Data Integrity in the Microbiology Laboratory: Why Plate Counts Are the Hard Case

Data integrity expectations were largely written with instrument-generated chemical data in mind — chromatograms, audit trails, electronic records. Microbiology fits that model badly, and it is the harder case precisely because so much of the data is a human being looking at a plate once and writing a number down.

Why Microbiology Is Different

Three properties make microbiological data uniquely difficult to defend after the fact.

It is transient. A plate is read and then discarded or autoclaved. Unlike a chromatogram, the original observation cannot be re-examined. Whatever was written down is all that survives.

It is subjective. Counting colonies on a crowded plate, deciding whether a colony is confluent, judging whether growth in a broth is turbidity or precipitate — these are judgements, and two competent analysts can differ.

It is largely manual. Most steps produce no instrument record. There is no audit trail behind a dilution series performed at a bench.

The consequence is that microbiological data integrity depends far more on procedure, contemporaneous recording and second-person review than on system controls.

ALCOA+ Applied to a Plate Count

Principle	What it means at the bench
Attributable	The analyst who counted is identified, not just the one who signed
Legible	Handwriting readable, corrections struck through not overwritten
Contemporaneous	Recorded at the moment of reading, not transcribed later from memory
Original	The bench sheet is the raw data, not the typed report
Accurate	Count, dilution and calculation all correct and checkable
Complete	Including results that were discarded and why
Consistent	Chronological, dated, with a defined sequence
Enduring	Retained on durable media for the required period
Available	Retrievable throughout the retention period

The two that fail most often in microbiology are contemporaneous and complete.

Contemporaneous Recording

Reading a batch of plates and writing the counts afterwards from a scrap of paper — or from memory — is the single commonest data integrity weakness in a microbiology laboratory, and it is usually done for convenience rather than dishonestly.

It matters because the transcription step is unverifiable. There is no way to establish afterwards whether the number on the bench sheet is the number that was on the plate.

Practical controls:

• Record directly onto the controlled bench sheet at the point of reading.
• Where gloves or the physical setup make that awkward, solve the setup rather than the record — a plate reader station with the sheet in reach, or a second person recording.
• Never use uncontrolled paper as an intermediate step. If a scrap sheet exists, it is the raw data and must be retained, which is worse than not creating it.
• Time-stamp readings where incubation duration matters.

Photograph the Plate

This is the most valuable single practice available to a microbiology laboratory and it is still uncommon.

A photograph converts a transient observation into a durable record. It makes a count re-examinable years later, it resolves disputes about whether a colony was present, and it demonstrates good faith more convincingly than any procedural control.

• Photograph any plate with a result that will be acted on — excursions, positives, out-of-specification counts.
• Photograph any plate where the count is difficult or the morphology unusual.
• Include the plate identifier in the frame so the image is attributable.
• Store images with the batch record, not on a shared drive nobody indexes.
• For sterility positives and Grade A recoveries, photograph as a matter of course.

The cost is a phone and a naming convention. The value appears the first time someone asks what was actually on that plate.

Completeness: Recording What Was Discarded

An audit finding that recurs is a laboratory whose records show only the results that were reported.

If a plate was overgrown and recounted at a higher dilution, both appear. If a test was aborted because the medium was cracked, that is recorded with the reason. If a result was invalidated, the original result, the investigation and the justification all remain in the record.

Records showing an unbroken series of clean, usable results from a laboratory handling real samples do not look thorough. They look curated, and an inspector reads them that way.

Second-Person Review

Because so much microbiological data is a judgement, independent review is doing more work here than in an instrument-based laboratory.

The reviewer should check the count against the calculation, the dilution factor applied, the incubation duration against the record, the media lot and its growth promotion status, the controls for that occasion, and whether the result is consistent with the historical trend for that sample point.

That final check is the one most often omitted and the most informative. A result well outside the historical distribution is worth a second look before release, even when it is within specification.

Electronic Systems

Where a LIMS or an electronic record system is used, the standard expectations apply — unique user access, no shared logins, an audit trail that cannot be disabled, and a validated system.

Two microbiology-specific points:

Transcription into a LIMS is not elimination of the paper record. If the count was written on a bench sheet and typed in afterwards, the bench sheet is the raw data and must be retained. A LIMS entry is the original only where the analyst enters directly at the point of reading.

Automated colony counters generate images and audit trails, which is a genuine data integrity advantage over manual counting and a reasonable argument for the capital where volume supports it.

Retention and Retrieval

Data integrity requires records to survive and to be findable, and microbiology generates a specific retention problem alongside the paper.

The isolate is a record. A culture archived under cryopreservation is the only way to re-examine an identification or compare against a future contamination. Discarding it destroys evidence that no document replaces — see facility flora library.

Retention periods should align to product shelf life plus the applicable regulatory record retention, and considerably longer for flora library entries where the value is historical comparison.

