (S) 0 Sig Q0 (S) 88 Sig Q (S) 6 U+Sig, Q0 (S) 80 C Q(0) (S) 08 U+Sig, Q (S6) 0 Frazier AL, et al. JAMA. 000;8:9-6. YOLS Choice Among Screening Strategies Which therapy should be adopted if the acceptability criterion is $0,000 / YOL Saved? $0,000 / YOL Saved? In what follows, demonstrate methods for selecting a single therapy from among these candidates All methods are based on selecting the therapy with an acceptable ratio All methods are transformations of one another they use same information in slightly different ways and all yield identical choices 7

8 Method : Efficient Algorithm (EA) for Choosing among Multiple Therapies (I) Suppose 6 therapies have the following discounted costs and life expectancies Treatment No screening (S) Sig Q0 (S) Sig Q (S) U+Sig, Q0 (S) C Q(0) (S) U+Sig, Q (S6) Cost YOLS Efficient Algorithm: Step Rank order therapies in ascending order of either outcomes or costs (the final ordering of the nondominated therapies will be the same which ever variable you choose) Treatment No screening (S) Sig Q0 (S) Sig Q (S) C Q(0) (S) U+Sig, Q0 (S) U+Sig, Q (S6) Cost YOLS Efficient Algorithm: Step Eliminate therapies that are strongly dominated (i.e., that have increased costs and reduced effects compared with at least one other alternative Treatment No screening (S) Sig Q0 (S) Sig Q (S) C Q(0) (S) U+Sig, Q0 (S) U+Sig, Q (S6) Cost YOLS

9 Efficient Algorithm: Step Compute incremental cost-effectiveness ratios for each adjacent pair of outcomes (e.g., between options and ; between options and ; etc.) Treatment No screening (S) Sig Q0 (S) Sig Q (S) C Q(0) (S) U+Sig, Q0 (S) U+Sig, Q (S6) Cost YOLS ICER 780 7,0 Dom 8,0,800 Efficient Algorithm: Step Eliminate therapies that are less effective (cost) but have a higher cost-effectiveness ratio (weakly dominated) than the next highest ranked therapy Rationale: Rather buy more health for a lower cost per unit than less health for a higher cost per unit e.g., eliminate S (sig,q), because: S is less effective than the next higher ordered S (U+sig,Q0) [7.87 YOLS vs. 7.0] AND The incremental ratio for moving from S to S (7,0) is greater than the incremental ratio for moving from S to S (8,0) Implies that moving from S to S is more costeffective than is moving from S to S Efficient Algorithm: Step Recalculate the ICERs (e.g., between options and ) Repeat steps and if necessary) Treatment No screening (S) Sig Q0 (S) Sig Q (S) C Q(0) (S) U+Sig, Q0 (S) U+Sig, Q (S6) Cost YOLS ICER 780 7,0 Dom,70,800 9

10 Efficient Algorithm: Step 6 Identify the acceptable therapy Maximum WTP < to,79 70 to,799,800+ Therapy S S S S6 Full Cost-Effectiveness Table Treatment Cost ΔC YOLS Δ Y S No screening S Sig Q S Sig Q S C Q(0) S U+Sig, Q S6 U+Sig, Q SD = strong dominance; WD = weak dominance ICER 780 WD SD,70,800 Reduced Cost-Effectiveness Table Treatment S No screening S Sig Q0 S U+Sig, Q0 S6 U+Sig, Q Cost ΔC 6 YOLS Δ Y ICER 780,70,800 0

11 Introduction to Method : Frontier Analysis (Geometry of Choice) Discounted Costs ($) CER = Slope Discounted Years of Life Saved We can also identify the optimal strategy using the costeffectiveness plane. In many cases, we focus on the upper right quadrant, where new therapies increase both costs and outcomes CHOOSING AMONG FRONTIER OPTIONS () Example $0,000 Discounted Costs ($) O O Discounted Options and both have acceptable average costeffectiveness ratios (i.e., below $0,000/YOLS) Choosing Among Frontier Options () Example Discounted Costs ($) O O $0, Discounted Years of Life Saved To evaluate the incremental ratios, shift the origin to the option with the lowest acceptable average costeffectiveness ratio, and reimpose the $0,000 acceptability criterion

