Why chronology anomalies need investigation
A concentration can appear to occur before a dose, at an implausible elapsed time after a dose, or in an unexpected place in a sequence of samples. These patterns deserve investigation, but they do not by themselves identify the cause. A shifted clock, a misassigned sample, an omitted dose, a genuine deviation from the planned schedule, and an unusual biological concentration can produce similar appearances in an analysis dataset.
The task is therefore to trace the records, not to make the timeline look pharmacologically familiar. Begin with the recorded administration and collection events, establish which sources support them, and test specific explanations against those sources. A concentration profile can help prioritize a review, but it should not substitute for evidence about what was administered, when a sample was collected, or which sample it was.
Establish the timeline before interpreting the profile
For each anomalous record, assemble the smallest useful timeline: the sample’s recorded subject and sample identifier; its recorded collection date and time; the corresponding concentration and units; and the administrations that could contribute to that concentration. Include the scheduled dose only as a comparison, not as proof that the dose was given. Distinguish an administration record from a planned administration.
Use source records appropriate to each event. Depending on the data collection process, these may include administration records, nursing documentation, clinic worksheets, specimen labels, laboratory accession records, or electronic system audit trails. Record whether each time means the event time, the time entered into a system, or a time converted during data processing. The harmonization of dates, clock times, and time zones, and reconstruction of elapsed time from visit timestamps, are useful foundations; the key question here is whether the particular dose and collection events can be linked reliably.
Before proposing a shift, check that the records refer to the same subject, study period, dose, and specimen. Consider date boundaries, relative-time resets, infusion start and end times, and whether the reported time is scheduled or actual. If an administration history begins partway through a dosing sequence, do not assume that the first dose in the analysis file was the first dose the participant received.
Worked Example: Testing a Suspected One-Hour Collection-Time Shift
All data in this example are hypothetical. For subject 018, the analysis file records a dose at 08:00 and a sample at 10:20, giving an elapsed time of 2 hours 20 minutes. The laboratory accession record shows that the same sample barcode was received from a collection documented at 09:20. The clinic worksheet independently records collection at 09:20. Suppose the study records establish that these source times are already on the same local clock basis.
Compare the times. The difference between the analysis-file collection time and both source times is 60 minutes: 10:20 − 09:20 = 1 hour. The elapsed time based on the analysis file is 10:20 − 08:00 = 2 hours 20 minutes. The elapsed time based on the two matching collection records is 09:20 − 08:00 = 1 hour 20 minutes.
Assess the evidence. Two independent records agree on 09:20, and the barcode links them to the concentration record. This supports a one-hour discrepancy in the analysis-file collection time for this sample. It does not establish why the discrepancy occurred, and it does not justify changing other samples by one hour. The proposed time should be treated as supported only to the extent that those source records are authoritative and correctly linked.
Test explanations against independent evidence
Several explanations may fit the same apparent anomaly. Test them in a disciplined order: verify identity and event linkage; compare the analysis record with source timestamps; check whether the dose history is complete for the relevant period; and only then consider whether the concentration itself remains unusual given the verified history. A timeline discrepancy is stronger evidence for a data-handling issue when it is replicated in independent records or follows a consistent, traceable system transformation.
A simple comparison makes the logic explicit. Let \(t_{\mathrm{file}}\) be a time in the analysis file and \(t_{\mathrm{source}}\) the corresponding time in a linked source record. Define the discrepancy as \(r=t_{\mathrm{file}}-t_{\mathrm{source}}\). If multiple matched events show the same \(r\), a shared shift is a candidate explanation. If the discrepancy varies among events, a single global shift is not supported. Even a constant difference is not proof of a time-zone or clock error: linkage, source definitions, and the recorded time basis still need to be checked.
This comparison also shows why the records used to calculate elapsed time matter. If dose and collection times are both shifted by the same amount, their difference is unchanged. If only one is shifted, the calculated elapsed time changes by that amount. Therefore, a plausible elapsed time does not guarantee correct timestamps, and an implausible elapsed time does not specify which event is wrong.
The equation applies only when the same offset \(\delta\) is applied to both times. It is a useful check on candidate explanations, not a method for choosing an offset from concentration data alone. For infusions, use the relevant documented start and end events; a single “dose time” may not adequately describe administration. For repeated dosing, a sample can reflect more than the immediately preceding dose, so the complete relevant history matters.
Worked Example: Finding an Incomplete Administration History
All data are hypothetical. A participant is prescribed a dose every 12 hours. The analysis file contains a dose at 08:00 on Tuesday and a concentration sample collected at 07:50 that day, with a concentration of 4.8 mg/L. The file shows the previous dose on Monday at 08:00, so the apparent interval from that recorded dose to the sample is 23 hours 50 minutes. The scheduled Monday evening dose is not present in the file.
Check the source administration record. A signed medication-administration record documents that the Monday dose was given at 20:00. The elapsed time from that documented dose to the Tuesday sample is 11 hours 50 minutes: 07:50 − 20:00 on the prior day = 11 hours 50 minutes. This is close to the planned 12-hour interval, but the important finding is not that the concentration “looks right.” It is that an administration supported by a source record is absent from the analysis file.
