Study Design
We conducted a retrospective cohort study based on a large US-based database called the Medical Information Mart for Intensive Care III (MIMIC-III) [14]. The MIMIC-III (v1.4) database contains comprehensive and high-quality data of well-defined and characterized ICU patients admitted to ICUs at the Beth Israel Deaconess Medical Center between 2001 and 2012. One author (HC) obtained access to the database and was responsible for data extraction (certification number 27252652). Our study complied with the Reporting of Studies Conducted using Observational Routinely Collected Health Data (RECORD) statement [15].
Selection of participants
Patients in the MIMIC-III who fulfilled the definition of sepsis were eligible for inclusion. Sepsis was diagnosed according to the sepsis-3 criteria [16]; in brief, patients with documented or suspected infection and an acute change in total Sequential Organ Failure Assessment (SOFA) score of ≥ 2 points were considered to have sepsis. Infection was identified from the International Classification of Diseases 9th Edition (ICD-9) code in the MIMIC-III. We excluded patients who were younger than 18 years or who spent less than 24 h in the ICU. Additionally, we analyzed only the first ICU stay for patients who were admitted to the ICU more than once. Included patients for whom initial CVP measurements were completed within 24 h after ICU admission were classified as the CVP group, and the rest of the patients comprised the no CVP group.
Variable extraction
The primary exposure was whether the patients underwent CVP measurements. The time to initial CVP measurement, the initial level of CVP and the duration of use of CVP were also collected. Baseline characteristics within the first 24 h after ICU admission were collected using structured query language (SQL), included age; sex; weight; ICU type; severity at admission as measured by SOFA score; the Simplified Acute Physiology Score II (SAPS II), and the Elixhauser comorbidity score. The use of mechanical ventilation, application of renal replacement therapy (RRT), and administration of vasopressors were also recorded. Vital signs included the MAP, heart rate, temperature (°C) and respiratory rate. Laboratory variables including white blood cell (WBC) count, haemoglobin, platelet counts, lactate, pH, partial pressure of oxygen (PO2) and partial pressure of carbon dioxide (PCO2) were measured during the first 24 h in the ICU. If a variable was recorded more than once in the first 24 h, we used the value related to the greatest severity of illness. The incidence of AKI was also extracted, and AKI was defined according to the Kidney Disease Improving Global Outcomes (KDGIO) criteria.
Comorbidities including congestive heart failure (CHF), atrial fibrillation (AFIB), chronic renal disease, liver disease, chronic obstructive pulmonary disease (COPD), stroke, and malignant tumour were also collected for analysis based on the recorded ICD-9 codes in the MIMIC-III database.
Outcomes
The primary outcome in the present study was 28-day mortality. Secondary outcomes included in-hospital and 1-year morality; the incidence of AKI within 7 days after ICU admission; the volumes (L) of intravenous fluid (IVF) in the first, second and third days in the ICU; the number of ventilator-free and vasopressor-free days within 28 days after ICU admission; and reduction in serum lactate (calculated as the difference between the maximum lactate level on day 1 and day 3).
Statistical analysis
Values are presented as the means (standard deviations) or medians [interquartile ranges (IQRs)] for continuous variables, and categorical variables are presented as total numbers and percentages. Comparisons between groups were made using the X2 test or Fisher’s exact test for categorical variables and Student’s t test, or the Mann-Whitney U test for continuous variables, as appropriate.
Multivariate regression was selected to characterize the relationship between CVP measurement and the primary outcome. Baseline variables that were considered clinically relevant or that showed a univariate relationship with the outcome (p < 0.10), including age, sex, weight admission period, severity score, use of mechanical ventilation, use of RRT, use of vasopressors, comorbidities, AKI, vital signs (MAP, heart rate, temperature and respiratory rate) and initial lactate level, were entered into a multivariate logistic regression model as covariates. To avoid bias induced by missing data, the analysis of the primary outcome was duplicated after multiple imputations.
Propensity score matching (PSM) and propensity score based inverse probability of treatment weighing (IPTW) were also used to adjust the covariates to ensure the robustness of our findings [17, 18]. A multivariate logistic regression model was used to estimate patient’s propensity scores for CVP measurement. One-to-One nearest neighbor matching with a caliper width of 0.05 was applied in the present study. An IPTW model was created using the estimated propensity scores as weights. The standardized mean differences (SMDs) were calculated to evaluate the effectiveness of the PSM and IPTW. Logistic regression was then performed on the matched cohort and weighted cohort, separately. Outcomes and therapeutic interventions were generated from the matched cohort.
CMA is a method to differentiate the total effect of a treatment into direct and indirect effects. The indirect effect on the outcome is mediated via a mediator. The analysis produces an average causal mediation effect (ACME), average direct effect (ADE), and total effect. To explore whether the effect of CVP measurement on the primary outcome is proportionally mediated by the reduction in serum lactate, we used CMA to characterize the causality relationship in our retrospective study.
Subgroups analyses of patients with positive blood cultures and septic shock were performed. Given that the effect of CVP measurement may vary according to the duration of CVP measurement and the initial CVP level, we first performed a series of multivariate logistic regressions based on different duration of CVP measurement (1 day, 2 days, 3 days, 4 days, 5 days, 6 days, 7 days and > 7 days), and then conducted sensitivity analyses comparing patients whose initial CVP level was below 8 mmHg or above 15 mmHg with patients in the no CVP group to evaluate the robustness of our findings.
All statistical analyses were performed using RStudio (version 1.2.5019), and p < 0.05 was considered statistically significant.