Why Regimen Changes Need Their Own Dates
A participant may start a study on one dose schedule, later have the dosing frequency changed, and then switch route or formulation. A data set that records only the starting regimen cannot show when those later instructions took effect. Conversely, putting a new regimen value on every observation can obscure whether it represents a change or merely repeats the current state.
The useful organizing idea is to treat each documented regimen change as an event with an effective time. Between two successive change times, the applicable regimen is constant unless another change is documented. This creates a clear link between the source record and the time interval during which a regimen is considered in force.
Use the study's established time origin and units for the event time, and apply consistent dose units and field conventions. A regimen-change time is not automatically the time of an individual administration. A regimen can describe the schedule intended or prescribed over an interval; separate dose-event records describe administrations. A documented change alone does not establish that every scheduled dose was taken.
Representing the Intervals
Let the distinct, ordered effective change times for one participant be \(t_0,t_1,\ldots,t_K\), where \(t_0<t_1<\cdots<t_K\). The regimen beginning at \(t_k\) applies on the half-open interval from \(t_k\) to \(t_{k+1}\). The final recorded state has no known end unless the data document one; for organizing the recorded history, it can be shown as continuing beyond \(t_K\) with an open end.
The half-open convention means the new state applies at the exact effective time. For example, if a change is effective at study time 8 days, the state before that change applies for \(t<8\), and the new state applies for \(t\geq 8\), subject to the next change. This convention makes adjacent intervals meet at a boundary without both claiming that boundary.
A change can update one attribute or several. If only frequency changes, the other documented attributes carry forward only when the source record and data specification support that interpretation. If a source gives a complete replacement regimen, preserve it as a complete state. In either case, retain enough information to distinguish the change event from the resulting regimen state; otherwise an omitted field might be mistaken for a discontinued value or a missing value.
Why the Interval Construction Works
After sorting and resolving the distinct effective times, the half-open intervals have two useful properties: no time is assigned to two adjacent states, and every time from the first recorded change onward is assigned to a state. This is a bookkeeping result, not proof that a participant actually followed the regimen.
The first equality follows because a time in the earlier interval must be less than \(t_{k+1}\), while a time in the later interval must be at least \(t_{k+1}\). The second follows because every time from \(t_k\) up to \(t_{k+2}\) is either below \(t_{k+1}\) or at least \(t_{k+1}\). Repeating this argument across the ordered change times gives a continuous partition of the recorded period. If the source has an unrecorded period, conflicting events, or an unknown initial state, the data may not justify such a partition without qualification.
Worked Example: A Frequency Change at a Known Time
Consider a hypothetical participant whose documented regimen is 100 mg orally as a tablet every 24 hours beginning at study time day 0. A source record documents a frequency change to every 12 hours, still 100 mg orally as a tablet, effective at day 8. Represent the states and decide which applies at day 8.
The distinct effective times are \(t_0=0\) and \(t_1=8\) days. The initial state applies on \([0,8)\) days: 100 mg orally as a tablet every 24 hours. The updated state applies from day 8 onward: 100 mg orally as a tablet every 12 hours. Thus, at exactly day 8, the updated frequency applies.
To check a record at day 6, test the interval conditions: \(0\leq 6<8\), so it falls under the every-24-hour state. At day 8, \(8\leq 8\) and the new interval has no documented end, so the every-12-hour state applies. The calculation identifies the documented regimen state; it does not establish the timing or occurrence of any individual administration.
Carry-Forward Changes and Complete Regimen States
Some source records document a single changed attribute, while others give a full replacement. The data specification should state how to interpret each form. For a single-attribute update, carry forward unchanged attributes only when that is supported by the record and its conventions. A resulting-state view can make the interpretation explicit, while the underlying change-event record preserves what the source actually said.
Worked Example: A Route and Formulation Change
In a hypothetical record, a participant's documented state from day 0 is 20 mg intravenously every 24 hours as a solution. At day 5, a complete regimen-change record specifies 40 mg orally every 12 hours as a tablet. The values are illustrative and do not imply that the two regimens are therapeutically equivalent.
The day 0 state applies on \([0,5)\) days. The day 5 state begins at day 5 and continues until another documented change or the end of the available record. At day 4.9, the earlier state applies; at day 5.0, the updated route, formulation, amount, and frequency apply. Because the day 5 record supplies a complete state, there is no need to infer which earlier attributes carry forward.
If the source had specified only “change route to oral” at day 5, the dose amount, frequency, and formulation would need to be handled according to documented rules. It would not be safe to assume that every unmentioned field either remains unchanged or becomes missing without checking what the source record means.
Checking a Regimen Timeline
A useful audit starts with each participant's effective change events, puts them in chronological order, and checks whether the resulting intervals agree with the source. Pay particular attention to duplicate timestamps, contradictory instructions at the same time, and records that cannot be assigned confidently because their date or time is incomplete. The earlier material on ordering simultaneous dose and sample events is relevant when an actual dose event and another event share a timestamp; it does not resolve conflicting regimen instructions on its own.
