OpenEvidence CAIHL draft report

HugoScore CAIHL Draft Report: OpenEvidence

Status: Draft for human review Last reviewed: 2026-06-10 Review method: Deep public-source review of official product, user guide, Visits, patient handout, after-visit summary, privacy, terms, and partnership announcement materials; no clinician account testing, patient-facing deployment review, vendor interview, or independent accuracy evaluation. Service: OpenEvidence Vendor: OpenEvidence Inc. Category: Clinical decision support / medical evidence platform

Summary

OpenEvidence is an AI medical knowledge and clinical decision-support platform for healthcare professionals. Public materials describe evidence-grounded answers using sources and partnerships that include NEJM, JAMA, NCCN, Wiley, Cochrane, and medical societies. The product also includes clinician workflow features such as Visits, clinical note drafting, after-visit summaries, patient handouts, patient profiles, uploaded documents, and patient-specific questions.

From a CAIHL perspective, OpenEvidence is patient-affecting but not patient-directed. It may improve clinician evidence access and help produce clearer patient-facing materials, but patients are not the primary users. The key agency issue is that a patient may be affected by OpenEvidence-informed documentation, treatment reasoning, or education without being able to use the same tool, inspect the same reasoning, correct the patient context, or challenge the output at the source.

Evidence Reviewed

• OpenEvidence homepage: https://www.openevidence.com/
• User guide: https://www.openevidence.com/user-guide
• Visits overview: https://www.openevidence.com/user-guide/visits-overview
• Patient handouts: https://www.openevidence.com/user-guide/patient-handouts
• After Visit Summary: https://www.openevidence.com/user-guide/visits-avs
• Patients guide: https://www.openevidence.com/user-guide/visits-patients
• Visits launch announcement: https://www.openevidence.com/announcements/visits-real-time-medical-intelligence
• Cedars-Sinai partnership announcement: https://www.openevidence.com/announcements/openevidence-partners-with-cedars-sinai-to-create-patient-aware-clinical-intelligence-with-agentic-clinical-ai
• Privacy policy: https://www.openevidence.com/policies/privacy
• Terms of use: https://www.openevidence.com/policies/terms

CAIHL Profile

• Who does this AI serve? Clinicians and health systems. Official materials frame the platform as intended for physicians, clinicians, verified HCPs, and healthcare-professional workflows.
• Can patients tell AI is involved? Unclear. Public user-facing pages are transparent to clinicians, but patient-facing notice for evidence answers, patient handouts, after-visit summaries, visit recording, or EHR-integrated use was not visible.
• Can patients meaningfully choose? Not established. Patients generally do not appear to choose whether a clinician uses OpenEvidence; recording, EHR-context use, and patient-specific AI analysis require deployment review.
• Can patients correct or challenge what the AI produces? Not established. Clinicians can review and edit outputs, but public materials do not show a patient-accessible route to inspect prompts, sources, reasoning, extracted patient context, or AI-shaped outputs.
• Does it help patients understand or act? Indirectly. Patient handouts and after-visit summaries can support understanding, but they remain clinician-mediated outputs rather than patient-controlled research or advocacy tools.

Agency Interpretation

OpenEvidence belongs near the institutional end of the HugoScore axis. It is patient-affecting because evidence answers, visit notes, summaries, record context, and clinical recommendations can influence the care a patient receives. It is not patient-directed because the primary user, account holder, evaluator, and editor is the clinician or health system.

The best case is that OpenEvidence helps clinicians bring better evidence into the room and produce clearer patient materials. The CAIHL risk is that the patient experiences the tool only as downstream authority: the answer, note, or recommendation arrives through the clinician, while the affected person cannot interrogate the same system, correct the context, or challenge the output at the source.

Key Unknowns

• Whether patients are told when OpenEvidence informs a clinical answer, after-visit summary, patient handout, note, diagnosis, or treatment discussion.
• Whether patients can refuse Visits recording, EHR-context use, or patient-specific AI analysis.
• Whether patients can see the sources, uncertainty, and patient-record facts used in a patient-specific answer.
• Whether patients can correct extracted patient data, generated notes, handouts, or after-visit summaries.
• How sponsored programs and advertising are separated from clinical answers, patient handouts, and clinician education.
• What independent evidence exists for accuracy, bias, completeness, specialty performance, patient communication quality, and safety.

Publication Recommendation

Ready for human review as a draft profile. Keep confidence at medium until patient-facing notice, refusal, correction, independent evaluation, and deployment-specific data workflows are verified.