Two-stage prediction model for in-hospital mortality of patients with influenza infection

摘要: Background Infleunza is a challenging issue in public health. The mortality and morbidity associated with epidemic pandemic influenza puts heavy burden on health care system. Most patients can be treated an outpatient basis but some required critical care. It crucial for frontline physicians to stratify by level of risk. Therefore, this study aimed create prediction model in-hospital mortality. Methods This retrospective cohort extracted data from the Chang Gung Research Database. included who were diagnosed between 2010 2016. primary outcome was illness. secondary analysis predict A two-stage-modeling method developed hospital We constructed multiple logistic regression illness first stage, then S1 score calculated. In second we used other construct backward model. area under receiver operating curve assess predictive value Results present study, 1680 met inclusion criteria. overall ICU admission 10.36% (174 patients) 4.29% (72 patients), respectively. stage I analysis, hypothermia (OR = 1.92), tachypnea 4.94), lower systolic blood pressure 2.35), diabetes mellitus 1.87), leukocytosis 2.22), leukopenia 2.70), high percentage segmented neutrophils 2.10) admission. Bandemia had highest odds ratio Stage 5.43). II C-reactive protein 1.01), urea nitrogen 1.02) model's assocaited 0.889 0.766, Conclusions two-stage efficient risk-stratification tool predicting mortailty. may optional than qSOFA SIRS