| name | chart-review-rucam |
| description | Score RUCAM (Roussel Uclaf Causality Assessment Method) for drug-induced liver injury (DILI) from EHR structured data + clinical notes. Triggers on: RUCAM, DILI causality, drug-induced liver injury, hepatotoxicity, R ratio, time to onset, rechallenge, liver injury scoring.
|
Procedure
RUCAM has 7 scored items whose scores sum to a total that maps to a causality
category (highly probable / probable / possible / unlikely / excluded).
You do NOT score the items. You extract the small components each item is
built from; the platform derives every item score, the total, and the category
deterministically from your components. Never call set_field_assessment on an
item_*, rucam_total_score, or rucam_causality_category field — those are
computed. list_criteria shows you exactly the component (leaf) fields to answer.
Several components/items depend on injury type (hepatocellular / cholestatic /
mixed) from the R ratio (ALT/ULN ÷ ALP/ULN) — extract that first.
Per-item logic (which component values map to which score) lives in
references/scoring/item-N-*.md — consult it for context, but commit the
components, not the score.
Tools
- Structured data (RUCAM plugin):
get_patient_summary, get_suspect_drug,
get_medications, get_drug_episodes (45-day-gap merge), get_lft_series,
get_lab_extremum, get_serology, get_conditions,
get_hepatotoxicity_category, compute_r_ratio. The patient (person_id) and
data dir are pre-bound — call these WITHOUT person_id/data_dir.
- Notes:
list_notes / read_note (platform MCP, faithfulness-gated).
- Write: one
set_field_assessment per component below (with a brief
rationale + evidence); set_review_status when every applicable component is done.
Components to extract
Read each field's criterion (read_criterion / list_criteria) for its exact
allowed values and evidence rules, then commit it. Grouped by the item it feeds:
Injury type (feeds items 1–3)
injury_track — from compute_r_ratio (hepatocellular / cholestatic / mixed).
Item 1 · time to onset
onset_path — initial_treatment / re_exposure / from_cessation / not_calculable (get_drug_episodes).onset_latency_days — integer days (drug-start→injury for path A; stop→injury for path B).
Item 2 · course after stopping
dechallenge_outcome — the anchor lab's course after the drug stopped (get_lab_extremum).
Item 3 · risk factors
rf_alcohol, rf_pregnancy, rf_age_ge_55 — each yes / no.
Item 4 · concomitant drugs
concomitant_worst_timing — suggestive / compatible / incompatible / none.concomitant_worst_hepatotoxic — yes / no (get_hepatotoxicity_category on the worst co-drug).concomitant_attribution — yes / no (is a co-drug clearly the cause?).
Item 5 · exclusion of other causes — one flag per cause, yes only if ruled out
by a negative test or an explicit note exclusion (per references/scoring/item-5-exclusion.md):
- Group I (6):
g1_hav_ruled_out, g1_hbv_ruled_out, g1_hcv_ruled_out, g1_biliary_obstruction_ruled_out, g1_alcoholism_ruled_out, g1_ischemia_ruled_out.
- Group II (5):
g2_autoimmune_ruled_out, g2_sepsis_ruled_out, g2_chronic_hbv_hcv_ruled_out, g2_pbc_psc_ruled_out, g2_cmv_ebv_hsv_ruled_out.
alt_cause_explains — yes / no (does a non-drug cause sufficiently explain the injury? the −3 override).
Item 6 · prior hepatotoxicity
hepatotoxicity_class — labeled / probable / none (get_hepatotoxicity_category on the suspect drug).
Item 7 · rechallenge
rechallenge_result — positive_alone / positive_with_codrug / below_uln / none_or_insufficient.
Steps
- Read
references/scoring/item-0-setup.md; call compute_r_ratio and commit injury_track.
- Work through the components above, gathering evidence with the tools (+ notes) and
committing each via
set_field_assessment. Answer every applicable component —
for the Group I/II exclusion flags, that means one flag per cause even when a cause
was not assessed (no), because a missing flag leaves item 5 Pending.
set_review_status when done. The item scores, total, and category compute
automatically from your components.
