Why Operational Definitions Matter
A protocol may name treatment discontinuation, rescue medication, or treatment switching as an intercurrent event, but a label alone does not tell a site when to record that event. For example, “discontinuation” could mean a participant’s stated intention to stop, a clinician’s decision, a missed dose, or the end of all protocol treatment. Different interpretations can produce different event dates and different event classifications for otherwise similar participants.
An operational definition resolves this ambiguity in advance. It tells investigators and data managers what observable evidence qualifies, when the event is considered to have occurred, and what to record when the evidence is incomplete or conflicting. The purpose here is to make event ascertainment reproducible. How an event is handled in defining or estimating a treatment effect is a separate question.
An operational definition should not depend on an investigator’s unstated impression, such as “the participant was no longer really taking treatment.” Nor should it make the event synonymous with a data-collection consequence. A participant can discontinue assigned treatment while continuing scheduled assessments; conversely, an assessment can be missed while treatment continues.
A Practical Specification
For each event type, specify the qualifying evidence, the event unit (for example, the first occurrence or each episode), the date rule, and the recording procedure. If multiple sources can establish the event, state how discrepancies are handled. If a date cannot be determined exactly, record the uncertainty rather than silently substituting a convenient date.
State which treatment, medication, or therapy is relevant, and whether the definition concerns a first event, any occurrence, or episodes that can recur.
Specify the records or actions that qualify, such as a documented decision, a dispensing record, or a dose administration. Avoid relying on an undefined judgment such as “substantial use.”
Define which date represents occurrence: for example, the effective date of a documented stop, or the date the first qualifying dose was taken. State how partial dates and uncertain sequences are handled.
List required fields, source documents, responsibility for entry, and a process for queries or conflicting evidence. Preserve the distinction between “no event” and “event status unknown.”
These choices are protocol-specific. There is no universal number of missed doses that constitutes discontinuation, no universal medication list that defines rescue therapy, and no single date convention suitable for every trial. The protocol must select criteria that fit the intervention and clinical context, then apply them consistently.
Operationalizing Three Common Events
| Event | Criteria to make explicit | Timing and recording questions |
|---|---|---|
| Treatment discontinuation | Which assigned treatment; whether permanent cessation is required; whether a documented decision alone qualifies | Is the date the decision date or the last dose date? Are temporary holds recorded separately? |
| Rescue medication | Which products, indications, routes, and uses qualify; whether use must be documented as rescue | Is the event date the first dose date? Are repeated administrations separate episodes or one course? |
| Treatment switching | Which treatment counts as a switch; whether the new treatment must replace or merely supplement assigned treatment | Is timing based on the decision, prescription, or first administration? How is overlap recorded? |
For all three event types, the protocol should define an authoritative record or a review process across records. A prescription does not establish that a medication was taken; a participant report of use may establish use if the protocol permits that source. Likewise, a clinician’s recommendation to stop treatment is not necessarily the date treatment actually stopped. These distinctions should be deliberate rather than resolved differently at each site.
Worked Example: Defining Treatment Discontinuation
Hypothetical protocol scenario. A trial administers assigned study tablets for 12 weeks. The protocol team wants to record permanent discontinuation separately from temporary interruptions. The following definition is one possible operational choice, not a universal rule.
Criteria. Permanent discontinuation is recorded when the participant or investigator documents that no further assigned study tablets will be taken. A temporary hold with a planned resumption is recorded as an interruption, not permanent discontinuation. A missed dose alone does not qualify.
Timing rule. The event date is the documented effective date of permanent cessation. If that date is unavailable but the last dose date is known, the last dose date is entered as the best available date and flagged as such. A decision date is recorded separately when it differs from the effective date.
Application. Suppose the participant takes the last dose on 12 May, is told on 13 May to hold treatment, and resumes on 16 May. The planned resumption means this is a temporary interruption, so it does not meet the permanent-discontinuation criterion. Later, on 2 June, the participant and investigator document that treatment will not resume, effective 3 June. The permanent discontinuation date is 3 June; the decision date is 2 June. Recording both dates avoids treating the decision as if it were the cessation itself.
The definition also makes the negative case explicit: a missed dose, without evidence of permanent cessation, is insufficient. That rule reduces the chance that sites classify brief nonadherence as treatment discontinuation.
Worked Example: Defining Rescue Medication
Hypothetical protocol scenario. Participants may use a specified short-acting inhaled medicine for acute symptoms. The protocol defines rescue medication as use of a product on the prespecified list, by the specified route, for acute symptom relief. Routine use of a different medicine is not classified as rescue under this definition, even if it is clinically relevant.
