Some olde hospital regulatory capture thoughts: there are workflows built into EMR’s and forced onto physicians by regulation detached from their clinical value. Healthcare maintenance alarms are infamously noisy to such an extent that they’re worthless. The oft-cited reason for the gap: the buyers are administrators solving for the regulation while ignoring the clinical impact. The healthcare provider gets incentive money from the JC for hitting the “standard”, a good chunk of which gets passed right through to the Cerners / Epics to pay for their solution.

…

Managed to wring my body out between the rain. Leg lifts using the bbell squat rack at the office, and finished it with a run up in Baldwin Hills on the edge of Culver City.

The hill is just about dead center on the way from the office to Mayumi, and it sprawls up from warehouses into hills pumped by oil derricks. I’ll bet the oil money offset occupying a chunk of LA by developing the parks and trails. Wonderful place to explore.

…

Rainy day keeping me pent up / not lifting, and dinner with the Regardians ate the evening (Vicky’s All Day). Did get a small core workout done in the hotel room.

My thoughts have been circling round ChatGPT writing clinical notes. From what I understand, Bing Chat exposes the very same GPT-4 that Microsoft Nuance is entreating to write physicians’ notes. I believe that it’d be possible to understand the fire Microsoft toys with using applying existing synthetic test cases mapping patient data -> clinical notes; whether it has any potential.

…

Staying at The Mayumi in LA for my Regard double work week. Pretty cute little place – feels more like a small LA apt building than a hotel, and is largely booked out by us Regardians.

Work-related link: Microsoft Nuance DAX is exploring using GPT-4 to draft clinical notes, announcement here. Given GPT is an impressive writer, I’m certain some LLM could be incarnated to write a clinical note. Yes, yes. However, I also have to believe that a tailored model / high fidelity patient input will be more effective than just feeding the patient data into chat GPT and asking it to spit out a note. That being said, the in-context few-shot approach does seem to allow a “super” model that just focuses in appropriately.

…

Today got to watch F1 qualifying and continued with WRC. Max might be off podium tomorrow, which I’m not not looking forward to it :) On the WRC side Guanjuato is desolately beautiful, and the gravel roads look super fun. Massive rock spires trailing to a 5ft wide top with a single cactus on top. Sad Ott tho :(

Dense day tomorrow, so here’s the plan:

• pack up and load the car tonight
• head to SF in the morning, take off at 8:30a
• basketball with RN
• brunch with EB
• drive down to LA

I’ve wrapped the workouts for the week, earned a rest day from lifting. I did put in a light 6mi on the bike to keep my legs and gooch warm, and’ll wring myself out tomorrow with pickup ball.

…