Interpret cautiously. The source record supports an incomplete dose history in the analysis file. It does not establish that the 4.8 mg/L concentration is correct, prove that the participant followed every other scheduled dose, or rule out additional undocumented doses. The event and its source should be recorded for reconciliation; missing administrations must not be invented from the concentration.
Distinguish a sample-label problem from a time problem
A concentration may be associated with the wrong subject, visit, or collection time even when the clock time itself is accurate. Investigate identity using the chain connecting the collection event to the specimen and the specimen to the laboratory result: identifiers on collection records, container labels, accession data, and any available audit trail. A concentration pattern that appears more plausible after swapping labels is not sufficient evidence for a swap.
Worked Example: Investigating a Possible Sample Misassignment
All data are hypothetical. A clinic worksheet for subject 026 records sample barcode B771 collected at 14:00. The analysis file assigns the concentration result of 6.2 mg/L to subject 062 at 10:00. The laboratory accession record links barcode B771 to subject 026 and records its collection time as 14:00. A separate result bearing a different barcode is assigned to subject 062.
Trace the identifiers. The analysis-file assignment conflicts with the clinic worksheet and the laboratory accession record on both subject and collection time. The matching barcode links the 6.2 mg/L result to subject 026’s 14:00 specimen; the separate barcode indicates that subject 062’s record should be investigated independently.
Limit the conclusion to the evidence. This evidence supports an apparent sample-assignment discrepancy in the analysis file. It does not justify swapping values between subjects unless the full specimen-to-result linkage confirms both assignments. Nor does it show that either concentration is biologically implausible. If the identifier chain cannot be resolved, the record remains uncertain rather than being repaired by choosing the assignment that best fits an expected profile.
Diagnostics, alternatives, and limitations
Useful warning signs include a sample before the only recorded dose despite being marked post-dose; repeated samples with an identical unexplained time offset; abrupt changes in elapsed time at a date or study-period boundary; dose records that stop while concentrations continue; and disagreement between specimen identifiers in analysis and laboratory records. Automated chronology checks can flag these patterns consistently, but they cannot determine the cause. A flag is a request to review the linked records, not an automatic correction rule.
The main approaches answer different questions:
| Approach | What it can establish | What it cannot establish alone |
|---|---|---|
| Chronology rule or automated screen | Which records violate a specified timing or ordering check | Whether the timestamp, dose history, or sample identity is wrong |
| Source-record reconciliation | Whether linked records support a particular event time, administration, or specimen identity | Whether every unrecorded event did not occur |
| Concentration-profile review | Which records may warrant priority review in context | The actual dose, collection time, sample identity, or biological cause |
These approaches are complementary. A profile-based review can prioritize a record; chronology checks can describe the discrepancy; source reconciliation can support or reject specific data explanations. The limitation is that source records may themselves be incomplete, transcribed incorrectly, or ambiguous. A linked record is useful evidence, not infallible truth.
Do not infer a biological cause just because an anomaly remains after a simple timestamp check. Conversely, do not label an unusual concentration a data error merely because it conflicts with a typical pattern. The conclusion should match the strength of the evidence: confirmed discrepancy, likely explanation requiring confirmation, or unresolved anomaly. A later correction should preserve traceability to the original value and its supporting evidence.
A practical investigation sequence
State which dose, sample, subject, or time relationship appears inconsistent, without assigning a cause.
Check subject, study period, specimen identifiers, and whether each time refers to the event itself.
Compare relevant source administration and collection records on a consistent time basis, including preceding doses.
Ask whether the evidence supports a time shift, sample misassignment, incomplete dose history, or none of these.
Separate verified facts from hypotheses and unresolved questions; do not force a correction to achieve a preferred concentration profile.
Check Your Understanding
Answer each question using the evidence, not the concentration pattern alone. Solutions follow.
- A dose and its linked collection time are both recorded 30 minutes later than their source times. What happens to their calculated elapsed time if both records share the same offset?
- An analysis file shows no dose between Monday morning and Tuesday morning. A reliable administration record documents a Monday evening dose. What is established, and what is not?
- A sample’s concentration seems more plausible when assigned to another subject. Is that enough to change the subject assignment? What evidence should be checked?
- Several matched source events show different time discrepancies. Does that support applying one constant shift to all records? Explain.
- A concentration remains unusual after the administration and collection records are reconciled. What is an appropriate data-quality conclusion?
Solutions. (1) The elapsed time is unchanged because the same offset cancels: \((t_{\mathrm{collection}}+30\text{ min})-(t_{\mathrm{dose}}+30\text{ min})=t_{\mathrm{collection}}-t_{\mathrm{dose}}\). (2) The analysis-file history is incomplete for that documented administration; this does not prove that all other doses are known or that the concentration is correct. (3) No. Check the specimen identifier, collection documentation, laboratory accession, and result linkage; change the assignment only when evidence supports it. (4) No. Different discrepancies do not support one constant shift; investigate the individual event linkages and time definitions. (5) Record that the anomaly remains unresolved or unusual after reconciliation. Do not call it a data error or infer a biological cause without supporting evidence.