Bolus Dose Encoding

What a Bolus Record Represents

A dose record is not merely a number placed beside a time. It represents an administration that occurred, with an amount and a time, for a particular subject and route or formulation. For a bolus, the defining feature is that the dose is represented as delivered instantaneously at the event time. This representation is different from an infusion, for which delivery takes place over an interval. The record should express the administration that occurred, not an arbitrary distribution of its amount across observation rows.

The earlier material on dose-event records establishes that a dose is represented as an event, distinct from an observation. The amount and route or formulation must also be interpreted consistently, as established in the earlier material on dose amount and dose units and route and formulation fields. Here the focus is how to encode the bolus event itself.

“Instantaneous” describes the data and model representation; it does not claim that a real-world administration takes literally zero time. It distinguishes a bolus from an administration that the model represents as being delivered over time. That distinction matters because the model applies input differently: a bolus produces an immediate change in the amount in the receiving compartment, while an infusion supplies input over an interval.

A bolus record contains the amount for that event, not a concentration, an amount remaining in the body, or a daily total unless the daily total is itself the amount of a single administration. The event time identifies when the dose was given, using the data set’s established time origin and units. A dose record should not be made into an observation record simply because a concentration was measured at the same nominal time.

How an Instantaneous Dose Enters a Model

Consider a one-compartment model for an IV bolus. Let \(A(t)\) be the drug amount in the compartment, let \(D\) be the administered dose, and let \(k\) be a first-order elimination rate constant. Between administrations, assume elimination follows \(dA/dt=-kA\). At the bolus time \(t_d\), the dose adds \(D\) to the amount present immediately before administration:

Here \(t_d^-\) and \(t_d^+\) denote the instants just before and just after the event. For a single IV bolus, the dose is assumed to enter the modeled systemic compartment directly, with no delay or loss between administration and entry. These are modeling assumptions, not consequences of the data format. If they are inappropriate for the route or formulation, the model must represent the input differently.

The jump condition can be justified by representing the bolus input as an impulse. For a dose at \(t_d\), write the amount model as \(dA/dt=-kA+D\delta(t-t_d)\), where \(\delta\) is a unit impulse at the event time. Integrate over a narrow interval containing \(t_d\), from \(t_d-\epsilon\) to \(t_d+\epsilon\):

As \(\epsilon\) approaches zero, the elimination integral approaches zero if \(A(t)\) remains finite over the interval, while the impulse integral equals one. The change in amount is therefore \(D\), giving the stated jump condition. After the event, if no further input occurs, the amount follows the elimination model:

This derivation explains why a single bolus is encoded as one event at one time: the model applies one discrete input there. For a route that places drug in a depot compartment, the instantaneous jump can instead be in that depot amount; it does not imply an instantaneous plasma concentration jump.

One Administration Is Not Several Dose Records

A single administration should ordinarily be represented once. Putting the same dose amount on multiple rows can cause a model to apply that amount multiple times. Distributing one dose across observation records is also unsafe: observations have their own role, and their number or timing should not determine how much drug the model receives.

Multiple records are appropriate when they represent multiple administrations. For example, two separate 50 mg administrations at different times are two dose events, even if they are part of one prescribed regimen. Conversely, two source rows that each repeat a 100 mg dose because of a data merge do not establish that two administrations occurred. Decide from the administration history and event identity, not by counting rows or assuming repeated values mean repeated doses.

If one administration is represented in the source system as several components—for example, separate entries for portions given at the same event time—the modeling data should not accidentally treat those entries as several complete doses. Determine whether they are genuinely separate amounts, duplicate representations, or components that should be combined into one event. Preserve source detail where needed for traceability, but make the dose-event record used by the model correspond to the intended administration. If portions were actually administered at distinct times, represent those as separate events at their respective times rather than collapsing them into one bolus.

Practical Checks and Common Failure Modes

A useful quality check asks whether every dose-event row represents one administration and whether every confirmed administration appears once. Check the event time and amount together with the subject identifier, event type, route or formulation, and dose units. Units must be consistent with the model’s expected units; a numerically plausible amount in the wrong unit can still produce a severely incorrect input. Event times must use the established time origin and scale, rather than mixing calendar times and elapsed study times.

Also check for duplicate event records, dose amounts copied onto observation rows, and implausible sequences that may indicate a merge or coding error. If a dose value is repeated on rows with different sample times, establish whether those rows refer to separate administrations or whether the dose was mistakenly carried forward. A visual review of a few subject timelines can reveal patterns that a simple range check misses. These checks identify data representation problems; they cannot establish that the documented administration actually occurred if the source record is incomplete.

An infusion is not a bolus merely because it has a total amount and a start time. Its input is delivered over time rather than applied as one instantaneous jump. Encoding such an administration as a bolus changes the timing of drug input and may change the predicted concentration profile. The distinction is conceptual here: the duration and rate fields needed to represent an infusion are separate data elements and are not substitutes for a bolus event time and amount.

The main alternative to a bolus representation is a model input that explicitly distributes administration over time, such as an infusion representation. A separate alternative may be required when the route or formulation implies delayed or otherwise non-instantaneous entry into the modeled compartment. The suitable representation depends on the documented administration and model assumptions, not on which format is easiest to enter. A bolus record alone cannot capture the duration of an administration, undocumented missed portions, uncertainty in event time, or a complex delivery history.

Bolus encoding is therefore a precise statement: this amount was administered at this event time, and the model treats its input as instantaneous into the specified receiving compartment. It is not a general claim that all of the administered drug is immediately present in plasma, nor is it a license to copy the amount across multiple records.