Economic Evaluation of Health Interventions

Introduction and context

Health-care decision makers routinely face choices among interventions that differ in clinical benefit, cost, risk, timing and distributional consequences. A new medicine may improve survival but require substantial expenditure; a screening programme may prevent disease years later; a digital intervention may reduce hospital admissions while increasing primary-care workload. Economic evaluation provides a structured way to compare these alternatives and determine whether the additional health produced is worth the additional resources used.

Economic evaluation is the comparative analysis of alternative courses of action in terms of both their costs and consequences. It is not simply cost accounting, nor is it a substitute for clinical evidence. Rather, it combines clinical epidemiology, health economics, statistics, decision analysis and managerial judgement. The central question is usually: What additional health outcome is obtained for the additional cost of one intervention compared with its relevant alternative?

This chapter develops the principal methods used in economic evaluation: cost-minimization analysis (CMA), cost-effectiveness analysis (CEA), cost-utility analysis (CUA), and cost-benefit analysis (CBA). It also explains quality-adjusted life-years (QALYs), disability-adjusted life-years (DALYs), incremental cost-effectiveness ratios (ICERs), discounting, willingness-to-pay thresholds, sensitivity analysis and budget-impact analysis.

Learning objectives

By the end of this chapter, the reader should be able to:

1. Distinguish the main forms of economic evaluation and select an appropriate method.
2. Identify, measure and value costs and health outcomes from a stated perspective.
3. Calculate QALYs, DALYs, incremental costs, incremental effects and ICERs.
4. Apply discounting and interpret willingness-to-pay thresholds.
5. Conduct and interpret deterministic and probabilistic sensitivity analyses.
6. Distinguish cost-effectiveness from affordability using budget-impact analysis.
7. Critically appraise equity, ethical, methodological and governance issues in economic evaluation.

1. Foundations of economic evaluation

1.1 Scarcity, opportunity cost and the counterfactual

Health-care resources are scarce. Spending on one programme means that money, staff, beds, equipment or managerial attention cannot be used elsewhere. The value of the best forgone alternative is the opportunity cost. Consequently, an intervention should not be judged only by whether it produces benefits; it should be judged against what could have been achieved with the same resources.

Every evaluation requires a comparator or counterfactual. The comparator may be current standard care, usual practice, no programme, watchful waiting, or another new intervention. An inappropriate comparator can make an intervention appear attractive or unattractive for reasons unrelated to its true value.

1.2 Perspective

The perspective determines which costs and consequences are included. Common perspectives are:

A societal perspective is theoretically comprehensive, but productivity valuation and informal-care measurement can be controversial. A narrow payer perspective may be appropriate for a reimbursement decision, but it can omit important costs transferred to patients, families or other sectors. The perspective must be declared before analysis and applied consistently.

1.3 Types of cost

Relevant costs may include:

• Direct medical costs: medicines, consultations, diagnostics, procedures, hospitalisation and rehabilitation.
• Direct non-medical costs: transport, accommodation, childcare and social-care services.
• Indirect costs: productivity losses caused by illness, disability or premature mortality.
• Intangible costs: pain, anxiety and suffering, which are important but difficult to value monetarily.
• Capital costs: buildings, equipment and information systems. These should generally be annualised rather than charged entirely to the first year.
• Future costs: costs occurring after the primary intervention, including treatment of adverse effects or unrelated future illness, depending on the analytic convention.

Economic evaluation should use opportunity costs where possible. The purchase price of a resource is often a practical proxy, but it may not equal its true economic value when there is substantial unused capacity, negotiated pricing or price discrimination.

2. Choosing the evaluation method

2.1 Cost-minimization analysis

CMA compares costs when two interventions are assumed to produce equivalent outcomes. The least-cost alternative is selected:

where C₁ is the cost of intervention 1, C₀ is the cost of the comparator, and ΔC is the incremental cost. CMA is valid only when equivalence in relevant outcomes has been established, not merely assumed. If clinical equivalence is uncertain, CEA or CUA is more appropriate.

