Machine Learning Algorithms to Predict In-Hospital Mortality in Patients with Traumatic Brain Injury

Abstract

Traumatic brain injury (TBI) can lead to severe adverse clinical outcomes, including death and disability. Early detection of in-hospital mortality in high-risk populations may enable early treatment and potentially reduce mortality using machine learning. However, there is limited information on in-hospital mortality prediction models for TBI patients admitted to emergency departments. The aim of this study was to create a model that successfully predicts, from clinical measures and demographics, in-hospital mortality in a sample of TBI patients admitted to the emergency department. Of the 4881 TBI patients who were screened at the emergency department at a high-level first aid duty hospital in northern Taiwan, 3331 were assigned in triage to Level I or Level II using the Taiwan Triage and Acuity Scale from January 2008 to June 2018. The most significant predictors of in-hospital mortality in TBI patients were the scores on the Glasgow coma scale, the injury severity scale, and systolic blood pressure in the emergency department admission. This study demonstrated the effective cutoff values for clinical measures when using machine learning to predict in-hospital mortality of patients with TBI. The prediction model has the potential to further accelerate the development of innovative care-delivery protocols for high-risk patients.

Enhancing hospital course and outcome prediction in patients with traumatic brain injury: A machine learning study

“…As in our study, GCS was the most or second most significant factor in three of these studies. 29,33,34 UCH-L1 and GFAP were among our study’s top five most predictive factors, again indicating these biomarkers' significance.…”

“…Furthermore, there are studies describing models to predict the functional outcome or The Glasgow Outcome Scale-Extended, 5,[25][26][27][28][29] and more recent studies using images as inputs. 8,30,31 Furthermore, there are studies in the literature similar to ours that describe models for predicting in-hospital mortality, 29,[32][33][34][35] early mortality, [36][37][38] discharge position, 39,40 need for hospital admission, 6 emergency neurosurgery, 41 and length of hospital stay. 4 In addition to contributing to the body of knowledge by describing the efficacy of incorporating ML into patient care to predict multiple outcomes simultaneously in TBI patients, this study is unique since it has used blood biomarkers such as GFAP and UCH-L1, and non-contrast CT CDEs as input variables.…”

“…32 Moreover, the best-performing model of Hsu et al in the test set yielded an accuracy of 0.93, a PPV (precision) of 0.93, and a sensitivity (recall) of 0.93. 33 Finally, the best-performing model of Rau et al was artificial neural network-based and yielded an accuracy of 0.92, a sensitivity of 0.84, and a specificity of 0.93 in the test set. 34 While it is not ideal to directly compare models based on these results, our model produced comparable outcomes, except for sensitivity, due to the small number of deceased patients in our test set.…”

An interpretable neural network for outcome prediction in traumatic brain injury

“…In recent years, several studies have adopted machine learning methods to predict mortality in patients admitted with TBI [ 1 , 2 , 10 , 14 , 17 , 19 ]. However, these studies focused on in-hospital or early mortality.…”

“…Sometimes "self-fulfilling prophecies" [4] are caused by early withdrawal of treatment when a prognostic factor is found, whose relevance is then reinforced by the outcome. In recent years, several studies have adopted machine learning methods to predict mortality in patients admitted with TBI [1,2,10,14,17,19]. However, these studies focused on in-hospital or early mortality.…”