Objective: Prediction of long-term functional outcome in spontaneous intracerebral and intraventricular hemorrhage (ICH/IVH) is typically based on a limited number of severity factors known on admission. We investigated whether factors known at day 30 including medical comorbidities and in-hospital events in poor outcome survivors could accurately predict good vs. poor functional outcome at day 365.
Methods: We explored the relationship between ICH severity factors, medical comorbidities, prospectively collected and adjudicated events during first 30 days post ictus, and outcomes using data from the Clot Lysis: Evaluating Accelerated Resolution of Hemorrhage with Alteplase (CLEAR III) trial. We defined outcome using blinded assessment of mortality and modified Rankin scale (mRS) at 30 and 365 days (>3 signifying poor outcome). Multivariable logistic regression was used to predict mRS at day 365 from poor outcome survivors at day 30.
Results: Of 500 patients included in CLEAR III with ICH and severe IVH requiring external ventricular drainage, 345 (69%) had mRS 4 or 5 on day 30. The best performing model for distinguishing between patients with and without good outcome at 365 days included: age (Odds Ratio [OR], 0.93; 95% Confidence Interval [CI], 0.90- 0.96), COPD (OR, 0.17; CI, 0.04-0.80), diabetes (OR, 0.50; CI, 0.25-0.98), hyperlipidemia (OR, 2.07; CI, 1.01-4.27), bacterial ventriculitis (OR, 0.34; CI, 0.13-0.91), cerebral perfusion pressure <50 mmHg during first week (OR, 0.08; CI, 0.01-0.50), absent IVH on day 30 CT scan (OR, 3.08; CI, 1.69-5.62), enrollment (e) eICH volume (OR, 0.91; CI, 0.87-0.95), eIVH volume (OR, 0.97; CI, 0.95-0.99), thalamic ICH location (OR, 0.29; CI, 0.14-0.60), and eGlasgow Coma Scale (OR, 1.15; CI, 0.1.06-1.26). Cross-validated area under ROC curve (95%CI) was 0.81 (0.77-0.86) vs. 0.78 (0.73-0.82) for standard ICH prediction model with only enrollment predictors.
Conclusions: Among patients with spontaneous ICH and IVH requiring external ventricular drainage, 30-day prediction models for long-term functional outcomes have substantial predictive capability. Addition of medical comorbidities and in-hospital events to models using well-known ICH severity factors is of some value.