For example, two generic formulations may have demonstrated bioequivalence and equivalent adherence. If annual costs are £420 and £500, the first formulation saves £80 per patient per year. However, a cheaper formulation with poorer persistence is not a legitimate CMA conclusion unless the difference in outcomes is incorporated.

2.2 Cost-effectiveness analysis

CEA expresses outcomes in natural clinical units, such as life-years gained, strokes prevented, cases detected or symptom-free days. It is especially useful when the outcome is clinically salient and common to the alternatives. Its limitation is that comparisons across disease areas can be difficult: cost per stroke prevented cannot be directly compared with cost per tuberculosis case prevented without an additional decision rule.

2.3 Cost-utility analysis

CUA uses a generic measure of health-related quality and quantity, most commonly the QALY. It permits comparisons among interventions affecting different diseases, although it embeds value judgements about health states and may not capture all socially important outcomes.

2.4 Cost-benefit analysis

CBA expresses both costs and benefits in monetary units. It can show whether total monetised benefits exceed total costs and can compare health programmes with non-health investments. Monetary valuation may use willingness to pay (WTP), stated-preference methods, revealed preferences or the human-capital approach. CBA is conceptually broad but ethically and methodologically sensitive, particularly when valuing life, disability and unequal ability to pay.

3. Measuring health outcomes

3.1 QALYs

A QALY combines length of life and health-related quality of life. A year in perfect health is conventionally assigned a utility of 1, death a utility of 0, and states worse than death may have negative values. For a sequence of health states:

where uᵢ is the utility weight for health state i, bounded by the valuation method, and tᵢ is time spent in that state in years. The formula assumes utility is constant over each interval or that the interval is sufficiently short for the approximation to be reasonable.

If a patient gains 2 years at utility 0.8 followed by 3 years at utility 0.5, QALYs are:

Utilities may be obtained using instruments such as EQ-5D, standard gamble, time trade-off or visual analogue scales. They are not the same as percentage survival or a clinician’s subjective rating. Analysts should specify the instrument, tariff, population and method of handling missing data.

3.2 DALYs

A DALY measures health loss rather than health gained:

where YLL is years of life lost due to premature mortality and YLD is years lived with disability. A basic formulation is:

where N is the number of deaths and L is the standard remaining life expectancy at the age of death. For prevalence-based disability measurement:

where P is time lived with the condition and DW is the disability weight between 0 and 1. Incidence-based calculations may instead multiply incident cases by expected duration and disability weight.

DALYs are valuable for population health and global burden-of-disease comparisons, while QALYs are frequently used for technology appraisal. They are not interchangeable: a QALY represents health gained according to a utility scale, whereas a DALY represents health lost relative to an ideal. Both require normative assumptions.

3.3 Life-years and clinical outcomes

Life-years gained are transparent and clinically meaningful but ignore quality of life. Disease-specific measures may be more sensitive to the intervention, but they restrict comparisons across programmes. Outcome selection should reflect the decision problem, evidence quality and intended use of the results.

4. Incremental analysis and the ICER

The key economic comparison is incremental rather than average. Let intervention 1 be the new intervention and intervention 0 the comparator:

The incremental cost-effectiveness ratio is:

The unit is cost per additional unit of effect—for example, £25,000 per QALY gained. The ICER is meaningful when ΔE ≠ 0 and when the comparator is appropriate. It should not be calculated against a comparator that is strictly dominated.

An intervention is dominant if it costs less and produces more health. It is dominated if it costs more and produces less health. Extended dominance occurs when a combination of other alternatives produces more effect for a lower incremental cost per unit of effect. Proper analysis therefore orders interventions by effect, removes simple dominance and sequentially assesses extended dominance before interpreting the final ICER.

Worked example: CUA

Suppose a hospital is considering a remote-monitoring programme:

Then:

If the decision maker’s threshold is £30,000 per QALY, the programme is cost-effective under the base-case assumptions. This conclusion does not mean that the programme is affordable, guaranteed to work in every setting or ethically superior for every patient.

