Entering a requisition into a laboratory information system is not clerical work in the regulator's eyes. It is transcription of a medical order, and CLIA treats it accordingly. Laboratories that automate any part of it — or that key it by hand and assume the manual step is self-evidently safe — remain accountable for showing that what reached the LIS matches what the ordering provider wrote.
This article covers what 42 CFR §493.1241 says, which parts bear on data entry, and what evidence looks like when a surveyor asks for it. It is written for compliance leads and laboratory directors, and it is deliberately specific about citations so every claim can be checked against the regulation itself.
What §493.1241 requires
The section is titled Standard: Test request, and it governs the requisition and the information the laboratory must have before testing. Two parts matter most for accessioning.
§493.1241(c) enumerates the information the test request must solicit. In practice this is the checklist a requisition form has to satisfy: the patient's name or unique identifier; the sex and age or date of birth; the name and address or other suitable identifier of the authorized person ordering the test; the test to be performed; the specimen source where applicable; the date and, where relevant, the time of specimen collection; and any additional information relevant to a specific test. Subsection (c)(7) adds conditional elements for Pap smears — the patient's last menstrual period, indication of pregnancy, and prior cytology history — which is why Pap requisitions cannot be validated with the same rule set as a chemistry panel.
§493.1241(e) is the part most often overlooked. It requires the laboratory to ensure the accuracy of test requisition information when it transcribes or enters that information into a record system or a laboratory information system. That is an explicit, standing obligation about data entry accuracy — not about the form, and not about the assay.
Read together: (c) tells you what must be on the requisition, and (e) tells you the laboratory must be able to demonstrate that what got typed in matches it.
Why this is the step that gets examined
Accessioning sits between two systems of record — the provider's order and the laboratory's LIS — and errors introduced there propagate silently. A mistyped date of birth becomes a patient-matching failure. A mis-keyed test code becomes a wrong result on the right patient, or a claim that does not match the order. A missing collection time invalidates a timed specimen. None of these are caught by analytical QC, because the analyzer performs perfectly on the sample it was given.
That is also why "our accessioners are careful" is not an answer to (e). Care is not evidence. The regulation asks the laboratory to ensure accuracy, which implies a mechanism that produces something checkable.
What usable evidence looks like
No format is prescribed, which is exactly why laboratories struggle to produce it. In our experience the evidence a surveyor finds satisfying has four properties.
It compares orders to the source image, not to itself
A report showing that the LIS record is internally consistent proves nothing about transcription. The comparison has to be against the requisition as received — the scanned or faxed page — field by field. That means the source image has to be retained and linked to the order record, not discarded after keying.
It is sampled continuously, not assembled before the survey
A QA process that runs once a year in the two weeks before an inspection describes the two weeks before an inspection. A defensible process samples orders on an ongoing basis and produces agreement rates over time, so the record shows the state of the laboratory rather than the state of its preparation.
It reports disagreements, not just agreement
An exception log — what disagreed, on which field, what the correction was, and who made it — is more persuasive than a high agreement rate alone. A rate with no exceptions attached invites the question of whether anything was actually checked.
It maps required elements by citation
When a validation rule says "date of birth is required," the useful version of that rule also says why, pointing at §493.1241(c) rather than at internal policy. It converts "how do you know this requisition is complete?" from a discussion into a lookup.
Where automation helps, and where it makes things worse
Automation is well suited to producing this evidence, because a system that reads a form can record what it read, from where, and with what confidence. The audit trail becomes a by-product of the work rather than an extra task. Manual keying produces no such record unless someone builds one deliberately.
But automation makes the picture worse in one specific circumstance: when the system fills in fields the form did not contain. A model that infers a diagnosis code, normalizes a test name to the nearest billable equivalent, or completes a truncated identifier is generating order content, and no transcription QA process can validate a value that was never on the page.
This is why our own rules are absolute on the point. Diagnosis codes, test selections, and ordering-provider identifiers are transcribed only. A blank field stays blank. An unreadable one goes to a person who reads the image.
The same logic drives the decision to read checkboxes with classical computer vision rather than a language model. A checkbox decision made by measuring pixel density in a registered region can be rechecked by hand, with the numbers behind it stored alongside the result. A checkbox decision made by a model can only be re-run, which is not the same thing when a surveyor asks how you know.
A practical starting point
If you are building this from nothing, three steps get you most of the way. First, retain the source image and link it to the order record — everything else depends on it. Second, define a sampling rate and start comparing, even at low volume, so that agreement rates accumulate history. Third, write required-field rules against the citation rather than against habit, including the conditional (c)(7) elements for cytology.
Nexession produces all three as a by-product of how it processes a requisition. The QA evidence page shows what the output looks like, and a pilot runs it against your own forms with the measures agreed in writing beforehand.
This article describes our reading of 42 CFR §493.1241 and is not legal or regulatory advice. Verify requirements against the current text of the regulation and your accrediting organization's checklist.