What Leads to Operating on a Spontaneous Intracerebral Haemorrhage: Secondary Analysis of a Prospective Cohort of 149 Patients at Bouaké University Hospital (Côte d'Ivoire)

Background. Whether to operate on a spontaneous intracerebral haemorrhage (ICH) is still debated, and the choice rests largely on the team’s judgement. Little is known about the criteria that actually tip the balance, especially where resources are limited. We looked at what leads neurosurgeons to recommend surgery, not at whether surgery works. Methods. We analysed 149 consecutive patients with spontaneous ICH from a prospective cohort of 250 strokes managed in the neurosurgery department of Bouaké University Hospital from 1 January 2022 to 31 December 2023. The outcome was the surgical indication recorded in the chart. Six factors were tested by logistic regression: age, Glasgow Coma Scale (GCS), haematoma volume, mass effect, intraventricular haemorrhage and hydrocephalus. The main model used the 127 patients with complete data. Haematoma location, almost never recorded, could not be studied. We assessed discrimination (AUC), calibration (Hosmer-Lemeshow test) and collinearity (VIF), then checked robustness by multiple imputation and in the whole cohort with a known GCS (n = 219). Results. A surgical indication was made in 77 patients (51.7%): 40 were operated on and 37 were not. Three factors remained associated with the indication after adjustment: mass effect (adjusted odds ratio [aOR] 3.19; 95% CI 1.33–7.63; p = 0.009), hydrocephalus (aOR 3.70; 95% CI 1.31–10.49; p = 0.014) and intraventricular haemorrhage (aOR 2.42; 95% CI 1.04–5.63; p = 0.041). Age (aOR 0.93 per 10 years; p = 0.59), GCS (aOR 0.91 per point; p = 0.20) and haematoma volume (aOR 0.88 per 10 mL; p = 0.32) were not. Discrimination was moderate (apparent AUC 0.76; 0.72 after bootstrap correction) and calibration satisfactory (Hosmer-Lemeshow p = 0.55; VIF < 1.4). After multiple imputation, mass effect (aOR 3.06) and hydrocephalus (aOR 5.02) remained significant. Intraventricular haemorrhage kept the same direction without reaching significance (aOR 1.96; 95% CI 0.90–4.28; p = 0.091). In the 219 patients with a known GCS, all three signs remained significant (AUC 0.75). Conclusion. In this Ivorian cohort, the decision to operate follows the signs of mechanical impact of the bleed — mass effect, hydrocephalus and intraventricular haemorrhage — more than haematoma volume, GCS or age. These findings show how the decision is made in practice when resources are limited.