5. Cost-benefit analysis and willingness to pay

In CBA, net monetary benefit (NMB) provides a convenient decision rule:

where λ is the WTP threshold per unit of health effect, E is the health effect and C is cost. An intervention is preferred to its comparator when incremental net monetary benefit is positive:

The intervention is cost-effective if ΔNMB > 0, equivalently if its ICER is below λ when incremental effects and costs are positive.

For the worked example at λ = £30,000 per QALY:

Thus, the programme has positive incremental net monetary benefit of £1,500 per patient. NMB is often preferable for statistical analysis because it avoids ratios with unstable denominators and supports regression, confidence intervals and probabilistic analysis.

A CBA may estimate monetary benefit directly through WTP. If 1,000 households each would pay £40 annually for a prevention programme, gross stated benefit is £40,000 per year. This estimate must be interpreted cautiously: stated WTP is affected by income, framing, strategic responses and understanding of risk. Ability to pay is not the same as social value, so distributional effects must be examined.

6. Time preference and discounting

Costs and health benefits occurring in different years are not directly comparable because people and institutions generally place greater value on benefits and costs occurring sooner. Discounting converts future amounts to present values. For a future amount X_t occurring in year t at annual discount rate r:

For a stream of costs and effects:

where Cₜ and Eₜ are costs and effects in year t, and r_C and r_E are discount rates. Some guidelines use the same rate for costs and effects; others recommend differential rates. The applicable national or institutional guidance must be followed and reported.

If a programme produces £100,000 of benefit in year 3 at a 3% rate:

Discounting is distinct from inflation adjustment. Costs should first be expressed in constant prices using an appropriate price index, and then future values should be discounted. Analysts should state the base price year, currency, discount rates and treatment of half-year or continuous-cycle corrections.

7. Decision modelling

Economic evaluations commonly use decision trees for short-term pathways and Markov or state-transition models for chronic disease. A decision tree represents mutually exclusive branches and probabilities. A Markov model represents health states—such as stable disease, relapse, post-event and death—and transitions between them during defined cycles.

For a branch with probability p_j, cost C_j and effect E_j, expected values are:

with ∑ⱼ pⱼ = 1. Probabilities must be mutually exclusive and collectively exhaustive. In Markov analysis, transition probabilities must be valid—between 0 and 1, with each row summing to 1—and the cycle length, state rewards, discounting and half-cycle correction must be justified.

Models are useful when trials do not observe all relevant lifetime outcomes. They also create risks: structural assumptions, extrapolation, treatment waning, competing mortality and parameter uncertainty can materially influence results. A model should be transparent, internally validated, externally checked against observed data and subjected to sensitivity analysis.

8. Uncertainty and sensitivity analysis

Uncertainty has several sources:

• Parameter uncertainty: sampling error or uncertain costs, probabilities and utilities.
• Structural uncertainty: alternative model structures or assumptions.
• Methodological uncertainty: perspective, discount rate, time horizon or valuation method.
• Heterogeneity: genuine differences across patients, settings and subgroups.

8.1 Deterministic sensitivity analysis

One-way sensitivity analysis changes one parameter at a time over a plausible range. A tornado diagram can rank influential parameters. Two-way analysis varies two parameters, such as treatment effect and price. Scenario analysis changes a coherent set of assumptions, such as the time horizon or inclusion of productivity costs.

Ranges should be justified using confidence intervals, credible intervals, published studies, expert elicitation or plausible policy scenarios. Arbitrary ranges create misleading reassurance.

8.2 Probabilistic sensitivity analysis

In PSA, uncertain parameters are assigned probability distributions and jointly sampled, commonly using Monte Carlo simulation. Typical choices include beta distributions for probabilities and utilities, gamma or log-normal distributions for positive costs, and Dirichlet distributions for sets of competing probabilities. The analyst generates a distribution of incremental costs and effects and reports the probability that an intervention is cost-effective at different thresholds.

