Why Dose-Time Conventions Matter
A dose record needs a time, but a timestamp copied from a source system does not explain what that time means. It may mark when administration began, when it ended, or when it was scheduled to occur. These alternatives are not interchangeable. For an infusion, recording its end time as though it were its start time shifts the represented administration by the infusion duration. For a bolus, start and end are often treated as the same instant, but a nominal scheduled time can still differ from the actual administration time.
The purpose of a dose-time convention is to make that meaning explicit and consistent across records. The convention must preserve the event represented by the source data and match the time interpretation expected by the model input. A time value without that interpretation is not a complete dose specification.
Define the Event Before Converting the Time
For an infusion with actual start time \(t_s\), actual end time \(t_e\), and duration \(D\), all expressed on the same time scale, the relationship is:
If the source records both endpoints, the duration implied by them is \(D=t_e-t_s\). If the source records an end timestamp and an independently documented duration, the start time can be reconstructed as \(t_s=t_e-D\), provided the timestamp and duration refer to the same administration and use compatible units. A negative duration, or a duration that conflicts with the source record, is a signal to investigate rather than a value to silently correct.
A model dataset commonly represents an infusion dose at its actual start, together with the infusion duration and dose amount. Under that convention, an end timestamp must be converted to a start time before it is entered as the dose time. The end timestamp can still be retained as a source field or used for a quality check. If a modeling system instead explicitly defines dose time as the end, follow that definition and document it; do not assume a software default from the column name alone.
For a bolus represented as an instantaneous administration, the actual administration time is the dose time. A source may label this timestamp “start,” “given,” or something else; the data definition should establish whether it records the actual administration or merely a planned time. For nominal-time records, the assigned value is the scheduled time, not evidence that the dose was actually administered at that time.
Worked Example: Converting an Infusion End Time to a Start Time
Suppose a hypothetical source record gives an infusion start of 09:12 and an end of 11:42 on the same date. The study time origin is 08:00 that date, and the target convention is actual infusion start time. Express elapsed times in hours from the origin.
First calculate the duration. From 09:12 to 11:42 is 2 hours 30 minutes, or \(2.5\) hours. The start is 1 hour 12 minutes after the origin, so its elapsed time is \(1+12/60=1.2\) hours. The end is 3 hours 42 minutes after the origin, or \(3+42/60=3.7\) hours.
The consistency check is \(1.2+2.5=3.7\) hours. The dose record therefore uses a dose time of 1.2 hours and a duration of 2.5 hours. If the hypothetical dose amount were 600 mg, the corresponding rate would be \(600/2.5=240\) mg/h.
Entering 3.7 hours as the start-time dose time would place the modeled infusion 2.5 hours later than the source administration. The end timestamp is correct as an end timestamp; it is incorrect only if assigned the meaning “start” without conversion.
Convert Clock Timestamps to the Chosen Time Scale
A calendar timestamp such as 09:12 is not yet necessarily the time value used in a dataset. The time origin and units must also be specified. If elapsed time is measured from a chosen origin \(t_0\), convert an actual event timestamp \(t\) to elapsed time by subtracting the origin on a consistent time scale. For an infusion, first determine which endpoint the event represents, then calculate elapsed time for the endpoint required by the target convention.
Worked Example: Reconstructing Start Time From an End Timestamp
In a hypothetical record, an infusion end timestamp is 10:00, its documented duration is 90 minutes, and the time origin is 07:30 on the same date. The target dataset represents infusions at actual start time and uses elapsed hours.
The reconstructed start clock time is \(10{:}00-1\) hour \(30\) minutes, which is 08:30. Relative to the 07:30 origin, the start time is \(60/60=1.00\) hour. The end is \(150/60=2.50\) hours after the origin, and the duration is \(90/60=1.50\) hours.
Check the relationship: \(1.00+1.50=2.50\) hours. The dose record uses time \(1.00\) h and duration \(1.50\) h. If the hypothetical dose amount is 450 mg, the rate is \(450/1.50=300\) mg/h. This conversion is justified only if the 90-minute duration belongs to this administration and the 10:00 timestamp really denotes its end.
Actual Time and Nominal Time Answer Different Questions
Actual time describes when administration occurred according to the source record. Nominal time describes when it was intended or scheduled to occur. A nominal time may be useful when the analysis explicitly concerns planned timing, or when a protocol-defined schedule is the intended input. It does not become an actual administration time merely because it is convenient or regularly spaced.
If both values are available, preserve their distinct meanings. Use the one required by the analysis and model specification, and retain enough information to identify the other when it is relevant to review. Replacing actual times with nominal times removes information about deviations. Conversely, using an observed actual time when the intended analysis requires nominal times changes the time basis of that analysis.
Worked Example: Keeping Nominal and Actual Bolus Times Distinct
For a hypothetical bolus, the scheduled time is 08:00 and the source records actual administration at 08:17. The study time origin is midnight, and the target convention is actual administration time in elapsed hours.
