The "A-Eye" Advantage in Ophthalmology

Source: ChatGPT modified by NostaLab.

A new study in Nature Medicine offers a glimpse of what happens when doctors work side by side with artificial intelligence. The trial compared ophthalmologists making diagnoses on their own to those using an AI copilot called EyeFM. The results were striking. Physicians with AI support reached a diagnostic accuracy of 92 percent, compared with 74 percent without it. This difference in diagnostic accuracy was reported as statistically significant with p < 0.001. And in this study, p might really stand for patient!

These differences, statistical and clinical, are not subtle. They suggest that AI can move clinical practice beyond incremental gains toward a fundamental shift in overall care. For patients, it means more confidence that their condition is being identified early and accurately. For clinicians, it means a trusted "tech partner" who can sharpen their clinical judgment in real time. And in the context of ophthalmology, this isn't simply the addition of a completely new dynamic but the expansion of an already tech-centric practice.

Beyond Accuracy

The trial also showed that AI changes more than the diagnostic endpoint. Patient behavior shifted, too. Those treated in the AI-assisted arm of the study were more likely to comply with follow-up visits and to act on referral suggestions. That detail may seem secondary, but in practice, patient compliance is critical. A correct diagnosis matters little if a patient fails to provide the appropriate level of self-management. Here, the presence of AI seemed to reinforce trust and engagement and created a subtle but important change in how patients responded to their doctors.

This is where psychology may enter the story. The effect of AI is not limited to algorithms and images. It reaches into the human domain to potentially include elements of trust and confidence. A physician’s words carry weight, but when paired with the quiet (and validated) authority of an AI, they may carry more.

An Unasked Question

The study tells us what happens when AI is used. What it does not tell us is what happens when AI is removed. Do physicians return to their baseline level of performance, or does accuracy fall below it? This is the question of AI Rebound, a concept I wrote about recently in the context of colonoscopy. In that field, evidence suggests that doctors can underperform once the AI is withdrawn. Rather than returning to normal, they dip lower.

There is also a practical dimension here. Many clinicians do not always work in the same fully equipped setting. A physician who is supported by advanced technology in one office may practice elsewhere without it. The same doctor could look very different depending on the environment. For patients, that variability can create both opportunity and vulnerability.

Could the same fragility emerge in ophthalmology? The EyeFM study did not address this, but the possibility remains. If it does, then AI is not only augmenting performance in the moment. It may also be reshaping how physicians think and act, with consequences when the technological augmentation is pulled away.

A Vision for the Future

The EyeFM trial is important, and I think it's fair to extrapolate this beyond eye disease, at least hypothetically. This is part of a broader transformation in medicine, as AI moves from an option to an imperative. It changes accuracy, behavior, and perhaps even cognition itself. The challenge is not simply to celebrate better numbers, but to understand how this new layer of intelligence affects the human beings who use it and the patients who live with its outcomes.

This is the beginning of a new dynamic in medicine. It's not physician versus machine, and not physician replaced by machine, rather, physician plus machine, creating a "composite intelligence" that is more than either alone.