Presentation on theme: "Terminology and Jargon Demystified"— Presentation transcript:
1 Terminology and Jargon Demystified Meg Franklin, PharmD, PhDFranklin Pharmaceutical ConsultingMarch 27, 2015
2 03 01 04 02 05 Intro table of contents Health Econ Why are we here?03Decision Analysis The life of a tree Calculations01Study Design Types of studies Metrics to report04Economic Models Definitions Trends02Cost Analyses Usual suspects Trends05Conclusions Lessons learned Looking ahead
3 Joke of the DayHow many pharmacoeconomists does it take to change a light bulb?Four1 to estimate the cost of the new light bulb1 to estimate the life expectancy of this new light bulb1 to estimate the QOL associated with the light from the new light bulb1 to package the information so that it convinces the healthcare decision- maker to take out the old light bulb and put in a new one.
4 Why are we here? Introduction Improve writing Communicate effectively Share ideas and experiencesDevelop strategies for publicationsFocus on:NomenclatureWhat to reportCommon pitfallsCalvin and Hobbes, by Bill WattersonIntroduction
6 Observational study designs Cross-sectionalCase-ControlCohortTodayHistorical ControlStudy DesignAdapted from Figure 2-5 in Basic & Clinical Biostatistics (4th Ed).
7 Observational Studies Study DesignObservational study designsObservational StudiesStudy DesignDefinitionCommon usesCase seriesReports characteristics of a small groupHypothesis generationCross-sectionalReports data on a group of subjects at one time rather than over a period of timeDescribes what is happening right now; hypothesis generatingCase-controlBegin with the absence or presence of an outcome and then look backward in time to try to detect possible causes or risk factorsWhat happened?CohortBegins with the exposure and looks forward longitudinally for the outcomeWhat will happen?Study DesignBasic & Clinical Biostatistics (4th Ed).
8 Observational Studies Study DesignObservational Study DesignsObservational StudiesStudyDesignPopulation sizeLongitudinalDirection of observationComparison GroupCase seriesSmallYes, but shortProspectiveNoCross-sectionalBigOne point in timeCase-controlYesRetrospectiveCohortUsuallyStudy DesignBasic & Clinical Biostatistics (4th Ed).
9 Commonly reported metrics Study DesignCommonly reported metricsP values and CIsIf possible, report bothConfidence intervals tell you everything you need to knowSignificanceIdea of the rangeRR and ORsWhen are they alike?Risk vs OddsSignificanceNNT/NNHResonate with cliniciansDichotomous data requiredClinical significance vs statistical significanceStudy Design
11 Common pitfalls Study Design Terminology Significance based on CIs Efficacy vs effectivenessSignificance based on CIsCrossing 0 or 1 (depending on measurement)When metrics can be calculatedType of dataSignificanceStudy Design
13 Comparison of methodologies Cost AnalysesComparison of methodologiesMethodologyCostOutcomeCMADollarsClinical measureCEACBACUAQALYsCCAMultiple(any of the above)Cost Analyses
14 A word on cost-effectiveness Cost AnalysesA word on cost-effectivenessNomenclatureOften times articles will contain cost- effectiveness in the title (or text), when in fact it is really another type of cost analyses.ReportingDetermining the cost-effectiveness threshold is still an issueIssues with the cost-effectiveness planeCost AnalysesDrummond et al (1987).
15 What should be reported Cost AnalysesWhat should be reportedAlternatives to the CE planeNet Health Benefit (NHB)NHB = QALYs – (Cost/ WTP)Net Monetary Benefit (NMB)NMB = QALYs*WTP – CostCost-effectiveness acceptability curve (CEAC)CEACsAllow for the comparison of multiple treatment strategiesWTP is unknown, and foreign concept to many health professionalsCost Analyses
16 Common pitfalls Cost Analyses Terminology Discounting Type of studyDiscountingTime periods beyond 1 year should be discountedDeterministic vs ProbabilisticTrend towards probabilistic analysesStudy Design
17 Decision AnalysisThe life of a treeCalculations
18 Anatomy of a tree Chance Node Terminal Node Choice Node Decision AnalysisAnatomy of a treeChance NodeTerminal NodeChoice NodeDecision Analysis
19 Example Example Scenario: Decision AnalysisExampleExample Scenario:Given the cost of an antibiotic, the probability of success, the probability of an adverse event, and the cost of treating the adverse event, we can construct a decision tree.Specify Possible Costs, Outcomes, and ProbabilitiesFor each option, info should be obtained for the probability of occurrence and the consequences of the occurrenceProbabilities are assigned for each branch of the chance nodes, and the sum must add up to 1Consequences are reported as monetary outcomes, health-related outcomes, or sometimes bothArticles should include a listing of inputs used in the analysis as well as where or how the estimates were obtainedBillomycinMegacillinProbability of Clinical Success90%80%Cost of Antibiotic per Course of Therapy$600$500Probability of Adverse Events10%15%Cost of Treating Adverse Events$1,000Decision Analysis
20 Example Decision Analysis Decision Analysis Rascati KL. Essentials of Pharmacoeconomics. Philadelphia, PA: Lippencott Williams & Wilkins
21 Example Decision Analysis Cost Probability Cost * Probability BillomycinSuccess with no adverse events$6000.81$486Success with adverse events$1,6000.09$144Failure with no adverse events$54Failure with adverse events0.01$16Total for Billomycin1$700Megacillin$5000.68$340$1,5000.12$1800.17$850.03$45Total for Megacillin$65021Decision AnalysisRascati KL. Essentials of Pharmacoeconomics. Philadelphia, PA: Lippencott Williams & Wilkins
22 What to report Decision Analysis Ideally, a picture of the decision tree is includedProbabilities and Costs should be transparentAssumptions and sources should be relevant and accessibleDecision Analysis
23 Common pitfalls Transparency Decision Analysis Assumptions Inputs SourcesCalculationsDecision Analysis
25 Evolution of models Economic Models Deterministic Probabilistic All data is known beforehandOnce you start the process, you know exactly what is going to happenExample: Predicting the amount of money in a bank account.If you know the initial deposit and the interest rate, then you can determine the amount of the account after one yearProbabilisticElement of chance is involvedYou know the likelihood that something will happen, but you don’t know when it will happen.Example: Roll a die until it comes up ‘5’.In each roll, the probability that it comes up ‘5’ is 1/6Don’t know exactly when it will be ‘5’, but we can predict this fairly well..Economic Models
26 Approximately 50- 60% of economic models now are Markov models. When should you use a Markov model?A decision tree becomes too complexThe timeframe for the analysis is lengthyTransitions between health states are possible (e.g. recurrent events)Modeling a complex diseaseProbabilities change over timeDecision models embed Markov processes. Monte Carlo simulations are often used to solve Markov models.Economic Models
27 What to report Economic Models At a minimum, the write-up should include:Type of modelAssumptionsInputsResultsLimitationsEconomic Models
28 Common pitfalls Economic Models Transparency Terminology Assumptions InputsSourcesCalculationsTerminologyRate vs probabilityEconomic Models