generated: '2026-09-04'
method: searched
source: https://xcures.com/trust/
url: https://xcures.com/trust/
http_status: 200
title: xCures trust & transparency
tagline: What we strive for, how we measure it, and where to verify it.
note: >-
  Upgraded from the automated probe, which detected the keywords "SOC 2", "ISO 27001" and "HIPAA" on this
  page and attributed all three to xCures. Reading the page shows that is wrong for two of them: SOC 2
  Type 2 and ISO 27001:2022 are AWS's, which xCures inherits controls from. The certification xCures
  itself holds is HITRUST e1, which the keyword probe missed entirely. Attribution is corrected below.
certifications_held_by_xcures:
  - name: HITRUST e1 Certification
    scope: xCures Clinical Clarity Engine
    status: certified
    year: 2025
    upgrade_in_progress: HITRUST i1 Certification (2026)
    quote: >-
      "The xCures Clinical Clarity Engine is HITRUST e1 certified (in the process of being upgraded to
      HITRUST i1 Certification)."
  - name: HIPAA
    scope: xCures and the Clinical Clarity Engine, operating as a business associate
    status: compliant
    validation: annual HIPAA evaluation plus assessments under the HITRUST certification program
certifications_inherited_from_aws:
  attribution_warning: >-
    NOT xCures certifications. The page states these are held by AWS, that xCures "is able to inherit
    selected AWS HITRUST r2 controls to include as evidence in xCures's HITRUST certification program",
    and that it "periodically reviews the AWS ISO 27001 certification and SOC 2 attestation report".
    Attributing them to xCures overstates its posture.
  certifications:
    - name: HITRUST r2 Certification
      holder: AWS
    - name: ISO 27001:2022 Certification
      holder: AWS
    - name: SOC 2 Type 2 Attestation
      holder: AWS
security_controls:
  - control: Encryption
    detail: In transit (TLS) and at rest
  - control: Access Controls
    detail: RBAC with least-privilege and minimum-necessary principles, MFA, and SSO
  - control: Audit Logging
    detail: Immutable logs via AWS + Datadog, SIEM monitoring
  - control: Data Deletion
    detail: In accordance with contractual requirements
  - control: Provenance / audit trail
    detail: Full CRUD audit logs maintained across all data access and modification events
data_governance_standards:
  - standard: FHIR R4
    detail: All extracted data mapped to FHIR R4 resources for downstream interoperability
  - standard: OHDSI / OMOP
    detail: Concepts normalized to OHDSI Standardized Vocabularies (SNOMED, LOINC, RxNorm)
  - standard: mCODE
    detail: Oncology-specific data elements aligned to the HL7 Minimal Common Oncology Data Elements profile
  - standard: HIPAA
    detail: HIPAA compliant with annual HIPAA evaluation and HITRUST certification program
ai_model_validation:
  note: >-
    Unusual for this catalog and worth recording: xCures publishes measured accuracy for the AI that
    produces its API's output, with a stated method and a named limitation set. Most AI-in-the-loop
    providers publish neither.
  approaches:
    - name: Schema-based extraction
      detail: >-
        NER + relation extraction over unstructured clinical documents into FHIR R4 and OHDSI-normalized
        structured data, with linkage to source verbatim preserved per element.
    - name: Checklist-based assertion
      detail: >-
        RAG over the full longitudinal record, returning structured outputs with source citations and
        evidence-hierarchy rules to resolve conflicting documentation.
  method: >-
    Validated against clinically trained human reviewers; each field classified TP/TN/FP/FN with
    third-reviewer arbitration on discrepancies; random 10% audit. Only explicitly stated, verifiable
    extractions count as true positives — correct inferences not present verbatim are counted as errors.
  deployment_threshold: accuracy and precision >= 95% before an extractor or checklist enters production
  published_results:
    - extractor: Medications
      accuracy: 95.7%
      precision: 97.5%
      recall: 95.0%
      f1: 96.3%
    - extractor: Surgical Procedures
      accuracy: 96.6%
      precision: 97.7%
      recall: 98.8%
      f1: 98.2%
    - extractor: Cancer Diagnosis
      accuracy: 98.2%
      precision: 98.7%
      recall: 99.4%
      f1: 99.0%
    - extractor: Lines of Therapy
      accuracy: 97.0%
      precision: 95.4%
      recall: 99.8%
      f1: 97.6%
  results_caveat: >-
    xCures states these are a retrospective analysis of a defined historical dataset and do not guarantee
    future performance.
  source_citation: >-
    Stuhlmiller TJ et al. "A Scalable Method for Validated Data Extraction from Electronic Health Records
    with Large Language Models." Submitted for peer review, 2026. Full methods, supplemental tables and
    raw counts available on request.
  stated_limitations:
    - >-
      OCR quality — accuracy of extraction from scanned or faxed documents depends on document quality;
      degraded scans may introduce errors or omissions.
    - >-
      Semantic search coverage gaps — checklists and schema LLMs rely on semantic search and use only the
      top N semantically matched documents, so relevant information in lower-ranked documents can be
      missed.
  version_control: Extraction models are version-controlled and support rollback; A/B testing guides refinement
infrastructure:
  hosting: AWS
  monitoring: AWS + Datadog, SIEM