Machine learning links unresolving secondary pneumonia to mortality in patients with severe pneumonia, including COVID-19

• Text
• PDF

Abstract

BACKGROUND. Despite guidelines promoting the prevention and aggressive treatment of ventilator-associated pneumonia (VAP), the importance of VAP as a driver of outcomes in mechanically ventilated patients, including patients with severe COVID-19, remains unclear. We aimed to determine the contribution of unsuccessful treatment of VAP to mortality in patients with severe pneumonia. METHODS. We performed a single-center prospective cohort study of 585 mechanically ventilated patients with severe pneumonia and respiratory failure, 190 of whom had COVID-19, who underwent at least one bronchoalveolar lavage. A panel of ICU physicians adjudicated pneumonia episodes and endpoints based on clinical and microbiologic data. Given the relatively long ICU length of stay among patients with COVID-19, we developed a machine learning approach called CarpeDiem, which groups similar ICU patient-days into clinical states based on electronic health record data. RESULTS.CarpeDiem revealed that the long ICU length of stay among patients with COVID-19 is attributable to long stays in clinical states characterized primarily by respiratory failure. While VAP was not associated with mortality overall, mortality was higher in patients with one episode of unsuccessfully treated VAP compared with successfully treated VAP (76.4% versus 17.6%, P < 0.001). In all patients, including those with COVID-19, CarpeDiem demonstrated that unresolving VAP was associated with transitions to clinical states associated with higher mortality. CONCLUSIONS. Unsuccessful treatment of VAP is associated with greater mortality. The relatively long length of stay among patients with COVID-19 is primarily due to prolonged respiratory failure, placing them at higher risk of VAP. FUNDING. U19AI135964

Authors

Catherine A. Gao, Nikolay S. Markov, Thomas Stoeger, Anna E. Pawlowski, Mengjia Kang, Prasanth Nannapaneni, Rogan A. Grant, Chiagozie Pickens, James M. Walter, Jacqueline M. Kruser, Luke V. Rasmussen, Daniel Schneider, Justin Starren, Helen K. Donnelly, Alvaro Donayre, Yuan Luo, G.R. Scott Budinger, Richard G. Wunderink, Alexander V. Misharin, Benjamin D. Singer