Traumatic Brain Injury at Bouaké University Hospital, Côte d’Ivoire: A Case-Mix-Adjusted Comparison of Two Consecutive Hospital Cohorts

Abstract

Background.
In sub-Saharan Africa, motorised two-wheeler transport has expanded faster than injury prevention and trauma-care capacity. Temporal changes in the profile and outcomes of traumatic brain injury (TBI) within a single referral centre remain poorly documented in sub-Saharan Africa, and crude mortality is often compared between periods without accounting for differences in case-mix.

Objective.
The primary objective was to assess whether risk-adjusted in-hospital mortality differed between two consecutive hospital cohorts. Secondary, descriptive analyses compared epidemiology, admission severity, therapeutic activity and functional outcome. We hypothesised that any rise in crude mortality would be principally attributable to a more severe casemix rather than to deteriorating care.

Methods.
Single-centre observational comparison of two consecutive hospital cohorts of patients with traumatic brain injury: a historical cohort (2016–2021; n=594) reconstructed from the departmental database, and a recent prospectively documented cohort (2025–2026; n=772). Categorical variables were compared by chi-square/Fisher tests and continuous variables by the Mann–Whitney test; analyses were exploratory and p-values interpreted cautiously. Because the two periods differed in documentation completeness, variables were pre-classified, before any comparison, as ascertainment-robust or ascertainmentsensitive. The primary mortality model adjusted for pre specified, ascertainment-robust covariates (age, sex, Glasgow Coma Scale [GCS], road-traffic crash as mechanism); a sensitivity model added anisocoria.

Results.
The core demographic characteristics remained stable between periods (median age 30 years; male predominance). The distribution of injury mechanisms shifted toward a greater predominance of motorisation-related road-traffic injuries: roadtraffic crashes rose from 75.3% to 89.4% and motorised two-wheeler involvement from 68.5% to 77.1%, with emerging threewheeler injuries, while falls and assaults declined (all p<0.001). Documented helmet use among two-wheeler users remained very low and unchanged (5.6% vs 5.2%). The recent cohort had greater neurological severity (median GCS 13 in the recent vs 14 in the historical cohort; severe TBI 7.9% vs 1.3%; p<0.001) and higher crude mortality (4.7% vs 2.0%; absolute difference 2.6 percentage points, 95% CI 0.8–4.5; p=0.009). After case-mix adjustment, no statistically significant difference in mortality remained between periods (adjusted OR 1.23, 95% CI 0.60–2.52, p=0.57), GCS being the dominant determinant (OR 0.63 per point, p<0.001); the model showed good discrimination (AUC 0.78), the Hosmer–Lemeshow test revealing no major calibration defect (p=0.47), and findings were unchanged in the prespecified sensitivity analysis (OR 1.20, 95% CI 0.58–2.50). Stratum-specific mortality was comparable for mild (1.1% vs 1.2%) and moderate (4.3% vs 5.4%) TBI. Large apparent increases in CT-documented lesions and surgical activity coincided with implausibly large increases in routinely recorded acute signs (e.g., anisocoria 0.2%→10.5%); these are reported descriptively only, as they are mainly compatible with more complete ascertainment in the recent period rather than with a real change in clinical profile.

Conclusion.
Across two consecutive periods, the profile of admitted patients shifted toward greater initial neurological severity (as assessed by the GCS) and toward motorisation-related injuries, with a rise in crude mortality. This rise was principally attributable to differences in case-mix: risk-adjusted mortality did not change. Persistently very low helmet use represents the most readily actionable target for injury prevention. Beyond its clinical findings, this study illustrates the importance of explicitly accounting for data-quality differences when interpreting historical comparisons based on hospital registries.