Hematology / oncology AI scribe 2026: chemo regimen tracking, transfusion logs, CTCAE toxicity, MASCC, and the multi-cycle workflow

May 9, 2026 · 7 min read

Hematology / oncology is one of the highest-stakes documentation environments in medicine. Each visit lives inside a multi-cycle treatment plan, where drift in dosing, missed toxicity grading, or an undocumented transfusion can change downstream care. The visit note is also where billing, prior authorization, clinical-trial eligibility, and quality-measure reporting all converge. Generic ambient scribes capture conversation but produce notes that miss the structure heme-onc actually runs on: cycle-and-day, planned vs received dose, CTCAE v5 toxicity tables, transfusion logs, MASCC risk on febrile neutropenia, and the infusion-suite handoff that closes the loop with nursing.

The 2026 heme-onc-aware AI scribe stack handles five things general scribes do poorly: regimen tracking (cycle / day / planned vs received dose with reductions), CTCAE v5 toxicity grading and dose-modification reasoning, transfusion documentation per AABB standards, MASCC risk stratification for febrile neutropenia, and the OncoEMR / iKnowMed / Aria-aware schema that aligns with the platforms heme-onc practices already run on.

Visit-type adaptation

Heme-onc has five core visit types:

• New patient consultation — comprehensive history, prior treatments, performance status (ECOG / Karnofsky), pathology review, staging, treatment plan with informed consent.
• Treatment / cycle-day visit — toxicity review, labs, fitness for cycle, dose reduction or hold reasoning, infusion orders.
• Surveillance / off-treatment — remission status, late effects, secondary cancer screening, survivorship issues.
• Acute symptom (febrile neutropenia, pain crisis, etc.) — MASCC scoring, admission vs outpatient algorithm.
• Hematology benign (anemia, cytopenias, coagulopathy) — broad differential, bone marrow indication if needed, transfusion planning.

The scribe should detect visit type from context and apply the appropriate schema.

The heme-onc system prompt

You are documenting a hematology / oncology encounter.

INPUT:
- Encounter audio transcript
- Patient profile: cancer type, stage, biomarkers, current regimen with cycle/day,
  prior toxicity log, baseline labs, performance status (ECOG/Karnofsky)
- Clinical-trial enrollment status
- Open quality measures (OP-30, OCM, OnctTrim, OQR)

DETERMINE visit type, then apply schema:

For new patient consultation:
1. Cancer history (diagnosis, biopsy, staging, biomarkers — ER/PR/HER2, EGFR,
   ALK, KRAS, MSI, TMB, BRCA, etc.)
2. Prior treatments with response and toxicity
3. Performance status (ECOG 0-4)
4. Comorbidities affecting treatment selection (cardiac, renal, hepatic, neuro)
5. Geriatric considerations if ≥65 (G8, CGA components)
6. Treatment plan: regimen name (e.g., FOLFIRINOX, R-CHOP, CHOP-21), cycles
   planned, palliative vs curative intent
7. Informed consent topics covered
8. Trial eligibility consideration

For treatment / cycle-day visit:
1. Cycle number and day (e.g., C3D1 of FOLFIRINOX, C5D1 of R-CHOP)
2. Interval toxicity by CTCAE v5 (each event: term, grade, attribution)
3. Performance status today
4. Today's labs vs prior nadir
5. Fitness for treatment: yes / hold / dose reduce — with reasoning
6. Planned dose vs delivered dose with % reduction if applicable
7. Anti-emetic and supportive care plan
8. Infusion / chemo order set with route and rate
9. Next cycle planning and follow-up timing

For acute febrile neutropenia:
1. Temperature and neutrophil count
2. MASCC score (burden of illness, hypotension, COPD, solid tumor / no fungal,
   dehydration, outpatient status, age)
3. MASCC interpretation: low risk (≥21) outpatient candidate vs high risk
4. Empiric antibiotic plan
5. Cultures sent
6. Disposition plan

