Efficiency Is Not an Outcome

At HRS2026, top EP physicians tackled the AI question the industry keeps dodging: when does saving time actually translate to better patient outcomes? In Part 1 of this three-part series, Sean Shoffstall shares his insights.

At HRS2026 in Chicago, I had the privilege of moderating a Heart Rhythm Society TV panel with two people I deeply respect: Dr. Ben Steinberg, Chief Medical Officer, PaceMate and a practicing electrophysiologist at Denver Health and the University of Colorado, and Dr. Jonathan Piccini, director of electrophysiology at Duke University and the Duke Clinical Research Institute. We gave the session a deliberately provocative title: Are we using AI to improve outcomes, or just gaining efficiencies?

I chose that framing on purpose, because in my experience the industry almost never holds those two ideas apart. Spend an afternoon on any exhibit-hall floor and you’ll hear “better outcomes” and “greater efficiency” used interchangeably, sometimes in the same sentence, often in the same breath. They are not the same thing. And the longer we allow them to blur, the easier it becomes for a product to slip past the one question that should govern everything we build in this space: did it make the patient healthier?

Two words that mean very different things

Dr. Steinberg drew the line precisely. In clinical and outcomes research, he reminded us, the targets have always been clear — are we improving mortality, are we improving morbidity, are we improving health-related quality of life? Those are outcomes. They are defined around the patient, and improving them is the job.

Efficiency is a different animal. It is also a genuine good — fewer wasted hours, faster turnaround, less administrative drag on already stretched teams. But as Dr. Steinberg pointed out, efficiency gains often accrue primarily to the benefit of the system, and only secondarily, if at all, to the patient. So his first question for any AI tool was not “does it work?” but “who is it for?” Identify the audience, and you can finally judge whether the win you’re being sold is the win the patient needs.

Dr. Piccini offered a model I keep coming back to: think of it as a gradient. On the far left sit the operational realities of clinical practice — directing faxes, handling refills, summarizing the firehose of incoming data. Over there, efficiency is almost everything that matters. Of course, if providers or patients are miserable, that’s its own kind of bad outcome — but broadly, operational AI earns its keep by saving time and attention. As you slide right, though, toward diagnosis and then toward treatment, the standard has to revert to what it has always been: cardiovascular events, survival, functionality, quality of life. The error so many vendors make — and so many of us are tempted to accept — is carrying the left-hand scorecard over to the right. A tool that’s allowed to be judged on efficiency when it’s actually making a treatment decision is a tool that’s escaped accountability.

Why we keep conflating them

If the distinction is this clear, why does the industry blur it so reliably? The honest answer is incentives.

Efficiency is cheap to measure and it shows up fast. You can stand in a booth and demonstrate, in real time, that a task that took ten minutes now takes ninety seconds. Outcomes are the opposite. They demand large cohorts, long horizons, and serious money — and even then, the results humble you. Dr. Piccini was candid about this: he has deep sympathy for anyone running these investigations, because they require enormous numbers of patients, cost a fortune, and still routinely come back with a sample size or a predictive capacity short of what you hoped for. When the efficiency claim takes a demo and the outcomes claim takes a multi-year trial, it’s not hard to predict which one ends up on the slide deck.

I want to say something here plainly, because PaceMate is a vendor and I’m not interested in pretending otherwise: beginning on the efficiency side is not the sin. It is, in fact, the responsible place to start. Dr. Steinberg made this point and I agree with it completely — many of us are appropriately starting with the question of whether we can add value without taking over, without replacing what clinicians do. That’s the right posture for a technology that still has to earn its place. The problem is never delivering efficiency. The problem is relabeling it as an outcome and hoping the audience doesn’t notice the swap.

Where efficiency earns the right to be called an outcome

Here’s the part that gives me genuine optimism, and it’s the part I think gets lost in the cynicism about AI hype. Efficiency and outcomes are not permanently separate categories. There is a specific, identifiable place where one becomes the other: when removing noise frees finite clinical attention for the signal that actually changes care.

Every device clinic in EP is living inside a data deluge. The volume of remote-monitoring transmissions has outrun the human capacity to review them, and a large fraction of what comes in is non-actionable — false positives, repetitive alerts, transmissions that generate work but never change a management decision. That noise isn’t just an inconvenience. It’s a clinical risk, because it buries the transmissions that matter under the ones that don’t, and it does so in front of a team that gets more fatigued with every dismissal.

This is exactly the seam where an efficiency gain becomes a patient outcome. A study we were proud to support, published in Heart Rhythm, examined AI algorithms applied to implantable loop recorder data and found they decrease alert burdens by 20% and reduce episode burdens by 40% among patients monitored with LINQ II devices. Read that as a pure efficiency result and it’s still impressive — fewer transmissions to process. But follow the chain one link further. When close to half of the episode burden is noise, stripping it out means the true positives surface sooner, reach a less exhausted team, and command more attention per patient who genuinely needs intervention. The faster, cleaner path to the patient who is actually in trouble is an outcome. The efficiency was the mechanism; the patient benefit was the result.

That’s the bar I’d ask all of us to hold ourselves to — and I mean vendors first. Efficiency is honest, valuable, and worth pursuing on its own terms. But it has to be named accurately, and the best tools should be able to show you the line that connects their efficiency to a patient who ends up better off.

A question to carry out of the exhibit hall

So let me leave you with the same challenge I left the panel with. The next time someone slides a number across the table, ask one thing: is this an efficiency metric or an outcome metric?

If the answer is efficiency, that’s a perfectly good answer — but then tell me how it reaches the patient. Show me the chain from “we saved your team time” to “your patients are better off.” And if the answer is an outcome, then show me the rest of it: the population it was measured in, the endpoint that was moved, and the statistical power behind the claim.

The tools that can answer honestly in both directions — the ones that don’t flinch when you ask which kind of win they’re selling — are the ones that have earned a place in your clinic. Everyone else is hoping you won’t ask.

Watch the full HRS-TV panel with Dr. Steinberg and Dr. Piccini