Missed, Held, and Partial Doses

Why Administration Status Must Be Explicit

A prescribed dose is not necessarily a dose a person received. A dose may be missed, withheld by a clinician, taken late, or only partly administered. If the data record shows only the planned amount, these different events can all appear to be completed doses. If it shows only that no dose record exists, an analyst may not know whether the dose was missed, intentionally held, or simply not documented.

The goal is to describe the observed administration faithfully. Keep the scheduled plan, the actual administration, and the reason for any departure distinct. This is important even before a model is selected: a downstream analysis cannot recover the actual dose and timing from a record that silently substitutes the prescription for what happened.

Definition: A completed dose is an administration for which the recorded actual amount matches the intended amount, within the available source information. A missed dose was scheduled but not administered. A withheld dose was not administered because it was intentionally held, commonly under a clinical decision or protocol rule. A delayed dose was administered later than its scheduled time. A partial dose was administered in an amount less than the intended amount. The event status describes what happened; the amount and time fields describe the observed administration.

A withheld dose is a reasoned kind of non-administration, while “missed” may describe an unintentional omission. Source terminology varies, so preserve the source wording or map it through a documented, controlled code list. Do not infer intention from the absence of an administration record. A dose with unknown status is not automatically a missed dose.

Represent the Plan and the Observation Separately

For each scheduled opportunity, useful fields include the scheduled time and planned amount, the administration status, the actual administration time and amount when applicable, and a reason when one is available. The plan supplies context; it does not replace the observed event. Use the established time and amount units consistently, and retain enough source information to trace a coded status back to its record.

A practical structure has one record per scheduled dose opportunity or documented administration event, with fields such as:

Scheduled time and planned amount: what was intended according to the active order or schedule.
Status: completed, missed, withheld, delayed, partial, or another defined source status.
Actual time and amount: the observed time and amount administered, if an administration occurred and these are known.
Reason and provenance: the documented reason, if any, and the source from which the information came.

The amount field needs careful semantics. A numeric zero can mean that the known amount administered was zero, as for a confirmed non-administration. A missing amount means the amount is unknown or unavailable. These are not interchangeable. If a status is “missed” but the data schema uses a numeric amount, an explicit zero may represent the known amount administered; the status must still identify the event as missed. In a model-ready file, whether and how to encode that status as a dose event depends on the supported event structure. Do not create a completed dose merely to fill a row.

Worked Example: A Confirmed Missed Dose

A hypothetical oral regimen calls for 100 mg at 08:00 and 20:00. A medication log confirms that the 08:00 dose was not taken. The 20:00 dose was taken as scheduled. Represent the two opportunities without treating the missed dose as a completed dose.

For the 08:00 opportunity, record the scheduled time as 08:00, the planned amount as 100 mg, and the status as “missed.” The observed amount administered is 0 mg if the source confirms that none was taken. The actual administration time is not applicable because no administration occurred; do not insert 08:00 as an actual dose time.

For the 20:00 opportunity, record status “completed,” actual time 20:00, and actual amount 100 mg. The planned amount and actual amount agree for that event.

The distinction can be checked arithmetically: for the missed opportunity, actual amount administered is 0 mg, not 100 mg. The 100 mg remains useful as the planned amount, but it is not evidence of an administered dose. If the log merely omitted an 08:00 entry and did not confirm non-administration, the status and actual amount would instead be unknown until clarified.

Partial, Delayed, and Withheld Administrations

For a partial administration, record the amount actually received, not the planned amount. If both amounts are known and comparable, the administered fraction is a useful quality check:

\text{administered fraction} = \frac{\text{actual amount administered}}{\text{planned amount}}

This fraction summarizes the difference; it does not replace the actual amount or the status. It is interpretable only when both amounts refer to compatible dose units and the planned amount is positive. If the actual amount is unknown, do not calculate a fraction by treating it as zero.

Worked Example: Recording a Partial Dose

In a hypothetical study, a participant is instructed to take 80 mg. The source documents that 30 mg was taken and the remainder was discarded. Record the status and amount, and calculate the administered fraction.

The status is “partial,” the planned amount is 80 mg, and the actual amount administered is 30 mg. The amount not administered is \(80-30=50\) mg. The administered fraction is:

\frac{30\ \text{mg}}{80\ \text{mg}} = 0.375 = 37.5\%

Thus 37.5% of the planned amount was administered. The dose record must use 30 mg as the actual amount. Encoding 80 mg would overstate the observed dose by 50 mg; encoding zero would incorrectly imply that no amount was administered. The calculation assumes the source amount is reliable and that both values use the same dose basis and units.

For a delayed dose, the actual administration time is the time the dose was given or taken, not the scheduled time. Preserve both times when known. Do not move the dose to its planned time simply to make the schedule regular, and do not assume that later scheduled doses were shifted or omitted unless the record documents that change.

Worked Example: A Dose Taken Late

A hypothetical dose of 150 mg was scheduled for 06:00 and documented as taken at 09:15. No other dose information is available. Calculate the delay and identify the event time to use as the observed administration time.

The elapsed time from 06:00 to 09:15 is 3 hours and 15 minutes:

3 + \frac{15}{60} = 3.25\ \text{hours}

Record status “delayed,” planned time 06:00, actual time 09:15, and actual amount 150 mg. The observed administration occurred 3.25 hours after the scheduled time. That arithmetic does not establish what happened to the next dose; its time and status require their own evidence.

