Clinical Work
How I practice.
The views below describe my personal clinical practice and posture toward specific technologies. They are not Essentia Health institutional policy and they are not medical advice for any individual reader.
Hospital medicine is mostly about three things: the differential, the conversation, and the discharge. The differential is the cognitive work — keeping the right list of possible diagnoses in the right order, and revising it as the labs come back and the patient's story tightens. The conversation is the human work — getting the history, explaining the plan, having the family meeting nobody wants to have. The discharge is the systems work — making sure the patient lands somewhere safer than the hospital, with the right medications, the right follow-up, and the right person to call when something feels wrong.
The concrete shape of the work: full-time hospital medicine at Essentia Health-St. Mary's in Duluth since September 2024 — approximately fourteen shifts a month across medical-surgical, telemetry, and step-down units, covering admissions, daily rounding, discharge planning, and the standing coordination loop with consultants, case management, nursing leadership, social work, and palliative care. Documentation lives in Epic.
I have a deliberate philosophy about which of these gets technology and which stays analog.
The differential is where I want better tools.
The cognitive load of a busy inpatient service is genuinely hard. A list that can easily climb above fifteen patients on a given day, each with a moving target. I use clinical decision support tools that have been validated in populations resembling mine — risk calculators with published external validation, imaging algorithms that flag findings I might miss on a tired Sunday afternoon, drug-interaction checkers built into the EHR. None of these replace the judgment. All of them sharpen it.
For the specific tools I have evaluated and where I have landed on each, see the case studies.
The conversation is where I keep the technology out.
When I am taking a history, the patient is telling me things they have not told their spouse. When I am running a family meeting about transitioning to comfort care, the room has a stillness to it that does not survive a microphone in the corner. I do not use ambient AI scribes at the bedside, and I have written about why. The shorter version: the patient did not consent to a transcript, the note is not the artifact, and the legal accountability does not survive the cognitive outsourcing.
The discharge is where AI could actually help patients.
Most patients leave the hospital with a list of diagnoses they cannot explain and a medication list they cannot remember. The handout we send them home with is, almost universally, not good enough. This is the problem the teaching lab is trying to address: anatomically accurate, narrated, validated 3D visualizations that can stand in for the whiteboard sketch I would draw if I had time. AI does not produce these — studios like Nucleus Medical Media do, and have been doing for years. My contribution is curation and a short note on what each one is for.
What this all comes down to.
Medicine has always been an applied discipline. The question is never "is this technology cool", it is "does this technology improve the specific outcome that matters for this specific patient." Most of the time the honest answer is "we do not know yet, and I am not going to find out on my patient." Sometimes the answer is "yes, and here is how I am using it." The work is sorting between those two cases, calmly, one tool at a time.