Association of a Composite Respiratory Support Index with Extubation Failure in ICU Patients: Insights from the MIMIC-IV Database

Objective To investigate the association between a newly derived Composite Respiratory Support Index (CRSI) and extubation failure (EF) among critically ill patients receiving invasive mechanical ventilation (IMV), and to evaluate its predictive performance compared with the conventional Rapid Shallow Breathing Index (RSBI). Methods This retrospective cohort study was conducted using the MIMIC-IV version 3.1 database, including adult patients who underwent IMV for at least 24 hours followed by their first planned extubation. Patients who died within 48 hours post-extubation without reintubation were excluded. Key demographic, clinical, laboratory, and ventilatory parameters were extracted within 24 hours before extubation. The CRSI was derived by principal component analysis based on positive end-expiratory pressure, fraction of inspired oxygen, observed tidal volume, invasive mechanical ventilation duration, and observed respiratory rate. Logistic regression models were used to examine associations between CRSI and EF, with adjustments for demographics, comorbidities, and SOFA score. Subgroup analyses explored consistency across age, gender, BMI, and comorbidity strata. Restricted cubic spline (RCS) analysis assessed nonlinear risk patterns. The predictive performance of CRSI was compared with RSBI. Results A total of 3,589 patients were analyzed, including 2,960 in the Extubation Success group and 629 in the Extubation Failure group. Patients in the Extubation Failure group demonstrated higher ventilatory demands, impaired gas exchange, and elevated CRSI and RSBI values compared with those successfully extubated. Across all models, higher CRSI was independently associated with increased risk of EF (fully adjusted OR = 1.44, 95% CI 1.34–1.54, P < 0.001). Subgroup analysis confirmed consistent associations across most strata, except among patients with COPD and cancer. Notably, the association was stronger in patients with prior stroke (OR = 2.03, 95% CI 1.32–3.41, P = 0.003). RCS analysis revealed a nonlinear escalation of risk beyond a CRSI threshold of -0.42. When jointly modeled with RSBI, only CRSI remained an independent predictor of EF. Conclusion The CRSI provides a comprehensive assessment of ventilatory dependence by integrating multiple respiratory support parameters. It demonstrated superior predictive accuracy and stability compared with RSBI, with potential to improve individualized extubation decision-making. Incorporation of CRSI into bedside tools may facilitate timely interventions, reduce extubation-related complications, and enhance clinical outcomes in the ICU setting.