Data source
We conducted a large-scale, single-center, retrospective cohort study using data collected from the MIMIC-III open source clinical database (version 1.3, released on December 10, 2015), which was developed and maintained by the Massachusetts Institute of Technology (MIT), Philips Healthcare, and Beth Israel Deaconess Medical Center (BIDMC) [13]. Information derived from the electronic medical records of 46,476 unique critical care patients admitted to the ICUs at BIDMC between 2001 and 2012 was included in this free accessible database. Use of the MIMIC-III database has been approved by the Institutional Review Boards of BIDMC and MIT, and waiver of informed consent was granted.
Patient population
All patients in the database were screened. The inclusion criteria in this study were as follows: (1) adults (≥18 years of age) at ICU admission, with complete medical records including available CVP measurement records; (2) ICU stay ≥72 h; and (3) consecutive CVP monitoring ≥12 h. For patients with multiple ICU stays, only data related the first ICU admission were considered.
All available CVP measurements recorded during ICU stay were extracted. Other variables including ICU type, demographic data, age, sex, Elixhauser comorbid conditions [14] and admission illness severity scores [including the Simplified Acute Physiology Score (SAPS) [15] and Sequential Organ Failure Assessment (SOFA) [16]] were extracted from the database. Additionally, data on use of vasopressors, mechanical ventilation, laboratory results related to organ dysfunction and length of ICU stay and hospitalization were extracted from the database.
Exposure
The primary exposure was the mean CVP during the first 72 h after ICU admission. We also calculated the duration of ECVP10 as the time of CVP above the level of 10 mmHg, which was considered as a relatively higher CVP level that might by unbeneficial for patients in critical care settings [17], and used it as an alternative indicator of the mean CVP level to estimate the association between ECVP load and outcome in the critical care settings.
Outcome
The primary outcome was 28-day mortality after ICU admission. The secondary outcomes included duration of mechanical ventilation and vasoactive drug use (epinephrine, norepinephrine, vasopressin, dopamine, dobutamine, milrinone and phenylephrine), laboratory results related to organ dysfunction and the length of ICU admission and hospitalization. For some patients, whose death occurred outside the hospital, the Social Security Death Index was linked to the database for investigations related to mortality.
Statistical analysis
Baseline characteristics were stratified by quartiles of mean CVP during the first 72 h after ICU admission. Cox regression analyses were undertaken to compare the 28-day mortality among different mean CVP quartiles. All other covariates, which comprised demographic characteristics, ICU type, SAPS II score at admission, comorbidities and mean duration of CVP measurement, were included into the multivariable regression model. We also compared the survivors and non-survivors for their ECVP10 duration with the outcome independently through the stratification on mean CVP.
The results are expressed as mean ± SD (standard deviation) for normally distributed data or median [interquartile range (IQR)] for non-normally distributed data. Continuous variables were compared using one-way analysis of variance for normally distributed quantitative data, and Mann–Whitney U test for non-normally distributed quantitative data to determine differences between groups. All statistical analyses were performed by using the IBM® SPSS® Statistics version 22 (SPSS Inc., Chicago, IL, USA). Any p value <0.05 was considered statistically significant.
Subset and sensitivity analyses
Since the mean CVP level can only describe the overall level of CVP, in sensitivity analyses, we replaced the mean CVP level with the duration of ECVP10 as an alternative indicator to estimate the association between ECVP load and outcome in the critical care settings.
We examined whether elevated mean CVP level in the highest quartile was associated with poor outcome compared to the other quartiles in patients with and without sepsis, which was defined as suspected or documented infection and an acute increase of ≥2 SOFA points infection [18], and AKI, as defined by either a greater than or equal to 0.3 mg/dL increase within 48 h or a greater than or equal to 50% increase within 7 days of ICU admission, or acute dialysis, in keeping with the Kidney Disease Outcome Quality Initiative guidelines [19]. We individually tested association between elevated mean CVP and these variables in adjusted analysis and provided graphical representation of the stratified risks.