Use the established study-time origin, units, and documented time precision. Do not silently convert a date-only change into a more precise time.
Sort events within participant and identify repeated times. Determine from source documentation whether records at the same time are duplicate entries, compatible simultaneous updates, or conflicting instructions.
Apply documented carry-forward rules, then assign each state an interval ending at the next distinct effective time. Keep an unknown start or end visibly unknown rather than presenting it as confirmed.
Check records immediately before a change, exactly at its effective time when time precision supports that comparison, and after it. Confirm that no record is assigned to two states or to an unjustified gap.
An interval table makes the audit inspectable. For example, a change at day 4 should result in an earlier interval ending at day 4 and a later interval beginning at day 4, not two intervals that both include day 4. A second event at day 4 requires source review before choosing a single state: sorting alone cannot determine which conflicting instruction is correct.
Worked Example: Auditing Boundaries and a Conflicting Timestamp
Suppose a hypothetical participant has a documented 50 mg oral capsule regimen every 24 hours from day 0. A record changes the frequency to every 12 hours at day 4. Another record says “every 8 hours,” also effective at day 4, but the available source does not indicate whether it replaces or corrects the first change. Determine what can be represented and classify records at days 3.9, 4.0, and 4.5.
The day 0 state can be assigned to \([0,4)\) days if the day 4 events are valid effective changes. At day 3.9, the documented 50 mg oral capsule every-24-hour state applies. At day 4.0 and day 4.5, the event times establish that the pre-change state has ended, but the resulting frequency is unresolved: one source entry specifies every 12 hours and another specifies every 8 hours at the same effective time.
It would be unjustified to select the later-listed row, average the frequencies, or create a zero-duration state merely to force a single answer. Preserve the conflicting source information, flag the affected interval for reconciliation, and use a clarified record or an approved data-resolution rule before assigning a definitive post-day-4 regimen. The exact boundary convention identifies when the former state stops applying; it cannot repair a contradictory source.
Alternatives, Diagnostics, and Limitations
An event-based representation stores each change once and derives the intervals from successive effective times. It is compact and preserves a traceable change history, but users must apply the interval rule correctly. A current-regimen snapshot stores the state at a chosen time, which is convenient for a point-in-time summary but cannot by itself recover earlier states or the change dates. An expanded schedule can create a record for every expected administration, but it can be large and may be mistaken for evidence that doses were administered. These representations answer different questions; none should silently replace the source change events.
Useful warning signs include an interval whose end precedes its start, multiple incompatible states at one effective time, a record falling outside all known intervals, or an abrupt attribute change unsupported by a dated source entry. Also investigate apparent changes caused by inconsistent units, spelling, route codes, or formulation labels. Standardized field values make comparison possible, but standardization must not conceal a real change or resolve an undocumented discrepancy.
A regimen timeline has limits. It represents documented instructions, not necessarily administrations, adherence, or the reason for a change. A date-only source may not establish whether the new regimen applied before or after an event later on that same date. The final recorded regimen should not be treated as continuing indefinitely in clinical reality; an open-ended interval means only that no later change is available in the records being represented. Keep these distinctions explicit when using the timeline in downstream work.
Common Mistake
Wrong reasoning: “The new regimen is recorded on day 8, so it applies only after day 8; a dose or sample at day 8 must still use the old regimen.” This leaves the boundary undefined and can assign the same instant inconsistently.
Correction: State the effective-time convention explicitly. With half-open intervals \([t_k,t_{k+1})\), the new state starts at its documented effective time, so a record at exactly day 8 is assigned to the state beginning on day 8. If the source does not say whether the change was effective before or after an event that day, the ambiguity is a data limitation to resolve or flag, not a reason to invent a time.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A regimen starts at day 2, changes at day 7, and changes again at day 10. Write the interval for the middle regimen and state which regimen applies exactly at day 10.
- A participant's frequency changes at day 5, but a record at day 5 gives no route or formulation. What must be checked before assigning those fields to the new state?
- Two incompatible regimen changes have the same effective timestamp. What can the interval convention resolve, and what still requires source review?
- Why is a list of expected dose times not interchangeable with dated regimen-change events?
Solutions. 1. The middle regimen applies on \([7,10)\) days. At exactly day 10, the state beginning at day 10 applies, assuming its effective time and content are documented. 2. Check the source and approved data rules to determine whether route and formulation carry forward, are specified elsewhere, or are unknown; do not infer them merely from omission. 3. The convention establishes that the pre-change interval ends at that timestamp and that a new state begins there. It cannot choose between incompatible instructions that begin at the same time; those require reconciliation or an explicit resolution rule. 4. Change events describe when a documented regimen state changes. Expected dose times expand a schedule into planned administrations and do not establish that any dose was actually administered.