Amazon Comprehend Medical & Transcribe Medical — HIPAA-eligible AI services for healthcare: extract medical entities from text and transcribe clinical conversations.
What are Comprehend Medical & Transcribe Medical?
Amazon Comprehend Medical is a HIPAA-eligible NLP service that extracts medical information (conditions, medications, dosages, procedures) from unstructured clinical text.
Amazon Transcribe Medical is a HIPAA-eligible speech recognition service that converts medical conversations and dictations into text with specialized medical vocabulary.
Key Insight: Both services are pre-trained on medical data, HIPAA-eligible, and require no ML expertise — just call the API.
Amazon Comprehend Medical
Key Features
| Feature | Description |
|---|---|
| Medical Entity Detection | Identifies conditions, medications, dosages, procedures, anatomy |
| Protected Health Information (PHI) | Detects PHI entities across HIPAA Safe Harbor categories (you perform redaction in your application workflow) |
| Ontology Linking | Maps entities to ICD-10-CM, RxNorm, SNOMED CT |
| Relationship Extraction | Links medications to dosages, conditions to treatments |
| Confidence Scores | Returns confidence levels for each detected entity |
| Batch Processing | Process large document collections asynchronously |
| HIPAA Eligible | Covered under AWS BAA for protected health information |
Use Cases
Clinical Documentation Extract structured data from physician notes, discharge summaries, and patient records for EHR integration.
Pharmacovigilance Identify adverse drug events from clinical trial reports and post-market surveillance.
Claims Processing Auto-extract diagnosis codes (ICD-10-CM) and procedure codes for insurance claims validation.
Clinical Research Identify patients matching clinical trial criteria by analyzing medical histories.
Revenue Cycle Management Extract billable procedures and diagnoses for medical coding automation.
Pricing & Free Tier
| Aspect | Details |
|---|---|
| Free Tier (first month) | 8.5 million characters |
| Entity Detection (NERe) | $0.01 per unit (first 1M units), $0.005 (1M-10M), $0.001 (10M+) |
| PHI Detection | $0.01 per unit (first 1M units), tiered pricing |
| ICD-10-CM Linking | $0.01 per unit (first 1M units), tiered pricing |
| RxNorm Linking | $0.01 per unit (first 1M units), tiered pricing |
| SNOMED CT Linking | $0.01 per unit (first 1M units), tiered pricing |
Note: 1 unit = 100 characters. Average 5-page chart = ~8,500 characters = 85 units.
⚠️ Pricing Disclaimer: AWS pricing is subject to change. Pricing shown is based on information available as of January 2026. Always verify current pricing at the official Comprehend Medical pricing page.
How It Works
# Extract medical entities
response = comprehend_medical.detect_entities_v2(
Text="Patient diagnosed with Type 2 diabetes. Prescribed metformin 500mg twice daily."
)
# Returns:
# Entities: [
# {"Text": "Type 2 diabetes", "Category": "MEDICAL_CONDITION", "Type": "DX_NAME"},
# {"Text": "metformin", "Category": "MEDICATION", "Type": "GENERIC_NAME"},
# {"Text": "500mg", "Category": "MEDICATION", "Type": "STRENGTH"}
# ]Amazon Transcribe Medical
Key Features
| Feature | Description |
|---|---|
| Medical Vocabulary | Pre-trained on clinical terminology, medications, procedures |
| Automatic Punctuation | Adds proper punctuation and formatting |
| Speaker Identification | Distinguishes between clinician and patient |
| Real-time Streaming | Live transcription during consultations |
| Batch Processing | Process recorded audio files |
| PHI Identification | Automatically detects protected health information |
| Specialty Support | Optimized for primary care, cardiology, neurology, oncology, radiology, urology, OB-GYN, pediatrics |
| HIPAA Eligible | Covered under AWS BAA |
Use Cases
Clinical Documentation Transcribe doctor-patient conversations in real-time for automated note-taking and EHR entry.
Medical Dictation Convert physician dictations into text for patient charts, reports, and prescriptions.
Telemedicine Automatically transcribe virtual consultations for documentation and compliance.
Medical Transcription Services Replace manual transcription with automated, cost-effective solution.
Quality Assurance Transcribe patient calls for training and compliance monitoring.
Pricing & Free Tier
| Aspect | Details |
|---|---|
| Free Tier (first 12 months) | 60 minutes/month |
| Batch Transcription | $0.075 per minute ($4.50/hour) |
| Streaming Transcription | $0.075 per minute ($4.50/hour) |
| Minimum Charge | 15 seconds per request |
Cost Comparison:
- Standard Transcribe: $0.024/min
- Transcribe Medical: $0.075/min (3.1x more expensive)
- Premium reflects specialized medical vocabulary and HIPAA compliance
⚠️ Pricing Disclaimer: AWS pricing is subject to change. Pricing shown is based on information available as of January 2026. Always verify current pricing at the official Transcribe pricing page (Medical section).
