PURPOSE: Ventilator-associated pneumonia (VAP) is most likely to develop in patients requiring prolonged mechanical ventilation (MV). Selective digestive tract decontamination and the use of specialized endotracheal tubes reduce the incidence of VAP. Those interventions should be used for patients who require prolonged MV, but little data exists to assist clinicians in selecting those patients. Objective: create a model to predict the need for prolonged MV by using patient information available at the time of intubation.
METHODS: This retrospective, cohort study includes randomly selected adult patients who required intubation in the medical intensive care unit (ICU) of Mayo Medical Center, Rochester, MN, from 2007-08. We collected data available at the time of intubation and performed analysis to identify factors associated with prolonged MV, defined as 2 days or more.
RESULTS: A total of 213 patients were included, and 131 (61.5%) required MV for >2 days. The median (interquartile range) duration of MV was 3.0 (1.2 - 6.9) days. There were no significant differences in age (p=.636) or gender (p=.594) between the prolonged and short-term intubation groups. Factors associated with prolonged MV at a p value of 0.1 or lower by univariate analysis included: pneumonia or acute lung injury as the medical conditions leading to MV, pulmonary infiltrates, respiratory rate, temperature, arterial pH, base excess, arterial oxygen tension to fraction of inspired oxygen (PF) ratio, current smoking status, and not requiring intubation for airway protection, COPD exacerbation, cardiogenic pulmonary edema, hemorrhagic shock, or upper airway obstruction. A multiple logistic regression model developed by entering these variables had AUC (95% CI) of 0.816 (0.755-0.878) and Hosmer-Lemeshow statistic 9.432 (p=0.307).
CONCLUSION: A model that includes medical condition leading to intubation, smoking status, chest x-ray findings, vital signs, and arterial blood gas values has very good discriminatory power and good calibration in predicting the need for prolonged mechanical ventilation.
CLINICAL IMPLICATIONS: A model predicting prolonged MV may assist clinicians in selecting patients who would benefit from interventions that reduce incidence of VAP.
DISCLOSURE: Craig Karpman, No Financial Disclosure Information; No Product/Research Disclosure Information