— Observational study points to high level of care for low-risk group with mild hemorrhage
October 13, 2026 • 3 min read
A CT classification system appeared to distinguish patients safe for early discharge from the emergency department after traumatic brain injury (TBI) despite a positive head CT, a large observational study showed.
It's standard practice to admit almost all such TBI patients for intensive care unit (ICU) or step-down observation and a repeat CT 12-24 hours later, and in the study, the rate of admission from the emergency department jumped from less than 35% with a benign or normal CT scan to 74.1% with even minor findings.
However, only 0.9% of these mild hemorrhage patients -- categorized as TBI-RADSopens in a new tab or window III due to minor subarachnoid hemorrhage, subdural hematoma less than 5 mm, and no shift -- went on to get any cranial neurosurgical procedure.
That rate was the same for the TBI-RADS I-II patients who had benign or normal scans after TBI, and "far below" the 11.3% rate in the category IV patients with moderate CT findings and the 61.7% rate in the more severe patients, Sumit N. Niogi, MD, PhD, of Weill Cornell Medicine in New York City, reported at the American Society of Functional Neuroradiologyopens in a new tab or window meeting in Newport Beach, California.
Nevertheless, the TBI-RADS III patients were hospitalized a median of 47 hours, adding up to 4,390 bed-days over 5 years for patients who never had a cranial procedure at the single center studied.
"These are people that have that minor hemorrhage that today would go to the ICU for monitoring that are probably safe to go home or safe to go to ... a step-down unit or regular hospital bed," said Niogi.
The TBI-RADS system isn't in routine use. This study was part of a proof of concept that it could be used to potentially separate patients into different risk groups that might be appropriate for early discharge.
It's a resource allocation problem, Niogi said. "You have these people that are otherwise pretty healthy and they're taking an ICU bed slot away from somebody that's probably more critical."
Just sending a group of patients home in which 0.9% would end up needing cranial procedures is too risky, he said.
More work needs to be done to tease out another metric to further stratify out the higher risk TBI-RADS III patients, perhaps based on cognitive deficit or family history, he acknowledged.
"Historically, the concern was that tiny amount of hemorrhage can balloon, expand into something very critical," he said. "When you see something like that expand, there's usually another underlying cause. They have a coagulopathy, they're on some sort of medication that puts them at higher risk for bleeding. So if you kind of take away those patients, there's really no good evidence for why they need something as elevated as an ICU bed."
His group examined data from 7,590 patients who came to a single academic health system via the emergency department from 2021 through 2025 with a TBI coded on the encounter and a non-contrast head CT ordered at that encounter. A secure artificial intelligence model was used to translate the initial CT report into TBI-RADS classification with two independent reads per report.
Of the patients, 50.6% had normal or benign CT findings classified as TBI-RADS I-II, while 11.9% had mild hemorrhage by TBI-RADS III criteria, and 36.2% had moderate or severe hemorrhage classified as IV-V. Category mix was stable from 2022 through 2025.
Limitations were lack of neurological exam data, whereas no neurological deficit would likely be required to use TBI-RADS in practice, and no information on procedures after transfer to another hospital.
The next phase will be expanding the retrospective study out to other institutions to confirm the findings and then "a very careful prospective trial where we're following people that are getting monitored anyway, but saying, 'Would this person have been safe if we discharged them?'" Niogi said. The group is already working to combine their single-center outcome data with quantitative analysis of the CTs to see if there's a threshold volume of hemorrhage that correlates with whether or not it is safe to discharge.