Slide 1

Cost-benefit andcost-effectiveness analysis: applications to health care programs and policies March 2013 Greg Mason & PRA Inc. Greg Mason & PRA Inc.1Program evaluation and cost- effectiveness/cost-benefit analysisHealth economics typically examines specific treatmentsHealth treatments, interventions, programs or policies (TIPP) the form the norm of economic evaluationProgram evaluation - framework for assessing the benefits of TIPP, especially for complex actionsProject evaluation focus on specific interventionValue for money - measures net-benefits from TIPPNet benefits depend on the scope:individual (patient, provider)payer (government, insurer, employer, individual)societyand time horizonshort or immediate (1 year)medium (1-5)long (>5-10)

2 Greg Mason & PRA Inc.2Performance for TIPP measured along the results chainActivitiesOutputsImmediateoutcomes(Typically within a year)Longer-termoutcomes(Up to 15 years)EconomyCost to complete activities(e.g., cost of recruitment, course development)EfficiencyCost per nurse trained(e.g., total training costsdivided by graduates)EffectivenessCost per client that becomes employed(e.g., total program costs divided bynumber of patients that achieve a specific health outcome)3 Greg Mason & PRA Inc.34Economy reflects the unit cost of engaging in activities, such as trainee assessment, case planning for clients, or creating and managing a clinic.

Example: Cost of developing courseware (wages + overhead + materials)Example: Cost of setting up flu clinics (location, staff training, media, notifying vulnerable groups) Greg Mason & PRA Inc.4Efficiency is the unit cost of all the activities needed to produce and output.5Example: Cost per nurse graduate (number of grads divided by the total program cost)Example: Cost per potential flu shot (before any shots are actually given)Example: Cost of awarding a dollar to third party delivery for screening program Greg Mason & PRA Inc.56Effectiveness (also termed cost-effectiveness) is the cost of obtaining one unit of outcome. Example: Number of flu shots (immediate) Reduction in the incidence of flu as a result of a vaccination promotion program (longer term)Example: Increase in screening incidence (short term)Increase in five-year survival rate for Stage 2 breast cancers (long term) Greg Mason & PRA Inc.6Economic valuation in health careCost analysis: This is the most elemental of all economic analysis of health interventions.Example: Comparison of the costs of laparoscopic versus standard surgeries. Essentially descriptive.

Cost minimization: Two or more options for producing identical outcomes are compared with a selection of least cost approaches. This approach only examines direct costs to payers, and does not often include consequences to patients, adverse outcomes, or indirect costs.Example: Search for the least cost pharmaceutical with a give given efficacy

Cost-effectiveness: Compares the ratio of results to costs for two or more options. Example: Cost savings of screening (early detection) for cancers,

7

Greg Mason & PRA Inc. Greg Mason & PRA Inc.7Incremental cost-effectiveness ratio

8BAICER reflects the point where people change their minds.

If the new treatment is at A, then it may be adopted. If the new treatment is at B, the existing treatment may be preferred.The dashed line shows the cost-effectiveness trade where new treatments are not adopted and the old is retained.A proposed treatment at C would result in the old being retainedA new treatment at D would result in the existing treatment being adopted (the old dropped).Existing preferred??New preferredCD Greg Mason & PRA Inc.8Lipitor treatment (hypothetical)9C (0,$0)A (250,$5ook)B (300,$2500k)Total CostTotal Effect (Life years gained) ACER for B = CB/EBACER for A = CA/EAICER for B vs. A= C/ E = $2000k/50= $40,000 per LF Greg Mason & PRA Inc.Cost-utility analysis: In this case, the result reflects the subjective value of the outcome, such as the extension of an additional year of life as some level of quality (QALY).Cost benefit: CEA does not place a money value on the result. CBA provides a method for determining the total social worth of an investment. Unlike CEA, which requires at least two alternatives, CBA can be applied to a single investment.Challenges to CBA:Valuing results (e.g., lives saved) present conceptual challengesAversion to using money to measure the social value of outcomes

