Viral chart on AI and hospital mortality not confirmed

Datafloq's fact-check found no original source for the viral Protege DataLab chart claiming reduced mortality in US hospitals with rapid AI adoption, while the actual Drexel study (Scientific Reports) shows much more modest observational associations — for example, a 9.9% reduction in county mortality with access to workflow AI.

Compiled by Sergey KostenchukPublished 2026-09-21Updated 2026-09-22

2026-09-21 Research

💊 Viral chart on AI and mortality not confirmed

A Protege DataLab chart spread on Telegram: in US hospitals with the fastest AI adoption, mortality allegedly dropped to ~1010 versus ~1620 per 100,000 patients. Datafloq's fact-check (27.08.2026) found neither the chart nor the figures on the Protege DataLab website or in the a16z.news newsletter.

🌍 The real figures are more modest: the Drexel study (6166 hospitals, 3143 counties; Scientific Reports) links workflow AI with a 9.9% reduction in county mortality, and routine automation with a 5.1% reduction in pneumonia. The authors warn: these are associations, not causality — AI adoption may reflect the clinic's resource level.

👤 The first noticeable effects of AI in medicine are currently routine: documentation, scheduling, protocols. A pretty chart without a source and methodology is not scientific evidence.

Source 1: https://www.nature.com/articles/s41598-026-70027-1 Source 2: https://www.medrxiv.org/content/10.64898/2026.03.10.26347904v1

Sources