# Every AI output logged in the record with input hash, version and clinician response

Source: https://onco.cc/ideas/idea-data-ai-audit-trail-in-ehr/  
OnCo record `idea-data-ai-audit-trail-in-ehr` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Whenever an AI tool gives a result about a patient, the hospital system would permanently record what it saw, which version it was, what it said and what the doctor did with it.

## Summary

Most AI outputs are transient and not stored, making retrospective audit, harm investigation and performance measurement impossible. The proposal is a standard for AI audit trails in the EHR (FHIR resources capturing model identifier and version, input references and hashes, output, confidence, timestamp, and the clinician's acceptance or override), required for all deployed cancer AI and feeding post-market performance reporting.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: Universal audit trails will make post-market performance measurable at near-zero marginal cost and will allow root-cause analysis of AI-related harms that are currently unexplainable.
- Rationale: Aviation's flight recorders and pharmacy's dispensing logs made safety investigation and improvement possible; AI in care has no equivalent record.
- Proposed test: Implement the audit trail standard at five sites; demonstrate quarterly performance reports derived from it; test its use in retrospective review of ten discordant cases.
- Maturity: early-clinical
- Actor: engineering

## Sources

- Bottleneck evidence (AI that is built but not validated or deployed): Wu et al., How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals (Nature Medicine 2021): https://doi.org/10.1038/s41591-021-01312-x

## Connected records

- ideas: [Mandatory post-market performance reporting for cancer AI](https://onco.cc/ideas/idea-data-ai-post-market-performance-reporting/)
- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Data silos](https://onco.cc/bottlenecks/b-data-silos/)
- key papers: [How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals](https://onco.cc/key-papers/paper-wu-nat-med/)

---
JSON: https://onco.cc/api/v1/entities/idea-data-ai-audit-trail-in-ehr.json