A CT-based scoring system called TBI-RADS may help identify traumatic brain injury patients who can be safely discharged from the emergency department even when their scan shows a small hemorrhage, according to an observational study presented at the American Society of Functional Neuroradiology meeting.
Among 7,590 patients seen at a single academic health system from 2021 through 2025, those classified as TBI-RADS III—defined by minor subarachnoid hemorrhage, subdural hematoma under 5 mm, and no midline shift—had a 0.9% rate of requiring any cranial neurosurgical procedure. That was identical to the rate for patients with benign or normal scans (TBI-RADS I-II) and far below the 11.3% and 61.7% rates for more severe categories. Nevertheless, the mild-hemorrhage group was hospitalized for a median of 47 hours, totaling 4,390 bed-days over five years.
The study's lead author, Sumit N. Niogi of Weill Cornell Medicine, said these patients are often admitted to the ICU for monitoring but might be safe to go home or to a lower level of care. However, he acknowledged that discharging a group with a 0.9% chance of needing surgery is too risky without better ways to identify the rare patient who deteriorates. The researchers used a secure AI model to translate CT reports into TBI-RADS categories, and they plan to validate the approach at other institutions and in a prospective trial.
The study lacked neurological exam data, which would likely be required in practice, and did not capture procedures after transfer to another hospital. The authors also noted that hemorrhage expansion often signals an underlying cause such as coagulopathy or medication, suggesting that excluding those patients could make early discharge safer.