Structured Reporting Foundations
- A structured report is just a radiology report with assigned seats — findings live under predictable headings instead of one rambling paragraph.
- The goal isn't beauty; it's that the right reader finds the right answer in the same place every time, and nothing important quietly falls off the page.
- The classic skeleton is Clinical history → Comparison → Technique → Findings → Impression, with the Impression being the part everyone actually reads.
- "Structured" lives on a spectrum: from just using consistent headings, all the way to fill-in-the-blank templates and standardized lexicons.
- Structure helps humans skim and lets computers parse — but a tidy template wrapped around a wrong observation is still a wrong report.
Imagine two ways to give someone directions. One is a stream-of-consciousness voicemail: "okay so you'll go past the gas station, oh and there's construction, anyway turn somewhere near the big tree, can't miss it." The other is a numbered list of turns. Both might get you there. Only one lets a stranger pick up where you left off. A structured radiology report is the numbered list.
What "structured" actually means
A free-text report is a single block of prose where the radiologist narrates whatever they noticed in whatever order it occurred to them. It can be excellent. It can also be a place where a tiny but critical finding gets sandwiched between two boring sentences and never surfaces again.
Structured reporting fixes the organization problem by giving findings a fixed home. At minimum, that means consistent section headings. At the fancier end, it means an actual template with labeled fields — sometimes even a controlled vocabulary so "tiny," "small," and "minimal" stop being three words for the same thing.
So think of "structured" as a dial, not a switch:
| Level | What it looks like | Trade-off |
|---|---|---|
| Consistent headings | Same sections, every report, free prose inside | Easy to adopt; still relies on the author |
| Itemized templates | Fill-in fields per organ/finding | Great completeness; can feel like a checklist |
| Standardized lexicon | Defined terms and categories (e.g. BI-RADS, LI-RADS) | Machine-readable; needs everyone trained |
Standardized reporting systems like BI-RADS (breast) and LI-RADS (liver) are structured reporting taken to its logical end: not just fixed sections, but agreed-upon terms and assessment categories so a "category 4" means the same thing in every reading room. They are real, widely used examples — not the whole of structured reporting, but its showcase.
The standard skeleton
Most reports, structured or not, follow the same bones. Knowing the order is half the battle.
- Clinical history / indication — why are we even doing this scan? The question we're being asked.
- Comparison — which prior studies (if any) we looked at. "New" vs. "unchanged" is meaningless without this.
- Technique — what was done: modality, contrast, protocol. The fine print.
- Findings — the observations, ideally organized by organ system or anatomic region.
- Impression — the synthesis. The answer to the question.
The Findings section is where structure earns its keep. Marching through anatomy in a fixed order — same regions, same sequence, every single time — is the report-writing version of a search pattern. It's the same discipline you'd use when describing a finding: say the same things in the same order so your brain can't skip a step. If you always end an abdominal CT with the bones, you stop forgetting the bones.
The Impression is the only section many referring clinicians read in full. Treat it like the back-of-the-box summary: it should answer the clinical question, flag anything urgent, and stand on its own even if no one scrolls up to the Findings.
Why bother — two audiences, one report
Structure pays off for two very different readers.
The first is the tired human at 2 a.m. who needs one number from your CT and shouldn't have to read a short story to find it. Predictable headings let people skim straight to what they came for.
The second reader is a computer. When findings live in labeled fields and use consistent words, software can actually pull data out — for registries, for tracking a nodule over time, or for feeding downstream tools. A paragraph that says "there is a smallish thing, probably nothing" is a brick wall to a parser. A field that says Nodule size: 6 mm is a doorway.
The trap: structure is not the same as substance
Here's the part that humbles everyone. A template is scaffolding, not a brain. It guarantees you mention the liver; it does not guarantee you looked at the liver.
The classic structured-reporting failure is the auto-populated normal. A template pre-fills "Liver: normal" and a busy reader leaves it untouched even though there's an obvious lesion sitting right there. The structure made the report look complete while quietly making it wrong. Always confirm every pre-filled statement actually matches the images.
There's a second, subtler cost: rigid templates can bury an unexpected finding. If something important doesn't fit any of the neat boxes — an incidental finding that has nothing to do with why the scan was ordered — it needs a clear place to land and a loud voice in the Impression, not a quiet death in a field labeled "Other."
Structure organizes the truth; it doesn't generate it. A well-formatted report built on a missed finding is still a miss — just a tidier one.
Where to go from here
Foundations are the why and the shape. The next step is the how: building and using the actual fill-in templates that put this into practice, which is its own craft. That's the focus of report templates.
The single thing to carry forward: a good report isn't the one that sounds the most impressive. It's the one where the next person — human or machine — finds exactly what they need, exactly where they expected it, and trusts that nothing was quietly left out.