Ethics approval and consent to participate
The MIMIC-III database was approved by the Massachusetts Institute of Technology (Cambridge, MA) and Beth Israel Deaconess Medical Center (Boston, MA), and consent was obtained for the original data collection.
Source of data
An open and free critical care database, which contained comprehensive clinical data of patients admitted to the Beth Israel Deaconess Medical Center in Boston, Massachusetts between June 2001 and October 2012, termed the Medical Information Mart for Intensive Care (MIMIC)-III v 1.4, was retrieved.9 This database was released on September 2rd, 2016, in which extensive and de-identified in-hospital information of over 40,000 patients was included. All data were classified into 26 tables, consisting of demographic characteristics, vital signs, laboratory test results, imaging examinations, and a data dictionary. Included patients were assigned a special code on each hospital and intensive care unit (ICU) admission, thus we could relate each table using these codes to obtain a complete hospitalization record. Hospital staff entered the final precise diagnosis according to the International Classification of Disease 9th (ICD-9) Edition code when patients were discharged. Because the present study was conducted using an anonymized public database that satisfied review committee agreements, the requirement for ethical consent was not necessary.
Participants
We conducted a retrospective cohort study including critically ill AMI (STEMI or non-STEMI) patients in MIMIC-III database. Exclusion criteria includes the time of entering ICU less than 24 hours, patients who had never obtained arterial catheters and the number of measured hemodynamic parameters less than 25%. Only the first medical record was selected. After excluding 3890 patients meeting exclusion criteria, a total of 1606 cases were selected in the investigation and randomly stratified into primary group and validation group according to the proportion of 7:3 (Figure 1).
Data collection
Baseline characteristics first measured after ICU admission were collected using structured query language (SQL), including demographic characteristics, biochemical measurement, comorbidities, scores, treatments and outcomes. Demographic characteristics include age, sex, body mass index (BMI) and ethnicity. Biochemical measurements refer to leukocyte, hemoglobin, platelet, hematocrit, sodium, potassium, albumin, glucose, bicarbonate, bilirubin, creatinine, blood urea nitrogen (BUN), alanine transaminase (ALT), aspartate transaminase (AST), lactate, N-terminal pro brain natriuretic peptide (NT-proBNP), troponin T, creatine kinase-myocardial band (CK), creatine kinase-myocardial band (CK-MB), high density lipoprotein-cholesterol (HDL-C), low density lipoprotein-cholesterol (LDL-C), total cholesterol (TC) and triglyceride (TG). Comorbidities recorded with ICD-9 codes in the MIMIC-III database that were collected for analysis including hypertension, diabetes mellitus, heart failure, atrial fibrillation, valve disease, cardiac arrest, peripheral vascular disease, liver disease, infection, acute respiratory distress syndrome (ARDS), organ dysfunction. Reference scores are the Simplified Acute Physiology Score II (SAPS II), sequential organ failure assessment (SOFA), Glasgow Coma Scale (GCS) and Angus. The appliance of furosemide, vasoactive drug, PCI, CABG and hemodialysis were taken into treatment consideration. The primary endpoint was 30 days all-cause mortality, death from any cause following entering ICU date. Survival time was calculated from entering ICU date to death date.
Hemodynamic indices
The initial 22 hemodynamic indices for analysis were SBP, heart rate (HR), oxygen extraction (O2ext), arterial oxygen content (CaO2), oxygen transport (DO2), oxygen transport index (DO2I), Oxygen Saturation of Mixed Venous Blood (SvO2), central venous pressure (CVP), stroke volume (SV), stroke volume index (SVI), systemic vascular resistance (SVR), systemic vascular resistance index (SVRI), cardiac index (CI), mean pulmonary artery pressure (MPAP), left ventricular stroke work index (LVSWI), left ventricular stroke work (LVSW), pulmonary vascular resistance index (PVRI), cardiac output (CO), right ventricular stroke work index (RVSWI), right ventricular stroke work (RVSW), Oxygen saturation (SO2), and content of mixed blood (CvO2). After ruling out indexes that are missing more than 20% of the observation quantity, the final hemodynamic indexes that we analyzed were SBP, HR, CVP, SV, SVI, SVRI, SVR, CI, LVSWI, LVSW, CO, SO2 and CaO2. After ruling out indexes that are missing more than 20% of the observation quantity, the final hemodynamic indexes that we analyzed were SBP, HR, CVP, SV, SVI, SVRI, SVR, CI, LVSWI, LVSW, CO, SO2 and CaO2. All indices were obtained by the arterial catheter, the venous catheter and the transthoracic ultrasound machine.
Statistical Analysis
The R 4.1.0 (R Foundation for Statistical Computing) was used for statistical computing. The missing covariates were less than 5% and filled by multiple imputation method. Normal distributed and non-normal distributed continuous variables are denoted as mean ± SD or median (interquartile range), respectively. Categorical variables are expressed as number (percentage). Continuous data were compared by Student’s t-test or the rank-sum test, while categorical data were compared using the chi-square test.
We applied Least absolute shrinkage and selection operator (LASSO) regression 10 and Cox regression to analyze all plausible variables. In the primary cohort, included hemodynamic indexes were screened out using LASSO regression. Indexes selected from LASSO regression were further analyzed in multivariate Cox proportional hazard model to identify the significant prognostic factors associated with 30 days all-cause mortality of AMI patients. Subsequently, factors with prognostic significance in the multivariate Cox regression analysis were utilized to build a survival-prediction model and a nomogram was used to visualize the model. A receiver operating characteristic curve (ROC), area under the ROC curve (AUC) and fixed coefficient time-dependent AUC curve based on nomogram for predicting 30 days survival are used to assess discrimination of the model, while the calibration plot is used to graphically evaluate the calibration of the nomogram in both primary and validation cohorts. Decision curve analysis (DCA) is used to evaluate the clinical utility of models. Patients are divided into favorable prognosis and poor prognosis groups according to the nomogram total points. The Kaplan-Meier method demonstrated differences in the mortality between the two risk groups sorted by the median cut-off value of the nomogram . For evaluation of both short-term and long-term predictive power of the model, the calibration curve, survival curve and DCA of 7 days, 14 days, 30 days and 1 year are displayed. A two-sided p < 0.05 was considered statistically significant.