Criteria and timing. A qualifying event requires evidence that at least one dose was taken; merely prescribing or dispensing the medicine is not enough. The event date is the date of the first qualifying dose in an episode. A new episode begins after a prespecified 48-hour period with no qualifying doses. The 48-hour rule is a study convention chosen here to make episode counting reproducible.
Application. A participant reports taking doses on 7 and 8 September, then none on 9 or 10 September, and takes another dose on 11 September. The gap from the last dose on 8 September to the next on 11 September is three calendar days, exceeding the 48-hour episode-separation rule. The records therefore show two episodes, beginning 7 September and 11 September. If the participant had taken the next dose on 10 September within 48 hours of the previous dose, it would remain part of the first episode.
The protocol should identify acceptable evidence, such as a participant diary or interview, and provide a field for source and date certainty. If the product was dispensed but use cannot be confirmed, the event status is not “yes” solely on the basis of dispensing.
Worked Example: Defining Treatment Switching
Hypothetical protocol scenario. In a trial of assigned therapy A, the protocol defines switching as the first administration of a non-study therapy B in the same therapeutic class that is intended to replace A. Adding a medicine without replacing A is recorded as concomitant treatment, not a switch. This distinction is an explicit choice for this hypothetical trial.
Timing rule. The switch date is the date B is first administered, not the date it is discussed, prescribed, or dispensed. The reason for the change and the status of A are recorded separately.
Application. On 4 November, a clinician discusses therapy B. B is prescribed on 6 November and dispensed on 7 November. The participant first takes B on 9 November, while the last dose of A was taken on 8 November. Under the stated rule, the switching event date is 9 November. The discussion, prescription, dispensing, and last A dose are useful supporting details, but none replaces the defined event date.
If the participant had taken B while continuing A, the event would not meet this protocol’s replacement criterion at that point. The records should instead show concurrent use; if B later replaces A, the switch date is the first administration that satisfies the definition.
Justification, Diagnostics, and Failure Modes
A definition is useful only if different trained users can apply it to the same source records and reach the same classification. The core justification is therefore procedural: observable criteria, a specified date rule, and explicit recording fields reduce avoidable disagreement. They also make later review possible, because a reviewer can trace a classification back to the evidence rather than infer what a label meant.
Before finalizing the protocol, test the definition against short scenarios: a treatment hold followed by resumption, an uncertain last-dose date, a medication prescribed but never taken, and a new therapy started while assigned treatment continues. Ask two reviewers to classify the same scenarios independently. Disagreement often reveals a missing rule—for example, whether “permanent” depends on stated intent, observed duration, or a documented clinical decision.
- Vague threshold. “Discontinued after prolonged nonuse” is not reproducible unless “prolonged” and the evidence for nonuse are defined.
- Decision confused with occurrence. A planned stop, prescription, or recommendation may precede the action that the event definition uses as its timing criterion.
- Event confused with outcome missingness. Failure to complete an assessment does not by itself prove discontinuation, rescue use, or switching.
- Unknown treated as absent. If the source records do not establish whether an event occurred, record the status as unknown or unresolved according to the data system, not “no.”
- Inconsistent episode counting. Repeated medication use needs a stated rule for whether doses form one episode or multiple episodes.
Operational detail also has limits. A highly elaborate definition can burden sites, create fields that cannot be completed reliably, or imply precision that source records do not support. The protocol should require information that is clinically meaningful and realistically ascertainable. It should also distinguish exact dates from estimated dates and document how unresolved discrepancies are adjudicated.
Operational definitions are not interchangeable with the later choice of how an event enters an estimand or analysis. Nor does a clear definition eliminate missing or misreported event data. It makes the classification rule transparent; data quality still depends on collection, follow-up, and review procedures.
Check Your Understanding
Answer each question, then compare with the solutions below.
- A participant misses two doses but has no documented plan to stop and resumes treatment. Under a definition requiring documented permanent cessation, is discontinuation established?
- A rescue medicine is prescribed and dispensed, but the participant reports never taking it. Does this meet a definition requiring a dose to be taken?
- A protocol defines switching by first administration of replacement therapy. A new therapy is discussed Monday, prescribed Tuesday, and first taken Friday. Which date is the event date?
- Why should a protocol distinguish an event that did not occur from an event whose status cannot be determined?
- Name two rules needed to make repeated rescue-medication episodes consistently countable.
Solutions. (1) No. Missed doses alone do not satisfy the stated criterion, and subsequent resumption is consistent with an interruption rather than documented permanent cessation. (2) No; prescription and dispensing do not establish use when the definition requires an administered or taken dose. (3) Friday, because the stated timing criterion is first administration. (4) “No event” is a substantive classification, while “unknown” indicates insufficient evidence; conflating them can misrepresent event occurrence. (5) Specify the qualifying products and uses, and define when a dose begins an episode and what dose-free interval separates episodes.