Patient in mechanical ventilation in ICU (12/04/17 - 05/04/18)
Description
Purpose: To compare the predictive ability of the most used system scores Sequential Organ Failure Assessment (SOFA), Acute Physiology and Chronic Health Evaluation (APACHE) II and Simplified Acute Physiology Score (SAPS) II for mortality in mechanically ventilated (MV) patients in ICU (Intensive Care Unit). We conducted a single-center prospective observational study. It was realized in 8-bed ICU in a tertiary University Hospital in South Brazil during 12/04/17 - 05/04/18. In this period 85 patients were admitted in ICU and 60 patients were submitted to mechanical ventilation for 48 hours or more. Patients admitted to the ICU submitted to mechanical ventilation for 48 hours or more during this period were included. Patients not submitted to MV or with less than 48 hours were excluded. 60 patients were in MV for 48 hours or more, however 2 of them were transferred to another ICU and thus was not included in this study. Data were collected daily, by a single researcher, and in a longitudinal way to obtain variables of the medical, physiological and laboratory history. Database was divided in variables of patients' comorbidities: chronic obstructive pulmonary disease, chronic cardiac arrhythmia, ischemic heart disease, sleep apnea syndrome, asthma, neuromuscular and chest wall diseases, previous infections, neoplasia, sepsis, stroke, diabetes, other endocrine diseases; physiological variables: vital parameters (respiratory rate, pulse, mean artery pressure, axillary temperature), oxygenation parameters (PaO2/FIO2 ratio, PCO2, AaDO2, arterial pH, mechanical ventilation, Non-invasive ventilation) and acute physiological factors (use of vasopressor/inotropic, ARDS, dialysis, urine output, water balance, diet, Glasgow coma scale); laboratory variables: blood count, cultures, bilirubins, creatinine, urea, potassium, sodium, bicarbonate, albumin, glucose. It was also collected by the researcher daily the appearance of the following complications: postoperative infections in the ICU (pneumonia, abdominal and urinary sepsis and wound infections), hematological complications (thrombocytopenia <100,000) and renal complications (urine output <500ml or serum creatinine >170 μmol/l or dialysis for acute renal failure). Demographic data, diagnosis of patient admission, duration of ventilation mechanics, length of stay in the ICU and total hospital stay, survivors, non-survivors were obtained and classified within the study. The value of variables that are part of the system scores were the most abnormal value for each variable during the last 24 hours. APACHE II, SAPS II and SOFA were calculated on the day of admission and on the last day in MV. VM1: All patients on the first day of MV VM2: All patients on the last day of MV vm10: Survivors on the first day of MV vm11: Non-survivors on the first day of MV vm20: Survivors on the last day of MV vm21: Non-survivors on the last day of MV
Files
Steps to reproduce
Statistical analyses were performed by using RStudio Statistical Software16, version 1.0.136. The level of statistical significance was set at 5% (p-value<0.05). First, Shapiro-Wilk test was performed in order to evaluate the normality distribution of all the quantitative variables. Quantitative variables are presented by their median and interquartile range (IQR), as most of the variables had an asymmetric distribution. Quantitative characteristics are presented by their mean and standard deviation (SD) when symmetric and by their median and interquartile range (IQR) when asymmetric. Variables were compared between two groups based on the outcome of the patients: survivors or non-survivors. T-student test was performed as a parametric test and Wilcoxon Mann Whitney test as non-parametric test to investigate differences in quantitative variables between the two groups. Fisher exact test was used between qualitative factors. The capacity of scores to discriminate using mortality as an independent variable was measured via the receiver operating characteristic (ROC) curves, which illustrate the sensitivity (true-positive cases) against 1 minus the specificity (false-positive cases), positive and negative predictive values. The area under each ROC curve was calculated to assess the discriminatory ability of the different scores to distinguish whether a patient died or survived. Cut-offs for the APACHE II, SOFA and SAPS II scores were calculated using the maximum value for Youden’s index, where Youden’s index = sensitivity – (1 – specificity). De Long tests have been conducted in order to compare the differences between AUC pairs17. The Spearman graphical correlation matrix was conducted to test the correlation between the scoring systems of mortality. In order to examine the potential association of each scoring system with the mortality , univariate logistic regression models were fitted 18. Calibration or the agreement between predicted mortality of the study population was examined with the Hosmer-Lemeshow goodness-of-fit test 19 .