Comparison of baseline investigations and hospital mortality among ICU patients with heart failure between the public MIMIC-II database and Chinese PLA general hospital

Xueliang Huang,Zhengbo Zhang,Pengjun Zhang,Lei Fu,Guanghong Guo,Xingwang Jia,Chunyan,Zhang,Qian Jia,Yurong Wang,Xinyu Wen,Yaping Tian

2017-01-01

Abstract:Background/Aim: Heart failure is a complex clinical syndrome with high prevalence and mortality rates worldwide. Comparison of studies between countries may facilitate improvements in therapy and prognosis of heart failure (HF). We aimed to determine the differences in risk factors for hospital death of patients with HF admitted to the Intensive Care Units (ICU) of the Beth Israel Deacons Medical Center (BIDMC) and Chinese PLA General Hospital (PLAGH) by comparing hospital mortality rates and baseline investigations between the public Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II database and PLAGH. Methods: We selected 2439 ICU patients hospitalized for HF from the MIMIC-II database and 637 patients from PLAGH for comparison of creatinine and urea nitrogen (UN) levels. Any-cause hospital mortality and 18 diagnostic parameters (creatinine, UN, glucose, pH, cTroponinT, uric acid, calcium, cholesterol, triglyceride, total protein, albumin, partial pressure of oxygen (PO2), partial pressure of carbon dioxide (PCO2), red blood cell (RBC) count, chloride, potassium, amino terminal B-type natriuretic peptide (NT-proBNP), sodium) examined in both hospitals were compared between BIDMC and PLAGH using Mann-Whitney U or chi-squared test. Multivariate binary logistic regression analysis was used to select indexes that differed significantly between patients that died in hospital and dischargers. Results: No significant differences in any-cause hospital mortality were evident between the two hospitals (P=0.091). In the Mann-Whitney U test, baseline median age, UN, pH, cTroponinT, chloride, potassium, PO2, PCO2 and NT-proBNP levels were higher while RBC, uric acid, calcium, total protein, albumin and cholesterol were lower in patients from MIMIC-II compared to PLAGH. Age, serum cTroponinT and albumin in both hospitals, in addition to serum creatinine, urea nitrogen, glucose, pH, RBC, calcium, potassium, sodium, PCO2 and cholesterol in MIMIC-II as well as total protein and NT-proBNP in PLAGH were statistical different between patients that died in hospital and those discharged, respectively. Further, age and albumin in both hospitals, moreover serum urea nitrogen, glucose, pH and potassium in BIDMC and serum cTroponinT in PLAGH were associated with hospital mortality. Conclusions: Comprehensive studies comparing laboratory parameters and mortality in ICU patients from teaching hospitals between the USA and China may facilitate effective HF prognosis and identification of risk factors of mortality. Similar hospital mortality rates of ICU patients with HF were observed between BIDMC and PLAGH. Our findings may contribute to the evaluation and establishment of effective prognostic and risk factors for hospital mortality in ICU patients with HF.