Why Audit the Records Against Laboratory Metadata?
A below-quantification flag is meaningful only when it corresponds to the laboratory result for the correct specimen and analytical method. A dataset may contain a flag, a concentration field, and a quantification limit that look individually reasonable but belong to different assays, runs, units, or specimens. Comparing those fields only within the analysis dataset can therefore miss a mismatch: the dataset may be internally consistent and still disagree with the laboratory record.
The audit asks a focused question: does the analysis record accurately represent what the laboratory reported for this specimen under the applicable assay and run conditions? It is not a screen for whether a concentration is biologically plausible. The relevant evidence is the link between the analysis record and its laboratory sources: specimen identifiers, assay identifiers, report qualifiers, reported values, units, quantification limits, and any documented rules for choosing among multiple runs.
The comparison depends on a few assumptions. First, the laboratory report and its metadata are the designated sources for the analytical result and assay conditions. Second, the analysis record can be linked unambiguously to the relevant specimen and report. Third, the rules for selecting a result when a specimen was tested more than once are known. Finally, units and assay-specific limits can be interpreted on a common basis. If any of these conditions fails, the audit should flag the uncertainty rather than infer a definitive classification.
What Must Agree?
A useful audit separates two questions that are easy to conflate. The first is whether the dataset points to the correct laboratory result. The second is whether the result has been represented correctly in the analysis fields. Check the identifiers before interpreting the concentration: a well-matched number from the wrong sample is still the wrong result.
| Comparison | What to verify | Why it matters |
|---|---|---|
| Specimen linkage | Subject, visit or collection time, specimen ID, and aliquot where available | A correct result attached to the wrong specimen misstates the observation. |
| Assay and run | Analyte, assay or method ID, run ID, and any applicable rerun selection rule | Limits and result status may differ across assays or runs. |
| Laboratory result | Reported value, units, and qualifier such as quantified or below quantification | The qualifier indicates how the laboratory classified the result. |
| Analysis record | Concentration field, below-quantification flag, units, and linked limit | These are the fields used downstream and must represent the selected report. |
| Assay metadata | Applicable limit, its units, and any documented qualifications | A limit from another assay or unit scale cannot validate this result. |
The laboratory qualifier is important evidence. A comparison of the numeric concentration with the limit is a consistency check, not a replacement for the reported qualifier. For example, rounded values may appear equal to a limit even when the unrounded laboratory result was classified as below it. Likewise, a result may have special report annotations or rerun context that cannot be recovered from a simple numeric comparison.
A Reproducible Audit Sequence
A consistent sequence helps prevent one field from silently deciding the interpretation of another. The following steps move from identity to result representation, then to disposition.
Match the dataset row to the laboratory report using the available specimen identifiers and collection context. If more than one report could match, stop and investigate the linkage.
Confirm analyte and assay, then identify the report or run selected under the documented laboratory or study rule. Do not select a result merely because it agrees with the dataset.
Check the reported concentration's units and the assay limit's units. Convert units explicitly when necessary, and verify that the limit applies to this assay and result.
Compare the report qualifier with the analysis flag, then compare any reported numeric value with the dataset value using the documented precision and representation rules.
Classify the record as a match, a resolvable mismatch, or unresolved. Specify which evidence supports that outcome; do not force a correction where source identity or precedence remains uncertain.
The logic can be stated compactly. Let \(B_{\mathrm{lab}}\) indicate that the selected report classifies the result as below quantification, and let \(B_{\mathrm{data}}\) be the analysis dataset's flag, coded as 1 for below quantification and 0 otherwise. A status match requires:
This equality is meaningful only after the report has been linked to the correct specimen and the applicable assay and run have been established. It is also only one part of the audit: a status match does not prove that the concentration value, units, or limit is correct. Those fields must be checked separately.
Worked Audits
Worked Example: A Below-Quantification Flag Conflicts With the Report
Consider a hypothetical sample with specimen ID S104. The linked laboratory report identifies the analyte and assay as expected, gives an applicable limit of \(0.80\ \mathrm{ng/mL}\), and reports the result as below quantification. The analysis record contains a numeric value of \(0.80\ \mathrm{ng/mL}\) and a below-quantification flag of 0.
Step 1: Verify identity. The specimen ID, collection context, analyte, and assay match the selected report. Assume the report is the applicable result under the study's documented selection rule.
Step 2: Compare status. The report says below quantification, so \(B_{\mathrm{lab}}=1\). The dataset flag is not set, so \(B_{\mathrm{data}}=0\). Thus \(1\ne 0\): the status fields disagree.
Step 3: Interpret the numeric field cautiously. The dataset's \(0.80\ \mathrm{ng/mL}\) equals the stated limit, but this equality does not overturn the report qualifier. The value could reflect a reporting or data-processing convention, rounding, or an incorrectly represented result. The available facts establish a flag mismatch, not which explanation produced it.
Audit outcome. This is a confirmed status discrepancy. The record should be referred for resolution against the laboratory result and the applicable dataset convention. Do not decide that the result is quantified merely because its displayed value equals the limit.
Worked Example: A Unit Difference Creates an Apparent Limit Mismatch
A hypothetical report for specimen S208 gives a quantified concentration of \(0.0040\ \mathrm{mg/L}\). Its assay metadata lists a limit of \(5.0\ \mathrm{ng/mL}\). The analysis record has a concentration of \(4.0\ \mathrm{ng/mL}\), a flag of 0, and a linked limit of \(5.0\ \mathrm{ng/mL}\).