12 Colorectal Cancer Screening Example 000 COL, Q0 years U & SIG, Q years 70 U & SIG, Q0 years $,800 Costs ($) 00 0 $0,000 per YOLS SIG, Q0 years SIG, Q years $,70 $7, Years of Life Gained The convex hull represents the therapies that for a given level of effect have the lowest cost (or for a given level of cost have the highest effect Strong Dominance S: COL, Q0 years S6 Strong Dominance S: U & SIG, Q0 years Costs ($) 00 S: SIG, Q years 0 S: SIG, Q0 years Years of Life Gained Weak Dominance 000 S: COL, Q0 years S6 Costs ($) S: SIG, Q years Weak Dominance S: SIG, Q0 years S: U & SIG, Q0 years Years of Life Gained

13 Sig,q and the Frontier Weakly dominated, but Uncertainty (i.e., confidence region) might be such that we may not be able to exclude it from the frontier Weakly dominated therapies that lie close to the frontier, "might be considered [a] reasonable alternative...if there were noneconomic reasons to prefer them, such as patient or physician acceptability, availability, or other factors." Mark D. JAMA. 87;0:8-9. Method. Choice Using a Predefined Maximum Acceptable C-E Ratio U & SIG, Q years COL, Q0 years U & SIG, Q0 years $,800 Costs ($) 00 0 $0,000 per YOLS SIG, Q years SIG, Q0 years $,70 $7, Years of Life Gained Choose the therapy with a tangency between frontier and the lowest line with a slope defined by the maximum willingness to pay for the health outcome Method Recommendations Choose the therapy with a tangency between frontier and the lowest line with a slope defined by our maximum willingness to pay Maximum WTP < to,79 70 to,799,800+ Therapy S S S S6

14 Introduction to Method : Net Benefits A composite measure (part cost-effectiveness, part cost benefit analysis), usually expressed in dollar terms, that is derived by rearranging the cost-effectiveness decision rule: W > ΔC /ΔQ where W = willingness to pay (e.g., 0 or 00K) Net Benefits (II) Two forms of the net benefit expression exist depending on the rearrangement of this expression Perhaps most naturally for economists, net monetary benefits can be expressed on the cost scale (NMB) (W * ΔQ) - ΔC Alternatively, net health benefits (NHB) can be expressed on the health outcome scale: ΔQ - (ΔC / W) A potential disadvantage of the latter transformation is that NHB is undefined when the CR equals 0 NMB Rationale Overcomes problems associated with parametric tests of the ratio Study result is a difference in means, not a ratio of means, and is always defined and continuous Substitutes a poor-person s willingness to pay measure (the acceptability criterion) for the more theoretically correct individually-measured willingness to pay Differs from cost-benefit analysis in that it does not aggregate individuals' willingnesses to pay All else equal, we should adopt programs with net monetary (health) benefits that are greater than 0 (i.e., programs with incremental cost-effectiveness ratios that are less than WTP

15 Net Benefits and the CE Plane (I) Difference in Costs Net monetary benefit, - $,000 Net monetary benefit, - $,00 Net monetary benefit, $0-700 Acceptable upper limit, $0, Difference in QALYS Net monetary benefit, $,00 On the CE plane, NMB is represented by a family of lines all with a slope equal to W Net Benefits and the CE Plane (II) Each line represents a single value of net benefits For NMB, -intercept (because at the origin, W ΔQ = 0 and the formula reduces to -ΔC For NHB, point where the line intersects the horizontal axis For the line passing through the origin, both NMB and NHB = 0 Lines below and to the right of the net benefit=0 line have positive net benefits (i.e., acceptable costeffectiveness ratios) Lines above and to the left have negative net benefits *** Method, above, is equivalent to selecting the therapy with the largest valued NMB *** NMB and the Multitherapy Example Returning to the previous multitherapy example: suppose 6 therapies have the following discounted costs and life expectancies Treatment No screening (S) Sig Q0 (S) Sig Q (S) C Q(0) (S) U+Sig, Q0 (S) U+Sig, Q (S6) Cost YOLS Which therapy should be adopted if the acceptability criterion is $0,000 / YOL Saved? $0,000?

