The Problem
Each year, an estimated 5 million Americans seek emergency medical care for traumatic brain injury (TBI)1, a major cause of death and disability overall and a leading cause in young adults2. TBI treatment is extremely time-sensitive, with better outcomes associated with expedited care, necessitating early diagnosis and categorization of injury3; if secondary injury (inflammation, bleeding, neuron death) is not arrested, patients can be left with profound mental and physical disability. While treatment guidelines are well-established for cases of severe trauma, they are less clear for mild to moderate TBI4. The latter are much more difficult to assess, as traditional CT evidence for TBI can take many hours to materialize5,6, if at all7, with studies suggesting that for every positive mild TBI diagnosis, one goes undiagnosed (a “silent epidemic”)8. A common strategy when diagnosing mild to moderate TBI is thus to admit borderline cases for further monitoring, but the sensitivity and specificity of this process is low9.
Our Solution
To reduce the impact of TBI misdiagnosis on patient outcomes and reduce readmission rates within DUHS, we aim to build a deep neural network model that predicts with high accuracy whether a patient will experience an admittible TBI complication, without sacrificing interpretability.
We believe that this will result in more directed, well-informed care for acute TBI needs; reduced rates of re-admissions for “false-negatives”.
References
- Korley FK, Kelen GD, Jones CM, Diaz-Arrastia R. Emergency Department Evaluation of Traumatic Brain Injury in the United States, 2009–2010. J Head Trauma Rehabil. 2016;25(5):1032-1057.
- Centers for Disease Control and Prevention USD of H and HS. Centers for Disease Control and Prevention (2019). Surveillance Report of Traumatic Brain Injury-related Emergency Department Visits, Hospitalizations, and Deaths—United States, 2014. 2019. www.cdc.gov/TraumaticBrainInjury.
- Moppett IK. Traumatic brain injury: Assessment, resuscitation and early management. Br J Anaesth. 2007;99(1):18-31.
- Carney N, Totten AM, O’Reilly C, et al. Guidelines for the Management of Severe Traumatic Brain Injury, Fourth Edition. Neurosurgery. 2017;80(1):6-15.
- Servadei G. Giuliani, A. Maria Cremonini, P. Cenni, D. Zappi, G. S. Taylor, F. MTN. CT prognostic factors in acute subdural haematomas: the value of the “worst” CT scan. Br J Neurosurg. 2000;14(2):110-116.
- Menditto VG, Lucci M, Polonara S, Pomponio G, Gabrielli A. Management of minor head injury in patients receiving oral anticoagulant therapy: A prospective study of a 24-hour observation protocol. Ann Emerg Med. 2012;59(6):451-455.
- Yuh EL, Cooper SR, Mukherjee P, et al. Diffusion tensor imaging for outcome prediction in mild traumatic brain injury: a TRACK-TBI study. J Neurotrauma. 2014;31(17):1457-1477.
- Bazarian JJ, Veazie P, Mookerjee S, Lerner EB. Accuracy of mild traumatic brain injury case ascertainment using ICD-9 codes. Acad Emerg Med. 2006;13(1):31-38.
- Nishijima DK, Haukoos JS, Newgard CD, et al. Variability of ICU use in adult patients with minor traumatic intracranial hemorrhage. Ann Emerg Med. 2013;61(5):509-517.e4.
