Factors Independently Associated with the Surgical Indication for Traumatic Extradural Haematoma in a Sub-Saharan Neurosurgical Centre: An Analytical Observational Study

Background. Extradural haematoma (EDH) is a reversible head injury whose outcome depends on timely surgery. It occurs in roughly 2–8% of head injuries and in 5–15% of fatal ones; about 56% of cases are operated worldwide. In sub-Saharan Africa it mostly affects young men injured on motorised two- or three-wheelers, yet what drives the decision to operate there is poorly documented. Objective. To describe EDH as the principal computed-tomography (CT) lesion at Bouaké University Hospital and to identify the clinical and radiological factors independently associated with the decision to operate. Methods. Retrospective single-centre study (STROBE) of an institutional database of consecutive traumatic brain injury admissions (2017–2021: 84 patients; 2025–2026: 36 patients). We analysed patients with EDH as the principal lesion (n = 120). The primary outcome was the surgical indication, defined as the operative decision recorded in the medical file before any operation (yes/no). Death and the Glasgow Outcome Scale (GOS) were described. Groups were compared with Mann–Whitney and Fisher tests. Multivariable logistic regression (Glasgow Coma Scale [GCS], CT mass effect, anisocoria, age) identified independent factors; we checked collinearity (variance inflation factors, VIF), discrimination (area under the ROC curve, AUC, bootstrap optimism-corrected) and calibration (Hosmer–Lemeshow test). Analyses used Python (statsmodels). Results. Patients were young (median age 24 years) and mostly male (90.0%); road-traffic crashes caused 84.2% of injuries, mostly involving motorised two- or three-wheelers. Median GCS was 13 and 63.3% of injuries were mild. A surgical indication was recorded in 77 patients (64.2%), 59 of whom had a documented procedure. A lower GCS (adjusted OR 0.70 per point, 95% CI 0.53–0.92) and CT mass effect (adjusted OR 4.40, 95% CI 1.59–12.17) were independently associated with the indication; age and anisocoria were not. Discrimination was modest (AUC 0.72; optimism-corrected 0.69; pseudo-R² 0.12), the Hosmer–Lemeshow test showed no miscalibration (p = 0.35, low power at this sample size) and all VIF were below 1.5. In-hospital mortality was 3.3% (4/120). Of 109 patients with a known GOS (11 missing), 104 (95.4%) had a favourable outcome (GOS 4–5) at the last available assessment. Conclusion. In this young population, mostly injured on motorised two- or three-wheelers, the recorded decision to operate was associated chiefly with initial severity (GCS) and CT mass effect, not with age or the pupillary examination. This fits guideline criteria and is not a new decision rule. The model explains little of the decision: haematoma volume, associated open injuries and reasons for non-operation were not modelled. Mortality was low and functional recovery good. Deaths were too few to model outcome, which is better done on the whole cohort. Multicentre prospective studies should validate this decision profile in other African neurosurgical centres.