Opinion: AI depends on medical records that don’t always tell the full truth about women — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Gallup-Pivotal Women's Health Survey reported in September that 34% of U.S. women strongly or somewhat agreed a health care provider minimized or dismissed their symptoms in the past five years, and 33% reported being misdiagnosed or unable to get a diagnosis, versus 21% of men reporting those experiences.
- Seen But Not Heard, the study by Annemarie G. Hirsch and Jennifer Silva, paired electronic medical records with detailed interviews with 87 women about pain and illness, work, family, poverty, violence and health care encounters.
- Electronic medical records documented mistrust in only seven of the 87 women interviewed, even though 53 described a trust-breaking encounter with a health system.
- Misdiagnosis appeared in the records of only eight of the 32 women who reported it in interviews, while 30 women who said clinicians treated them with disrespect had that recorded nowhere.
- Kate, a pseudonym, spent more than a decade with pelvic pain while notes blamed her weight and then her mental health; only after a provider recorded her new husband's questions — "Painful intercourse for the past year" — did doctors perform the exams that diagnosed endometriosis.
- AI scribes are being adopted by clinics and hospitals across the country to turn recorded doctor-patient interactions into clinical notes, alongside tools that search and synthesize records for diagnostics and treatment plans; the authors say such tools are only as good as the data in the record.
- Hirsch and Silva say AI may learn and reproduce how women are framed in the record when symptoms are filtered through assumptions about weight or mental health; they call for clinical spaces where women feel safe disclosing symptoms and for clinicians to document delayed or incorrect diagnoses without fear of penalty.
Why it matters: For the clinics and hospitals now adopting AI scribes, the 8-of-32 gap in documented misdiagnoses is the whole ballgame: notes that omit women's symptoms give AI less to work with, and the authors say that pattern can reinforce the framing that left Kate without an endometriosis diagnosis for 10 years.
Ask SkimNews
