Outcomes of Traumatic Brain Injury and External Validation of CRASH Prognostic Model

Introduction As per World Health Organization (WHO) 2015 report, road injury is the tenth cause of mortality in the world[1]. 90% of these occur in Low and Middle-Income countries (LMICs)[2]. The complex interaction of human, vehicle and environmental factors along with lack of sustainable preventive programs has contributed to this 'silent epidemic' of injuries. Amongst Injuries, Traumatic Brain Injuries is the leading cause of morbidity and mortality.

The young working men of age 20-40 years are the most vulnerable group to suffer a Traumatic Brain Injury[3]. Around 60% of those with TBI do not resume work[3]. A study done in Delhi, India showed that patients with severe TBI who were unconscious at discharge, 50% of them died at six months and another 30% remained in vegetative state[4].

Clinicians have to answer about the prognosis of the injured patient to the anxious near ones on arrival as well as throughout the course of treatment. Researches have shown prognostic models are statistical models that combine data from patients to predict outcome are likely to be more accurate than simple clinical predictions[5]. A multicenter randomized control trial published a prediction model for traumatic brain injury patients[6]. Unlike other prediction models this model was based on data from High Income and Low Middle-Income Countries. However, the external validation of this prediction model was not done in LMICs data[7]. Hence to fill this gap we aim to study the outcome of patients with Traumatic Brain Injury and also validate the CRASH trial prediction model in traumatic brain injury patient. e.g. A 45 year old male , came with road traffic accident leading to traumatic brain injury. His Glasgow Coma Scale was E4V4M5, both the pupils reacting to light and CT scan brain suggestive of Sub-dural hematoma of thickness approximately 3 mm along left basi-frontal superior and middle frontal convexities without significant underlying mass effect. These details inserted in CRASH PROGNOSTIC MODEL calculator as shown below. As per this the mortality at 14 day is 4.4% and morbidity at 6th month is 12.6% i.e. unfavorable outcome.

Data collection Data collection will be done by a resident and a medical student of the same hospital. Patients will be prospectively followed in the ward till discharge or death. Information will be entered in the case record form based on patients file. Written, valid informed consent will be taken to follow up patients on day 14th day and 6 months after the date of injury by telephonic conversation.

Outcomes:

Outcome of the study is mortality at 14th day and mortality and morbidity at 6th months after head injury. This morbidity at 6th months will be measured by using GLASGOW OUTCOME SCALE by telephonic conversation

Variables:

Demographic Variables, mechanism of injury, transfer status, vital parameters on admission like systolic blood pressure, heart rate, Glasgow coma scale will be noted from case records. Patients pupillary response will be collected from case record sheet. Ct scan findings and whether or not patient was operated (neurosurgery) will be noted. Other injuries will also be recorded.

Sample Size The sample size for an external validation study should have at least 100 patients with the outcome and 100 without the outcomes[8,]. The short-term outcome is mortality at 14 days. The mortality in all groups of TBI is 23%. So, to have 100 events we will need a sample size of minimum 500 patients. An interim analysis will be done at one year to see if the sample size is met and a further decision on the study duration will be made.

Analysis Data will be entered in Microsoft Excel 2016 and statistical analysis will be done using SPSS version 17. We will assess the performance of the models in our dataset in terms of discrimination and calibration. Discrimination describes how well a model distinguishes between patients with and without the outcome of interest. Calibration indicates how closely predicted outcomes match observed outcomes.

To assess discrimination, we will calculate the area under the receiver operating characteristic curve (AUC). An AUC of 1 implies perfect discrimination, whereas an AUC of 0.5 implies that a model's discrimination is no better than chance.

To assess calibration, we will plot observed versus predicted outcome. Ethics Informed consent will be taken by the co-investigator of this study from the patient/relative (legal representative), after patient is admitted in the ward. The informed consent will be to follow up the patient telephonically at 14 days and 6 months from the date of injury. During course in ward or at discharge if patient is conscious and is in a state to give consent, re-consent will be taken from patient himself. If the patient refuses to give consent then his/her data will not be recorded and no follow up will be done for the purpose of study.