Synoptic Reporting
Structured, checklist-based cancer reporting where every required data element appears as a named field — which improves completeness and, as a side effect, makes reports machine-readable.
What it is
A narrative pathology report is prose: complete in the hands of a careful pathologist, variably complete otherwise, and hard to parse reliably. Synoptic reporting replaces or supplements prose with an explicit list of required data elements — tumour type, grade, size, extent, margins, node counts, lymphovascular invasion, and so on — each reported as a labelled field.
The two dominant standards are CAP (College of American Pathologists) protocols and ICCR (International Collaboration on Cancer Reporting) datasets, organ-specific and versioned. Version matters: required elements change between editions, so a compliance check is always against a stated edition.
The benefits divide into two kinds. Clinically, completeness improves measurably — required fields cannot be silently omitted, which is the main argument for the format. Computationally, a synoptic report is already structured data, so downstream extraction becomes reliable rather than probabilistic. That second benefit is what makes synoptic reporting a prerequisite for most report-mining work rather than merely a quality initiative.
The honest counterpoint: rigid templates can fit unusual specimens badly, and a field can be filled in without being considered. Completeness is necessary, not sufficient.
What it looks like at national scale
Siebers 2026 — The Dutch Nationwide Pathology Databank (Palga) is the largest documented national implementation, and it turns this page from a description of a format into a set of numbers about adoption, which is where the real difficulty lives.
The Netherlands connects all 39 of its pathology departments to one infrastructure. Its standardised structured reporting (SSR) protocols began with colorectal and breast resections in 2009, moved onto a dedicated protocol module in 2013, and have been CE-certified since 2019. As of mid-2026 there are 34 operational protocols covering nearly every tumour type with a national treatment guideline, plus a few non-oncological ones such as placental pathology and urine cytology. Protocols are built from ICCR datasets plus national guidelines and UICC TNM, with mandatory core elements and department-configurable non-core ones, and each item within a protocol is bound to SNOMED CT — not just the report as a whole.
The adoption numbers are the part worth carrying:
- more than 900,000 SSR reports in 2025;
- more than one third of all pathology reports;
- use is not mandatory in general, but it is mandatory for colon biopsies and cervical smears in the population screening programmes, and strongly recommended by national guidelines for lung, endometrial, breast, ovarian and prostate cancer.
So a mature national programme, seventeen years in, with certification and guideline backing, sits at about a third. Read the other way, roughly two thirds of reports are still narrative, and adoption varies by department and by protocol — which means any dataset built from structured fields is drawn from a non-random subset. That is a coverage problem in the sense of Coverage Bias and the source does not quantify it.
The named obstacles are not technical, and this is the most useful thing on the page for anyone proposing synoptic reporting locally: clinician resistance to changing established reporting routines, incompatibility with existing workflows, and the perception of increased workload. Tooling does not address any of the three.
One claim is worth reading but not repeating as established: the same source attributes improved patient outcomes, including survival, in colorectal cancer to the introduction of SSR, citing Sluijter 2019 (JCO Clinical Cancer Informatics). That is a pointer, not an appraisal — the paper is not read here, and a before-and-after national comparison is exactly the design where such a claim is won or lost. [unverified]
Why it matters for my work
This is the target standard for the group’s report-quality tooling. Pathology Report Checker Skills scores synoptic elements for CAP/ICCR compliance, and pathology-report-colon-QA does narrative-to-synoptic gap analysis for colon resections — that is, measuring how much a free-text report would have contained had it been synoptic.
It also sits upstream of Extract Report Text: the more synoptic the source corpus, the less the extraction pipeline has to infer, and the lower the silent error rate feeding downstream cohorts.
How it connects
Pathology Report Checker Skills — the tooling that operationalises compliance scoring against these standards.
Report Text Extraction — the method that has to work harder the less synoptic the source report is.
Interobserver Agreement — structured reporting reduces disagreement by removing ambiguity about what is being asked, which is one of its main justifications.
LLM Report-Quality Checking — the LLM layer that scores and rewrites reports against these standards.
Siebers 2026 — The Dutch Nationwide Pathology Databank (Palga) — the national implementation this page’s adoption figures come from, and the source of the one control that actually forces the format: a validator that blocks report authorisation until the coded summary is well-formed.
Coverage Bias — if only a third of reports are structured and adoption varies by department and protocol, a dataset of structured fields describes a self-selected subset, not the caseload.
Open questions
- Which CAP/ICCR editions does the checker validate against, and is the edition recorded in its output? Editions change; an unversioned compliance score ages badly. [unverified]
- What proportion of Memorial’s cancer reports are currently synoptic? That baseline would make the quality work measurable.
- Does the checker handle Turkish-language reports, given that the group’s corpus is bilingual? Not recorded. [unverified]