03 · What You Need to Know
Choosing Between Trial-Embedded and Standalone Economic Evaluation
What Is an Embedded Economic Evaluation?
An embedded economic evaluation is conducted alongside an intervention trial or another primary research study, with economic measurements incorporated into the research protocol.
For example, a randomized trial may collect healthcare utilization, intervention costs, and health outcomes from participants assigned to different treatment conditions.
In education, a trial comparing two instructional support programs may collect instructor time, technology expenditure, participant expenses, and learning outcomes.
When the data support an appropriate comparison, researchers may estimate incremental costs and consequences within the study population.
However, embedding economic measurement does not automatically make the analysis valid. The economic question must still be consistent with the trial's design, follow-up, population, and available information.
What Is a Standalone Economic Evaluation?
A standalone economic evaluation is conducted as a distinct study rather than as an integrated component of the original intervention trial.
It may use existing trial results, observational datasets, administrative records, published literature, or decision-analytic models.
For example, researchers may evaluate the long-term economic consequences of an intervention after several effectiveness studies have been completed.
A standalone evaluation may also examine a decision problem that differs from the original trial, such as implementation across an entire health system or university network.
The term standalone describes the organization of the economic study, not a single required methodology. A standalone evaluation may be model-based, observational, or otherwise designed according to its economic question.
How Do Embedded and Standalone Economic Evaluations Differ?
| Feature |
Embedded evaluation |
Standalone evaluation |
| Relationship to intervention research |
Integrated into the primary trial or study |
Conducted as a distinct economic investigation |
| Typical evidence |
Prospectively collected trial costs and outcomes |
Existing studies, records, models, or new economic data |
| Population |
Often follows trial eligibility and recruitment |
May address a different decision population |
| Comparator |
Usually aligned with trial conditions |
May include alternatives not studied in the original trial |
| Time horizon |
Often supported directly by trial follow-up |
May use modeling to address longer horizons |
| Important advantage |
Coordinated measurement of costs and outcomes |
Flexibility to address a broader economic decision |
| Important limitation |
May inherit trial-specific constraints |
May depend more heavily on external evidence and assumptions |
Neither approach is universally superior. A well-designed embedded evaluation may provide strong comparative evidence, while a carefully developed standalone model may better represent the consequences relevant to a longer-term policy decision.
When Is Embedding Economic Evaluation Within a Trial Appropriate?
Embedding is particularly useful when the trial's population, intervention, comparator, and follow-up correspond closely to the intended economic question.
For example, a pragmatic intervention trial may provide an opportunity to observe both outcomes and resource utilization under conditions relevant to routine practice.
Economic data can be collected using compatible assessment schedules and participant identifiers, subject to appropriate governance safeguards.
Prospective planning can also reduce the risk that important resource quantities are unavailable when the trial concludes.
Researchers should consider whether economic evaluation should be planned prospectively when trial-based economic analysis is an important objective.
When Might a Separate Economic Evaluation Be Preferable?
A separate evaluation may be appropriate when the economic question cannot be answered adequately within the original trial.
For example, the trial may have a short observation period while the intervention influences costs and outcomes over many years.
Alternatively, the trial may use a highly selected population, intensive research support, or a comparator that does not reflect the alternatives available to decision-makers.
A standalone evaluation may integrate evidence from several studies or use administrative data to address a broader population.
However, combining evidence introduces assumptions about comparability, transferability, and causal relationships that must be examined explicitly.
Does the Intervention's Development Stage Matter?
Yes. The maturity of intervention evidence influences the value of economic evaluation.
An early efficacy study may focus on whether the intervention produces its intended effect under controlled conditions.
Detailed economic evaluation may be premature if the intervention is still undergoing substantial development or if its eventual implementation model remains uncertain.
Nevertheless, basic resource documentation may still be useful for future costing.
When sufficient effectiveness evidence has accumulated, a standalone evaluation may be appropriate for comparing adoption strategies or estimating economic consequences under routine conditions.
The choice should reflect the research uncertainty rather than a rigid requirement to complete a fixed number of trials first.
Can an Embedded Economic Evaluation Use Modeling?
Yes. Embedded economic evaluations are not necessarily limited to descriptive analysis of directly observed trial costs and outcomes.
Researchers may combine trial data with decision-analytic modeling when important consequences extend beyond follow-up.
For example, a trial may provide short-term effectiveness and resource-use estimates, while a model projects later outcomes using relevant external evidence.
The distinction between embedded and standalone evaluation therefore should not be reduced to observed data versus modeled data.
The central difference concerns whether the economic evaluation is integrated into the original study or organized as a distinct investigation.
How Does the Time Horizon Influence the Choice?
The economic time horizon should capture the important differences in costs and consequences between alternatives.
If these differences occur within trial follow-up, an embedded evaluation may provide sufficient evidence.
If important consequences arise later, researchers may require extended follow-up, record linkage, modeling, or a separate evaluation.
