As a leading global health burden, community-acquired pneumonia (CAP) continues to account for substantial morbidity and mortality, particularly among patients with pre-existing comorbidities. This study investigated the prognostic impact of comorbidities on in-hospital mortality in adult patients with bacterial CAP. Secondary diagnoses were considered comorbidities if present at admission and reflective of chronic conditions. Multidimensional comorbidity assessment included: Charlson Comorbidity Index (CCI), Elixhauser Comorbidity Index (ECI), condition counts of comorbidities in the Charlson list and Elixhauser list, and binary variable for every medical condition including the comorbidities in the Charlson list and Elixhauser list, every Clinical Classifications Software Refined (CCSR) category, International Statistical Classification of Diseases and Related Health Problems-10th Revision (ICD-10) chapter, block, category, and subcategory. Univariate and multivariate logistic regression analyses were used to evaluate the relationship between comorbidity and in-hospital mortality of CAP. Using split-sample validation, prediction models for in-hospital mortality were developed on a randomly selected derivation cohort (60%), with performance metrics rigorously assessed on an independent validation cohort (40%). Of the total of 11,164 patients, 380 (3.40%) died in the hospital. The most frequent comorbidities in the Charlson and Elixhauser lists are hypertension (44%), Congestive heart failure (19%), diabetes (15%) and cerebrovascular disease (14%). All comorbidity indicators are independent risk factors for in-hospital death. Alcohol abuse, tumor, cerebrovascular disease, congestive heart failure, dementia, neurological disorders, and renal diseases are independent risk factors for in-hospital death. All the multivariate models show good performance in predicting in-hospital death in the validation group. The area under the receiver operating characteristic curve (AUROC) ranges from 0.753 to 0.871. The best performing model based on AUROC includes binary variable for each comorbidity in the Charlson and Elixhauser comorbidity list. Comorbidities significantly contribute to CAP mortality risk. The multivariate model using age, sex and comorbidities can predict the risk of in-hospital death of CAP with good performance.
Cheng et al. (Tue,) studied this question.