Predictors of prolonged use of mechanical ventilation in patients with acute respiratory failure and acute heart failure in the CVCU RSUD Dr. Saiful Anwar Malang

Abstrak

Background Acute respiratory failure (ARF) is a critical condition that often complicates hospitalization and commonly arises from cardiopulmonary dysfunctions such as acute heart failure. Prolonged mechanical ventilation (PMV) in these patients is associated with increased morbidity, mortality of about 30%, and greater healthcare resource utilization. Identifying predictors of PMV is essential to improve outcomes and optimize management strategies.   Methods A retrospective cohort study was conducted on all patients who underwent endotracheal intubation in the Cardiovascular Care Unit (CVCU) of RSUD Dr. Saiful Anwar Malang from 2015 to 2021. Patients with incomplete medical records or who died within 14 days of mechanical ventilation were excluded. Univariate and multivariate logistic regression analyses identified independent predictors of PMV. Receiver operating characteristic (ROC) curves were generated to assess model discrimination using the area under the curve (AUC), with corresponding sensitivity and specificity. Data were analyzed using SPSS 22.0.   Results Five independent predictors of PMV were identified: tachycardia (p = 0.013), metabolic acidosis (p = 0.002), impaired renal function (p = 0.009), shock (p = 0.006), and major bleeding (p = 0.002). Multivariate analysis showed the following odds ratios(OR, 95% CI): tachycardia 2.06 (1.09–5.99), metabolic acidosis 2.03 (1.09–6.33), impaired renal function 2.87 (1.28–6.46), shock 2.83 (1.13–7.06), and major bleeding 1.36 (1.18–2.15). The model demonstrated good discrimination with an AUC of 0.83 (95% CI 0.77–0.88), sensitivity 0.87, and specificity 0.73.   Conclusion In patients with respiratory failure due to acute heart failure, tachycardia, metabolic acidosis, impaired renal function, shock, and major bleeding were independent predictors of prolonged mechanical ventilation. The predictive model showed high sensitivity and acceptable specificity, supporting its clinical usefulness for early identification of high-risk patients and targeted intervention.