If you’ve ever submitted a quality measure and then wondered later whether your abstraction would hold up under a CMS audit, you already understand the problem.
It’s not that the evidence isn’t there. It’s that finding it takes too long. And by the time your team surfaces it, the window to act has already closed.
After working with more than 1,800 healthcare organizations, we’ve seen this pattern repeat regardless of hospital size, EHR system, or staffing model. Quality teams don’t struggle most with understanding measure logic. They struggle with finding the evidence required to apply it.
Chart-abstracted measures have always been expensive. But the stakes have changed.
Four forces are making the evidence problem harder than anything quality teams have dealt with before:
The evidence problem was manageable when quality was primarily a compliance exercise. It’s much harder when quality determines reimbursement, accreditation, and public reputation all at the same time.
A 2023 JAMA study at Johns Hopkins found that chart-abstracted measures cost $33,871 per metric per year in personnel alone. Electronic measures? $1,902. That’s an 18-fold difference, and it has nothing to do with measure complexity. It’s about where the evidence lives and how hard it is to reach.
But the personnel cost is only the visible part.
The invisible costs are just as significant:
eCQMs delivered on their promise, for the measures they cover. Labor dropped. Cost per metric fell dramatically.
But some questions can’t be answered with structured data. Consider the difference:
Was a lactate drawn? Structured data answers that in seconds.
Was a lactate drawn within three hours of the moment a patient first met severe sepsis criteria, on a case where three physician notes contradicted each other over six hours? That requires reading the chart.
The first question is an eCQM. The second is SEP-1, the only chart-abstracted clinical process measure required for CMS IQR in 2026, now in the HVBP safety domain. Trained abstractors agree on SEP-1 time zero in fewer than 40% of cases. That one disagreement cascades to every downstream bundle element. Two abstractors. Same chart. Wildly different compliance outcomes.
The measures staying chart-abstracted longest are exactly the ones where clinical narrative can’t be replaced. The evidence problem isn’t going away. It’s concentrating.
We hear a version of the same story from quality teams across the country.
A team of two or three people. Abstracting only the measures required for CMS reporting. Sampling where regulations allow it. Staying current, barely, by accepting every accommodation available. And calling that a success, because by the standards the field has set, it is.
But consider what stays hidden. The documentation patterns that won’t surface until a validation request arrives. The institutional knowledge that walks out the door when someone leaves. The bar that has quietly dropped to clearing the minimum.
That’s what a decade of unsolved retrospective quality looks like from the inside.
AI-assisted abstraction separates the retrieval problem from the judgment problem.
The result isn’t just faster abstraction. It’s earlier visibility into performance, more consistent results across reviewers, less validation risk, and abstractors doing the work only they can do.
One thing matters more than speed: defensibility. An AI-suggested value you can’t trace back to its source document isn’t an improvement over manual abstraction. It’s a different kind of risk, one that goes undetected until CMS or TJC decides to check.
The right model is AI that surfaces evidence, links every suggestion to its source document, flags uncertainty rather than papering it over, and keeps the abstractor in control of the final decision. The audit trail gets built before submission, not assembled after a challenge arrives.
Medisolv has supported quality reporting across more than 1,800 healthcare organizations for more than 25 years. Through our acquisition of Health Elements AI, we’re building AI-assisted abstraction directly into the workflows quality teams already use.
AI-assisted registry abstraction for CathPCI, STS, and GWTG Stroke, with additional registries supported.
ENCOR for Hospital Abstracted Measures AI-assisted chart abstraction enters Early Access, starting with the measures where the evidence problem is hardest.
In both cases, the design principle is the same: AI surfaces the evidence, the abstractor validates the result, and every decision is traceable. Because that’s what defensible quality abstraction actually requires.
Many quality leaders think they’re managing measures. Increasingly, they’re managing evidence. Measures are simply the output.
The organizations that shift from retrospective evidence retrieval to earlier visibility won’t just report quality better. They’ll improve it. And that’s what quality was always for.