10Economic valuation in health care(continued) Greg Mason & PRA Inc. Greg Mason & PRA Inc.10CEA, CUA, and CBA comparedMonetary value of net change to welfareof all outcomes for all stakeholders ($)Cost of outcome ($)Outcomes (outputs, inputs)(actual changes measured in natural units, not $)Social cost ($) (tangible and intangible)CBACEACUASubjective value of outcomes (subjective value of outcomes - adjusted natural units, not $)Cost of outcome ($)11 Greg Mason & PRA Inc.11Application of result chain analysis: the logic modelA logic model is the core tool for conceptualizing causal relationships between program interventions and outcomes.Health screening using third party delivery (e.g., diabetes screening in First Nation communities)Electronic medical records to improve service delivery by providers (currently in primary care clinical settings)

12 Greg Mason & PRA Inc.12

13Results chain for a health screening program Greg Mason & PRA Inc.13Cost-effectiveness of EMRs The electronic medical records storylineElectronic medical records (EMRs) comprise a technical change in health care delivery (primary care/chronic disease management) that supports behaviour change by providers and patients. Primary care physicians record patient care data on an electronic record than replaces the previous paper records.EMR system support a variety of elements (functions)Patient appointment schudulingCentralization of information from all care-giversEliminating drug interactionsTheory of changeEMR produces changes in behaviour that lead to enhanced delivery of health care that improves health outcomes (lowered morbidity/mortality, extension of life quality, and lowered costs)EMRs support changes in health outcomes causally associated with a specific element of an EMR system.

14 Greg Mason & PRA Inc.14Causal models and EMR benefit indicatorsPhysician investment in EMR software (e.g., technology, human and other resources)

Activity (e.g., planning, learning the system, training clinic staff)

Outputs (e.g., implemented systems)

Immediate outcomes (e.g., changed behaviour by clinic staff, increased information flows, coordination of lab information)

Longer-term outcomes (e.g., changes in health, social, environmental, and economic status)

15 Greg Mason & PRA Inc.15EMRs causal logicEMRs comprise the adoption/implementation of technologies that support the delivery of health care by providers.The EMR causal logic comprises: technology adoption (software/hardware features), and activities such as training that... results in the implementation of management systems (outputs) that...support behaviour change by providers and patients, (immediate outcome) that...in turn, support the realization of changes in health status (longer-term outcomes)EMR benefits support behaviour change by providers and patients these behaviour changes lead to changes in health outcomes.EMR feature Behavioural response (provider, patient, system)Immediate outcomeLonger-term outcomeEMR causal logic-specific benefit16 Greg Mason & PRA Inc.16The EMR benefit logic model Note: There are seven general classes of EMR benefit other categories can be used. Within each class, many specific EMR features exist, all of which support health outcomes and all of which can potentially be measured using a specific benefit indicator.

17 Greg Mason & PRA Inc.17Foundations of cost-effectiveness and cost-benefit analysisMeasuring outcomesMeasuring costsCEA and CBA compared 18Three core ideas Greg Mason & PRA Inc.18OutcomesTypical health TIPP outcomes:increase in quality adjusted life years (QALY)avoided costsextension of life reduction of mortalityreduction of morbidityrecovered productivity (earnings power and working life)19 Greg Mason & PRA Inc.19Cost-utility analysis: toward subjective valuation of benefitsusing QALY and health-related quality of lifeMany treatments do not restore perfect health.Subjective measures score perceived health out of 10: health-related quality of life (HRQL) and QALY. Therefore, if patients score their quality of life as .6 (60% of perfect health), then for one year the patient has .6 QALY, and outcomes are valued at 60%.QALY for a patient compared to healthy personYearPatientHealthy1.712.613.514.415.31Total2.55Cost-utility analysis (CUA) uses self-report data (e.g., surveys, focus groups) to measure quality of life and outcome (years of life after medical treatments).20 Greg Mason & PRA Inc.20Measurement of the value of lifeValuation of life:willingness to pay for insurancediscounted present value of expected lifetime earningswage differential for hazardous jobsQALY:combines measure of incremental time and incremental qualitymethodspre-scored instruments (indirect utility)direct elicitation (standard gamble, time trade-off, and rating scale)21 Greg Mason & PRA Inc.21

