Encyclox

What A 1968 Paper Asks Of Healthcare AI

· curiosity

The Forgotten Patient in AI-Powered Healthcare

In 1968, Lawrence Weed published a paper in the New England Journal of Medicine that would revolutionize medical records. His concept of the problem-oriented medical record (POMR) organized patient charts around specific problems rather than departmental or data source silos. This innovation was simple yet profound.

Weed’s solution was the SOAP note system – Subjective, Objective, Assessment, and Plan. Each entry in the chart followed a single line of reasoning: what the patient reported, what exams and labs showed, what clinicians concluded from it, and what came next. These elements coalesced around a problem list, allowing clinicians to grasp the patient’s situation at a glance.

The POMR gained traction, becoming an integral part of every certified Electronic Health Record (EHR). However, its original purpose was lost in translation as EHRs became optimized for billing and reimbursement rather than clinical understanding. Problem lists were reduced to coders translating them into billing codes, often failing to capture the complexity of patient conditions.

This oversight has significant implications for healthcare AI today. Systems designed with a focus on downstream use cases neglect the reasoning they’re meant to preserve – the essence of what makes medical records valuable in the first place. Warner Slack, a pioneer in patient-centered care, foresaw this potential pitfall and implored me to remember that “the most important thing in the exchange of patient information is to preserve the clinical intent.”

The Lost Art of Clinical Understanding

In 1988, I attempted to address this issue through a paper called “A Feature Dictionary Supporting a Multi-Domain Medical Knowledge Base.” This effort aimed to capture clinical language with enough precision that machines could understand it as physicians do. Nearly sixty years after Weed’s pioneering work, the mission remains unchanged: preserve meaning, protect purpose, and serve patients.

The problem-oriented approach has been largely abandoned in today’s AI-driven healthcare landscape. Instead of capturing a clinician’s reasoning, systems prioritize efficiency and scalability over accuracy. This trade-off may be efficient for administrators but neglects the patient’s fundamental right to receive care that respects their unique situation.

A Cautionary Tale: When Technology Fails Patients

The POMR’s decline serves as a cautionary tale about what happens when technology is optimized for the wrong purpose. In this case, it was not designed to serve patients but rather to facilitate billing and reimbursement. This phenomenon is not unique to healthcare; we see similar patterns in other industries where technology prioritizes profit over people.

Healthcare AI’s primary function should be to augment human decision-making, not replace it. By neglecting this fundamental principle, we risk exacerbating existing healthcare disparities and creating new ones. The stakes are high: every patient deserves care that is informed by their individual needs, not just their billing codes.

Preserving the Patient in a World of AI

In recent years, there has been a surge of interest in developing AI systems that can interpret medical records with greater accuracy. While these efforts hold promise, it’s essential to remember why we’re pursuing this goal: to serve patients better. Let us not forget the problem-oriented approach pioneered by Weed and Slack – an approach that prioritized patient-centered care above all else.

The future of healthcare AI should be built on a foundation of clinical understanding, not just data analysis. By reorienting our priorities toward preserving the meaning and purpose of medical records, we can create systems that serve patients with dignity and respect. The challenge ahead is clear: to develop technologies that honor the patient’s voice, not just their billing codes.

The Patient-Centered Imperative

As healthcare continues its rapid march toward AI-driven decision-making, we must remain vigilant about the values that underpin our endeavors. By preserving the problem-oriented approach and prioritizing clinical understanding, we can ensure that patients receive care tailored to their unique needs – not just optimized for administrative efficiency.

The patient’s voice must always come first in healthcare. As we forge ahead in this uncharted territory, let us remember the words of Warner Slack and Lawrence Weed: every patient deserves care that respects their individuality.

Reader Views

  • HV
    Henry V. · history buff

    The 1968 paper's emphasis on clinical understanding over billing codes is a crucial reminder for today's healthcare AI. However, we must also acknowledge that POMR's limitations are rooted in its paper-based design, which often relied on manual transcription and fragmented care records. In the era of EHRs, we've traded one set of problems for another: how to integrate disparate digital systems while preserving the clinical intent of patient data? The article glosses over this issue, but it's a vital consideration as AI-powered healthcare continues to evolve – can we truly digitize human judgment and care without sacrificing its essence in the process?

  • IL
    Iris L. · curator

    While the resurgence of interest in Weed's POMR is welcome, we must be cautious not to romanticize its original purpose as solely about enhancing clinical understanding. The truth is, healthcare AI has always been driven by economic imperatives, and the POMR was adapted to fit this reality rather than being the driving force behind it. Until we acknowledge and address these underlying motivations, our attempts to reclaim the clinical intent of medical records will be hampered by a fundamental contradiction: can we truly preserve the value of healthcare data when its primary purpose is to generate revenue?

  • TA
    The Archive Desk · editorial

    The POMR's downfall is a cautionary tale for AI developers: when clinical intent is lost in translation, so too are the benefits of innovation. What's often overlooked is how this loss affects patient data long after they've left the hospital. When problem lists become mere billing codes, valuable insights into treatment efficacy and potential side effects go up in smoke. To truly harness healthcare AI's potential, we need to rethink how these systems ingest and utilize POMR data, prioritizing longitudinal analysis over snapshot billing reports.

Related articles

More from Encyclox

View as Web Story →