The actual time is \(8+17/60=8.2833\) hours after the origin. The nominal time is \(8.00\) hours. Their difference is \(17/60=0.2833\) hours, or 17 minutes.
Under the stated actual-time convention, the dose record uses 8.2833 h. Using 8.00 h instead would represent the scheduled time, not the recorded administration. If the analysis instead specifies nominal times, 8.00 h is the applicable value, but it must be labeled and interpreted as nominal. Neither value should be silently substituted for the other.
Assumptions, Diagnostics, and Failure Modes
Timestamp conversion is valid only when the source fields have interpretable meanings. Before converting, confirm that timestamps refer to the same dose, that their date and clock components are complete enough for the calculation, and that any duration is expressed in compatible units. Clock timestamps also need a consistent time zone and a clear rule for daylight-saving transitions. A local time that is ambiguous during a clock change may not identify a unique elapsed time.
When converting to elapsed time, apply the specified origin and units consistently. Do not subtract clock readings while ignoring a date change. For example, an infusion starting at 23:50 and ending at 00:20 the following day lasts 30 minutes, not a negative 23 hours 30 minutes. Date-aware timestamp arithmetic gives \(D=0.5\) h. This check is especially important when records are stored as separate date and time fields.
Useful checks include comparing calculated durations with recorded durations, checking that reconstructed starts precede ends, and checking that the converted start plus duration recovers the converted end. A mismatch may indicate a wrong endpoint label, a unit error, an incorrect time zone, a date rollover problem, or a mismatch between records. A systematic offset across many records can indicate that the wrong convention has been applied; offsets that vary by duration may specifically suggest that infusion ends were treated as starts.
The main alternatives are not simply different formatting choices. They represent different event definitions:
| Convention | Timestamp used | What it represents | Main risk |
|---|---|---|---|
| Actual start time | Time administration began | Start of actual delivery | Using an end timestamp without subtracting duration |
| Actual end time | Time administration finished | Completion of actual delivery | Supplying it to a model that interprets dose time as start |
| Nominal time | Scheduled or target time | Intended timing | Interpreting it as proof of actual administration |
A start-time convention is often practical for infusion records when the model expects a dose event at infusion initiation and uses duration to describe delivery. An end-time convention may be suitable only when the analysis and software explicitly use that meaning or when the required representation is derived accordingly. Nominal time may be appropriate for a schedule-based analysis, but it cannot recover actual timing deviations that were not recorded. No convention can resolve an ambiguous source timestamp or manufacture a missing administration time without additional evidence.
A Reproducible Preparation Workflow
A defensible conversion is a documented data rule, not a one-off adjustment. For each dose type, identify the source timestamp field and its meaning, define the target convention, state the time origin and units, and specify the transformation. Keep source values available for audit, and report how records with missing or contradictory timing information are handled. Do not infer an actual administration time from a nominal time unless the analysis makes that assumption explicit.
Determine whether each timestamp denotes actual start, actual end, or nominal time. Resolve definitions from the source documentation where possible.
State what the model’s dose-time field represents, and identify whether the conversion differs for infusions and instantaneous doses.
Use duration to convert an end to a start when justified, then apply the time origin and units. Check endpoint arithmetic and record-level consistency.
Preserve source timing information and document assumptions, exceptions, and any records that could not be unambiguously converted.
Common Mistake
A common mistake is to reason, “The source time is the dose time, so I can copy it into the model dataset.” That is valid only if the source and target fields refer to the same event and use the same time scale. An infusion end timestamp copied into a start-time field shifts the infusion later by its full duration. A nominal timestamp copied into an actual-time field misstates what happened. The correction is to define the event represented by each field before copying or converting it.
A second mistake is to treat a plausible-looking elapsed time as evidence that the conversion is right. A value can fall within the study period and still refer to the wrong endpoint. Check the arithmetic against the source dates, duration, and documented convention rather than relying on plausibility alone.
Check Your Understanding
Answer each question, then compare with the solutions below.
- An infusion starts at 13:10 and ends at 15:40. The target convention is actual start time. What start-time value should be used if the origin is 12:00, and what duration should accompany it?
- A record contains an infusion end time but no start time or duration. Can the actual start time be reconstructed? Explain.
- A nominal bolus time is 14:00, while the recorded actual administration is 14:06. Which time belongs in a dataset whose convention is actual administration time?
- Why can an infusion duration help diagnose a systematic difference between source times and model dose times?
Solutions. (1) The start is 1 hour 10 minutes after the origin, so use \(1+10/60=1.167\) h; the duration is 2 hours 30 minutes, or 2.50 h. The endpoint check gives \(1.167+2.50=3.667\) h after the origin, corresponding to 15:40. (2) No. An end time alone does not determine a start time; a valid duration or actual start timestamp is also required. (3) Use 14:06, because the specified convention is actual administration time. The 14:00 value is nominal. (4) If end timestamps are mistakenly treated as starts, the difference will be approximately the infusion duration. A pattern in which the offset varies with duration supports that diagnosis, although source definitions and other data issues must also be checked.