Skip to main content

Table 1 examples of popular TBI prognostic models

From: Prediction of in-hospital mortality in patients with post traumatic brain injury using National Trauma Registry and Machine Learning Approach

Model Applies to Objective(s) Variables Performance
Trauma Injury Severity Score (TRISS) Trauma patients treated at hospitals with or without TBI [5] Calculates the probability of survival Age, revised trauma score (GCS, systolic blood pressure, respiratory rate), trauma type and Injury severity score (ISS) [5] • Good discrimination power
• Not specifically designed for TBI [6]
• Prone to poor performance in severe TBI [5]
• AUC in previous studies: 89% [5], 90% [7] and 92% (8)
The International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT) Adult patients (age ≥ 14 years) with TBI and GCS ≤ 12 Predicts the 6-month mortality and unfavorable outcomes [5] Age, GCS motor scale, pupils reactivity, hypoxia, hypotension, CT results (epidural or subarachnoid hemorrhage), lab values (blood glucose level and hemoglobin concentration) [8] • Good discrimination power
• Accurate outcome prediction when large sample size is utilized [5, 8]
• Poor precision at the individual patient level [9]
• AUC in previous studies: 80% [10], 83% [11], 85% [7] and 86% [8].
Corticosteroid Randomization After Significant Head injury (CRASH) Adult patients (age ≥ 16 years) with TBI and GCS ≤14 (9) Predicts the probabilities of 14-day mortality and 6-month unfavorable outcome [8] Age, GCS, Pupils reactivity, major extracranial hemorrhage and CT findings (midline shift, obliteration of third ventricle, subarachnoid hemorrhage, petechial hemorrhage, and non-evacuated mass) [10] • Good discrimination power [8]
• Accurate outcome prediction when large sample size is utilized [8, 10]
• Poor precision at the individual patient level [9]
• AUC in previous studies: 86% [7], 87% [8] and 89% [10]
Marshall scale Patients who sustained TBI Grades the TBI and predicts the TBI outcomes on the basis of CT scan findings Presence of mass lesion, midline shift, and status of the peri mesencephalic cisterns • Simple to use
• Reasonable discrimination power
• Narrow scope (limited to 3 variables)
• Limited applicability to clinical practice [12]
• AUC in previous studies: 71% [13], 63.5% [14] and 78% [12]
Rotterdam CT scoring Patients who sustained TBI Grades the TBI and predicts the TBI outcomes on the basis of CT scan findings Presence of mass lesion, midline shift, status of the peri mesencephalic cisterns and the presence of traumatic intra-ventricular or sub-arachnoid hemorrhage (tSAH) [13] • Reasonable discrimination power
• Does not differentiate between the type and size of the mass lesion [12]
• AUC in previous studies: 69.8% [14] 84% [12] and 85% [15]
Helsinki Computerized Tomography Score Chart Patients who sustained TBI Grades the TBI and predicts the TBI outcomes on the basis of CT scan findings Mass lesion type, Mass lesion size, presence of intraventricular hemorrhage, suprasellar cistern • Superior to Marshall and Rotterdam scales
• Good accuracy and discrimination power
• Lower performance when used alone as a predictive method [12, 14]
• Reported AUC: 71.7% [12] and 74.6% [14],