.svg)
.svg)
.svg)
.svg)
Already in the chart
Why hospitals keep losing patients they already diagnosed
3 min read
David Kirk, MD
Chief Medical Officer,
Regard
Regard
"A few weeks ago, we had a patient admitted who quickly died from aortic stenosis. My hospital had actually diagnosed the patient at least two years prior, when fixing it was still easy. But I never got the chance to see that patient. Regard is exactly what we need to never miss a patient like him again."
A physician told me this recently, and it's not anything new. Ask any service line leader or CMIO for a version of this story and they'll have one ready to go.
A murmur turns up during an urgent care visit, and the covering physician orders an echo. It comes back showing moderate-to-severe aortic stenosis, preserved heart function, and a normal right ventricle, making the patient a straightforward candidate for TAVR. The report recommends a referral to structural heart, but the patient has no primary care doctor to route it through, and the physician who ordered the echo has already moved on to the next patient. The referral sits in the report and goes nowhere.
Over the next eighteen months, the patient shows up three more times: twice to the emergency department for syncope, charted as dehydration, and once for overnight observation for heart failure that resolves quickly with diuresis. Each of those encounters is clinically defensible on its own. The teams in the room made reasonable calls based on what a single visit showed them, but the disease was written across four visits, and no single visit gave anyone a reason to go back and read the other three.
Why qualified patients don't get the lifesaving pocedures they need
What's worth examining is why a finding this clear doesn't reliably turn into a referral or a follow-up.
The first reason is that these findings are often unstructured data: "Moderate-to-severe aortic stenosis with preserved right ventricular function" is a sentence in an echo report, not a code in a structured field. The dehydration read on the syncope visit and the diuresis note on the heart failure admission are free text too. Registries, quality flags, and most population health tools run on structured data: problem lists, ICD codes, discrete lab values. A sentence in a radiology impression or a nurse's note doesn't show up there, no matter how many times it's charted.
The second reason is that finding these patients has depended on manual chart review: care navigators reading through a stack of notes, one patient and one condition at a time, to painstakingly piece together fragmented, unstructured data.
And while manual review works, it doesn't scale. Teams can only cover a handful of high-priority conditions this way, and even for those, they typically catch only a fraction of the patients who'd qualify for a TAVR procedure. Once a referral clears that bar, tracking whether it actually happens is a problem of its own. A referral doesn't have a deadline the way a diagnosis does. Nothing on the calendar forces the consult to get scheduled, and no one is assigned to check back if it doesn't, so it simply sits until someone happens to notice it never happened. By the time many of these patients reach a structural heart team, they arrive with severe pulmonary hypertension, a failing right ventricle, and frailty, and get labeled too sick for a procedure that would have been routine two years earlier.
How missing TAVR candidates translates to financial loss for health systems
Fragmented care does not only cost patients a missed diagnosis. It costs health systems the procedure. And the leak is larger than most leaders assume:
The loss is downstream. When a patient slips away before the procedure, the system forfeits not just the visit but the imaging, the follow-ups, and the intervention itself, which is where the margin lives.
The patients already in your building are the volume strategy.
What changes at population scale
A chart reviewer can work through unstructured notes carefully for one patient and one disease. No hospital can staff that same process across its entire population and every condition buried in free text.
The technology required should read the unstructured record the same way a clinician would, at the scale of an entire population and across many conditions at once, rather than one flagged patient and one condition at a time.
Two industry developments are raising the stakes on this at the same time:
01
Medicare is proposing a rewrite of the TAVR coverage determination that would let the heart team evaluate candidates asynchronously, by chart review, with a single required in-person visit, treating the record itself as substantive evidence for a coverage decision rather than just a starting point for one.3 A final decision is expected in September.
02
Commercial payers are tightening from the other direction. CDI leaders tell us they're seeing denials on diagnoses that appear only in an answered query, on the logic that a condition serious enough to matter should show up elsewhere in the record too.
Regulators and payers are starting from opposite ends, but landing on the same standard. The full record, not any single note or query, is what counts.
Some health systems have already built for this:
01
Kaiser Permanente's SureNet has run electronic surveillance across its membership since 2006, growing to more than 80 distinct programs with a measured success rate of 61% on the care lapses they flag.4
02
Brigham and Women's built ambulatory safety nets on the same model, routing abnormal lung imaging and colonoscopy results into tracked, navigator-driven follow-up.5
Most health systems already have someone responsible for this kind of work: a navigator, a quality nurse, a CDI specialist. The job is staffed. It's the volume that's impossible: reading every note, for every patient, across every condition.
See the patients your screening missed. Free.
Count of missed cases, dollar value attached, evidence on every finding. Under 4 hours of IT, results in under two weeks, zero commitment.
References
- Weiner M, Perkins AJ, Callahan CM. Errors in completion of referrals among older urban adults in ambulatory care. Journal of Evaluation in Clinical Practice. 2010;16(1):76–81. doi:10.1111/j.1365-2753.2008.01117.x
- Patel MP, Schettini P, O'Leary CP, Bosworth HB, Anderson JB, Shah KP. Closing the referral loop: an analysis of primary care referrals to specialists in a large health system. Journal of General Internal Medicine. 2018;33(5):715–721. doi:10.1007/s11606-018-4392-z
- Centers for Medicare & Medicaid Services. Proposed decision memo: transcatheter aortic valve replacement (CAG-00430R2). June 15, 2026
- Imley TM, et al. An electronic clinical surveillance system to reduce failure to follow up abnormal lab results and failure to order needed tests: description and methods. The Journal of Applied Laboratory Medicine. 2025;10(5):1362. doi:10.1093/jalm/jfaf073
- Emani S, Sequist TD, Lacson R, Khorasani R, Jajoo K, Holtz L, Desai S. Ambulatory safety nets to reduce missed and delayed diagnoses of cancer. The Joint Commission Journal on Quality and Patient Safety. 2019;45(8):552–557. doi:10.1016/j.jcjq.2019.05.010