A withheld dose should likewise remain distinct from a missed dose if the source identifies a clinical or protocol decision to hold it. Record the planned opportunity, status “withheld,” and documented reason when available. Since no administration occurred, the actual amount administered is zero only when non-administration is confirmed. A reason such as “nausea” or “laboratory criterion” should be retained as documented or mapped using a defined reason code; it should not be invented from other fields.

Worked Example: A Dose Held by a Clinician

A hypothetical 40 mg dose scheduled for 10:00 was held following a clinician’s decision because of a documented safety concern. The record confirms that it was not given. Show the relevant encoding.

Record scheduled time 10:00, planned amount 40 mg, status “withheld,” and the documented safety concern in the reason field. Record actual amount administered as 0 mg if the schema requires a numeric amount and confirms no administration. Do not record 40 mg as an administered dose, and do not label the event merely “missed” if the data distinguish intentional holds from unintentional omissions.

If the source says only “hold considered” but does not establish whether the dose was actually withheld, the status is unresolved rather than confirmed withheld. The coding must reflect the evidence, not the most plausible interpretation.

Why These Distinctions Affect Model Input

For a bolus-dose representation, the administered dose amount and its time determine the input event. In a simplified notation, a sequence of administered boluses can be written as:

u(t)=\sum_j A_j\,\delta(t-t_j)

Here \(A_j\) is the amount actually administered at time \(t_j\), and \(\delta\) denotes an impulse at that time. This representation assumes bolus input and is used here only to show why event coding matters. A confirmed missed or withheld dose contributes no administered bolus. A partial dose contributes its observed amount, and a delayed dose contributes at its observed time. If a planned amount is entered as though it were administered, the input is too large; if a delayed dose is assigned its scheduled time, the input timing is wrong.

This justification does not mean every analysis file should contain a zero-amount dose event for each non-administration. Some formats encode non-administration in a separate status or disposition record; some model-ready formats may support an explicit zero amount. The correct choice depends on the target schema and its event semantics. The essential requirement is that the data preserve confirmed non-administration and do not present the planned amount as observed input.

Checks, Alternatives, and Failure Modes

Before using the data, check that status, amount, and time agree. A record marked “missed” with a positive actual amount needs clarification; a “partial” record with an actual amount equal to the planned amount may indicate a miscoded status or a different meaning in the source. A delayed event should have a scheduled and actual time that support the stated delay. Check that amounts use consistent units, and review cases where actual time or amount is missing. A missing value is not evidence of zero, and no record is not proof of non-administration.

The main alternatives have different consequences:

Encoding choice	What it communicates	Main risk
Status plus observed amount and time	What happened, with the plan retained for context	Requires a clear code list and consistent source interpretation
Planned amount entered as the administered amount	What was prescribed, not necessarily what was received	Overstates or mistimes input after a missed, held, partial, or delayed dose
No row for a non-administration	Nothing explicit about that scheduled opportunity	Cannot distinguish confirmed non-administration from absent documentation
Zero amount without a status	A numeric zero, if its meaning is defined	May hide whether the event was missed, withheld, or another kind of zero

The first choice is generally the most informative source representation. A downstream file may need a different structure, but transformations should retain the distinction in an auditable way. Do not collapse missed and withheld events merely because both have zero administered amount: their dose input may be the same, but their documented meanings differ.

Common Mistake: Treating the Prescription as the Administration

A common error is to copy the scheduled amount and time into the actual dose fields, then assume that the dose was administered because those fields are populated. This confuses intent with observation. The correction is to enter the planned values as plan fields, enter actual amount and time only when supported by evidence, and code non-administration or partial administration explicitly. If the evidence is insufficient, retain an unknown status or amount rather than filling it from the prescription.

Another pitfall is encoding every missed dose as an unlabelled zero. A zero may be a valid actual amount for a confirmed non-administration, but without a status it loses the distinction between a missed opportunity, a clinician-directed hold, and a source or data-processing problem. Conversely, leaving the dose out entirely can make a confirmed missed dose indistinguishable from incomplete data capture. Use the target schema’s supported event structure while preserving the underlying meaning.

Key takeaway: The prescribed dose describes what was intended; the administration record describes what was observed. Preserve status, actual amount, actual time, and reason separately when available. Never infer an administered dose from the planned regimen alone.

Check Your Understanding

Answer each question, then compare your reasoning with the solutions.

A scheduled 25 mg dose is confirmed not to have been taken. What status and actual amount should be recorded when the schema requires an amount?
A planned 60 mg dose was documented as partially administered: 45 mg was received. What fraction of the planned dose was administered?
A dose scheduled at 07:30 was actually given at 10:00. How late was it, and which time is the observed administration time?
A record has no entry for a scheduled dose, and no source confirms what happened. Should it be coded as missed? Explain.
Why is an unlabelled zero amount less informative than a confirmed non-administration with an explicit status?

Solutions. (1) Record status “missed” and actual amount 0 mg, because non-administration is confirmed; retain 25 mg as the planned amount. (2) The fraction is \(45/60=0.75\), or 75%. Record 45 mg as the actual amount and status “partial.” (3) The delay is 2 hours and 30 minutes, or 2.5 hours. The observed administration time is 10:00; retain 07:30 as the scheduled time. (4) No. The evidence supports an unknown or unresolved status, not a confirmed missed dose. (5) A zero alone does not say whether the dose was missed, intentionally withheld, or affected by an encoding or documentation problem. An explicit status preserves that meaning.