Abstract
Globally, traumatic brain injury (TBI) is a leading cause of death and disability. Adults aged 60 years and over are increasingly contributing to the TBI patient population, typically due to falls at standing height. Most adults over the age of 60 with a TBI present to the emergency department (ED) via the ambulance service. However, the care received by older adult TBI patients is suboptimal, caused by undertriage due to presentations that do not reflect the underlying injury severity. This body of work aims to develop a clinical decision rule (CDR) that could support ambulance paramedics in identifying older adults with a neurosurgically important head injury.
A systematic literature review with narrative synthesis was conducted to determine which variables were suitable for building the CDR. A number of variables were identified, and a unique dataset was created to derive the CDR by retrospectively linking ambulance and ED patient care records. Patients in the dataset were adults aged 60 years and older who presented to one regional ambulance service in the United Kingdom with a head injury between the 1st of January and 31st of December 2020. Data were collected on the preidentified clinical variables, and stepwise variable selection was performed to determine which variables should be used in the CDR. A binary logistic regression model using cross-validation resampling methods trained and tested the CDR.
The final dataset included 3,545 patients who presented to the ambulance service in 2020 with a head injury; 2,111 were conveyed to the hospital, and 162 were found to have a TBI, of whom eight were accepted by neurosurgery. The low number of outcome events results in a change of the primary outcome to a TBI on an intracranial bleed found on a head computed tomography (CT) scan. Clopidogrel, previous myocardial infarction, chronic kidney disease, focal neurological deficit and falls from more than two metres were associated with an increased likelihood of a TBI. A small proportion (38.27%, n=62) of patients with a TBI presented with no head injury symptoms.
Following analysis, the CDR was derivated and internally validated, and it was found that the model could correctly classify most patients as not having a TBI. However, it could not detect patients with a TBI classified as false negatives. This resulted in the CDR producing reasonable performance metrics but casts doubt on its clinical utility.
Identifying older adults with a TBI remains challenging; not all older adults with a TBI may present with symptoms. While clinical guidelines are cautious regarding patients taking anticoagulant medication, consideration should be given to patients taking Clopidogrel. These new insights presented in this thesis could impact clinical guidelines and policy, influencing which older adults with a head injury are conveyed to the ED.