Welcome to Optometry Simplified.
In this weekly newsletter, I've curated the best resources to help you grow personally and professionally.
My mission is to find what's best for my patients and my practice.
Here's what I've found...
Links I Liked
Prescribe warm compresses wisely.
Not all dry eye is MGD, not all MGD is obstructive, and not all obstructive MGD responds to heat. However, warm compresses get recommended almost reflexively across all of it. This piece breaks down the actual thermal physiology behind why they work when they work, and more importantly, when they don't. Worth a read before your next OSD patient. Optometry Times
Your best diagnostic instrument still doesn't need a power cord.
A new article by Michela Kenning, OD and Maggie Ho, OD rounded up 10 clinical pearls from Pacific University faculty on diagnostic tools in optometry. These are the kind of hands-on exam skills that AI and OCT simply can't replicate. Which are your favorites? Review of Optometry
Research I'm Reading
This will be the single best thing you do for your dry eye patients.
You earn the title of dry eye specialist when you can clearly delineate when dry eye is not present, ie., when it's some other OSD comorbidity or masquerader rather than true DED. That distinction changes everything. Because you'll be treating the actual problem. This DEWSIII paper is not new, but if you haven't read it yet, it is oh so worth your time. American Journal of Ophthalmology
Deep Thoughts
How do we use AI without losing our clinical mind?
By "clinical mind," I mean something specific. It's not just medical knowledge. It's the judgment that accumulates through thousands of patient encounters. It's the pattern recognition that fires before you can fully articulate why. It's the instinct that tells you something is off even when the numbers look fine and the human attunement that lets you read a patient's fear, hesitation, or confusion in real time and respond to it.
The clinical mind is built slowly, through repetition under uncertainty, and it is what separates a competent technician from a truly excellent clinician.
That distinction matters more now than ever, because there is solid evidence that automation erodes it.
Studies from aviation and surgery have shown that over-reliance on automated systems measurably degrades skill and pattern recognition over time. When the system does the reasoning, you stop building the intuition that only comes from doing the hard work yourself.
Pilots who lean on autopilot lose manual proficiency. Surgeons who follow robotic prompts without independent judgment make different errors than surgeons who don't. We are not exempt from this dynamic.
The risk isn't that AI replaces us. It's that we gradually outsource the thinking that makes us worth trusting in the first place.
And there's a second reason the clinical mind matters that no amount of technology changes: patients need a human being.
Not a diagnosis. Not a risk score. A person they can look at, ask questions of, and feel genuinely seen by. This doctor-patient relationship is not a nice-to-have. It is part of how good outcomes happen. AI has no access to that. It never will.
So how do we think clearly about where AI fits?
I keep coming back to Evidence-Based Medicine (EBM) as the right framework. David Sackett defined it as the integration of three things: the best available research evidence, your clinical expertise, and your individual patient's values and preferences. All three, together, inform every significant decision.
What I love about EBM is that it's a process, not an outcome. Annie Duke, in her work on decision-making, makes the point that the best decisions are evaluated by the quality of the process used to make them—not by how they turned out.
A good decision can still have a bad outcome. A bad process can accidentally produce a good one. EBM gives us the process. It's a filter for making wise decisions despite uncertainty, not a rigid algorithm, and not a mandate to wait for a perfect RCT before you act.
When you understand EBM that way, AI's role becomes much clearer.
Let me unpack what I see as AI's role in each of the EBM pillars.
AI helps most in the first pillar: research evidence. And for that, it is genuinely useful. For example, here are some of the tools I've been using:
NotebookLM is where I start. You upload the AOA Clinical Practice Guidelines, AAO Preferred Practice Patterns or any guidelines you trust and interrogate them in plain language. Ask it a specific clinical question and it answers from within the documents you gave it. No hallucinated citations, no wandering off into the internet. It stays in your sources. You can even create visuals or a podcast from the material.
Perplexity is a good starting point when you're trying to get oriented on a topic quickly. It searches the web in real time, cites its sources, and gives you a conversational answer you can actually read. Think of it as a smarter Google, not a replacement for the literature.
Elicit is where you go when you want to know what the peer-reviewed research actually says. It searches over 138 million academic papers and structures the results like a mini systematic review, real citations, real studies, no hallucinated references. I'm newer to it than Perplexity, but it's built specifically for evidence synthesis in a way that nothing else I've used matches.
Consensus synthesizes answers from peer-reviewed research and shows you the degree of scientific agreement on a question. Similar to perplexity, it is useful for quickly getting oriented before you go deeper.
The rule I'd give you for all of them: AI finds, you still appraise. The tool retrieves. You still evaluate by comparing sources, checking for consistency, and accounting for chance. That step does not get outsourced.
What about EBMs second pillar: clinical expertise?
AI can play a supporting role as a thinking partner by pressure-testing a differential, drafting patient education materials, and working through a complex case out loud.
But form your own impression first. Every time. If AI output becomes the first frame on a clinical problem, you are building a habit that works against you.
What about EBMs third pillar: patient preferences and values?
This is all you, my friend. Rise and shine (and do your job). AI has minimal help to offer. It can retrieve OHTS data that applies to your glaucoma suspect but it doesn't know that her mother went blind from glaucoma, and she is terrified. Or conversely, that she prefers a very minimalistic approach to intervention for not only glaucoma but any other disease risk. That conversation is yours.
So, use the tools. Use them aggressively for research. But cherish and keep your clinical mind.
That's not a limitation of AI. That's just what it means to be a doctor.
Practice Performance Partners Pick
Coding changes are easy to ignore until they cost you.
In this article, Peter Cass, OD, walks through the 2026 updates to 92137 and what they actually mean for your day-to-day billing.
If OCT is part of your clinical workflow (and it should be), this is worth a careful read.
Can you do me a favor? If you found any of these resources helpful, share this newsletter with one of our colleagues!
See you next week!
--Kyle Klute, OD, FAAO