Retrievability matters as much as retention. Records in a box in a warehouse that nobody can index are retained and not available. An inspector asking for the bench sheet behind a specific result three years ago is testing retrieval, and a laboratory that takes two days to find it has a finding regardless of whether the record existed.

Media and reagent traceability must survive alongside. A result is only defensible if the growth promotion status of the medium used can be established, which means the media lot number on the bench sheet and the growth promotion record must both be retrievable and linkable.

Contract laboratory records are your records. Where testing is outsourced, the agreement should specify retention period, retrievability on request, and transfer of the full record set including isolates on termination.

What Inspectors Look For

• Bench sheets completed in a single pen and hand for an entire shift, suggesting later transcription.
• Results recorded in a sequence that does not match incubation times.
• Correction fluid, overwriting, or corrections without initial, date and reason.
• No record of discarded, aborted or repeated tests.
• Out-of-specification results with no corresponding investigation record.
• Media lot numbers absent, so growth promotion status cannot be linked to a result.
• Analyst identified only by a signature at the bottom of a page covering many tests.
• An implausibly clean record set from a laboratory processing difficult samples.

How DSTechnoverse Helps

Data integrity in microbiology is mostly about practice at the bench rather than about systems, and that is where we focus.

That covers bench documentation design that makes contemporaneous recording the easy path, plate photography protocols and retention, second-person review criteria including trend consistency checking, media and reference culture traceability linking every result to its controls, out-of-specification record structure, gap assessment against ALCOA+ for microbiology specifically, and analyst training on why the practices exist rather than only what they are.

We are based in Indore, Madhya Pradesh and work with laboratories across India.

Behavioural Causes, Not Just Procedural Ones

Most data integrity problems in microbiology are not deception. They are shortcuts taken by people under pressure, in a system that made the shortcut easier than the correct path.

Recognising that changes the remediation. Retraining a person who transcribed counts from a scrap sheet does not help if the plate reading station is across the room from the documentation and the analyst has forty plates to read before the shift ends.

• Workload that makes contemporaneous recording impractical is a management problem presenting as a compliance one.
• Physical layout that separates the reading position from the record encourages transcription.
• Pressure to release creates an incentive to under-report a marginal count, and the pressure is usually implicit rather than stated.
• Fear of consequence for reporting a positive discourages reporting. A laboratory where an excursion triggers blame will have fewer recorded excursions and the same number of real ones.
• Unclear procedures leave analysts making judgement calls inconsistently, which reads as unreliability rather than ambiguity.
• No deputy means the sole competent analyst works when ill and rushes when behind.

The diagnostic question worth asking is whether the correct behaviour is the easy behaviour. Where it is not, expect the shortcut, and fix the system rather than the person.

Frequently Asked Questions

Why is microbiology harder than chemistry for data integrity?

Because the data is transient, subjective and manual. A plate is read once and discarded, the count is a judgement, and most steps generate no instrument record — so the controls have to be procedural rather than systemic.

Should we photograph every plate?

Not every plate, but every plate whose result will be acted on — excursions, positives, out-of-specification counts, unusual morphology. It converts a transient observation into a durable record for very little cost.

Is a LIMS entry the original record?

Only if the analyst entered it directly at the point of reading. Where a bench sheet was completed first and typed in later, the bench sheet is the raw data and must be retained.

What if a plate is overgrown and has to be recounted?

Record both — the overgrown plate, the reason, and the recount at higher dilution. Completeness means the record shows what happened, not only what was reported.

How should corrections be made?

A single strike-through leaving the original legible, with the correct value, the initials of whoever changed it, the date, and a reason. Never overwritten, never obliterated.

Does an electronic system solve the problem?

It solves part of it and creates new obligations. A LIMS with unique logins, an uneditable audit trail and direct entry at the point of reading genuinely improves attributability and contemporaneousness. What it does not solve is the subjectivity of the count itself, and it introduces validation, access control and backup obligations that a paper system does not have. A poorly implemented LIMS with shared logins and a disabled audit trail is worse than a well-run paper system, because it carries the appearance of control without the substance.

How should we handle a result an analyst believes is wrong?

Record it, then investigate it. The temptation to discard a result that looks implausible and repeat the test is exactly the behaviour data integrity expectations exist to prevent. The correct sequence is to record the observed result, document why it is questioned, investigate through the out-of-specification procedure, and retain both the original and any subsequent result with the reasoning connecting them. An analyst's judgement that a result is wrong is a hypothesis requiring evidence, not a licence to delete.

What does an implausibly clean record look like?

An unbroken run of usable, in-specification results from a laboratory handling real samples, with no aborted tests, no recounts and no investigations. Real laboratories generate untidy records, and the absence of untidiness reads as curation.

---

The strongest microbiological programmes are the ones where the reasoning is written down. DSTechnoverse provides industrial microbiological testing with risk-based programme design, statistical trending and investigation support. Based in Indore, Madhya Pradesh, serving manufacturers across India. Talk to our microbiology team.