A cost-effectiveness acceptability curve (CEAC) shows:

for alternative values of λ. PSA quantifies decision uncertainty; it does not eliminate it and should not be confused with patient-level treatment-effect prediction.

9. Budget-impact analysis

Cost-effectiveness asks whether an intervention provides sufficient value relative to its opportunity cost. Budget-impact analysis asks whether the payer can absorb the financial consequences over a specified period. The two analyses answer different questions and should not be substituted for one another.

A simplified annual budget impact is:

where N_t is the number of eligible patients receiving the intervention in year t, and C₁,t and C₀,t are per-patient costs under the new and current scenarios. A fuller analysis includes uptake, population growth, treatment discontinuation, offsets, administration and implementation costs.

For example, if 2,000 patients are treated, additional annual cost is £3,000 per patient, and implementation costs are £200,000:

Budget impact should specify the eligible population, uptake assumptions, time horizon, current treatment mix, price agreements, uncertainty range and whether costs are cash expenditures or economic opportunity costs. A highly cost-effective intervention may be unaffordable in the short term; a budget impact can support phased adoption, price negotiation, capacity planning or managed entry.

10. Illustrative case: a digital heart-failure monitoring service

This is a fictional illustrative case, not a real-world recommendation.

Background and problem

A regional health system is considering a digital service for adults recently discharged after heart-failure admission. Patients receive remote weight and symptom monitoring, nurse review and escalation protocols. The service may reduce readmissions but requires technology, training and staff time.

Analysis

The evaluation adopts a health-system perspective and a two-year horizon. Costs include the platform, nurse time, emergency visits and admissions. Outcomes are QALYs derived from a validated health-status instrument. The comparator is current discharge planning. A decision model combines trial readmission rates with longer-term mortality and quality-of-life estimates.

Base-case results are:

Thus:

At a threshold of £30,000 per QALY, incremental NMB is:

The service is cost-effective in the base case. However, PSA shows that its probability of being cost-effective is only 68%, principally because the readmission effect and staff workload are uncertain. A two-year budget-impact analysis estimates 4,000 eligible patients by year two, with 70% uptake. If the incremental cash cost is £2,700 per patient and implementation costs are £350,000, year-two impact is approximately:

Approach and solution

The health system could adopt the service with conditions: a staged rollout, prospective monitoring of readmissions and patient-reported outcomes, predefined stopping rules, workforce assessment and renegotiation if usage exceeds assumptions. Equity monitoring should examine digital access, language, disability, deprivation and the needs of patients unable to self-monitor.

Outcome and learning points

The case demonstrates that a favourable ICER does not by itself establish affordability, implementation feasibility or equity. Decision makers should integrate cost-effectiveness, budget impact, operational capacity, evidence maturity and distributional consequences.

11. Ethics, equity and governance

Economic evaluation necessarily makes value judgements. Maximising total QALYs may disadvantage people with severe disability, older people or those requiring high-cost treatment. Equal QALY valuation can conflict with severity-based or priority-to-the-worst-off principles. Analysts should therefore report distributional consequences rather than presenting efficiency as the only legitimate objective.

Important governance practices include:

• declaring perspective, comparator, time horizon, discount rates and price year;
• preregistering or documenting the analysis plan where feasible;
• disclosing funding, conflicts of interest and manufacturer involvement;
• separating observed evidence from assumptions and expert judgement;
• protecting confidential patient data used in modelling;
• testing subgroup results without using inappropriate statistical stereotypes;
• avoiding double counting of costs or benefits; and
• explaining uncertainty to decision makers rather than hiding it behind a single ICER.

Technology and AI can assist with electronic health-record costing, automated identification of resource use, synthetic control construction and model calibration. Yet algorithmic predictions may reproduce unequal access, coding bias or historical under-treatment. AI-supported evaluations require data governance, external validation, explainability proportionate to the decision, monitoring for drift and clear human accountability.

