Use case · Cardiology · Quality documentation
TAVI documentation in 3 to 4 minutes instead of 20
Tertiary-care hospital, Germany · Cardiology · 2025 · n = patients · hospital named only with approval
- 85%
- less documentation time per patient
- From 20 minutes to 3 to 4 minutes
- Department
- Cardiology
- Hospital
- Tertiary-care hospital, Germany
- Use case
- Structured quality documentation for every TAVI patient
- Data
- 48 fields from 5 document types
- Platform layers
- Data foundation and LLM extraction
- Period and sample
- 2025 · n = patients
The problem
For every TAVI patient, the team had to fill 48 structured quality fields by hand. The information was spread across five document types. One patient took about 20 minutes of skilled clinical time.
What was implemented
The documents already flowed into the Firemetrics data foundation. On top of it, a language model extracts the 48 fields, with OCR for legacy documents and a clinical feedback loop: clinicians confirm or correct, and the extraction learns from it. The result feeds mandatory registry and quality reporting.
Systems connected and time to go live:
The measured result
Documentation time per patient fell from 20 minutes to 3 to 4 minutes, a saving of 85%.
Documentation time per patient Minutes, before and after Firemetrics
- 20 min
- After
- 3 to 4 min
- 0 5 10 15 20
Source: Firemetrics time measurement, Tertiary-care hospital, Germany, cardiology, 2025, n = patients. After: range 3 to 4 minutes.
Project phases
- 01Field catalogue and document types agreed with cardiology
- 02Extraction tuned and validated on past cases
- 03In daily use, measured against the old process
- Extracted fields 6 of 48
- Aortic valve gradient severe ✓
- Atrial fibrillation persistent ✓
- LVEF 60 % ✓
- NYHA class III Review
- STS score 4.2 % ✓
- Access route transfemoral ✓
Illustration of the extraction view with sample values. Real, anonymised screenshot.
Use cases
Related use cases
Browse all use cases
