Predictive accuracy of clinicians estimates of death and recovery after acute intracerebral hemorrhage: pre-specified analysis in INTERACT3 study

Introduction: Accurately predicting a patient’s prognosis is an important component of decision-making in intracerebral hemorrhage (ICH). We aimed to determine clinicians’ ability to predict survival, functional recovery, and return to premorbid activities in patients with ICH. Methods: Pre-specified secondary analysis of the third intensive care bundle with blood pressure reduction in acute cerebral hemorrhage trial (INTERACT3), an international, multicenter, stepped-wedge cluster randomized controlled trial. Clinician perspectives on prognosis were collected at hospital admission and Day 7 (or before discharge). Prognosis questions were the likelihood of (i) survival at 48 h and 6 months, (ii) favorable functional outcome (recovery walking and self-care), and (iii) return to usual activities at 6 months. Clinician predictions were compared with actual outcomes. Results: Most clinician participants were from neurosurgery (75%) with a median of 8 working years (IQR 5–14) of experience. Of the 6,305 randomized patients who survived 48 h, 213 (3.4%) were predicted to die (positive predictive value [PPV] 0.99, 95% confidence interval [CI] 0.99–0.99). Of 5,435 patients who survived 6 months, 209 (3.8%) were predicted to die (PPV 0.93, 95% CI: 0.92–0.93). Predictions on the favorable functional outcome (PPV 0.54, 95% CI: 0.52–0.56) and satisfied ability to return to usual activities (PPV 0.50, 95% CI: 0.49–0.52) were poor. Prediction accuracy varied by working years and region of practice. Conclusions: In patients with ICH, clinician estimates of death are very good but conversely they are poor in predicting higher levels of functional recovery and activities.