Step 1: Confirm the unit conversion. Since \(1\ \mathrm{mg}=10^6\ \mathrm{ng}\) and \(1\ \mathrm{L}=10^3\ \mathrm{mL}\), \(1\ \mathrm{mg/L}=10^3\ \mathrm{ng/mL}\). Therefore:
The converted laboratory value agrees with the dataset value. Both the laboratory report and the dataset indicate a quantified result, so \(B_{\mathrm{lab}}=0\) and \(B_{\mathrm{data}}=0\). The status fields agree. The converted value is also below the listed \(5.0\ \mathrm{ng/mL}\) limit, which conflicts with the report's quantified qualifier.
Step 2: Do not resolve the conflict by assumption. The specimen and assay identifiers still need to be checked, as does whether the metadata limit applies to this report and whether an annotation, dilution, or other documented assay condition affects interpretation. The arithmetic identifies the conflict; it does not establish that the qualifier or metadata is wrong.
Audit outcome. The concentration and units reconcile, but the report qualifier and listed limit are inconsistent. Mark the record unresolved until the laboratory source or assay metadata clarifies the applicable classification. A unit conversion can explain apparent numeric differences, but it cannot settle a genuine conflict between status and metadata.
Worked Example: A Documented Rerun Determines the Result to Audit
For hypothetical specimen S315, the laboratory system contains two reports for the same analyte. The initial run reports below quantification with a limit of \(1.0\ \mathrm{ng/mL}\). A documented procedure states that the later valid rerun supersedes the initial result. The later report is quantified at \(0.72\ \mathrm{ng/mL}\), with an applicable limit of \(0.25\ \mathrm{ng/mL}\). The analysis record contains \(0.72\ \mathrm{ng/mL}\), the later run ID, and a below-quantification flag of 1.
Step 1: Apply the selection rule. The documented procedure selects the later valid rerun, so the initial report is not the result to compare with this analysis record.
Step 2: Check the later result against its limit. The reported value is \(0.72\ \mathrm{ng/mL}\), and the applicable limit is \(0.25\ \mathrm{ng/mL}\). Their difference is:
The reported value is \(0.47\ \mathrm{ng/mL}\) above the listed limit, consistent with the later report's quantified status. The dataset's value and run ID agree with that later report, but its flag does not: \(B_{\mathrm{lab}}=0\), whereas \(B_{\mathrm{data}}=1\).
Audit outcome. This is a resolvable flag mismatch. The later run is selected by the stated rule, and its report supports a quantified status. The initial below-quantification report explains how an outdated flag might have entered the dataset, but does not justify retaining that flag for the selected result.
Interpretation, Diagnostics, and Limitations
An audit is only as reliable as its linkage and source hierarchy. A record may be impossible to classify confidently when specimen identifiers are duplicated, a report is missing, the run-selection rule is undocumented, or the limit's assay and unit are uncertain. These are not minor inconveniences: each prevents a key comparison from being made. The appropriate outcome is unresolved, not an inferred match.
Several patterns are useful diagnostics for a failing audit process:
- Many apparent mismatches disappear after unit conversion. This suggests that units are not being harmonized before comparison.
- Flags disagree mainly on specimens with multiple reports. Check whether the selected-run rule is applied consistently.
- A limit is absent or identical across records from distinct assays without a documented reason. Verify the assay-to-metadata linkage rather than filling in a limit by guesswork.
- Values and flags match, but specimen IDs or collection context do not. This is a linkage failure, not a validated result.
- Values appear equal to a limit after rounding. Inspect the report qualifier and available precision; displayed equality alone is inconclusive.
An automated comparison can efficiently identify differences in identifiers, flags, values, units, or limits, but it cannot resolve every source conflict. A person must evaluate exceptions using laboratory conventions and the documented result-selection rule. Manual review alone can supply context, but is less reproducible if different reviewers apply different comparison criteria. A practical approach is therefore a structured comparison followed by targeted review of exceptions. Neither approach licenses choosing whichever field produces the preferred classification.
This audit has a deliberately narrow scope. It establishes whether the analysis record corresponds to the applicable laboratory evidence; it does not determine whether a concentration is biologically plausible, select a statistical model for below-quantification observations, or establish the cause of a laboratory discrepancy. If source materials conflict, the audit identifies the conflict for resolution rather than claiming certainty beyond the evidence.
Check Your Understanding
For each question, decide what the audit can conclude from the stated evidence.
- A linked report says below quantification, but the dataset flag is 0. The specimen and selected assay match. What is the status comparison?
- A report gives \(0.006\ \mathrm{mg/L}\), while the dataset gives \(6\ \mathrm{ng/mL}\). Are the numeric values consistent after conversion?
- A dataset value is below the listed limit, but the laboratory report says quantified. Is the record automatically confirmed as below quantification?
- Two runs exist, and no rule identifies which result is applicable. Can the auditor select the run that agrees with the dataset?
Solutions. (1) It is a confirmed status discrepancy: \(B_{\mathrm{lab}}=1\) and \(B_{\mathrm{data}}=0\). (2) Yes. Using \(1\ \mathrm{mg/L}=1000\ \mathrm{ng/mL}\), \(0.006\times1000=6\ \mathrm{ng/mL}\). (3) No. The numeric comparison conflicts with the report qualifier, so verify the assay, units, applicable limit, and any report annotations; do not infer status from the number alone. (4) No. Without a documented selection rule or other authoritative evidence, the applicable result is unresolved and should be investigated.