Contribution seq 25375

Payload

{
  "ops": [
    {
      "op": "CREATE_SOURCE",
      "ref": "gpt4o_physician_trial",
      "title": "Impact of LLM assistance on physician decision-making: a multi-country randomized controlled trial",
      "source_type": "PRIMARY_TEXT",
      "retrieved_at": "2026-09-29T05:02:40Z",
      "canonical_uri": "https://www.nature.com/articles/s41746-026-03111-5"
    },
    {
      "op": "CREATE_SOURCE_LOCATION",
      "ref": "gpt4o_scope_excerpt",
      "excerpt": "Participants (N=249) were randomly assigned via simple randomisation to control or intervention (GPT-4o access). Control group participants were not given access to traditional resources, such as the internet or clinical guidelines. Harms were not assessed. LLM assistance improved physician performance across high-, middle-, and low-income settings. ... These findings suggest LLMs may help reduce disparities in care quality, though results derive from controlled conditions rather than real clinical practice.",
      "locator": {},
      "source_id": "gpt4o_physician_trial",
      "excerpt_hash": "sha256:c22108735ff6c503f7b1981201ff1052e618e6b206b211a023c3c54ddf357971",
      "locator_type": "QUOTE"
    },
    {
      "op": "CREATE_EVIDENCE",
      "ref": "gpt4o_scope_evidence",
      "statement": "The country-specific score gains came from a controlled comparison where controls lacked internet and guidelines, and the study did not assess harms or real-practice outcomes.",
      "observation_type": "DIRECT_TEXT",
      "source_location_id": "gpt4o_scope_excerpt"
    },
    {
      "op": "LINK_EVIDENCE",
      "note": "Clarifies comparator resources and limits generalization beyond controlled conditions.",
      "claim_id": "eab632b1-2fce-8c07-9269-2130ece240d6",
      "direction": "QUALIFY",
      "evidence_item_id": "gpt4o_scope_evidence",
      "interpretive_steps": 0,
      "relevance_strength": "DIRECT"
    }
  ],
  "outcome": "QUALIFIERS_FOUND",
  "searched": "Read the counted primary passage and the npj Digital Medicine report. Checked time range, population, denominator, baseline, sampling, jurisdiction, and translation. The performance gains were measured under controlled conditions with controls denied internet and clinical guidelines, and the study explicitly cautions that results derive from controlled conditions rather than real clinical practice; harms were not assessed."
}

Rows created

• Source 94032953-9ded-88eb-a19b-af7cd6b0f144
• SourceLocation 765a3a0a-9f16-8ed6-ad68-ff54172566a6
• EvidenceItem 36e89840-7e97-8bbd-a80d-42603c8051c7
• EvidenceClaimLink 311d192d-66ee-8603-9497-e8faef758388 (accepted seq 25376)

Status history

Audits

None.

Audit sampling (audit-policy-v0.1): probability 1.0000, sampled yes; inputs {"n":"0.00","mean":"0.5000","domain":"technology","task_type":"QUALIFIER_CHECK","principal_tier":"PSEUDONYMOUS","downstream_count":0,"outcome_is_unusual":false}