For example, a short-term educational trial may observe the cost of introducing a digital platform but not its maintenance requirements across subsequent academic years.
Researchers should distinguish the economic consequences missed by short-term trials from those directly observed.
A longer horizon alone does not require a standalone study, but it may strengthen the case for one when substantial additional evidence is necessary.
What Happens When the Trial Population Differs From the Decision Population?
Trial eligibility criteria may restrict participation to a population that differs from those who would receive the intervention in routine practice.
For example, an educational intervention trial may recruit students with reliable internet access, while institutional adoption would include students with more varied technological resources.
These differences may influence both costs and outcomes.
A standalone economic evaluation may incorporate evidence relevant to the broader population, but extrapolation requires appropriate assumptions and data.
Researchers should not assume that trial-based economic estimates automatically generalize to excluded groups.
How Does the Comparator Affect Economic Evaluation Design?
Economic evaluation concerns choices between alternatives. The comparator should therefore reflect the decision the analysis intends to inform.
A trial may compare an intervention with usual care or existing practice at one institution. Decision-makers elsewhere may face different alternatives.
For example, a university may need to choose among several digital support platforms rather than between one platform and no intervention.
A standalone evaluation may combine suitable evidence to compare alternatives not evaluated directly within one trial.
However, indirect comparisons require careful consideration of differences in populations, settings, outcomes, and potential confounding.
Can a Standalone Evaluation Combine Evidence From Multiple Studies?
Yes. A standalone economic evaluation may use systematic reviews, meta-analyses, administrative records, and other relevant sources.
Combining evidence may improve coverage of intervention effects, resource use, and longer-term consequences.
However, different studies may use incompatible outcome definitions, resource categories, or implementation conditions.
Researchers should assess heterogeneity and explain how evidence is selected, synthesized, and incorporated into the economic model.
Using more sources does not automatically produce more reliable estimates if their underlying assumptions are incompatible.
What Is the Role of Decision-Analytic Modeling?
Decision-analytic modeling represents possible pathways, events, resource consequences, and outcomes associated with alternative decisions.
Models may be useful when relevant outcomes extend beyond direct observation or when evidence from several sources must be combined.
Common structures include decision trees and state-transition models, although the appropriate approach depends on the decision problem.
Models require assumptions about relationships between variables, event probabilities, intervention effects, and future resource use.
Researchers should justify the model structure, validate it where feasible, and examine uncertainty.
Modeling is not a substitute for credible effectiveness evidence.
Does a Standalone Evaluation Require Randomization?
No. Economic evaluation is not defined by randomization.
A standalone evaluation may use results from randomized trials, observational studies, or other suitable evidence.
However, causal estimates derived from nonrandomized comparisons require appropriate attention to confounding, selection, and other sources of bias.
Researchers should distinguish uncertainty in resource valuation from uncertainty in the estimated causal effects of the intervention.
An elaborate economic model cannot compensate for systematically biased effectiveness inputs.
Can Trial Conditions Distort Implementation Cost Estimates?
Potentially. Trials may provide additional training, supervision, equipment, or personnel not available during routine implementation.
Some of these resources are necessary to deliver the intervention, while others exist solely to support research procedures.
Researchers should distinguish routine delivery resources from research-specific expenditure.
For example, trial outcome assessors may not be required during ordinary program delivery, whereas continuing technical support may remain essential.
A standalone evaluation may estimate costs under a different implementation model, but those estimates should be supported by evidence rather than arbitrary assumptions.
How Do Implementation Costs Differ From Intervention Delivery Costs?
Implementation costs concern activities required to introduce, adopt, and integrate an intervention into practice.
They may include training, facilitation, workflow redesign, and initial infrastructure development.
Delivery costs concern resources consumed in providing the intervention to recipients.
For example, faculty training may be an implementation cost, while routine formative feedback is a delivery activity.
Researchers should distinguish these categories when estimating initial adoption and ongoing operation.
The selection of resource-use data for intervention studies should reflect this distinction.
Can an Embedded Economic Evaluation Increase Participant Burden?
Yes. Additional resource-use questionnaires, cost diaries, productivity measures, and follow-up assessments may increase participant demands.
Some information may be obtained through administrative or financial records, reducing the need for direct reporting.
However, existing records may not capture participant expenses or unpaid time.
Researchers should consider how economic data collection affects participant burden and study quality.
Where the economic component would substantially disrupt the primary trial, a separate study or more focused measurement strategy may be preferable.
Can a Standalone Evaluation Be Conducted After the Trial?
Yes. Researchers may conduct a separate economic evaluation after trial results become available.
However, retrospective analysis may be limited when important resource quantities were not recorded and cannot be reconstructed reliably.
For example, instructor preparation time or participant connectivity expenses may be absent from institutional financial records.
A later evaluation may use external evidence or additional data collection, but the resulting assumptions and limitations should be transparent.
Researchers should distinguish the organizational decision to conduct the evaluation separately from the methodological decision to collect certain data prospectively.
