Healthcare Calibration
Public overview of UAIF-CAL-HEALTH-v1.0; the operational package is controlled.
Purpose and scope
Healthcare AI incident classification with sector-specific harm context and clinical/regulatory boundaries.
The package calibrates UAIF classification for sector context without replacing the UAIF Technical Specification or applicable law.
Status and validation boundary
Version 1.0 is provisional pending empirical inter-rater validation. The source document states that provisional weights must not be used as the sole basis for regulatory reporting, certification or consequential decisions.
Public information
- Document reference: UAIF-CAL-HEALTH-v1.0
- Version: v1.0 provisional
- Public scope and exclusions
- Method based on public regulatory material, research and recognised severity frameworks
- Known validation and legal-review limitations
Sector risk context
Healthcare and life-sciences AI operates where errors are directly life-threatening. Diagnostic, clinical-decision-support, medication-dosing, patient-monitoring, and surgical-robotic AI act on patients in real time, and a single model failure can cause death or serious injury before clinicians detect it. Privacy exposure is extreme because the sector processes special-category health data (GDPR Art. 9 / HIPAA). The dominant harm vectors are physical (clinical harm), rights/patient-safety integrity, and privacy; financial and reputational effects are secondary.
1.1 Applicable Regulations
• EU AI Act — Annex III high-risk; safety-component AI in medical devices (MDR 2017/745, IVDR 2017/746) [T1]
• FDA 21 CFR Part 803 medical-device reporting; FDA SaMD / AI-ML guidance [T1]
• UK MHRA vigilance and device-safety reporting [T1]
• India CDSCO medical-device and software guidance [T1]
Extracted from the source sector profile. Detailed calibration and operational material remains controlled.
Public sector orientation
Representative classification concerns include patient safety, clinical workflow, diagnostic or treatment support, privacy, vulnerable populations and professional oversight. These themes help teams scope incident intake and analysis; they are not calibration weights, legal conclusions or an exhaustive risk list.
Use with the UAIF layers
- Establish incident identity, provenance and remediation state.
- Separate cause, manifestation, acute harm and chronic-harm proxies.
- Apply severity and context logic with the provisional calibration status visible.
- Record sector and jurisdiction metadata for human review and routing.
- Preserve uncertainty, contrary evidence and the version of every referenced artifact.
Controlled resource
The detailed package, calibration weights, scoring implementation, evidence logic and maintenance material are confidential commercial resources governed by UAIF-LIC-003 and applicable GEL terms.
Notably Absent
No empirical validation result, regulatory approval, certification status, universal sector applicability or legal-reporting determination is claimed.