http://www.medicine.ox.ac.uk/bandolier/painres/download/whatis/QALY.pdf22 Greg Mason & PRA Inc.22Problems with HRQLThe scales used to collect HRQL clearly have potential to bias the measure.As much as possible, patients need to assess the quality of life in their own terms and not in some objective measure of health.People are remarkably adaptable.Methods for HRQL include:preference weighting combines many attributes of health into a single score. Techniques such as factor analysis are common to create a single index of HRQL.large health surveys offer the benefit of large samples to isolate specific subpopulations, but often have restricted dimensions of health.Issues in HRQL:age affects how people think about health disease stage affects perceived HRQLnature of the intervention (many wish to avoid chemo, may opt for surgery as opposed to extended treatment)23 Greg Mason & PRA Inc.23Problems with QALYsPayers are much more interested in actual changes in cash QALYs do not include, among other things, costs of treatment and benefits to the economy by increasing productivity. QALYs need to be updated and are undoubtedly cultural.The value of a life tends to become rule of thumb QALYs now appear to lie between $50,000 and $100,000 this is not often updated and studies appear to use rules of thumb.Chronic diseases where decline is gradual are difficult to express in terms of QALY.Valuations of quality change (increase) as a condition persists. People become accustomed to their new state (eg. Being wheel chair bound or blind)Subjective assessments remain variable and circumstances other that the disease/accident progression.24 Greg Mason & PRA Inc.24CBACEACUAScopeGlobal multiple outcomes valuedLocal single output/outcomeLocal single output/outcomeUnit of measure for outcomeMoney equivalentNaturalUtility or perceived value of outcomeTimeframeExtended Immediate/short termImmediate/short termPrimary decision purposeProspectiveRetrospectiveRetrospectiveApplicationOutcomes or impacts onlyActivities-outputs-outcomesAdjusted cost-effectivenessReferenceNo reference neededAt least one alternativeAt least one alternativeCBA, CEA, and CUA compared25 Greg Mason & PRA Inc.25CEA, CUA, and CBACEA is a specific calculation of the costs needed to produce a unit change in one measurable indicator (input, output, or outcome).CUA is an extension of CEA, and places a subjective valuation on the measured outcome indicator. This is the quality adjusted life year (QALY).CBA measures the net change of all outcomes for all stakeholders by using a ratio of social benefits to social costs. The term social encompasses all private costs to all economic agents as well as the external benefits and costs associated with that activity.CUA depends on measuring QALYs (very flaky) CBA requires a measure of social gains and social costs (very hard) CEA is strongly recommended26 Greg Mason & PRA Inc.26Cost-effectiveness analysisoutputs and outcomes must be strategic(and valid and reliable)Outputs and outcomes should reflect the core goals of the program: An output/outcome selected for CEA should be selected to reflect the activity of a program. For this reason, outputs often replace outcomes in CEA because they are easier to measure.Outputs and outcomes must be strategic:A single output/outcome, if used alone, must represent a central goal of the program. Variation in a strategic outcome should correlate with other results:Increases in that output/outcome should also be correlated with other outcomes.27 Greg Mason & PRA Inc.27Cost-effectiveness analysis example vaccination programs for at-risk groups Outputs include the creation of vaccination sites or alternative delivery (e.g., pharmacies, places of work, physician offices, shopping malls).Outcomes include:the numbers vaccinatedreduction in sick days avoided deathsThe delivery approach with the lowest cost of organizing per potential vaccination delivered is the most cost-effective in terms of outputs.The delivery approach with the lowest number of sick days per dollar cost is the most cost-effective in terms of outcomes.28 Greg Mason & PRA Inc.28Rationality and timeTime flows in one direction. The results chain shows transformation of inputs into outputs into outcomes over time.A trade-off exists between enjoyment now and enjoyment (of more) tomorrow.Interest rates reflect willingness to delay enjoyment (patience).In general, we prefer faster results.The willingness to delay a result (investment, consumption) is time preference.Any economic decision must account for the time preference of all stakeholders.29 Greg Mason & PRA Inc.29Time value of money(discounting)The future value F of an amount P ($100), saved for k years at an interest rate i of 10% is (assuming interest is paid on the last day of each year and the savings are deposited on the first day)With prevailing interest rates of 10%, people should be indifferent between receiving $100 now or $133 in three years. The present value P of receiving $133 in three years (assuming interest at 10%) is $100.The basic idea behind discounting is that we will always choose present consumption to investing/saving for future consumption unless a reward exists for deferring consumption. We need to be paid interest to defer present consumption.Interest expresses time preference.

30 Greg Mason & PRA Inc.30Discounting a future stream of benefits

Imagine you were to receive a regular annual payment (rent or royalty) of $100.

