Determining Reintubation Risk Factors and Effects on Outcomes in the Surgical ICU.

Abstract

PurposeReintubation (RI) in surgical intensive care unit (SICU) patients is associated with significant morbidity and mortality. Reported RI rates are lower in surgical populations (6-9%) than in medical populations. This study evaluated the rate, risk factors, and outcomes associated with RI in a SICU cohort.MethodsA retrospective cohort study of adults admitted to a Level-1 trauma center SICU who received mechanical ventilation between 2019 and 2020 was performed. Exclusion criteria included age <18 years, self-extubation, tracheostomy before extubation, admission to a nonsurgical service, or death before extubation. RI was defined as unplanned intubation within 96 h of extubation. Demographic, clinical, and peri-extubation variables were analyzed using multivariable logistic regression. Post hoc exploratory analyses evaluated operative and nonoperative intubation subgroups.ResultsOf 667 patients, 52 (7.8%) experienced RI. In the full-cohort model, atrial fibrillation, spinal cord injury, alcohol withdrawal, ventilator-associated pneumonia (VAP), combined propofol-dexmedetomidine sedation, and admitting service were independently associated with RI; VAP timing relative to RI was uncertain. RI occurred in 5.8% of patients intubated for an operative procedure and 11.2% of those intubated for a nonoperative indication (P = .011). Compared with successful extubation, RI was associated with longer mechanical ventilation (11 vs 1 day), intensive care unit (ICU) stay (15 vs 2 days) and hospital stay (25 vs 10 days), as well as higher rates of tracheostomy (63% vs 2%), delirium (77% vs 33%), VAP (29% vs 2%), and in-hospital mortality (23% vs 1%, all P < .001).ConclusionsThe incidence of RI in this SICU was 7.8% and was associated with markedly worse outcomes. Patient-specific risk assessment may help guide risk-stratified post-extubation monitoring and respiratory support. The exploratory subgroup findings require prospective validation.