Pathology Report Extraction: Fields to Capture

Pathology report extraction turns a pathology report into a structured record per specimen: identifiers, procedure, dates, the signing pathologist, the final diagnosis text copied word for word, and any synoptic checklist items as question and answer pairs. Extraction captures what the pathologist wrote. It never summarizes, codes or interprets a diagnosis on the reader's behalf.

Pathology report fields to capture for every specimen

The pathology report fields to capture for every specimen fall into three groups: who and where (patient, accession number, performing lab), what (specimen label, site and procedure as printed), and when and by whom (collection, receipt and sign-out dates, and the signing pathologist). One report can hold several specimens, so each specimen becomes its own record.

Some of these fields come straight from the CLIA test report rule that covers every lab report, which the guide to lab report header fields and the CLIA report elements lists item by item. For tissue work, 42 CFR 493.1273 on histopathology adds that tissue pathology reports must be signed by a qualified individual, and that labs must use acceptable terminology of a recognized system of disease nomenclature.

A practical core record for one specimen, with fictional example values:

Keep the specimen label as the lab prints it. A report with parts A, B and C is three records sharing one accession number, and the label is what ties each diagnosis back to its specimen.

Dates are easy to swap. Collection, receipt and sign-out often sit close together in the header, and they can fall on different days. Store all three with their printed labels rather than a single "date" field.

What are synoptic pathology report data fields?

Synoptic pathology report data fields are the items in a structured checklist section, printed as a label followed by a chosen answer, such as "Histologic Type: Adenocarcinoma". They differ from narrative text because each item is already separate. Extraction copies each label and answer as a pair, keeping the order and wording the report uses.

The checklist content usually comes from published protocols. The College of American Pathologists says its cancer protocol templates provide guidelines for collecting the essential data elements for complete reporting of malignant tumors, and that synoptic reporting is done using the CAP electronic Cancer Protocols. Its release notes mark some elements as required ("core"), and some answer lists include options such as "Not Performed".

The table below shows a fictional synoptic section captured as question and answer pairs. Every value is made up for the example and copied exactly as it would print.

Rules for capturing synoptic items:

• Copy the question text too. Labels change between protocol versions, and the label tells a later reader which question the answer belongs to.
• Keep answers as text. "14" is a count the report printed, so it can also go in a number field, but keep the text beside it. "Cannot be determined" is a real answer, not a blank.
• Keep nested items nested. Some items have sub-items, such as a distance under a margin item. Store the parent and child together.
• Record the protocol version when the report prints it, since items differ across versions.

Synoptic sections are the easiest part of a pathology report to extract, because the structure is on the page. The narrative parts are harder, and are covered next.

Narrative pathology report extraction: final diagnosis text and gross description

Narrative pathology report extraction captures free text, mainly the final diagnosis and the gross description, as complete blocks of text tied to their specimen. The rule is to copy, not condense. A summary of a diagnosis is a new statement that the pathologist did not sign, so the extracted field holds the exact wording, line breaks included.

The final diagnosis section is usually organized by specimen label. A fictional example:

A. Colon, right, resection: Adenocarcinoma. See synoptic report. Fourteen lymph nodes, negative (0/14).

B. Appendix: No diagnostic abnormality.

Each lettered paragraph becomes the diagnosis text for that specimen's record. Splitting on the label is safe; splitting inside a paragraph is not. If a diagnosis says "See synoptic report", store the sentence as written and link the specimen's synoptic pairs to the same record. Do not copy synoptic answers into the diagnosis field.

The gross description is the lab's account of the specimen as received: what arrived, how it was labelled, its size and how it was sampled into blocks. Capture it as one text field per specimen, plus the block key if the report prints one, such as "A1-A4: tumor; A5: margin". The block key is often useful to pull into a small table, because it ties slides back to parts of the specimen, but keep the original text too.

Other narrative sections to capture as text:

• Microscopic description, when present.
• Comments and notes, which may refer to other cases or to tests sent elsewhere.
• Clinical history as supplied with the request, labelled as such so nobody mistakes it for the lab's finding.
• Intraoperative consultation results, which some reports print with their own time stamp.

Turning narrative text into coded concepts is clinical abstraction, a different job with its own training and rules. The clinical records category covers free-text abstraction in general. Extraction stops at faithful text with a link to where it sits on the page.

How does pathology report data reach research registries?

Pathology report data reaches cancer registries through electronic reports sent directly from laboratories to state central cancer registries. The CDC's National Program of Cancer Registries describes that path: labs send HL7 version 2.5.1 messages, following the NAACCR Volume V standard, and registries receive them through CDC tools and shared reporting platforms.

The CDC page on advancing electronic reporting to cancer registries gives the details. It lists the ability to report using HL7 2.5.1 with an SPM segment as one of the items a lab is assessed on, and says that guidance for electronic reporting of anatomic pathology reports from labs to state cancer registries comes from NAACCR Volume V, and that NPCR built eMaRC Plus to receive and process cancer data from labs to state registries. It describes California using CAP electronic Cancer Checklists to keep data consistent, and the AIMS cloud platform as a single point for reporting.

This matters for anyone handling lab report data for research registries from paper or PDF. The path CDC describes runs on structured HL7 messages and checklist content, not page images. If you extract pathology reports for a registry or a research project, three habits keep the data usable:

• Keep the specimen as the unit of record, with the accession number and specimen label on every row.
• Keep synoptic items as question and answer pairs, with the protocol version, so they can be matched to the registry's expected items.
• Keep the original diagnosis text, so a registry abstractor can code it under their own rules.

Registries and research use also raise privacy questions, since pathology reports carry identified patient data. The rules that apply depend on who is reporting and why, and this post does not cover them.

Some fields on pathology-adjacent lab reports, such as biomarker results with numeric values, also carry units and reference ranges. The guides to extracting reference ranges from lab reports and normalizing lab result units with UCUM cover those fields.

Pathology report extraction pitfalls: addenda, amended reports and multiple specimens

Pathology report extraction goes wrong most often on three edge cases: addenda that add to a signed report, amended reports that replace part of one, and multi-specimen reports where text drifts onto the wrong specimen. Each one changes which record a piece of text belongs to, so each needs a rule decided before extraction starts.

Addenda. An addendum adds information after sign-out, such as results of a test that finished later. It does not erase the original report. Store it as a separate dated text block attached to the same accession number, with its own signer and date, and keep the original diagnosis unchanged.

Amended reports. A corrected or amended report replaces earlier content. CLIA's rule on corrected reports in 42 CFR 493.1291 says that when errors in reported results are detected, the lab must issue corrected reports promptly and maintain duplicates of the original report as well as the corrected report. Mirror that in your data: keep both versions, mark which one is current, and never overwrite the original record in place.

Multiple specimens. A report with parts A through D can put a long diagnosis for A over two pages, with B starting mid-page. Text from A's continuation can land on B. Check that every specimen label in the header has exactly one diagnosis paragraph, and that every diagnosis paragraph starts with a label that exists.

Two smaller traps are worth a rule each:

• Outside consultations. A report may quote another lab's diagnosis. Store it as quoted text with the other lab's name, never as this lab's finding.
• Duplicate deliveries. The same report can arrive by fax and by portal. Match on accession number, status and sign-out date before creating a second record.

These are also the cases to route to a person. Pathology reports are not one of the document types Schema Care lists; it reads lab reports and medical records, scoring each field and sending uncertain ones to review. It accepts no real patient data until a business associate agreement is signed, as the compliance section on the homepage sets out with the status of each safeguard.