Structures cost-effectiveness and health economic analyses with QALY calculations and model validation. Use when conducting health economics research, calculating QALYs, or building economic models.
Structures cost-effectiveness and health economic analyses with QALY calculations and model validation. Use when conducting health economics research, calculating QALYs, or building economic models.
Health economics and outcomes research (HEOR) provides the evidence that payers, HTA bodies, and policy-makers use to determine drug coverage, pricing, and reimbursement. A well-structured cost-effectiveness analysis (CEA) can secure formulary placement and market access; a poorly designed one can exclude an otherwise effective therapy from coverage. Regulatory bodies (NICE, CADTH, PBAC, ICER, AMNOG) each have specific methodological requirements, and analyses that fail to meet these standards are rejected. This skill implements the CHEERS 2022 reporting standards, ISPOR modeling guidelines, and major HTA-body requirements to produce analyses that withstand technical and methodological scrutiny.
Checkpoint A — Intake and Scoping
Required Intake Questions
What is the target decision-maker (NICE, CADTH, PBAC, ICER, US payer, hospital formulary committee)?
What is the intervention and comparator(s)?
What is the perspective (societal, healthcare-system, payer, patient)?
What is the time horizon (trial-duration, lifetime, fixed period)?
What type of economic evaluation is required (cost-effectiveness, cost-utility, cost-benefit, cost-minimization, budget-impact)?
What clinical data are available (pivotal trial results, systematic review, observational data)?
What are the primary outcome measures (QALYs, LYs, clinical events avoided)?
What discount rate applies (3% US/ICER, 3.5% NICE, 1.5% CADTH)?
Is a budget-impact analysis (BIA) also required?
What is the submission deadline?
Required Source Documents
Clinical trial data (efficacy and safety results for intervention and comparator)
Systematic review of clinical evidence (for model inputs)
Published cost data (drug acquisition costs, administration costs, monitoring costs, hospitalization costs, AE-management costs)
Health-state utility values (EQ-5D, SF-6D, or disease-specific utility studies)
Epidemiologic data (incidence, prevalence, natural history)
Target HTA body's reference case methodology and submission template
CHEERS 2022 checklist
ISPOR modeling guidance documents
Step 1 — Define the Decision Problem
Structure the economic question using the PICOS-T framework adapted for HEOR:
Element
Specification
Population
Target patient population; align with label indication and clinical-trial population
Intervention
Drug/device/procedure with dose, schedule, and duration
Comparator(s)
Current standard of care; include all relevant comparators per HTA-body requirements
Outcomes
QALYs (primary for CEA/CUA), LYs, clinical events, costs
Setting
Healthcare system, country, care setting
Time Horizon
Sufficient to capture all relevant costs and outcomes (lifetime for chronic conditions)
Perspective
As specified by target decision-maker
Discount Rate
Per target-country guidelines
Document the decision problem explicitly — this is Section 1 of most HTA submissions.
