Defining outcomes
Define an outcome so another researcher can apply the same rule to the same observations and obtain the same value.
Fit
When it matters, and when it does not
It matters when
- You turn a broad construct, such as recovery, into an endpoint you can measure.
- You compare studies whose outcomes share a name but may use different rules.
It matters less when
- You have a complete definition and need the treatment contrast. Specify the estimand, including events that change interpretation. Read: The estimand
Before you start
What to check in your study
- The clinical idea and why it mattersAsk patients and clinicians which aspects matter. A core outcome set gives a minimum, not every study-specific detail.
- A rule for measurementName the instrument or event criteria, assessor, units, aggregation and assessment schedule.
- The time origin and horizoncompetingEventssummaryMeasureDefine follow-up, death, treatment changes and missing observations before choosing a summary.
- The measure represents the intended construct in this setting.How to check: Examine validity and reliability evidence. A convenient code or surrogate is not automatically a patient-important outcome.
- Groups are assessed using comparable rules.How to check: Check schedules, blinding where feasible and access to records. Different surveillance can mimic an outcome difference.
Plain language
How it works
Start with a construct: the clinical idea you want to study. Turn it into an operational definition: the rule for observing it. The variable stores the resulting value. A testable hypothesis states a contrast in that variable for a defined population. Recovery might include symptoms, function and readmission; one measure does not exhaust it.
COS-STAD supports defining scope and involving relevant stakeholders when developing core outcome sets. CONSORT-Outcomes asks trial reports to define outcomes and explain how and when they were assessed. Chan and colleagues found discrepancies between protocols and published outcomes. Preserve the planned definitions and disclose changes. For observational contrasts, specify confounders as well as measurement rules.
Worked example
A worked example
- Setting
- An invented discharge study uses unplanned readmission as one aspect of recovery in 120 adults.
- Question
- What proportion has an unplanned readmission within 30 days?
- Approach
- Follow each participant for 30 days after discharge. Count the first unplanned inpatient readmission, excluding planned care. All outcomes are observed and nobody dies before follow-up ends.
- Result
- 18 of 120 participants are readmitted, 15 in 100.
- What it means
- That is a fixed-horizon binary outcome. Time to first readmission asks a different question; death before readmission needs a prespecified strategy.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
An outcome name is not an operational definition. Specify what counts, how and when it is measured, and how death and missing observations affect the question before seeing results.
- Calling recovery an outcome without a measurement rule.Instead: Move from construct to operational definition to recorded variable. State a testable hypothesis about the chosen contrast.
- Choosing the primary outcome after looking at results.Instead: Prespecify primary and secondary outcomes and their analysis roles. Disclose changes, timing and reasons.
- Coding an unobserved outcome as no event.Instead: Name the missingness assumption and analysis plan. For plausible missing not at random, add a delta or tipping-point sensitivity analysis.
- Interpreting an adjusted association as the causal effect of treatment.Instead: State the causal assumptions and confounder plan. A precise outcome definition alone does not control confounding.
Inside Mallard
How Mallard uses this
Mallard checks whether secondary endpoints have declared analysis roles when a prespecification record is present.
concept.defining_outcomes, the id a plan cue records.Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Core Outcome Set-STAndards for Development: The COS-STAD recommendations (opens in a new tab) · PLoS Med · doi 10.1371/journal.pmed.1002447
- Guidelines for Reporting Outcomes in Trial Reports: The CONSORT-Outcomes 2022 Extension (opens in a new tab) · JAMA · doi 10.1001/jama.2022.21022
- Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles (opens in a new tab) · JAMA · doi 10.1001/jama.291.20.2457