PURPOSE

Risk factors and prognostic model of fatal outcomes need to be investigated for the increasing number of immunocompromised hosts (ICHs) who are hospitalized for severe pneumonia with high hospital mortality.

PATIENTS AND METHODS

In this single-center, retrospective study, we recruited 1,933 ICHs with severe pneumonia who were admitted to the intensive care unit (ICU) in West China hospital, Sichuan university, China between January, 2012 and December, 2018. Clinical features, laboratory findings, and fatal outcomes were collected from electronic medical records. Patients were randomly separated into a 70% training set (n=1,353) and a 30% testing set (n=580) for the development and validation of a prediction model. All data within 24 hours of ICU admission were compared between survivors and non-survivors. The risk factors were identified through LASSO and multivariate logistic regression analysis, and then used to develop a predicting nomogram. The nomogram for predicting hospital mortality of ICHs with severe pneumonia in the ICU was validated by C-index, AUC (area under the curve), calibration curve, and Decision Curve Analysis (DCA).

RESULTS

Eight risk factors, including age, fever, dyspnea, chronic renal disease, platelet counts, neutrophil counts, PaO/FiO ratio, and the requirement for vasopressors, were adopted in a nomogram for predicting hospital mortality. The nomogram had great predicting accuracy with a C-index of 0.819 (95% CI=0.795-0.842) in the training set, and a C-index of 0.819 (95% CI=0.783-0.855) in the testing set for hospital mortality. Additionally, the nomogram had well-fitted calibration curves. DCA demonstrated that the nomogram was clinically beneficial.

CONCLUSION

This study developed a novel nomogram for predicting hospital mortality of ICHs with severe pneumonia in the ICU. Validation showed good discriminatory ability and calibration, indicating that the nomogram was expected to be a superior predictive tool for doctors to identify risk factors and predict mortality, and might be generally applied in clinical practice after more external validations.