Database
This retrospective study utilized the Medical Information Mart for Intensive Care IV (MIMIC-IV) dataset[16] .According to the database, investigators had to pass an examination on 'Protecting Human Research Participants' on the NIH website and sign a data use agreement. A member of the team, Meiling Lu, passed the exam and was granted access (Record ID: 11818066).
This study was conducted in accordance with the tenets of the Helsinki Declaration. This study did not require informed consent or ethical approval.
Data analysis and statistical methods
Data extraction and study population identification were accomplished by screening the 'long_title' field of the 'mimiciv_hosp.procedures_icd' table of the MIMIC-IV database for patients that underwent TAVI surgery. In view of the fact that patients may be admitted to the ICU more than once, we included only the length of the first ICU stay for each patient.
Data extraction using PostgreSQL software: age, sex, weight, ethnicity, length of ICU stay, length of hospitalization, vital signs (respiratory rate, heart rate, DBP, MBP,SBP, temperature, and SpO2), coexisting comorbidities (chronic pulmonary disease, congestive heart failure, myocardial infarct, peripheral vascular disease, cerebrovascular disease, dementia, diabetes, rheumatic disease, renal disease, peptic ulcer disease, liver disease, malignant cancer ), laboratory tests (RBC 、RDW 、MCH 、MCHC 、MCV 、hematocrit 、hemoglobin 、platelet 、WBC 、PT 、PTT 、glucose 、AG 、bicarbonate 、Bun 、Ca2+、Cl−、Cr、 Na+、K+), scoring systems (APSIII,GCS,CCI,LODS,SAPSII,SOFA, OASIS, and SAPSII), dialysis, mechanical ventilation, vasoactive agents. Laboratory data and vital signs were extracted within 24 hours of ICU admission. Owing to potential bias, missing values for parameters greater than 20% were excluded. The mean of the nonmissing values was used to interpolate missing data in continuous variables.
Normality of continuous variables was assessed using the Kolmogorov-Smirnov test. If the variables were normally distributed, they were presented as mean ± standard deviation (M ± SD). Independent samples t-test were used for comparisons. When the distribution was not normal, we used the Mann-Whitney U test and presented the continuous variables as median interquartile ranges (IQRs). Categorical variables are expressed as percentages and numbers. Chi-squared tests and Fisher's exact tests were used for analysis. Z-tests were used to compare the AUCs of the subjects' working characteristic (ROC) curves.[17]Perform a goodness-of-fit test. In order to determine the net benefit of the six scoring systems, a decision curve analysis (DCA) was carried out[18, 19] .