The further in the future a benefit or cost, the less its present value.The present value of the $100 to be received in three years is worth less than the value of the $100 received right now.At 0% we are indifferent between $1 now and $1 in a yearAt 10% we value future earnings much less31 Greg Mason & PRA Inc.31Valuing costs and benefits over time(vaccination)32 Greg Mason & PRA Inc.32Case example: HPV vaccination for girlsapplication of discountingThis is a partial (economists) model.It assumes that the benefit of disease prevention is the avoidance of lost working time*.Preserved income is the only benefit of health policy.Cumulative discounted wages (income) represent the economic value of a life.Life tables and cancer incidence models support the prediction of when death will occur.There are no other costs in this model*.The costs of illness and then recovery (time off due to illness, direct costs of treatment, indirect costs of caregiver time) are not included*.33 Greg Mason & PRA Inc.33HPV vaccinationBelieved to provide protection against 80% of the HPV viruses that cause cervical cancer.The model assumes 100% efficacy*.Cervical cancer is a low incidence cancer that manifests itself (first cases) when women are in their late twenties.Probability of contraction believed to be a function of sexual activity.Vaccinations are given at age 12 and cost $300.Unclear whether a booster is needed.First death occurs at age 20 (according to life tables).In the 2008 cohort of Alberta girls age 12 (n=21,993), 13 are expected to die from cervical cancer between 2008 and 2045.The benefits of the vaccination are the wages for these women calculated between time of death and age 50.34 Greg Mason & PRA Inc.34Key issues on HPV vaccinationWhat outcomes could be monitored?What other costs need to be included in the model?If you were the Minister of Health, what evidence would you need to cancel the program?extend vaccination to boys (who carry HPV, but clearly do not get cervical cancers)?How does HPV vaccination differ from vaccination for an infectious disease?

35 Greg Mason & PRA Inc.35Structure of vaccination CEABenefitreduced short-term cost due to illnessreduced long-term cost for caring for the small number of catastrophic incidentsaverted loss of incomes for those who are disabled/dyingaverted cost of lost time at work and playCosteconomic loss for the small number who experience adverse reactions to vaccine vaccination program

36 Greg Mason & PRA Inc.36Vaccination for measles(Axnick et al., 1969)

Axnick, N.W., Shavell, S.M., and Witte, J.J.,(1969). Benefits Due to Immunization Against Measles, Public Health Reports , Vol. 84, No. 8. 37 Greg Mason & PRA Inc.37

38 Greg Mason & PRA Inc.38

39 Greg Mason & PRA Inc.39Annual stream of benefits and costsAxnick (measles)

40 Greg Mason & PRA Inc.40Summary of measles CEA

41 Greg Mason & PRA Inc.41Partial Cost Benefit Analysis Cancer Screening 42 Greg Mason & PRA Inc.42Decision modellingDecision modelling is rapidly becoming a central methodology.The case paradigm (diagnosis/treatment/death) Markov ModelsDisease (cancer) may be modelled as a series of decision points with knowable probabilities. Each case presents as decision points over time.The typical cycle is one year and the disease is charted as a series of tests (with positive/negative results).Once diagnosis is confirmed, the decision points become treatment points with three outcomes (cured, continued treatment, death).A case cycles for an extended period.Death terminates the cycle.Costs are attached to each step in the cycle (cost of screening, cost of treatment).The disease paradigm (Monte Carlo Models)Many trials (thousands) are conducted reflecting the various outcomes.This builds a profile of the disease in a population some never get the disease and pass through others die at the first cycle.This builds a profile of costs.

43 Greg Mason & PRA Inc.43Estimating the cost-effectiveness of primary health care(cancer screening and diabetes management)

44 Greg Mason & PRA Inc. Greg Mason & PRA Inc.44Primary health careThe ultimate goal of primary health care is better health for all. The World Health Organization has identified five key elements to achieving that goal: reducing exclusion and social disparities in health (universal coverage reforms)organizing health services around people's needs and expectations (service delivery reforms)integrating health into all sectors (public policy reforms)pursuing collaborative models of policy dialogue (leadership reforms)increasing stakeholder participation

World Health Organizationhttp://www.who.int/topics/primary_health_care/en/45 Greg Mason & PRA Inc.45Prospective cost-effectiveness analysis of screening for cancerPrimary care Reduces the net costs to the payer (health insurance, government, and the individual) by avoiding entirely or intercepting disease at a more easily treatable stageAverts disease, yielding to economic benefits to individuals and society because of greater functional life expectancy and the consequent capacity to earn and play other productive rolesIncreases the quality of life, reducing treatment burden on patients and families, and extending the interaction of the patient with family and society

