write-report-with-ehr

write-report-with-ehr

Combine the patient record with descriptor measurements, then write a comparison-aware report. This skill writes the report itself; it does not delegate to write-report-simple.

How to use

1. Look up the record. Invoke lookup-patient-history with image_path. If it returns "Not on file," continue without history and write COMPARISON: None available.
2. Pick the prior image. Use the most recent study in IMAGING HISTORY, before the current study, that has a retained image. Resolve the current and prior image paths to absolute paths.
3. Measure. Invoke write-imaging-factsheet with the current image and that prior image, or the current image alone if no prior image is retained. Keep its factsheet and each scan's audit_dir.
4. Write the report in the format below.

Reading the change

• Check what happened between the prior image and the current exam. If PROCEDURES or TREATMENTS fall between them, such as a resection, the factsheet's change lines compare across that event. Say so, and do not describe the change as growth of the same lesion. Compare with later summary-only studies, such as a post-operative baseline, in words.
• Read change against the treatment timeline. After radiation, enhancement and edema can increase without tumor growth (pseudoprogression or radiation necrosis). When the record allows both treatment effect and tumor, say that imaging alone cannot separate them; do not call progression outright.
• Expected findings are expected. Describe items from EXPECTED POST-TREATMENT FINDINGS in one sentence; do not call them new.
• Answer the order. The IMPRESSION addresses the question in the current study's pending order.
• Medications matter. Mention steroids when describing edema or mass effect.

Sources

• Measurements and change come from the factsheet only, copied verbatim with units. Do not compute, round, or convert numbers. Report on all available measurements, both volumetric and linear.
• Values from summary-only studies, such as midline shift on a CT, are quoted with that study's date; the tools did not measure them.
• Location and side come from the record or your own read of the image; the descriptors assign none.
• Do not invent priors. A finding absent from every prior report and summary is new.
• If LAST RECORD ENTRY NOTE says the record stops before the current exam, do not assume what happened in the gap.
• The report must not mention tools, models, masks, descriptors, or the record's JSON.

Output format

Return the two trace sections first, then the report block last.

## Prior information used

Up to 6 one-line bullets: the record fields and prior studies you relied on, including the prior image you measured against and any procedure between it and the current exam.

## Clinical details used in the report

Up to 10 one-line bullets, one per substantive report phrase, each paired with its source: a record field (`diagnosis`, `procedures`, `medications`, `pending_orders`, `prior report _NNN`, `study _NNN summary`) or a factsheet descriptor ID (`[L1.bidimensional]`). Flag anything with no source.

<confidence>{confidence}</confidence>
<report>
EXAM: {exam from the current pending order}

INDICATION: {indication from the current pending order}

COMPARISON: {the prior image's exam, date, and study_id; then other relevant studies by date and study_id}

TECHNIQUE: {series, plane, and contrast; or "MRI of the brain; sequence details not provided."}

FINDINGS:
{One or two prose paragraphs: the lesion and its change first, with what happened in between; then edema and mass effect; then expected post-treatment findings in one sentence.}

IMPRESSION:
1. {Answer to the order, with the change in context.}
2. {Next finding, if needed.}
</report>

• {confidence}: low, medium, or high, from how clearly the factsheet and the record support the findings.
• The report is plain text, at most 250 words: no markdown, bullets, or bold. Impression numbering is literal text.