Multivariate Model to Predict Survival in Community-Acquired Pneumonia

Abstract Background & Aims: Pneumonia continues to be a leading source of respiratory complications in emergency medical settings. Limited research has been conducted on constructing predictive models utilizing biomarkers to estimate the in-hospital mortality rates among patients with Community-Acquired Pneumonia (CAP). Our study aims to develop a comprehensive nomogram to project the survival probabilities at 7, 14, and 28 days for individuals afflicted with CAP. Methods: Utilizing the Medical Information Mart for Intensive Care (MIMIC) - III database, we selected 1,433 patients. These individuals were subsequently segregated into training set and validation set. Variables were chosen through the Cox regression approach, subsequently crafting a prognostic nomogram. The predictive capacity of this novel model was appraised using the receiver operating characteristic (ROC) curve, concordance index (C-index), calibration plot, net reclassification index (NRI), and integrated discrimination improvement (IDI), and was juxtaposed against the Acute Physiology Score III (APSIII) and the Sequential Organ Failure Assessment (SOFA). Results: The constructed nomogram incorporated the following variables: APSIII, Age, Temperature, WBC (White Blood Cell Count), Glucose, INR (International Normalized Ratio), Hemoglobin, Sodium, SOFA, Religion, Ethnicity, and Gender. Notably, this nomogram demonstrated superior performance compared to both the APSIII and the SOFA score, as evidenced by the ROC curve, C-index, NRI, and IDI evaluations. Conclusion: We have harnessed a diverse array of biomarkers to construct a nomogram that surpasses the accuracy of APSIII and SOFA. This tool holds the potential to assist healthcare professionals in enhancing treatment strategies and prognosticating patient outcomes.