This CEA focuses on the first of these benefits.46 Greg Mason & PRA Inc.46Key cost-effectiveness analysis question for cancer screeningDoes the short-term cost of increasing the uptake of screening for cancer in the general population result in health systems savings because cancers are detected at an earlier stage for a small number of people?Maybe47 Greg Mason & PRA Inc.47Markov modelling of CEAA Markov model creates a probability structure of disease outcomes over an extended period (50 years).This can be applied to many medical situations, with vaccination, and primary health screening being the most common.The simplest screening/vaccination model uses binomial probabilities at each decision point (true-false probabilities).Presumably, the vaccination reduces the severity of getting the disease (P2 < P3)by a lot.Also, cancer screening should detect cancers at an earlier stage.FOBTno testtestno diseasediseasediseaseno diseaseP11-P1P21-P2P31-P348 Greg Mason & PRA Inc.48Markov model of cancer screeningThe figure gives a general overview of how a decision tree dealing with fecal occult blood test (FOBT) screening for colorectal cancer appears.

Each node ( ) is a binary decision, determined by a simple coin toss (but with unequal probabilities).

49 Greg Mason & PRA Inc.49The decision tree of primary health cost-effective analysis:Markov model Monte Carlo simulationThe P1, P2, and P3 values guide the patient through the tree.

The values of Pi come from incidence/prevalence data.

The bottom cancer or state are those who are diagnosed through symptoms and not screening. They are assumed to have more advanced stages of cancer and therefore are more costly to treat.startFOBTno FOBTcancer diagnosiscancer diagnosisno cancer diagnosisno cancer diagnosisP11 -P1P2P31 -P31 -P2healthyhealthyhealthycancercancercancercancer50 Greg Mason & PRA Inc.50Screening cost-effectiveness three cancers

X > Y > ZIn general, population-wide FOBT and mammograms are cost-effective.Population-wide Pap tests are not.51 Greg Mason & PRA Inc.51CautionThis analysis says nothing about the advisability of cancer screening for those at risk (mother/sister/aunt) or for those who have had cervical cancerPap tests are indicated in those individual cases.New therapies (HPV vaccine) may reduce the need for Pap tests.This model only examines the first three categories of cost savings/benefits associated with primary health screening.52 Greg Mason & PRA Inc.52Cost-effectiveness analysis applied to primary health care - summaryUseful estimates that guide prospective CEA are possible. CEA rests on establishing model analogies (the decision tree model).Data to calibrate the models can be drawn from the literature and in the future will be supported by increasing availability of electronic medical records.The adage, an ounce of prevention is worth a pound of cure, needs to be taken with a grain of salt.53 Greg Mason & PRA Inc.53From CEA to CBA and backWhy is CBA so hard?54 Greg Mason & PRA Inc.54

Inputs/activities OutputsImmediate outcomes Final outcomes Resources are mobilized to: Provide counsellingDesign promotional materialTrain primary care providers in the distribution of the screening kitOutreach programs designed and implementedKits distributedStaff trained Lower morbidity and mortalityIncreased life yearsReduced health system costsIncreased social and economic contribution of survivorsAre we getting inputs/resources at the lowest cost?Are resources deployed at least cost?Arethe processes well organizedand coordinated? EconomyAre the activities all needed for the outputs?Are we creating outputs at the lowest cost?Are the outputs available on time and to the required quality?EfficiencyWhat is the cost per unit outcome?Has the project/program produced outcomes at a cost consistent with other approaches?Is this the best/least cost way to get desired outcomes?EffectivenessIntermediateoutcomes Client awareness Increased use of screeningIncreased participation in treatment and prevention Cancer cost-effectiveness results (current trend screening versus Canadian Institute for Health Information screening)

FOBT(colorectal cancer)Mammography (breast cancer) Pap test(cervical cancer)

DescriptionCost per individualCost per individualCost per individual

Current trend cancer screening simulation over 25 years$1,360$2,047$283

Full screening simulation over 25 years$1,327$2,042$381

Difference$33$5$98

Count of simulated individualsTotal estimated savings or costsTotal estimated savings or costsTotal estimated savings or costs

Simulation across population $X million savings$Y million savings$Z million cost