16 Numeric Net Monetary Benefit Methods Can follow a modified version of method to calculate incremental NMB Modifications for calculation of Incremental NMB: In step, calculate NMB rather than costeffectiveness ratios In step, eliminate therapies that are less effective (cost) but have a smaller NMB than the next highest ranked therapy (weakly dominated) In step, recalculate the NMBs Select the therapy that has BOTH the greatest effectiveness AND a positive incremental NMB Comparing 6 Strategies' Monetary Benefits Alternatively, can calculate the monetary benefit (MB) for each therapy based on its own costs and effects rather than incremental costs and effects Step. Calculate each therapy s MB by multiplying the therapy s average (NOT incremental) effect times WTP and subtracting the therapy s average cost MB i = WQ i - Ci Select the therapy with the greatest MB Yields the same conclusions as the other methods for selecting a therapy Method Cost YOLS NMB, $0K NMB, $0K No Scr (S) ,868 * 866,8 Sig, Q0 (S) ,8 867,6 Sig, Q (S) ,9 867,8 U+Sig, Q0 (S) ,70 868,90 C,Q0 (S) ,8 867,77 U+Sig,Q (S6) ,6 868,6 * (0,000 * 7.8) = 69,90, subtracting 0 = 69,868 6

17 Exercise: Selecting a Therapy Suppose you evaluated therapies and observed the following costs and effects Using method, which strategy would you recommend if WTP = 0,000? If WTP = 7,000? Step.??? Step Rank Order Step. Rank order the therapies by increasing cost or effect

18 Step.??? Step Dominated Therapies Step. Eliminate any strongly dominated therapies There are no strongly dominated therapies Step.??? Step

19 Calculate ICERS Step. Calculate incremental cost-effectiveness ratios ICER , , , ,000 Step.??? Step ICER , , , ,000 Weakly Dominated Therapies Step. Eliminate any weakly dominated therapies ICER , , , ,000 Eliminate strategy with an ICER of 0k because strategy is more effective and has a lower ICER 9

20 Step.??? Step ICER , , , ,000 Recalculate the ICERS Step. Recalculate the ICERS ICER , , , ,000 Step 6.??? Step 6 ICER , , , ,000 0

21 Therapy Selection Step 6. Select the option with the largest ICER that is lower than the maximum WTP ICER , , , ,000 # if WTP=0,000; # if WTP=7,000 Recommendation? Maximum WTP <0,000 0,000 to <,000,000 to <0,000 0,000+ Therapy S S S S Simultaneous Comparison Description of the selection algorithm may suggest that we take a path through different options, which assumes we will adopt lower cost/effect pairs before we will adopt higher cost/effect pairs Instead, all algorithms are simply step-by-step procedures that simultaneously compare all of the options As done by identifying the tangency between the NMB lines and the "health production" frontier, or By comparing MBs

