In-hospital mortality among intracranial tumor admissions at a tertiary hospital in Luanda, Angola: a retrospective cohort study

Abstract

Abstract Background In-hospital death among patients admitted with intracranial tumors is a clinically important short-term outcome because it reflects acute neurological deterioration, treatment timing, and the capacity of hospital care pathways. Outcome reporting from African neuro-oncology settings remains limited. Methods We performed a retrospective cohort study of 93 consecutive admissions recorded between January 2020 and December 2025 at the Neurosurgery Service of Hospital do Prenda, a tertiary hospital in Luanda, Angola, which was the sole data source for this study. The unit of analysis was the admission; because the anonymized registry does not allow reliable patient-level linkage, the number of unique patients and any repeat admissions could not be determined. The primary outcome was in-hospital death. Univariable analyses examined age, sex, surgery, length of stay, registry-based diagnostic group, and admission year. Because only 21 deaths occurred (fewer than ten events per candidate variable), a pre-specified, restricted Firth penalized logistic regression sensitivity analysis was limited to three predictors: age (per 10-year increase), surgery, and a high-risk registry diagnosis category (glioblastoma or metastatic disease versus other registry diagnoses). Reporting followed STROBE and RECORD principles. Results Twenty-one of 93 admissions ended in in-hospital death (22.6%). Age ≥ 50 years was associated with death (odds ratio [OR] 4.51, 95% confidence interval [CI] 1.61–12.67; p = 0.004), and admissions ending in death were older on continuous analysis ( p = 0.019). Sex, surgery, and length of stay were not associated with death in univariable analysis. Mortality differed across registry diagnostic groups overall ( p = 0.010). In the penalized complete-case model (89 admissions, 21 deaths), high-risk registry diagnosis remained associated with death after adjustment (OR 5.70, 95% CI 1.58–20.52; p = 0.008), whereas age (OR 1.32 per 10 years, p = 0.117) and surgery (OR 0.79, p = 0.669) were not; with only 21 events, these estimates are exploratory and susceptible to residual confounding. Conclusions In this exploratory Angolan cohort, older age and a high-risk registry diagnosis were associated with in-hospital death in univariable analysis. After restricted adjustment, the association with high-risk registry diagnosis persisted, whereas the age estimate became imprecise and crossed the null; with only 21 deaths, these findings are hypothesis-generating and do not establish a ranking of determinants. The observed null association for surgery should not be interpreted causally, as patients selected for surgery likely had better performance status and more resectable tumors.

The authors' abstract, as published at the source. BMC Cancer, 2026 · DOI ↗