Can a Trial Collect Economic Data Without Conducting a Full Embedded Evaluation?
Yes. A trial may collect selected resource-use information to support future economic analysis without attempting a complete comparative economic evaluation within the original study.
For example, researchers may document staff time, equipment use, training, and participant expenses while postponing longer-term economic modeling.
This can preserve useful information that would otherwise be difficult to recover.
However, collecting economic data does not guarantee that a future cost-effectiveness analysis will be feasible.
The eventual analysis still requires suitable comparative outcomes, an appropriate perspective, and defensible methods.
How Do Budget Impact and Cost-Effectiveness Questions Differ?
Cost-effectiveness analysis compares costs and consequences of alternatives, while budget impact analysis examines the financial consequences of adopting an intervention for a particular budget holder.
For example, a university may need to know whether an intervention offers good economic value and whether it can afford the initial expenditure during its annual budget cycle.
These are related but distinct questions.
A standalone budget impact analysis may be appropriate even when a trial-based cost-effectiveness analysis has already been completed.
Researchers should specify the decision being informed rather than assume that one economic analysis answers every financial question.
How Should Statistical Uncertainty Be Addressed in Embedded Evaluations?
Trial-based economic evaluations should account for uncertainty in costs, outcomes, and their relationship.
Cost data may be skewed, and missing observations may differ systematically across participants.
Appropriate statistical methods should reflect the trial's design, clustering, allocation, and analytical estimand.
For example, an economic evaluation alongside a cluster-randomized trial should not ignore correlation among participants within the same institution.
Researchers should avoid interpreting a nonsignificant difference in costs as proof that the alternatives have equal economic consequences.
How Should Uncertainty Be Addressed in Standalone Evaluations?
Standalone evaluations may involve parameter uncertainty, structural uncertainty, and uncertainty about transferring evidence to the decision population.
Deterministic sensitivity analyses can examine alternative assumptions, while probabilistic sensitivity analysis may characterize uncertainty in model parameters.
Scenario analyses may examine different implementation arrangements, time horizons, or population characteristics.
Researchers should distinguish uncertainty supported by empirical distributions from assumptions introduced because evidence is unavailable.
A precise model output should not be interpreted as a precise prediction when the underlying evidence is weak.
What Are the Main Advantages and Limitations of Each Approach?
Embedded Economic Evaluation
Potential advantages: Coordinated cost and outcome measurement, direct alignment with trial comparisons, and prospective collection of difficult-to-reconstruct resources.
Potential limitations: Trial-specific populations, short follow-up, additional participant burden, and economic questions constrained by the original comparison.
Standalone Economic Evaluation
Potential advantages: Flexibility to address broader populations, alternative comparators, longer horizons, and evidence from multiple sources.
Potential limitations: Dependence on external evidence, assumptions about transferability, retrospective data limitations, and potential model uncertainty.
The choice should depend on which limitations matter most for the intended decision rather than an assumption that one approach is methodologically superior in all circumstances.
When Is a Combined Approach Most Appropriate?
A combined approach may be useful when a trial can provide credible short-term cost and effectiveness data but the economic decision requires a longer horizon or broader implementation context.
Researchers may collect relevant resource-use information prospectively, conduct a trial-based analysis, and subsequently use those findings within a model-based evaluation.
For example, a university may estimate one-semester intervention costs directly and model future delivery across additional cohorts using justified assumptions about enrollment and maintenance.
However, the model should not assume that observed short-term benefits persist indefinitely.
Researchers should explain how trial evidence and external information contribute to the final economic estimates.
How Should Researchers Decide Which Approach Is Feasible?
Feasibility depends on the research objectives, available data, analytical expertise, participant burden, institutional access, and funding.
An embedded evaluation may be practical when the trial already collects most relevant outcomes and economic measurements can be added without substantial disruption.
A standalone evaluation may be more feasible when suitable datasets already exist or when the economic question requires substantial modeling expertise.
However, separating the study does not automatically reduce total research costs or eliminate data limitations.
Researchers should evaluate the full information requirements of each approach before selecting one.
How Should the Chosen Economic Evaluation Design Be Reported?
Reports should identify whether the economic evaluation was embedded, standalone, or combined, and explain the methodological rationale.
Researchers should specify the perspective, intervention, comparator, population, time horizon, outcomes, resource-use measures, valuation methods, and uncertainty analyses.
For embedded evaluations, the relationship between the economic analysis and the trial protocol should be clear.
For standalone evaluations, evidence sources, synthesis methods, modeling assumptions, and validation procedures should be described.
CHEERS 2022 provides relevant reporting guidance for health economic evaluations, although compliance with reporting guidance does not independently establish methodological validity.
Watch Out
Do not assume that embedding economic evaluation automatically produces more credible evidence, or that a standalone model can compensate for missing or biased effectiveness data. The economic design must match the decision problem, and conclusions should reflect the strengths and limitations of the evidence used.