Study population
This was a restrictive observational study using data from the Medical Information Mart for Intensive Care IV (MIMIC-IV version 1.0) database from 2008 to 2019 [14]. MIMIC-IV, an update to the MIMIC-III, is a real-world publicly available clinical database maintained by Beth Israel Deaconess Medical Center, listing more than 60,000 ICU incidents. The database was accessed by an individual who has completed the Collaborative Institutional Training Initiative examination (Certification number 39022265 for Hao). All data were extracted using the SQL programming language. This longitudinal, single-center database included 76,540 patients who were admitted to an ICU. We included only patients admitted to the ICU for the first time. All intensive care patients diagnosed with septic shock were screened and identified by the “long_title” in the “d_icd_diagnoses” table of MIMIC-IV database. Our study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [15]. The code used for data extraction is available at GitHub (https://github.com/MIT-LCP/mimic-iv).
The inclusion criteria of our study were: adult patients (> 18 years old) with septic shock and treatment with vasopressors. The exclusion criteria were: (1) age less than 18 years; (2) diagnosis of malignant arrhythmia, which can lead to cardiogenic shock; (3) history of ischemic heart disease; (4) diagnosis of cardiomyopathy; (5) hepatic-related diseases such as liver cirrhosis; (6) history of chronic renal disease; (7) shock other than septic shock, and (8) pregnancy.
Data extraction
We extracted patient parameters, including age, sex, weight, race, type of ICU, the first 24 h Sequential Organ Failure Assessment (SOFA) score, the Oxford Acute Severity of Illness Score (OASIS), the Logistic Organ Dysfunction Score (LODS), the Charlson comorbidity index, interventions [i.e., renal replacement therapy (RRT), mechanical ventilation use, and VPs use], vital signs (i.e., MAP, SBP, DBP, HR, and temperature), initial lactate level, initial arterial pH, net fluid balance, fluid input at day 1, and urine output. Notably, we also extracted vital signs including MAP, SBP, DBP, and HR at 1 h, 2 h, 4 h, 8 h, 12 h, 24 h, 48 h, and 72 h after treatment initiation with VPs.
Study covariates and outcomes
The SI was calculated as the quotient between HR and SBP. MSI was calculated as the ratio between HR and MAP. DSI was calculated as the ratio between HR and DBP. These three parameters were calculated before VP treatment, and at 1 h, 2 h, 4 h, 8 h, 12 h, 24 h, 48 h, and 72 h after VP treatment. The patients could receive VP treatment before or after admission to the ICU. The parameters of pre-VP treatment were calculated as the average values of the patient data that were within 6 h before admission to the ICU until VP treatment. The primary endpoint of interest was the 3-day mortality rate of patients with septic shock requiring vasopressors. The secondary endpoint was in-hospital mortality.
Statistical analysis
Continuous variables were described using the least-squares mean and 95% confidence intervals (CIs). Categorical variables are presented as percentages. Logistic regression analyses were used to compare patient characteristics and outcomes according to the quartiles of pre-VP SI, DSI, and MSI. Odds ratios (ORs) with 95% CIs of the relationships between pre-VP SI, DSI, and MSI and 3-day/in-hospital mortality rates were calculated using logistic regression models. When pre-VP SI, DSI, and MSI were modeled as a quartile-based categorical variable, the lowest quartile was set as the reference group in each model. The significance of the linear trends of 3-day/in-hospital mortality rate across categories of pre-VP SI, DSI, and MSI was examined by assigning the median value to each quartile, and these variables were analyzed as a continuous variable in multivariate models. Multivariable logistic models were prepared as follows: Model 1 was adjusted for age, sex, race, and type of ICU care unit. Model 2 was further adjusted for the SOFA score on day-1 based on Model 1. Repeated-measures ANOVA was used to evaluate time-course differences in SI, MSI, and DSI pre-VPs and at 1 h, 2 h, 4 h, 8 h, 12 h, 24 h, 48 h, and 72 h after VP treatment between 3-day/in-hospital septic shock requiring vasopressors survivors and non-survivors. The performance of SI, MSI, and DSI pre-VPs, and at 1 h, 2 h, 4 h, 8 h, and 12 h after VP treatment were described by the area under the receiver operating characteristic (ROC) curve (AUC) and compared with DeLong analysis. Youden’s index was used to determine optimal cut-off values. Statistical significance was set at P < 0.05. All analyses were conducted using the SAS version 9.4 software (SAS Institute, Cary, NC, USA).
