Factors associated with in-hospital mortality after stroke at a tertiary hospital in Angola.

In-hospital mortality after stroke at Hospital do Prenda was 15.7% among 479 episodes.

• 75 patients died among 479 hospitalisation episodes (15.7%; 95% CI 12.7-19.2).
• Single-centre retrospective cohort at a public tertiary hospital in Luanda, January 2023 to December 2025.
• Four episodes ending in transfer were excluded; the unit of analysis was the hospitalisation episode.

Decedents were older than survivors and had a much shorter hospital stay.

• Median age was 58.5 years among decedents versus 52.0 among survivors (p = 0.049).
• Median length of stay was 3 versus 6 days (p < 0.0001).
• The authors state the length-of-stay difference reflects early mortality rather than a prognostic exposure.

Mortality rose in a graded manner with age and remained graded after adjustment.

• Mortality increased from 6.7% at age 45-54 years to 28.0% at ≥75 years.
• For ≥75 years, adjusted OR was 5.92 (95% CI 2.28-15.38; p = 0.0003).
• Adjusted model was exploratory, covering age, sex, year, stroke subtype, recorded comorbidities, and head-CT status.

Patients younger than 45 years also had elevated mortality, concentrated in haemorrhagic stroke.

• Mortality among patients <45 years was 17.2% (adjusted OR 2.97, 95% CI 1.28-6.87).
• Within this group, mortality was 22.2% for haemorrhagic versus 10.8% for ischaemic stroke.
• Age-by-subtype interaction p = 0.0009.

Sex, year, stroke subtype, recorded hypertension, and recorded diabetes were not associated with in-hospital mortality.

• These variables were assessed in univariable and exploratory multivariable analyses.
• Hypertension and type 2 diabetes were extracted from free-text fields using predefined string rules.

Absence of head CT was associated with higher crude mortality, though the estimate was imprecise.

• Crude mortality was 30.0% without head CT versus 15.0% with it.
• OR 2.42 (95% CI 0.90-6.52), with a confidence interval including 1.

Recorded hypertension prevalence was low, indicating substantial under-ascertainment of comorbidities.

• Recorded hypertension prevalence was only 19.8%.
• Comorbidities were extracted from free-text fields, which limits interpretation of their null associations.

Statistical and data limitations mean adjusted estimates should be viewed as exploratory associations rather than independent prognostic effects.

• Events-per-variable ratio was low (~7) and calibration was poor in some models.
• No data on stroke severity, functional status, ICU care, ventilation, or reperfusion.
• Single-centre retrospective design limits causal inference.

Most Angolan stroke patients reach hospital beyond the thrombolysis window and many deaths occur within 72 hours of admission.

• This is the stated rationale for investigating routinely recorded admission variables associated with mortality.
• The short median stay among decedents (3 days) is consistent with early deaths.