SESSION TITLE: Non Pulmonary Critical Care
SESSION TYPE: Original Investigation Slide
PRESENTED ON: Monday, October 28, 2013 at 07:30 AM - 09:00 AM
PURPOSE: To describe the incidence of delirium, identify factors and create a predictive model of the development of delirium in ICU.
METHODS: We conducted an observational prospective study. For six months we evaluated the presence of delirium using the CAM-ICU test. Demographic data, history, APACHE II score, and predisposing factors associated with ICU admission (undercurrent infection, medication received, electrolyte disturbances, mechanical ventilation). Admitted 648 patients of whom 100 patients were excluded because the stay was less than 24 hours, 9 due to drug intoxication, 3 for being under 18, 18 per RASS -4, 4 for schizophrenia and 38 for loss of information
RESULTS: Of the 476 patients included in this study, 95 patients developed delirium, 52 developed hypoactive delirium (54.7%, 95% CI = 44.2-65), 21 (22.1%, 95% CI = 15.2-31.8) and hyperactive 22 (23.3%, 95% CI = 15.1-32.9) mixed type. Statistically significant variables were: age over 65 years, APACHE II> 12, patients with brain injury, fever, hypernatremia, hyponatremia, hypokalemia, use of midazolam, beta blockers, steroids, remifentanil, propofol, norepinephrine, dopamine ,mechanical and noninvasive ventilation, tracheostomy, acute renal failure and nosocomial pneumonia. A model was constructed using multiple logistic regression to identify the best predictors of delirium: Mechanical Ventilation (OR: 12.30 ;P <0,0001), Hypernatremia OR: 4.17; P:<0,009 ) ,use beta blocker (OR 2.34; P: <0,0012) age over 65 years (OR: 2.23 P:<0,008) , Neurocritical patients (OR:2.24;P:0.009) and hypokalemia (OR: 2.11, P:0,009) The probability calculated by the model for developing delirium was 3.77% when none of the variables was present and 95.63% when all were present.
CONCLUSIONS: With these variables identified as predictors of delirium, a probabilistic model was constructed using multiple logistic regression.
CLINICAL IMPLICATIONS: The analysis of results using predictive models can be useful in the early diagnosis and prevention of delirium, decreasing its short and long term incidence.
DISCLOSURE: The following authors have nothing to disclose: Florencia Ballestero, Emiliano Descotte, Micaela Ferreyra, Fernando Grassi, Estefania Requesens, Miguel Blasco, Elias Soloaga, Sebastian Chiapela, Felipe Chertcoff
No Product/Research Disclosure Information