The Prescription Got to the Pharmacy. So Why Can’t It Be Filled?

Prescription abandonment is often blamed on the price tag. But millions of scripts never even get a chance to be rejected for cost. Instead, they get caught in a maze with no clear exit between the EHR and the pharmacy counter.

The clinical decision was correct, but the transaction failed. Now you've got a frustrated patient, an overburdened pharmacist, and a provider doing rework for a script they thought was handled. This breakdown happens millions of times a day, and it’s entirely preventable.

What’s at the root of this problem? How is machine learning changing the equation? And what does the latest round of federal regulation mean for EHR vendors navigating these issues? DrFirst experts Colin Banas, M.D., M.H.A., Chief Medical Officer, Nick Barger, PharmD, VP of Provider Solutions, and Tyler Wince, VP of Specialty Solutions, tackled these questions and more.

What’s a “dirty prescription?”

Barger: A dirty prescription is one that can’t be dispensed. Even though it is clinically correct, there’s something about it that doesn’t meet the downstream requirements of the pharmacy or payer. It could be a missing diagnosis code, the wrong unit of measure, or a quantity that exceeds what the insurance plan allows for a given period. None of those are clinical errors, but every one of them can stop a prescription cold.

Dr. Banas: From a physician’s perspective, it’s deeply frustrating. You’ve done everything right. You’ve made the decision, written the script, moved on. Then the patient shows up at the pharmacy and is told there’s a problem—and now you’re being asked to fix something you didn’t know was broken.

The rules governing prescription formatting have almost nothing to do with clinical decision-making. Providers are being asked to play a game they haven’t been told the rules to, and patient outcomes are on the line.

What about the prior authorization problem?

Wince: It’s the same root problem, just further downstream. Some prior authorizations get triggered because a diagnosis code was missing, the wrong one was selected from the patient’s chart, or the units of measure didn’t match what the PBM was expecting. It could be a specialty medication that got sent to a retail pharmacy that can’t fill it. None of these are clinical problems. It’s a data problem that piles up into a massive administrative burden for everyone involved.

And then, after all that work, it turns out the PA wasn’t required in the first place. The claim and the real-time benefit check both indicated that it was necessary, but it wasn’t. That’s a lot of wasted effort that traces back to something that could have been caught much earlier in the process.

Barger: And that’s really a key insight. Most of this friction is preventable if you catch it upstream. The dirty prescription, the unnecessary PA, the pharmacy routing issue—none of these things are inevitable. They’re the result of information that exists somewhere in the system not being available to the provider in the moment, when they’re making the prescribing decision.

What does solving this look like?

Wince: At DrFirst, we’re focused on assisted prescribing. Machine learning helps surface the right information during the prescribing process itself, not after the claim hits the pharmacy. The models are trained to identify the specific patterns that generate dirty prescriptions and prior authorization triggers: mismatched codes, formulary exceptions, dosage inconsistencies, pharmacy routing issues. That intelligence gets pushed forward into the prescribing workflow, so providers have what they need to get it right the first time.

Barger: Health IT systems have historically been built to tell doctors what they can’t do, like those yellow yield signs that pop up in the workflow to flag potential errors. That creates alert fatigue. Providers start ignoring alerts, including valid ones, because the signal-to-noise ratio is so poor. Machine learning lets us flip that model: Instead of flagging what’s wrong after the fact, we guide the workflow toward a clean prescription before the problem exists. I’d rather doctors practice medicine than have to learn the rules of pharmacy dispensing just to avoid rework.

Dr. Banas: That’s exactly the right framing. If it’s good for doctors, it’s probably good for patients. Reduce the friction, surface the right information at the right moment, and let providers practice medicine instead of managing administrative fallout.

The HTI-5 comment period just closed, so a Final Rule is imminent. What’s the headline for EHR vendors?

Barger: There are actually two headlines, and vendors need to pay attention to both. The first is that HTI-5 is a signal that this administration is reducing certification burdens. This means EHR vendors can focus resources on innovation rather than compliance checkboxes. A lot of what’s being removed, such as drug-interaction checking or certain privacy and security requirements, are already the de facto market standard anyway. They don’t need to live in a certification framework anymore because everyone already does them.

The second headline is that the HTI-4 e-prescribing deadlines didn’t move. Formulary and Benefit (F&B) version 60 and real-time prescription benefit (RTPB) certification are due by the end of 2026. The NCPDP SCRIPT 2023 standard upgrade follows in 2027, with a 2028 deadline. If you were hoping HTI-5 gave you a reprieve on any of that, it didn’t. Those are hard stops, and vendors who aren’t already working on them need to be.

Dr. Banas: The cumulative roadmap is real. You’re going to be in some phase of e-prescribing upgrade between now and the middle of 2028. For EHR vendors whose core competency is clinical workflow, not prescription transaction infrastructure, that’s a significant drain on engineering and product resources. The smarter path for many of those vendors is to partner with specialists who have already built, certified, and maintained this infrastructure. That frees their teams to focus on what they do best.

HTI-5 proposes removing the so-called “AI nutrition label” requirement for AI algorithms. Is that a concern?

Barger: The nutrition label model was intended to give users visibility into how an AI algorithm reached a recommendation. It is a transparency layer baked into certified health IT, but its removal makes sense in context. The AI ecosystem is moving so fast that locking down a specific transparency format was always going to create drag, and the requirement was arguably already outdated by the time it was written.

But removing the label doesn’t remove the trust problem. Healthcare organizations are not going to deploy AI just because the certification burden is lighter. Every health system we interact with has an AI governance function. Nothing gets through the door without a rigorous review—is it accurate, is it safe, is it timely? The market is self-regulating in ways the rule doesn’t need to mandate.

Wince: And from a technical standpoint, the removal creates real opportunity. Large language models have unlocked a new category of data that traditional ML couldn’t meaningfully touch, things like clinical notes, ambient listening, complex chart relationships. Being able to deploy those models faster, without a prescriptive certification framework slowing things down, is significant. That’s where the next generation of unstructured workflow automation in the EHR is going to come from.

So what should EHR vendors do with this moment?

Barger: Don’t mistake the regulatory pause for a finish line. The vendors who win in this environment aren’t just the ones who clear the compliance bar, they’re the ones who use the breathing room to build something genuinely better.

Wince: The dirty prescription problem is solvable, and so is the unnecessary prior authorization and the alert fatigue that’s eroding physician experience. The technology exists and the regulatory environment creates space to solve them.

Dr. Banas: The opportunity here is to get the compliance work off your plate so you can focus on the innovation. The regulation is clearing the runway, but the question is: What are you going to build on it?

Want to learn more?

Read more about HTI-5 and HTI-5 in Barger’s latest Regulatory Talk article published by Healthcare IT Today. If you’d like to learn more about our prescription orchestration platform and electronic prior authorization solutions, reach out for a demo here.