Medical Device DHF Traceability Audit

What it is

For a medical-device client, 46 controlled design history file documents (requirements, test protocols, executed records, hazard analyses, FMEAs, a cybersecurity worksheet) were pulled from Drive and parsed into a relational SQLite database of 506 requirements, 362 trace links and 1,824 test steps. Automated checks surfaced 200 findings such as orphans, wording drift, verification closed without evidence and citations to sections that do not exist, each traceable to a document, table and row. Deliverables include an offline single-file HTML explorer, a final report and a Phase 1 impact analysis.

The problem

Regulated documentation spread across dozens of Word, Excel and PDF files made it impossible to see which requirements were actually verified.

How it works

1. Python scripts extract .docx, .doc, .xlsx and PDF sources into structured tables (extract, build_db, records, cyber, versions).
2. analyze.py runs bidirectional trace agreement, orphan, wording-drift, design-section and verification-evidence checks and stores findings as data.
3. consistency.py fails the build if any published number disagrees with the database or if the explorer contains a network call; the explorer is one self-contained HTML file because the content is confidential.
4. Findings were re-attacked by independent reviewers working from the raw documents, which corrected fifteen factual errors and cut one headline claim from 7 broken design references to 1.

Stack

Disciplines

For your business

Document digitization and audit pipelines that turn a pile of controlled documents into a queryable database with checked findings.