Data source
The Massachusetts Institute of Technology's Laboratory for Computational Physiology maintains the Multiparameter Intelligent Monitoring in Intensive Care III (MIMIC III, V.1.4) database, which includes data on over 50,000 patients admitted to the intensive care unit at Beth Israel Deaconess Medical Center between 2001 and 2012 (13). We attended a training course on 'protecting human subjects' in order to apply for access to the database.
The establishment of the database was approved by the institutional review boards of the Massachusetts Institute of Technology (Cambridge, MA) and Beth Israel Deaconess Medical Center (Boston, MA). The author Jiang extracted the data for this study after passing the National Institutes of Health's online training course (certification number: 9322422).
Population selection criteria
In total, 58976 intensive care unit (ICU) patients were recorded in the MIMICIII database, of these, we included in our study patients who were older than 16 at the time of their initial admission and who underwent invasive ventilation for a minimum of 48 consecutive hours. Patients were excluded if they met the criteria: had incomplete ventilatory variables to calculate MP and norMP, had > 1% missing data, and/or were died or extubated during the first 48 h. We used only data from the patient's initial ICU admission or initial hospitalization.
Data extraction
The structured query language (SQL) was used to extract data from the database, and included tidal volume (Vt), positive end–expiratory pressure (PEEP), peak pressure (PIP), RR, and the inspired fraction of oxygen (FiO2). The following equation was used to calculate mechanical power (7, 11):
MP(J/min) = 0.098 × Vt × RR × (PIP – ΔP × 0.5), where the driving pressure (ΔP) = PIP – PEEP (14).
norMP (×10 − 3 J/min/kg) = MP/PBW (15), where PBW was the predicted body weight calculated by using the equation as used in previous studies of ventilation (16):
PBW = 50.0 + 0.91(height [cm] – 152.4) in males,
PBW = 45.5 + 0.91(height [cm] – 152.4) in females.
Due to the fact that the patients provided multiple measurements, the mean of the highest and lowest values obtained during the second 24 hours was used. The norMP in the second day of ventilation was chosen because during the first 24 hours usually mechanical ventilation is subjected to several changes and may result in more noise.
The following demographic data (using first 24 hours of admission data) were collected: age, gender, ethnicity (white, black, or other), height, weight, comorbidities, and disease severity scores (Acute Physiology and Chronic Health Evaluation [APACHE] III) (17, 18). Vital signs and laboratory measurements were captured as lowest and the highest values in the first day of ventilation.
Clinical outcome
To gather information about ICU patients' status, the follow-up followed from ICU admission and ended at death. The major endpoint was ICU mortality, the secondary endpoints included 30-day, 90-day mortality; ICU length of stay (ICU_LOS), the number of ventilator-free days at day 28 (VFD_28, specified as the days from effective weaning to day 28; patients who died prior to weaning were considered to have no ventilator-free days).
Statistical analyses
Continuous variables are presented in the tables as the median with interquartile ranges. The required Wilcoxon test, or Kruskal– Wallis test, was applied. Chisquared test or Fisher's exact test was used for categorical variables, which are presented as a percentage. Patients were categorized into groups according to ICU mortality.
The median and interquartile range of norMP was used to classify all patients. For all outcomes, univariate and multivariate regression were used to account for potential confounding variables. Additionally, subgroup analyses were conducted to determine the relationship between norMP and the primary outcome according to the Vt and PIP levels.
Statistical significance was described as a two-sided p < 0.05. SPSS software was used for all statistical analysis (SPSS-22.0; IBM Corp., Armonk, NY, USA).