Published August 30, 2026 | Version V1

JMIR 105947 Named Professional Assessments Dataset

Description

This dataset contains 570 citation-level records from a paired validation study of an AI-generated subspecialty anesthesiology knowledge base across 19 sampled subspecialties. Each record includes a stable citation identifier, subspecialty, named professional validator, four criterion-level AI ratings, four corresponding human ratings, AI and human overall verdicts, and two post hoc PI-adjudicated issue-label fields. The assessments were provided by 10 named professional validators. The package includes the 15-column dataset, data dictionary, README, CC BY 4.0 license notice, and SHA-256 integrity manifest. It supports recomputation of the reported agreement statistics but not regeneration of the original model-plus-retrieval executions. Human ratings are an operational reference, not ground truth. The released files contain no patient data, clinical records, completed authorization forms, signatures, contact information, institutional fields, form dates, free-text comments, or provider/tool records.

Files

JMIR_105947_CC_BY_4_0_LICENSE_NOTICE_v1.txt

Files (71.0 kB)

Name Size
md5:5d5cf6c980cc777ce874dc00f22472ca
463 Bytes Preview Download
md5:6e4ea1b230b59eac61d81ce0503f7567
979 Bytes Preview Download
md5:40a748e0ce8a862adcb719c11b121594
3.8 kB Preview Download
md5:438150cd56368993a6e3c20b1b7d8a95
2.6 kB Preview Download
md5:483388fbd7bb05fbe1494a76fb4ea55d
63.2 kB Preview Download