03 · What You Need to Know
Selecting Outcomes That Make Pragmatic Trial Findings Useful
What Is a Clinically Relevant Outcome?
A clinically relevant outcome provides information about a consequence of treatment that matters for clinical decisions or the people receiving care.
Examples may include mortality, symptoms, functional ability, health-related quality of life, treatment complications, and important healthcare events.
However, clinical relevance is not determined solely by whether an outcome is commonly recorded in medical practice.
A laboratory measurement may be clinically useful without directly representing how a patient feels, functions, or survives.
Likewise, a patient-reported measure may be highly relevant even when it is not routinely recorded in administrative databases.
Researchers should distinguish outcomes that directly matter to patients from indicators that are useful because they predict, explain, or accompany those outcomes.
Why Are Meaningful Outcomes Particularly Important in Pragmatic Trials?
Pragmatic trials are intended to inform decisions under conditions relevant to ordinary practice.
Decision-makers need evidence about whether adopting an intervention produces benefits that justify its costs, burdens, risks, and implementation requirements.
For example, an intervention that improves patients' knowledge about medication may be promising. Yet a healthcare system considering adoption may need to know whether patients experience better symptom control, fewer complications, or improved functioning.
Knowledge improvement may help explain how an intervention operates, but it does not necessarily establish that the intervention achieves its intended practical benefit.
The primary outcome should therefore correspond to the consequence the trial is intended to evaluate.
What Does PRECIS-2 Say About Primary Outcomes?
The PRECIS-2 framework includes the primary outcome as one of nine domains used to assess the explanatory or pragmatic orientation of trial design.
A more pragmatic outcome is generally one that is directly relevant to participants and the decisions made in usual practice.
A more explanatory outcome may be selected because it is sensitive to a proposed mechanism or useful for clarifying an intervention's biological or behavioral effects.
However, these orientations are not mutually exclusive.
An outcome can be scientifically informative and practically meaningful at the same time.
PRECIS-2 should be used to assess alignment with the intended trial purpose, not as a rule requiring every study to select the most distal or difficult-to-measure outcome.
What Is the Difference Between Patient-Important and Surrogate Outcomes?
Patient-important outcome
An outcome directly relevant to how patients feel, function, survive, or experience treatment-related consequences.
Surrogate outcome
A measurement used in place of a patient-important outcome because it is expected to predict or represent treatment effects on that outcome.
For example, a physiological biomarker may be used as a surrogate for a later clinical event.
However, an intervention's effect on a biomarker does not necessarily establish its effect on the clinical outcome.
Surrogate validity depends on the intervention, disease context, and evidence linking treatment-induced changes in the surrogate to changes in the outcome of interest.
Researchers should not assume that a strong individual-level association between a surrogate and a clinical outcome proves that treatment effects on one reliably predict treatment effects on the other.
Are Intermediate Outcomes Always Inappropriate?
No. Intermediate outcomes can provide valuable evidence about intervention mechanisms and early effects.
For example, a digital health intervention may improve patients' understanding of treatment instructions before changes in longer-term health outcomes become detectable.
Such findings may support intervention development and theory testing.
However, intermediate improvement should not be described as definitive evidence of a consequential benefit that was not measured.
A pragmatic trial may include intermediate outcomes as secondary measures while prioritizing a primary outcome relevant to the intended adoption decision.
The appropriate balance depends on the research question, existing evidence, and feasible follow-up period.
How Do Meaningful Outcomes Differ Across Research Fields?
Clinical relevance is a field-specific expression of a broader principle: research outcomes should correspond to consequences that matter for the intended decision.
In education, meaningful outcomes may include independent learning, demonstrated competence, retention, course progression, or the ability to transfer skills into authentic tasks.
In organizational research, they may involve sustained work performance, employee wellbeing, or service quality.
In clinical research, they may involve patient functioning, symptoms, adverse events, or survival.
Researchers should not assume that all such outcomes are equally important or that every study requires multiple outcomes.
The priority depends on the intervention's purpose and the people affected by the decision.
