Why Modern Healthcare AI Must Remember the 1968 Problem-Oriented Record
An essay warns that today’s AI tools risk repeating the same patient-centering failures of early medical records, citing Lawrence Weed’s 1968 paper and Warner Slack’s advocacy.
Drawing on the influence of Warner Slack, the author stresses that patient records should serve patients, not regulations or reimbursements. Slack’s early observation that records lacked indexes and relevance foreshadowed today’s AI challenges. Lawrence Weed’s 1968 New England Journal of Medicine article introduced the problem-oriented medical record and SOAP note format, reorganizing charts around specific health issues instead of departmental silos.
While the problem list later became a legal requirement for certified electronic health records, its original purpose has been diluted by billing imperatives. Modern healthcare AI, the author warns, may repeat this mistake by focusing on downstream tasks and neglecting the clinician’s reasoning. He recounts his own 1988 work on a feature dictionary aimed at teaching machines the clinical language needed to understand patient records. The overarching message is that AI must preserve clinical intent, protect the record’s purpose, and ultimately serve patients.
Why it matters
If AI ignores clinical reasoning, health data may become less useful for patient care and more driven by billing.
In this story