22 Goal of Selection Process The goal of the selection process is to choose options with acceptable average and incremental costeffectiveness ratios Choose options whose ratios with all other options are acceptable Implication: We cannot ignore the economic value of U and Sig every 0 years and U and Sig every years when evaluating Sig every years or colonoscopy every 0 years What Is the Maximum Acceptable Ratio? Traditionally, cost-effectiveness ratios less than $0,000 to $0,000 per quality-adjusted life-year saved (or net monetary benefit cost lines defined using these ratios) have been considered acceptable Little analytic attention has been given to identifying an appropriate acceptability criterion There has been a growing debate about whether the acceptability criterion in the U.S. has increased (e.g., at a minimum to $00,000 per QALY) Not clear that acceptable levels derived for the point estimate of the cost-effectiveness ratio should be used to determine the acceptable levels for the upper limit of the confidence interval for the cost-effectiveness ratio What Is the Maximum Acceptable Ratio? US Gov t EPA: 9. M / life (~K / undiscounted YOLS) FDA: 7.9 M / life (~76K / undiscounted YOLS) DOT: 6 M / life (~K / undiscounted YOLS) Australia: $AU K - 76K /YOLS Italy: 60,000/QALY Netherlands: /QALY Sweden: SEK 00,000 (,000) / QALY UK: 0-0K / QALY WHO report: times GDP per DALY

23 Are All Ratios of Equal Value? Mortal, relatively incurable diseases vs. diseases that principally affect quality of life Are acceptable ratios for the former higher than for the latter? NICE, appraisal committees can consider giving greater weight to achieved in the later stages of terminal diseases (Nature, 09/009) As more treatments become available and the disease appears less incurable, does the acceptable incremental ratio for new therapies begin to approach the "standard" acceptable ratio? Small budgetary impact Are All Ratios of Equal Value? (II) Identifiable individuals Do individuals have a set of social preferences that differ from their individual preferences $,000,000 to cure 00 blind invalids $,000,000 to cure 00 blind healthy individuals Compensation for risks imposed by society Acceptability and the Lower Left Quadrant? Economists usually treat ratios in the upper right and lower left quadrants symmetrically If we would not spend more than $0,000 per QALY saved for a more costly and more effective new therapy in the northeast quadrant I, then we would not spend more than $0,000 per death averted for a more costly and more effective alternative therapy in the southwest quadrant i.e., we would adopt a less costly and less effective new therapy if its ratios of savings per QALY lost were greater than $0,000 compared with the alternative

24 Acceptability and the Lower Left Quadrant? (II) Some have suggested that preferences for gains and losses of health are asymmetric Common assumption is that people need to be paid more to give up health than they are willing to pay to gain health (possibly an income effect) Such asymmetries can be incorporated into decision making for individual therapies, but complicates NMB calculation, construction of acceptability curves, and league-table decision making Negative Cost-Effectiveness Ratios If the point estimates for the differences in costs and effects are of opposite signs (either increase costs and decrease effectiveness or decrease costs and increase effectiveness), the resulting cost- effectiveness ratio will be negative The magnitude of negative point estimates for ratios in the same quadrant does not provide information about the relative preferability of these different therapies Negative Ratios (II) When reporting on the cost-effectiveness of a therapy (e.g., if you are comparing only two options), and the resulting cost-effectiveness ratio (or the CI of the ratio) is negative, do not report the negative value (because the magnitude conveys little if any information) Instead simply report that the ratio represents that the therapy is dominant/dominated If the lower and upper limits of the confidence interval (CI) for the CER are both negative, the relative magnitude of the two limits provides information about whether or not the CI includes the Y axis of the CE plane (return to this idea when we discuss sampling uncertainty for CERs)

25 Take Home Messages (I) Decision making using cost-effectiveness ratios requires attention to average and incremental cost-effectiveness ratios To make decisions using these ratios, they must be compared to: (Most common:) Other accepted and rejected interventions (e.g., against league tables), or (Growing in use:) A predefined standard (i.e., an acceptability criterion) against which they can be compared (e.g., $0,000 per year of life saved might be considered the largest acceptable ratio), or (Rarely or never:) Utility curves trading off health and cost Take Home Messages (II) Use of a predefined standard (e.g., $0,000 per year of life saved) equates decision making using costeffectiveness ratios and decision making using net monetary benefits Do not report the magnitude of negative point estimates of cost-effectiveness ratios