What Makes an Outcome Convenient but Potentially Misleading?
A convenient outcome is one that can be collected relatively easily, cheaply, or consistently.
Convenience is not inherently problematic. Routinely collected data may support large and efficient pragmatic trials.
Problems arise when the available measure does not adequately represent the intended benefit.
| Convenient measure |
Potential limitation |
More decision-relevant question |
| Platform login frequency |
Does not establish meaningful participation or benefit. |
Did participants achieve the intended improvement? |
| Knowledge quiz score |
May not represent behavior or performance outside the assessment. |
Can participants apply the knowledge appropriately? |
| Laboratory biomarker |
May not reliably represent treatment effects on patient-important outcomes. |
Did the intervention improve health or functioning? |
| Number of consultations |
May reflect access, need, or service use rather than improvement. |
Did care produce better patient outcomes? |
| Training completion |
Does not necessarily establish competence. |
Can participants perform the intended task? |
The table does not imply that process indicators lack value. Their usefulness depends on whether they are interpreted according to what they actually measure.
Should the Primary Outcome Always Be Patient-Reported?
No. Patient-reported outcomes can capture symptoms, functioning, and experiences that may not be adequately represented by clinical records.
However, some important outcomes are appropriately measured through clinical assessments, validated instruments, or administrative records.
For example, mortality does not require a patient-reported scale, while pain severity often depends on patient reporting.
The measurement approach should correspond to the construct of interest.
Researchers should consider validity, reliability, responsiveness, accessibility, and the potential influence of treatment knowledge on reporting.
Can Routinely Collected Data Be Used for Pragmatic Outcomes?
Yes. Electronic health records, registries, administrative databases, and institutional information systems may provide efficient outcome data.
Such sources can reduce participant burden and support follow-up within ordinary practice.
However, data availability does not guarantee measurement validity.
Researchers should examine whether the relevant events are consistently recorded, whether coding practices differ across sites, and whether treatment assignment influences the likelihood of an event being documented.
Missing information should not automatically be interpreted as absence of an event.
Routinely collected data are most useful when their quality is adequate for the outcome and estimand being evaluated.
How Does Outcome Measurement Affect Risk of Bias?
Outcome measurement can introduce bias when assessment procedures differ between treatment groups or when knowledge of assignment influences judgments.
For example, instructors who know which students received an educational intervention may rate their work differently if assessment criteria are subjective.
Independent blinded assessment may reduce this concern.
In clinical studies, unblinded participants may report subjective symptoms differently because of expectations or experiences associated with treatment.
However, the absence of blinding does not automatically establish bias.
Researchers should evaluate the plausibility and consequences of assignment-related influences on the specific outcome.
Should Outcomes Be Identical Across Study Sites?
For a pooled primary analysis, researchers generally need sufficiently consistent outcome definitions and measurement procedures to support meaningful comparison.
However, identical data collection systems are not always necessary.
For example, different institutions may use distinct electronic records while applying a common operational definition of an event.
Researchers should establish whether the available measures are sufficiently comparable.
Differences in coding, assessment frequency, or data completeness may create measurement heterogeneity that requires attention.
Standardizing outcome definitions can be important even when intervention delivery remains flexible.
How Should Stakeholders Contribute to Outcome Selection?
Patients, practitioners, policymakers, and other relevant stakeholders may identify outcomes that researchers would otherwise overlook.
For example, clinicians may prioritize a physiological measure, while patients place greater importance on fatigue, daily functioning, or treatment burden.
Neither perspective should automatically displace the other.
Stakeholder involvement can clarify which outcomes are consequential for the intended decision.
However, researchers must also evaluate whether proposed outcomes can be defined and measured credibly.
Stakeholder importance and measurement quality are complementary considerations.
What Are Core Outcome Sets?
Core outcome sets identify a minimum collection of outcomes recommended for measurement and reporting in studies addressing a particular health condition or research area.
The Core Outcome Measures in Effectiveness Trials initiative, known as COMET, supports the development and use of such sets.
