Written by Clay Smith
Spoon Feed
Emergency medicine practice is going to be increasingly influenced by artificial intelligence (AI). This article discusses some of the ways we can expect it to impact our practice.
AI is jolly good fun!
The author explains that AI will move through three phases:
- Mapping – What problems should AI solve?
- Measuring – How do we validate these solutions?
- Management – How will we integrate AI into our daily clinical workflow and make sure it stays accurate and equitable?
Here is what I am excited about in each phase.
Mapping: AI is already capable of and will reliably: 1) read imaging studies, 2) read ECGs, 3) do our H&P, 4) make diagnostic and treatment recommendations, 5) check for drug interactions, 6) predict complications and clinical deterioration risks, 7) predict resource needs, 8) predict crowding and staffing needs, 9) help determine triage priorities, 10) help make disposition recommendations. And this is just the beginning.
Measuring: The best solutions are what the AMA calls augmented intelligence. We use AI and validate it with skilled physicians, who can spot when AI is being stupid and add can add their common sense and real world experience caring for patients.
Management: I’m optimistic that AI can automate and do many complex, repetitive and onerous tasks for us, making us more efficient. It will be essential to have human oversight to make sure AI solutions remain current and equitable.
Tomorrow, we will discuss some of the ethical aspects of AI in medicine in more detail.
How will this change my practice?
It’s hard to think of an area of my practice that AI won’t change. I’m also extremely excited about AI for education. It will allow us to make JournalFeed smarter by improving and expanding our search capabilities, allowing us to help more specialties, and enabling readers to engage in self-guided learning.
Source
The AI Future of Emergency Medicine. Ann Emerg Med. 2024 May 22:S0196-0644(24)00043-X. doi: 10.1016/j.annemergmed.2024.01.031. Online ahead of print. PMID: 38795081
