Risk Factors Of In-Hospital Mortality In Patients Treated For Pneumonia At A Tertiary Care Centre In Switzerland (September, 10.1159/000508666, 2020)

Background:Little is known about risk factors upon hospital admission that are associated with in-hospital death of patients hospitalized for bacterial pneumonia. Identifying such factors may help to optimize the treatment and lower the mortality of these patients.Objectives:The aim of the study was to characterize baseline characteristics of patients hospitalized for bacterial pneumonia in Switzerland and to identify risk factors associated with all-cause in-hospital mortality.Methods:Routinely collected electronic health record data of patients discharged from a large Swiss tertiary care hospital between August 2009 and 2017 were analysed. Potential risk factors such as patient demographics, physical examination findings, vital signs, laboratory results, and comorbidities were considered within +/- 24 h of admission. Univariable and multivariable logistic regression models identified risk factors for in-hospital death. The area under the receiver operating characteristic (ROC) curve was used to compare the identified factors to existing pneumonia scoring systems.Results:Out of 1,781 hospital stays with initial and main diagnosis of bacterial pneumonia, 85 patients (4.85%) died (33.9% female, median age 62.3 years [interquartile range, 52-75]). Age, low systolic blood pressure, underweight, a missing value for body mass index, decreased haemoglobin level, raised C-reactive protein, high urea, high lactate dehydrogenase, concomitant pleural effusion, and cancer were independently associated with in-hospital death. The area under the ROC curve was 0.89 for the multivariable model containing the identified predictors.Conclusions:Our data are consistent with previous trials characterizing patients hospitalized for pneumonia. Additionally, we identified new and independent risk factors associated with in-hospital death among patients treated for bacterial pneumonia. Findings need to be further validated in larger multicentre cohorts.