How It Works
# Start medical transcription job
response = transcribe_medical.start_medical_transcription_job(
MedicalTranscriptionJobName='consultation-123',
LanguageCode='en-US',
MediaFormat='mp3',
Media={'MediaFileUri': 's3://bucket/audio.mp3'},
OutputBucketName='bucket',
Specialty='PRIMARYCARE', # or CARDIOLOGY, NEUROLOGY, etc.
Type='CONVERSATION' # or DICTATION
)Comprehend Medical + Transcribe Medical Pipeline
Common Workflow:
- Transcribe Medical → Convert doctor-patient audio to text
- Comprehend Medical → Extract structured medical data from transcript
- Integration → Push structured data to EHR, billing system, or analytics
Audio Recording
↓
Transcribe Medical (speech → text)
↓
Comprehend Medical (text → entities)
↓
Structured Data (conditions, meds, procedures)
↓
EHR / Billing System
When to Use Medical Services
| Use | Don’t Use |
|---|---|
| Healthcare applications with PHI | Non-medical applications (use standard services) |
| HIPAA compliance required | Public data or non-regulated industries |
| Medical terminology | General domain text/speech |
| Clinical documentation | Consumer health content |
Medical vs Standard Services
Comprehend Medical vs Comprehend
| Aspect | Comprehend Medical | Comprehend |
|---|---|---|
| Training Data | Medical text | General text |
| HIPAA | ✅ Eligible | ✅ Eligible |
| Entities | Medical-specific | General (people, places, orgs) |
| Pricing | ~10x more expensive | Standard pricing |
Transcribe Medical vs Transcribe
| Aspect | Transcribe Medical | Transcribe |
|---|---|---|
| Vocabulary | Medical terminology | General vocabulary |
| HIPAA | ✅ Eligible | ✅ Eligible (with BAA) |
| Pricing | $0.075/min | $0.024/min |
| Specialty Support | 8 medical specialties | None |
Important Notes
Both Services:
- HIPAA Eligible: Covered under AWS Business Associate Agreement (BAA)
- No Data Retention: AWS doesn’t store your audio or text after processing
- Not a Substitute: These services assist clinicians but don’t replace medical judgment
- US English Only: Both services support only English (US) currently
- Confidence Scores: Review low-confidence outputs before clinical use
Comprehend Medical Specific:
- Not De-identification: PHI detection doesn’t meet HIPAA de-identification requirements alone
- Requires Review: Use human review (A2I) for critical applications
- No Training: You cannot train custom models — pre-trained only
Transcribe Medical Specific:
- Not Real-time Enough: ~2-5 second latency — not suitable for live captions in surgery
- Specialty Matters: Choose the right specialty for best accuracy
- PHI in Transcripts: Transcripts may contain PHI — handle accordingly
Integration with Other AWS Services
AWS HealthScribe: Single managed API for ambient clinical documentation workflows
Amazon A2I: Add human review loops for low-confidence entities or transcriptions
Amazon S3: Store audio files (Transcribe Medical) and documents (Comprehend Medical)
AWS Lambda: Build automated processing pipelines
Amazon SageMaker: Train custom models on top of extracted entities
TL;DR
Comprehend Medical
- What: Medical NLP service for extracting entities from clinical text
- Features: Medical entities, PHI detection, ontology linking (ICD-10-CM, RxNorm, SNOMED CT)
- Free Tier: 8.5M characters first month
- Pricing: ~$0.01 per unit (100 chars), tiered discounts
- Best for: EHR data extraction, clinical research, claims processing
Transcribe Medical
- What: Medical speech-to-text service for clinical audio
- Features: Medical vocabulary, 8 specialties, speaker ID, PHI detection
- Free Tier: 60 minutes/month for first 12 months
- Pricing: $0.075/min (3x standard Transcribe)
- Best for: Clinical documentation, medical dictation, telemedicine transcription
Together
- Powerful Pipeline: Audio → Text → Structured Medical Data
- HIPAA Compliant: Both covered under AWS BAA
- AWS HealthScribe: All-in-one solution combining both services + generative AI
Resources
Amazon Comprehend Medical Official product page and overview.
Amazon Transcribe Medical Official product page and overview.
Comprehend Medical Pricing Detailed pricing breakdown.
Transcribe Medical Pricing Detailed pricing breakdown (scroll to Medical section).