Step 2 — Select and Build the Economic Model
Choose the appropriate model structure:
Decision Tree
Best for: Short-term decisions with few health states (acute events, surgical choices, diagnostic strategies)
Structure: Branching pathways with probabilities at each node; terminal nodes with costs and outcomes
Limitation: Cannot model recurrent events or time-varying transitions
Markov Cohort Model
Best for: Chronic diseases with distinct health states and transitions over time (oncology, cardiovascular, autoimmune)
Structure: Health states (e.g., stable disease, progressed disease, death) with transition probabilities per cycle
Cycle length: Typically 1 month or 1 year; half-cycle correction required
Limitation: Memoryless (transition probabilities do not depend on how long a patient has been in a state — unless tunnel states are used)
Partitioned Survival Model (PartSA)
Best for: Oncology (most common model type for NICE/CADTH oncology submissions)
Structure: Area under the curve between overall-survival and progression-free-survival KM curves defines health-state occupancy
Limitation: Transition probabilities are implicit, not explicit; PFS gains must translate to OS gains for the model to be coherent
Microsimulation / Discrete Event Simulation
Best for: Complex patient pathways, heterogeneous populations, treatment sequences
Structure: Individual patients simulated with attributes that affect transitions and outcomes
Limitation: Computationally intensive; harder to validate and debug
Model Selection Criteria
Structural assumptions must be clinically plausible and documented
Model structure should be validated against published models in the same disease area
Target HTA body preferences should inform model choice (NICE prefers PartSA for oncology; CADTH is more flexible)
Step 3 — Populate the Model with Data
Source and document all model inputs:
Clinical Inputs
Treatment efficacy: Hazard ratios, response rates, or survival curves from pivotal trials or meta-analysis
Survival extrapolation (for models extending beyond trial data): Fit multiple parametric distributions; select based on AIC/BIC, visual fit, clinical plausibility, and external validation against registry data
Adverse events: Incidence rates from clinical trials; include only AEs with meaningful cost or utility impact (grade ≥3, or frequent grade 1-2 AEs with significant QoL impact)
Treatment duration and discontinuation: Model treatment duration from trial data; define stopping rules
Cost Inputs
Drug acquisition: WAC, ASP, or net price per dosing cycle; account for vial sharing, wastage, weight-based dosing
Administration: Infusion costs, office-visit costs, self-administration training
Monitoring: Routine lab tests, imaging, office visits per treatment guidelines
Disease management: Costs by health state (stable vs. progressed disease); distinguish direct medical costs, direct non-medical costs (transportation, home care), and indirect costs (productivity loss) based on perspective
Terminal care: End-of-life costs for fatal conditions
Utility Inputs
Health-state utilities: EQ-5D values mapped from clinical-trial data (preferred by NICE, CADTH); SF-6D or direct elicitation as alternatives
Utility decrements: For AEs (applied for duration of the AE) and disease progression
Disutility of treatment: If treatment itself affects QoL (e.g., injection-site reactions, infusion time)
Mapping: If EQ-5D was not collected in the trial, use validated mapping algorithms from disease-specific instruments
Document every input with: value, distribution (for PSA), source, and justification.
Step 4 — Run the Base-Case Analysis
Execute the model and report:
Total costs: Per arm, broken down by component (drug, administration, monitoring, AE management, disease management)
Total QALYs (or LYs): Per arm, broken down by health state
Report meets the target HTA body's submission requirements
Quality Audit
Model is internally validated (extreme-value testing, trace checks, hand calculations for simple cases)
Model is externally validated against observed data (trial results, registry data, published models)
Half-cycle correction is applied (for Markov models) or justified as unnecessary
Discount rate matches the target jurisdiction's reference case
QALYs are calculated correctly (utility × time in health state, summed across all health states)
Cost-year and currency are stated; costs are inflated to a common year using appropriate index (CPI Medical)
PSA runs ≥1,000 iterations with stable results
No double-counting of costs or outcomes across model components
All [VERIFY] flags have been resolved or escalated
Guidelines
The model is a simplification of reality — document every simplifying assumption and its potential impact
Survival extrapolation is the most scrutinized element of oncology models — present all fitted curves, goodness-of-fit statistics, and clinical plausibility assessments
Use net prices (not list prices) when available — HTA bodies will apply their own price assumptions if not provided
QALYs should be based on patient-reported utility values (EQ-5D preferred by NICE); vignette-based utilities are less preferred
Budget-impact and cost-effectiveness analyses answer different questions — do not conflate them
Model transparency is essential — HTA bodies expect executable models with all inputs visible (no black boxes)
For multi-indication products, conduct separate analyses per indication — cross-indication averaging is not appropriate
Validate against competing published models; explain and justify any differences in conclusions
Escalate to health economist when structural uncertainty dominates (multiple plausible model structures yield different conclusions)
This skill produces HEOR analyses and reports — final conclusions and pricing/reimbursement recommendations require health-economics, clinical, and market-access team review