12. Best-practice workflow

A defensible evaluation generally follows these steps:

1. Define the decision problem, population, setting and decision time.
2. State the perspective and select the relevant comparator.
3. Identify all relevant costs, outcomes and resource consequences.
4. Select the evaluation type and outcome measure.
5. Assemble clinical, epidemiological, utility and cost evidence.
6. Choose an appropriate model and justify its structure.
7. Discount future costs and outcomes according to applicable guidance.
8. Calculate incremental costs, effects, ICERs and net monetary benefits.
9. Examine dominance and validate the model.
10. Conduct deterministic, scenario and probabilistic sensitivity analyses.
11. Conduct a separate budget-impact analysis where affordability matters.
12. Report limitations, equity implications, implementation requirements and uncertainty.

A high-quality evaluation is transparent enough for another analyst to reproduce, clinically credible enough for practitioners to trust, and decision-relevant enough for managers and policy makers to use.

Chapter Summary

Economic evaluation compares alternative health interventions by relating their resource use to their consequences. CMA is appropriate only when outcomes are demonstrably equivalent. CEA uses natural clinical outcomes, CUA commonly uses QALYs, and CBA monetises both costs and benefits. QALYs combine utility and survival, while DALYs combine years of life lost and years lived with disability.

Decision making is based on incremental costs and effects, not averages. The ICER expresses additional cost per additional unit of health, while net monetary benefit converts health effects into monetary decision terms using a willingness-to-pay threshold. Discounting places future costs and outcomes on a comparable present-value basis. Sensitivity analysis addresses parameter, structural and methodological uncertainty. Budget-impact analysis assesses affordability and cash-flow consequences, which are distinct from cost-effectiveness.

Professional practice requires explicit perspective, credible comparators, transparent assumptions, model validation, appropriate uncertainty analysis and attention to equity, ethics, privacy and implementation capacity. Economic evidence should inform—rather than mechanically replace—clinical, organisational and societal judgement.

Key Terms

Opportunity cost: Value of the best alternative use of resources forgone.

Perspective: Viewpoint determining which costs and consequences are included.

Comparator: Alternative intervention or current practice against which a new intervention is assessed.

Cost-minimization analysis: Evaluation selecting the least costly option when outcomes are equivalent.

Cost-effectiveness analysis: Evaluation relating costs to outcomes in natural clinical units.

Cost-utility analysis: Evaluation using a preference-weighted generic health measure, commonly the QALY.

Cost-benefit analysis: Evaluation expressing both costs and benefits in monetary terms.

QALY: Quality-adjusted life-year combining health-related utility with duration.

DALY: Disability-adjusted life-year measuring years of healthy life lost.

Utility: Preference weight representing the value of a health state.

ICER: Incremental cost divided by incremental effect.

Dominance: A situation in which one option is more costly and less effective, or less costly and more effective, than another.

Net monetary benefit: Monetary value of incremental health minus incremental cost.

Willingness-to-pay threshold: Maximum acceptable cost per unit of health gain.

Discounting: Converting future costs or effects into present values.

Sensitivity analysis: Examination of how results change when assumptions or parameters vary.

Probabilistic sensitivity analysis: Joint simulation of uncertain parameters using probability distributions.

Budget-impact analysis: Estimation of the financial effect of adopting an intervention within a defined population and period.

QALY threshold: Decision benchmark for the monetary value assigned to one additional QALY.

Markov model: State-transition model used to represent disease progression over repeated cycles.

Review Questions

5 Short Answer Questions

1. Why is opportunity cost central to economic evaluation?
2. State one condition required for valid cost-minimization analysis.
3. Define a QALY and explain the meaning of a utility weight.
4. What is the difference between an ICER and net monetary benefit?
5. Why should budget-impact analysis be conducted separately from cost-effectiveness analysis?

