Como avaliamos o risco de alucinação da IA clínica

A prova de Notat não é uma percentagem mágica de precisão. É uma metodologia transparente: extrair primeiro os factos clínicos, gerar notas a partir desses factos, mostrar o contexto bruto ao clínico e rever cada resultado face às evidências.

Método de avaliação

1. Build the reference facts

A reviewer reads the encounter material and creates a reference set of clinical facts: symptoms, negatives, medication changes, assessment, plan, safety-netting, and coding-relevant details.

2. Compare transcript-direct output

We generate a note from transcript-style input and mark unsupported statements, missing high-salience facts, incorrect attribution, and invented certainty.

3. Compare FactsContext output

We generate documentation from extracted facts and review whether each sentence is supported by a visible fact. The clinician-facing fact list is evaluated as part of the output, not hidden.

4. Record limits, not magic numbers

We do not publish fake accuracy percentages. Each evaluation note includes dataset scope, review date, known limitations, and examples of what the system still requires clinicians to verify.

O que os revisores assinalam

Unsupported clinical assertion

Wrong medication, dose, frequency, or route

Missing red-flag negative or safety-net advice

Wrong diagnosis certainty: possible vs established

Wrong speaker attribution

Unsupported ICD-10 suggestion

Clinician-visible evidence for each key statement

Facts reusable for notes, codes, referrals, and patient instructions

Nota interna datada

Nota de avaliação atual, 2026-07-01: esta página documenta o método e os exemplos qualitativos usados para avaliar a arquitetura FactsContext. Não reivindica validação externa nem precisão universal. O próximo marco de prova deve ser uma revisão cega, estratificada por especialidade, com o tamanho do conjunto de dados, a concordância entre revisores e a taxa de afirmações não sustentadas reportados abertamente.

Extração de factos vs. geração direta a partir da transcrição

Medication change discussed twice

Risco da transcrição direta

“Increase amlodipine to 10 mg and stop lisinopril.” The transcript contained a correction: the clinician first considered stopping lisinopril, then decided to continue it after reviewing renal function.

Resultado do FactsContext

Facts: amlodipine increased to 10 mg daily; lisinopril continued; renal function normal; review in 6 weeks. Note generated from those facts only.

Porque é que importa

Transcript-direct generation can smooth over corrections. FactsContext preserves the final decision as a discrete fact before writing.

Negative finding matters

Risco da transcrição direta

“No neurological symptoms.” The actual encounter only documented no saddle anesthesia and no bladder symptoms; leg radiation was present.

Resultado do FactsContext

Facts: left leg radiation to calf; SLR positive left; no saddle anesthesia; no bladder or bowel symptoms. Note keeps the negatives specific.

Porque é que importa

Broad negative statements are risky. The fact list keeps the clinical context granular and reviewable.

Code suggestion requires evidence

Risco da transcrição direta

Suggested J44.1 for COPD exacerbation without showing the symptom or treatment evidence.

Resultado do FactsContext

Facts: increased breathlessness, purulent sputum, prednisolone burst, antibiotics started. Suggested J44.1 with those facts as evidence.

Porque é que importa

The code is easier to verify because the reason for the suggestion is visible, not buried in prose.

Limitações conhecidas

Clinician review remains mandatory. Notat drafts; clinicians verify and sign.

The method reduces unsupported statements by architecture, but no clinical AI should claim zero hallucinations.

Small internal evaluations are useful for engineering direction, not a substitute for external clinical validation.

Specialty, language, audio quality, speaker overlap, and local coding rules can change performance.

Leia a prova e depois verifique os seus próprios factos.

O método de avaliação é simples porque o produto foi concebido para ser inspecionável: factos primeiro, nota depois, revisão clínica sempre.

Experimente o Notat gratuitamente