Health professor Neil Wenger was deep right into a years-long examine on seriously ill primary care patients when he uncovered a special however persistent problem: Many sufferers who had been focused for follow-up interventions had really died, and their hospitals didn’t learn about it.

It wasn’t what Wenger had got down to analysis. But it was unacceptable, he stated, for well being methods to lose observe of critically sick major care sufferers who’d been handled on-site for years. To be clear, these had been not one-time out-of-state sufferers who’d are available in for a single go to.

“When I found a bunch of them were dead and we didn’t actually know it, that’s when I said we need to describe it,” he stated. “This is a cohort where we need to know about everyone who is dead.”

In an try to ascertain how widespread the issue was, he examined roughly 12,000 critically sick major care sufferers from UCLA Health, considered one of three California well being methods in his earlier examine. Wenger and his co-investigators discovered that a number of hundred — a fifth of deceased sufferers in his pattern — had been marked alive in digital information, however useless in state public well being recordsdata.

Most of these sufferers had additionally been focused for various sorts of contact after they’d died: greater than 2,000 mixed cellphone calls, affected person portal messages, notices about excellent appointments and mailings about preventive care like flu photographs and most cancers screenings, amongst different messages.

Now, Wenger and his group are main a cost to eradicate these errors, urging well being methods to replace their verification course of and pressuring legislators to make dying knowledge extra accessible. He says hospitals haven’t had a lot incentive to take action till now, as the difficulty solely arises amongst a subset of sufferers — although it’s doubtless extra widespread than in his examine, which solely coated critically sick major care sufferers, suggests.

The creation of insurance coverage contracts rewarding hospitals for wholesome sufferers — referred to as value-based care in business jargon — might push them to gather extra correct knowledge, if solely to show that sufferers died elsewhere. And incomplete affected person knowledge might critically impair well being methods’ cost towards AI and predictive algorithms.

There’s no nationwide benchmark for the way typically these mix-ups occur, and Wenger seems to be among the many first to quantify them. But well being leaders STAT spoke with stated it doubtless occurs at each well being system, although the prevalence varies in line with inside backstops and state knowledge rules.

“This is a totally solvable problem, which is actually the reason we wrote the [JAMA Internal Medicine] research letter,” Wenger stated. “It’s a health system’s responsibility to be accountable for their patient and that means being able to follow when someone has passed away.”

The drawback could be solvable, but it surely’s not simple. While some medical notes do reference a affected person’s dying, the file usually isn’t labeled “deceased” till one other division verifies it, well being leaders stated. It’s particularly simple to lose observe of sufferers who’ve solely visited a well being system as soon as — although these fell exterior the scope of Wenger’s analysis — or those who died at house or in an outdoor facility or out of state.

Cross-referencing dying databases can also be a problem. The Social Security Administration’s grasp file solely accounts for individuals with social safety numbers, and states usually keep their very own dying information. Hospitals depend on these state information, and charges and accessibility range by jurisdiction. While California does keep a dying file, well being IT leaders there instructed STAT that what well being methods are given entry to isn’t detailed sufficient to totally confirm the dying of some sufferers with widespread names.

“In California, this file is available to health care facilities only in a form that contains inadequate specificity to inform clinically about death,” Wenger’s analysis letter reads. “In contrast, a more complete file is available ‘for purposes of law enforcement or preventing fraud.’”

For sufferers who aren’t in these databases however nonetheless have digital well being information, hospitals should discover different methods to determine their dying — typically by utilizing expertise that mines the web for obituary knowledge or dying information, flagging a affected person as doubtless deceased, after which manually verifying their dying by acquiring a dying certificates. Companies like LexisNexis and Veritas Data Research promote these providers.

But many hospitals merely wouldn’t have a sturdy verification course of, stated Eric Cheng, chief medical informatics officer at UCLA, the place Wenger’s analysis was carried out.

“If a patient were to call the clinic or a doctor and say a family member died, we don’t necessarily do the best job in documenting that the same way. Physicians don’t know whether that’s stored, the front desk clinic may not be comfortable if they’ve never heard of the patient — they would all document in the note, but not in the official place where it should be,” stated Cheng, who oversees UCLA’s medical information division.

“We know we should do it, but we need a little push to do it — and I think this [research] will be helpful for us to demonstrate and prioritize work on this for the rest of the organization,” he stated.

Fixing the issue may require inventing a wholly new workflow for medical and administrative employees, he stated. “We have to make it easier for them to fill out a form saying [a patient] is likely deceased, and then someone who does this more often independently confirms and does the final changing the status,” he stated. “The front desk, or provider or clinician wouldn’t bear the responsibility of getting this important decision right.”

Health system leaders instructed STAT that the difficulty has grave implications for synthetic intelligence and machine studying methods that draw on affected person information.

“We’ve tried to develop models to predict who’s likely to be deceased and those models rely on accurate data — we don’t have good data to train our own prediction models [and] they’re going to be off,” Cheng stated.

“If the system doesn’t know the patient has passed away, they’re still included in various population health cohorts,” stated Paul Fu, chief medical info officer at City of Hope most cancers middle and co-chair of informatics commerce group the American Medical Informatics Association’s well being methods committee.

Lawmakers have tried to handle the difficulty, however with little success to this point: A bill introduced early in Covid, which died within the House Energy and Commerce committee, aimed to implement higher knowledge sharing — together with dying knowledge — between the federal authorities, states and well being methods, stated Michael Hogarth, a UCSD researcher who labored with Wenger on the examine. Industry groups have also lobbied for a nationwide affected person identifier to higher match information, however the effort has confronted resistance in Congress from Republicans apprehensive about surveillance.

The march towards value-based care, and the promise of insurance coverage payouts for more healthy sufferers, might nudge hospitals to shore up their document protecting, Hogarth stated.

But till it’s fastened, “you have to read the obituary of your patient in the newspaper to know that they’ve died,” stated Deepti Pandita, chief medical info officer at UC Irvine Health and a member of AMIA’s board. “It happens more often than you think.”