Core outcome sets can improve comparability across studies and reduce selective emphasis on outcomes that happen to produce favorable results.
However, a core outcome set does not necessarily determine the single primary outcome for every trial.
Researchers may need additional outcomes relevant to their intervention and decision.
They should also verify that a proposed set is appropriate for the population and research context.
How Should Researchers Balance Relevance and Feasibility?
A highly meaningful outcome may be difficult to measure because it requires long follow-up, specialized assessment, or access to incomplete records.
Researchers should not automatically replace it with an easier measure.
Instead, they should consider whether the study can be redesigned to collect the meaningful outcome credibly.
Possible approaches include validated short instruments, routine records with established measurement properties, or a feasible follow-up schedule.
However, compromises may be necessary when resources or participant burden make certain assessments impractical.
Such compromises should be justified transparently rather than presented as if the convenient measure were equivalent to the desired outcome.
Can a Convenient Outcome Be the Best Primary Outcome?
Yes, when it is valid, meaningful, and appropriate to the decision.
For example, an accurately recorded hospitalization outcome may be both routinely available and directly relevant to patients and healthcare systems.
Similarly, an independently assessed course completion outcome may be practical to collect and consequential for educational policy.
The problem is not convenience itself.
The problem is selecting an outcome primarily because it is available while overlooking whether it represents the intended benefit.
Should Pragmatic Trials Use Composite Outcomes?
Composite outcomes combine multiple events or measures into one endpoint.
They may be useful when the components collectively represent a meaningful clinical or practical consequence.
However, components may differ in importance, frequency, or responsiveness to treatment.
For example, a composite combining a minor service event with a serious clinical outcome may be difficult to interpret if the overall effect is driven mainly by the less important component.
Researchers should justify the composite, define its components, and report component-level findings where appropriate.
Combining outcomes does not automatically improve their relevance.
How Does Follow-Up Timing Affect Outcome Relevance?
Outcome timing should correspond to when the intervention is expected to produce meaningful effects.
For example, a learning intervention may improve immediate knowledge but require additional time before changes in independent performance become apparent.
Similarly, a clinical intervention may influence short-term symptoms before affecting longer-term functioning or complications.
Researchers should distinguish outcomes that measure early mechanisms from those that evaluate sustained benefits.
However, longer follow-up may increase missing data, costs, and exposure to other influences.
The appropriate schedule depends on the intervention's causal theory and intended decision.
How Does Outcome Selection Affect Sample Size?
Different outcomes have different distributions, event rates, measurement errors, and expected treatment effects.
A rare but important clinical event may require a larger sample or longer follow-up than a frequently observed intermediate measure.
However, researchers should not automatically select a less meaningful outcome simply because it produces a smaller sample size requirement.
They should consider whether the available resources can support a credible study of the decision-relevant outcome.
Sample size planning should use defensible assumptions about the chosen outcome and the smallest effect considered important.
How Should Researchers Interpret Discordant Outcomes?
An intervention may improve an intermediate measure without improving a patient-important outcome.
For example, participants may demonstrate better knowledge of disease management without experiencing measurable improvement in functioning.
Such findings do not necessarily mean the intervention has no value, but they limit claims about the outcomes that did not improve.
Conversely, a meaningful outcome may improve without the proposed intermediate measure changing as expected.
This pattern may challenge the assumed mechanism or reflect limitations in its measurement.
Researchers should interpret outcomes according to their prespecified roles and avoid selectively elevating favorable secondary findings after an unfavorable primary result.
How Should Outcomes Be Reported?
Reports should identify the prespecified primary outcome, its operational definition, measurement instrument or data source, and assessment timing.
Researchers should explain why the outcome matters to the intended decision and describe relevant limitations in validity or completeness.
Effect estimates should be accompanied by appropriate uncertainty measures and interpreted in relation to practical importance.
Secondary and surrogate outcomes should be clearly distinguished from direct evidence of patient-important or otherwise consequential benefits.
Transparent reporting helps readers determine whether the trial demonstrates an effect that matters, rather than merely an effect that was convenient to measure.