For transfusion documentation (when applicable):
1. Indication and lab thresholds met
2. Product (PRBC / platelets / FFP / cryo) and dose
3. Pre-transfusion vitals and verification per AABB
4. Reaction monitoring plan
5. Post-transfusion lab plan

For surveillance:
1. Disease status (NED, stable, partial response, progression)
2. Imaging review with RECIST 1.1 if applicable
3. Late effects (cardiac, neuro, fertility, second malignancy screen)
4. Survivorship care plan elements

CTCAE v5 vocabulary (consistent grading):
- Grade 1 = mild
- Grade 2 = moderate
- Grade 3 = severe
- Grade 4 = life-threatening
- Grade 5 = death related to AE
Use exact CTCAE term where applicable.

CITE transcript or labs. Surface RECIST measurements; surface dose reductions
with reasoning; flag missed quality-measure opportunities (pain assessment,
chemo plan in chart, NCCN compliance).

CTCAE v5 as the toxicity language

Common Terminology Criteria for Adverse Events v5 is the universal toxicity grading framework. Heme-onc notes that lack consistent CTCAE grading produce charts that read inconsistently across visits and across providers, and that fail audit when a dose reduction or treatment hold needs to be defended. A heme-onc-tuned scribe enforces CTCAE term and grade for each adverse event extracted from the transcript, producing an auditable toxicity table at each cycle visit.

MASCC for febrile neutropenia triage

The Multinational Association of Supportive Care in Cancer (MASCC) score stratifies febrile neutropenia patients into low-risk (score ≥21, candidate for outpatient management with oral antibiotics) vs high-risk (score <21, inpatient IV antibiotic management). Documenting the score components plus the risk-aware disposition is both clinically defensible and quality-measure-friendly.

RECIST and response documentation

For solid tumors, Response Evaluation Criteria In Solid Tumors (RECIST 1.1) governs how response is measured: target lesion sum, non-target lesion qualitative, new lesions, derived response (CR / PR / SD / PD). A scribe with RECIST awareness extracts the imaging-reported measurements and produces a response paragraph that aligns with what trial protocols and payors expect.

The oncology platform layer

Heme-onc practices typically run a community-oncology platform on top of a general EHR: OncoEMR, iKnowMed, Aria, Mosaiq. These platforms have native fields for regimen, cycle / day, toxicity, dose modification, and infusion orders. A heme-onc-tuned scribe should produce note text that maps cleanly into those fields rather than free-text that the oncologist re-types into discrete entry. The DIY stack handles this; vendor scribes generally do not.

The volume economics

A community heme-onc practice with three physicians sees 60-80 patient visits per day across new consults, treatment days, and surveillance. Audio per visit averages 15-30 min for treatment days, 45-60 min for new consults. Per-day audio: ~20-25 hours × $0.05 = ~$60-75/day with the LessRec DIY stack. Vendor subscription scribes at $200-300 per provider per month = $600-900/month for the practice. Variable cost wins meaningfully when treatment volume drops in low-census weeks (post-holiday, pandemic disruption, etc.).

Vendor and DIY paths

Vendor ambient scribes capture conversation. They underdeliver on cycle-and-day awareness, CTCAE grading, MASCC scoring, RECIST documentation, and OncoEMR / iKnowMed-aligned discrete-field output. The DIY stack — LessRec Whisper API + a heme-onc-tuned system prompt with CTCAE / MASCC / RECIST awareness + your oncology platform's discrete fields — produces audit-ready documentation across NCCN / ASCO QOPI / OCM / OQR quality programs.

BAA chain

Practice + general EHR + oncology platform (OncoEMR, iKnowMed, Aria, Mosaiq) + transcription vendor + LLM vendor.

When to start

Community heme-onc practices on quality programs (OQR, OCM, ASCO QOPI) have the strongest case for a CTCAE / MASCC / RECIST-aware DIY stack. The audit risk on dose modification reasoning and toxicity grading is real; the tooling that produces consistent vocabulary at variable cost is the right answer.