5 Long Answer Questions

1. Compare CMA, CEA, CUA and CBA, including their outcomes, strengths and limitations.
2. Explain how perspective, comparator and time horizon can change the conclusions of an economic evaluation.
3. Describe the construction and interpretation of QALYs and DALYs, including important normative assumptions.
4. Explain dominance, extended dominance, discounting and the appropriate interpretation of an ICER.
5. Discuss deterministic and probabilistic sensitivity analysis and their roles in communicating uncertainty.

3 Analytical / Application Questions

1. A new programme costs £14,500 per patient and produces 2.10 QALYs; usual care costs £12,000 and produces 2.00 QALYs. Calculate the incremental cost, incremental QALYs and ICER. At a threshold of £30,000 per QALY, calculate incremental net monetary benefit and interpret the result.
2. Design a budget-impact analysis for introducing a high-cost medicine over five years. Specify the eligible population, uptake, comparator, implementation costs, treatment displacement, uncertainty and outputs required by a payer.
3. Critically appraise the fictional digital heart-failure case. Identify at least four assumptions that could change the decision and propose an appropriate sensitivity or validation method for each.

Multiple Choice Questions with Answers

Question 1

Which condition is essential for a valid cost-minimization analysis?

A. The interventions must have demonstrably equivalent relevant outcomes.

B. The new intervention must have a lower ICER than all alternatives.

C. All benefits must be expressed in monetary units.

D. The intervention must produce at least one additional QALY.

Correct Answer: A. The interventions must have demonstrably equivalent relevant outcomes.

Question 2

An intervention costs £11,000 rather than £8,000 and produces 3.35 rather than 3.20 QALYs. What is its ICER?

A. £2,000 per QALY

B. £20,000 per QALY

C. £22,000 per QALY

D. £30,000 per QALY

Correct Answer: B. £20,000 per QALY

Question 3

Which expression correctly defines a QALY for constant utility over each time interval?

A. Utility ÷ time

B. Cost × utility

C. Utility × time

D. Time ÷ mortality

Correct Answer: C. Utility × time

Question 4

What does a DALY measure?

A. Monetary benefit from treatment

B. Health loss represented by years of life lost plus years lived with disability

C. The price of one additional QALY

D. Only the number of deaths in a population

Correct Answer: B. Health loss represented by years of life lost plus years lived with disability

Question 5

At a willingness-to-pay threshold λ, which intervention is preferred on net monetary benefit grounds?

A. The one with the highest total cost

B. The one with negative incremental net monetary benefit

C. The one with positive incremental net monetary benefit

D. Always the intervention with the highest QALYs, regardless of cost

Correct Answer: C. The one with positive incremental net monetary benefit

Question 6

What is the present value of £100,000 received in year 3 at a 3% annual discount rate, approximately?

A. £91,514

B. £97,000

C. £100,000

D. £109,273

Correct Answer: A. £91,514

Question 7

Which statement best distinguishes budget-impact analysis from cost-effectiveness analysis?

A. Budget impact measures affordability, whereas cost-effectiveness measures value relative to health gained.

B. Budget impact always includes QALYs, whereas cost-effectiveness never does.

C. Budget impact is only used for clinical trials.

D. Cost-effectiveness analysis excludes costs.

Correct Answer: A. Budget impact measures affordability, whereas cost-effectiveness measures value relative to health gained.

Question 8

What is probabilistic sensitivity analysis designed primarily to assess?

A. Whether a comparator has been clinically approved

B. Joint uncertainty in model parameters and the resulting decision uncertainty

C. The exact future price of every resource

D. Only the effect of changing one parameter at a time

Correct Answer: B. Joint uncertainty in model parameters and the resulting decision uncertainty

Question 9

An intervention that costs less and produces more health than its comparator is described as:

A. Dominated

B. Extendedly dominated

C. Dominant

D. Cost-neutral

Correct Answer: C. Dominant

Question 10

Which perspective generally includes productivity losses and informal care when relevant?

A. Societal perspective

B. Hospital pharmacy perspective only

C. Laboratory perspective

D. Medicine acquisition perspective

Correct Answer: A. Societal perspective