Conference posters and abstracts, pilot evaluation results, the assessment instrument we built to measure AI against human first drafts, and the Evidence Mastery book series.
Results reported at ISPOR Europe 2026 from 20 structured evaluations by 7 respondents from 7 organisations, covering 8 modules. Each respondent completed the EvidenceAi Assessment Survey (EAS) immediately after generating a document.
Teal bar = 95% confidence interval; grey line = human-equivalent reference (5.0) on comparative domains. Scores below 5.0 indicate inferiority to a human first draft; above 5.0, superiority. Absolute domains are not compared with a human reference.
A prospective mixed-methods evaluation comparing first drafts generated by the EvidenceAi™ Suite with human-generated first drafts, using a ten-domain assessment survey embedded at the point of generation. All eight modules evaluated scored at or above the human-equivalent threshold across the comparative domains.
Development of the EvidenceAi™ Assessment Survey (EAS), a human-anchored 10-domain instrument for evaluating generative AI tools in pharmacoepidemiology, and its pilot application: calendar-time and human-effort savings per document alongside scientific and editorial quality.
A 10-domain, human-anchored instrument for evaluating generative AI tools in pharmacoepidemiology: five comparative domains (0–10, with 5.0 = human equivalent), three absolute domains, two resource measures (calendar time and human effort saved) and a free-text impression. It is available for non-commercial research use.
Volume 1 of a five-volume series: principles, data sources, tools and templates, methods, and diseases and drugs. Volumes 2–5 are scheduled from December 2026 to May 2028.
More than 50 peer-reviewed publications in pharmacoepidemiology, real-world evidence and health data.