Delayed intubation and increased hospital mortality in critically ill patients: a dose-response analysis of a multicentre retrospective cohort

Background. Whether delaying invasive mechanical ventilation in favour of a trial of non-invasive support harms critically ill patients remains contested, a debate sharpened by the COVID-19 pandemic. Most evidence anchors intubation timing to illness onset in selected syndromes. We examined the association between time from intensive care unit (ICU) admission to first charted invasive ventilation and hospital mortality. Methods. Multicentre retrospective cohort study using the eICU Collaborative Research Database v2.0 (208 US hospitals, 2014–2015). Of 29,098 mechanically ventilated adults in their first ICU stay, 4,398 had a plausible charted ventilation-onset timestamp (− 72 h to + 14 days) and known discharge status. Patients were classified as pre-ICU intubation ( < − 1 h), immediate (− 1 to + 2 h, reference), early delay (2–24 h) or delayed (> 24 h). The primary outcome was hospital mortality. Multivariable logistic models with hospital-clustered robust standard errors adjusted for demographics, illness severity, Glasgow Coma Scale, sepsis, comorbidities, organ support and hospital characteristics. Sensitivity analyses included multiple imputation, exclusion of deaths within 48 h and modified Poisson estimation of adjusted risk ratios; prespecified strata were sepsis, Sequential Organ Failure Assessment (SOFA) score and age. Results. Crude hospital mortality was 22.4% (pre-ICU), 25.6% (immediate), 44.1% (early delay) and 34.9% (delayed). Despite the delayed group being least severely ill (median SOFA 4 vs 6–8), delay was associated with graded adjusted mortality: early delay odds ratio (OR) 2.16 (95% confidence interval 1.22–3.83) and delayed OR 2.48 (1.53–4.04) versus immediate intubation (P for trend < 0.001). Pre-ICU intubation was associated with lower mortality (OR 0.48, 0.38–0.61). Estimates were robust to all sensitivity analyses (delayed OR 2.31–2.87; adjusted risk ratio 1.61, 1.17–2.20) and broadly consistent across prespecified strata. Among survivors, delayed intubation was associated with longer ICU (ratio 2.75, 2.17–3.48) and hospital (ratio 1.98, 1.66–2.36) stays. Conclusions. Intubation delayed beyond ICU admission was associated with a graded, dose–response increase in adjusted hospital mortality despite lower baseline severity and biases favouring the delayed group. The findings support protocolised, time-limited non-invasive trials with early escalation and identify pre-ICU controlled intubation as a distinct, lower-risk pathway. Prospective validation with respiratory parameters at the decision point is warranted.