Can a prognostic model using variables available at ED intubation (age, GCS score, admission lactate, and vasopressor use) accurately predict in-hospital mortality in elderly patients undergoing emergency intubation and invasive mechanical ventilation?
A parsimonious 4-variable model (age, GCS score, admission lactate, and vasopressor use) accurately predicts the high risk of in-hospital mortality in elderly patients requiring emergency intubation, which may facilitate early shared decision-making.
Background: Elderly patients requiring emergency intubation and invasive mechanical ventilation (IMV) in the emergency department (ED) face substantial mortality risk. However, prediction models utilizing only ED-obtainable variables for early prognostication remain limited. Objective: To develop and internally validate a prognostic model using variables available at ED intubation to predict in-hospital mortality in elderly patients. Methods: This retrospective cohort study included 273 patients aged ⩾ 65 years who underwent emergency intubation and IMV initiation in the ED of a tertiary hospital between June 2019 and June 2024. Candidate predictors included demographics, comorbidities, Glasgow Coma Scale (GCS), vasopressor use, admission lactate, PaO2/FiO2 ratio, shock index, and baseline laboratory values. Multiple imputation addressed missing data. Least absolute shrinkage and selection operator (LASSO) logistic regression identified optimal predictors, followed by multivariate logistic regression. Internal validation employed bootstrap resampling (1000 iterations). Model performance was assessed via discrimination ( C -statistic), calibration (calibration plot, Brier score), and clinical utility (decision curve analysis). Results: The overall in-hospital mortality rate was 72.2% (197/273). After LASSO selection and clinical adjudication, the final model included age, GCS score, admission lactate, and vasopressor requirement as independent predictors. The model demonstrated excellent discrimination ( C -statistic 0.834, 95% CI: 0.784–0.884), good calibration, and superior net benefit across threshold probabilities of 0.3–0.8 compared to default strategies. Bootstrap-corrected optimism was minimal (optimism-corrected C -statistic 0.827). Sensitivity analyses confirmed model robustness. Conclusions: A parsimonious model incorporating age, GCS score, admission lactate, and vasopressor use accurately predicts in-hospital mortality in elderly patients undergoing emergency intubation and IMV, using only variables readily available at the time of ED presentation. This tool has the potential to facilitate early, evidence-informed shared decision-making.
Liu et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: