I attended my first CHEST conference with low expectations. I wasn’t presenting anything. I was there to observe, learn, and follow whatever caught my interest.
Before arriving, I marked the sessions I wanted to attend. Unsurprisingly, that meant plenty of AI talks, along with quality improvement and other systems-focused sessions. Here are my takeaways.
AI adoption looked different in the room
I spend a lot of time reading about AI in medicine. Between LinkedIn, journal articles, and using these tools myself, it’s easy to assume everyone else is experimenting with them, too.
CHEST reminded me how much that shapes my expectations.
During a demonstration of using OpenEvidence to answer clinical questions and Claude to draft patient letters, I swear I heard people gasp and say “whoa.” Tools I’d grown familiar with were clearly new to at least some people in the audience.
The specialty-specific uses caught my attention
I was particularly interested in the sessions on pulmonary function tests. The examples included tools to help interpret flow–volume loops and support a clinician’s differential diagnosis. One study discussed in the session found that pulmonologists’ diagnostic accuracy improved with explainable AI assistance, although selected-case performance doesn’t establish better patient outcomes.
What interested me was how specific the task was. I could picture where an additional interpretation might fit into reviewing a PFT, and why I’d want to understand which findings drove the suggestion before accepting it.
Another presentation described Northwestern’s use of an AI caller to contact patients with existing low-dose CT screening orders and help them schedule. The results didn’t blow me away, but the use case felt practical, since an order has already been placed, and someone still needs to help turn it into a completed scan.
I kept wondering how it would compare with a text message offering specific appointment times. Would patients be more likely to respond if they could simply choose a slot? I’d be interested in that comparison before assuming the AI caller is the best approach.
Tracheostomy: thinking through what comes afterward
Outside the AI sessions, a talk on tracheostomy brought the focus back to decisions we make in the ICU.
The session emphasized the course after placement: the possibility of ongoing ventilator dependence, time in post-acute facilities, caregiver responsibilities, and the need for a clear plan for weaning and decannulation. Those considerations belong in the conversation before the procedure, alongside the patient’s goals.
For me, that’s a useful way to organize the discussion. When we talk about a tracheostomy, how clearly are we describing what the following weeks and months could involve? And how realistic is the recovery plan given the support available to that patient?
Inhaler costs belong in the clinical conversation
I also attended a session on the cost of inhalers and other pulmonary medications. I didn’t take detailed notes, but here’s point stuck with me:
The delivery device deserves attention when we talk about patents and competition, alongside the medication itself.
The discussion also left me with a question about the transition toward lower-emission inhalers: how will changes to devices and propellants affect patents, competition, and what patients pay? I’d want to look more closely at the specific products and policies before drawing a firm conclusion.
It’s a topic I want to keep following. We can choose an appropriate inhaler in clinic, but affordability still determines whether a patient can fill it. I’ve covered the broader pricing system in The Physician’s Guide to the Drug Supply Chain.
I left my first CHEST with a better sense of what I want to pay attention to: how colleagues are actually using AI, whether a new tool helps complete a task, and what happens to patients after we make a recommendation. I came to observe and learn. Having a few of my assumptions challenged made the trip worthwhile.
