Evidence explorer

Every claim we hold, and how far we can stand behind it.

142 claims across 39 products. 30% are confirmed against the source text by the evidence gate; 39% come from a class with no vendor interest. Everything below is browsable, filterable, and available as JSON.

What this corpus is made of

Evidence class	Claims	Confirmed	What it can support
Peer-reviewed study	19	5	Independent review before publication. The only class that survives a sceptical committee unaided.
Regulatory filing	4	0	A statement made to a regulator, where being wrong has consequences.
Independent report	32	9	Analysis by someone with no stake in the sale.
Customer case	4	3	A real deployment, but selected and framed by someone with an interest.
Press release	63	23	The vendor's own announcement. Useful for dates and facts, not for outcomes.
Vendor page	20	2	Marketing copy. Establishes what is claimed, never that it is true.

Showing 8 of 8 matching claims (142 in the full corpus) · 50% of this selection is confirmed · clear filters

Real-world AI scribe study (JMIR 2026) participant on Heidi: My hospital has been piloting Heidi, and I have used it extensively in the emergency department (ED)
Heidi ScribePeer-reviewed studyConfirmed2026-08-13Open source ↗
Provenance
Claim ID
clm_07d8295a7406b9118434
Version
1
Workflow
Ambient Clinical Documentation
Dimension
evidence strength
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://pmc.ncbi.nlm.nih.gov/articles/PMC13472498/
Peer-reviewed Nature Health study (SAFEMOM) of Ada integrated into South Africa's government-run MomConnect WhatsApp maternal health platform found safe health-seeking behavior rose nearly a third and appropriate care-seeking rose more than 40%; an independent physician panel rated Ada's recommendations safe in 98% of cases
Ada Health Assessment (AI symptom assessment & care navigation)Peer-reviewed studyUnverified2026-06-30Open source ↗
Provenance
Claim ID
clm_e424d6e33e196547ba29
Version
1
Workflow
Clinical Evidence & Decision Support
Dimension
evidence strength
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://doi.org/10.1038/s44360-026-00125-x
JMIR Med Inform 2026 prospective study: Abridge was chosen as the AI platform and integrated with the Epic EMR at Hawaii Pacific Health Medical Group.
AbridgePeer-reviewed studyConfirmed2026-06-23Open source ↗
Provenance
Claim ID
clm_a5941d994b2af01d6698
Version
1
Workflow
Ambient Clinical Documentation
Dimension
evidence strength
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://pmc.ncbi.nlm.nih.gov/articles/PMC13289844/
JMIR Med Inform 2026: Abridge was chosen as the AI platform and integrated with the Epic EMR.
AbridgePeer-reviewed studyConfirmed2026-06-23Open source ↗
Provenance
Claim ID
clm_00b51ce1cd925912e9d6
Version
1
Workflow
Ambient Clinical Documentation
Dimension
interoperability
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://pmc.ncbi.nlm.nih.gov/articles/PMC13289844/
JMIR Med Inform 2026 prospective Abridge study: Signal metrics found a 21% decrease in time spent on notes per day among high users.
AbridgePeer-reviewed studyConfirmed2026-06-23Open source ↗
Provenance
Claim ID
clm_2d77987ee04fa4cb752f
Version
1
Workflow
Ambient Clinical Documentation
Dimension
economic value
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://pmc.ncbi.nlm.nih.gov/articles/PMC13289844/
Prospective multi-centre study (Open Forum Infectious Diseases, 2026) of qXR version 3 AI-based chest X-ray screening across 3 public health facilities in Chhattisgarh, India: of 2,745 CXRs screened, 363 patients were presumptive for TB and 162 confirmed cases were found (44.63% positivity among presumptive cases), and an 80.21% increase in TB case notifications was observed during the AI-implemented period versus baseline (P < .001).
qXR (Chest X-ray AI Suite)Peer-reviewed studyUnverified2026-05-01Open source ↗
Provenance
Claim ID
clm_326a114cde5f51400f91
Version
1
Workflow
Imaging & Diagnostic AI
Dimension
evidence strength
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://pmc.ncbi.nlm.nih.gov/articles/PMC12810203/
Multi-country head-to-head accuracy comparison (Annals of the American Thoracic Society, 2026) of 7 automated chest X-ray algorithms on 3,901 chest X-rays from India, Madagascar, the Philippines, South Africa, Tanzania, Uganda, and Vietnam found qXR achieved the WHO minimum target accuracy for a tuberculosis triage test (>=90% sensitivity with >=70% specificity), with specificity of 70% or greater at 90% sensitivity.
qXR (Chest X-ray AI Suite)Peer-reviewed studyUnverified2026-05-01Open source ↗
Provenance
Claim ID
clm_36833a1aa15ec821791a
Version
1
Workflow
Imaging & Diagnostic AI
Dimension
evidence strength
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://doi.org/10.1093/annalsats/aaoag011
Applied Clinical Informatics (2026): early implementation of DAX Copilot in a pediatric hematology-oncology division across 11,544 outpatient encounters (Jan-Jul 2025) found that for encounters of 40 minutes or less, median documentation time was 11 minutes with DAX versus 24 minutes without - an approximately 30% lower documentation time (95% CI 19.9-38.4%; p<0.001).
DAX CopilotPeer-reviewed studyUnverified2026-03-01Open source ↗
Provenance
Claim ID
clm_61a71bb995cf683609ed
Version
1
Workflow
Ambient Clinical Documentation
Dimension
evidence strength
Retrieved
2026-08-27
Reviewer
healthit-gate-v1
Source URL
https://pmc.ncbi.nlm.nih.gov/articles/PMC13192280/

Know a source we are missing?