Objectives: In critically ill patients, the assessment of neurologic function can be difficult because of the use of sedative agents. It is not known whether neurologic signs observed under sedation can predict short-term outcomes. The objective of this study was to assess whether abnormal brainstem responses within the first 24 hrs of sedation are associated with mortality and altered mental status postsedation. Design: Observational prospective study including an initial single-center and a subsequent multicenter study to develop and then validate the prognostic models. Setting: Three mixed and two medical intensive care units. Patients: Mechanically ventilated intensive care unit patients sedated with midazolam (± sufentanyl). Interventions: Neurologic examination including the Glasgow Coma Scale, the Assessment to Intensive Care Environment score, cranial nerve examination, response to noxious stimuli, and the cough reflex was performed. Measurements and Main Results: Seventy-two patients were included in the initial group and 72 in a subsequent validation study. Neurologic responses were independent of sedative dose. Twenty-two patients in the development cohort and 21 (29%) in the validation group died within 28 days of inclusion. Adjusted for Simplified Acute Physiology Score II score, absent cough reflex was independently associated with 28-day mortality in the development (adjusted odds ratio [OR], 7.80; 95% confidence interval [CI], 2.00–30.4; p = .003) and validation groups (adjusted OR, 5.44; 95% CI, 1.35–22.0; p = .017). Absent oculocephalic response, adjusted for Simplified Acute Physiology Score II score, was independently associated with altered mental status after the withdrawal of sedation in the development (adjusted OR, 4.54; 95% CI, 1.34–15.4; p = .015) and validation groups (adjusted OR, 6.10; 95% CI, 1.18–25.5; p = .012). Conclusions: Assessment of brainstem responses is feasible in sedated critically ill patients and loss of selected responses is predictive of mortality and altered mental status.
