[show abstract] [hide abstract]
ABSTRACT: Glycemic alterations are known as a risk condition of death in several diseases, such as ischemic cardiovascular and neurological disorders. The fact that its tight control under narrow normality levels decreases mortality and morbidity have led to further studies seeking to confirm the results and expand them to other disease areas.
To determine whether glycemic changes by themselves are a mortality risk factor in a sample of patients within an Intensive Care Unit (ICU), among which predominates traumatic-surgical patients.
Demographic and analytical characteristics were revised, as well as common monitoring variables in an ICU, among a sample of 2,554 patients from admissions between 1st January 2004 and 31 December 2008. Data were obtained from a database which endorsed records compiled with the monitoring ICU patients program "Carevue". They were processed with dynamics sheets included in the Excel software with the following variables: initial glycemia, mean glycemia during the first 24 hours and number of determinations performed. We used the mean value in the admission day of the remaining analytical and monitoring variables and the number of test performed on this first day. The sample was stratified in two groups for the statistical analysis: a) General Sample (MG) and b) sample excluding patients admitted after a programmed surgery (EQP). In both cases the effect of initial and averaged glycemia was checked. Group b was divided in two, according to the number of determinations b1) a single blood glucose determination group (EQP1) and b2) a multiple determination group (EQPM). From this group of non-programmed surgical patients the study was repeated in those patients who stayed at the ICU 3 or more days (EQP3D). Chi-square and Mantel-Haenzel test for the ODD ratio determination were performed for qualitative variables; quantitative variables were examined with the Mann-Whitney test. At each analysis level, logistic regression was performed using mortality as the dependent variable, including those variables with p-values < 0.05 which represented more than 60% of the data. An initially saturated model with backward till the final equation was used. A p-value of 0.05 (i.e. p < 0.05) was set as the significant threshold for all statistical analysis. They were performed with SPSS and GSTAT 2 statistical software.
A total of 2,165 of the 2,554 admitted patients during the study period were included (96.5%). Exclusion criteria were absence of plasma glucose determinations. In the bivariate analysis, first and mean glucose blood levels showed significant differences in mortality rates in absolute figures and also when data were classified stratified in three levels (< 60 mg/dl; 60-110 mg/dl or > 110 mg/dl) or in two (normal values 60 to 110 mg/dl and unusual figures < 60 mg/dl or > 110 mg/dl). These significant differences were lost when a logistic model was applied. From the remaining variables, renal function and NEMS showed to be mortality risks factors in this sample.
Hyperglycemia is a predominant phenomenon in critically ill patients. Hypoglycemia is less frequent and is associated with higher mortality rates. Initial glucose blood level, by itself, was not a mortality risk factor in the multivariate study and at none of the studied levels. Average glycemia did not add any prediction power. The changes in glucose blood levels seemed to be an adaptation process, which determined by itself a risk for the patient's discharge, at least in the first 24 hours period after ICU admission.
Nutricion hospitalaria: organo oficial de la Sociedad Espanola de Nutricion Parenteral y Enteral 06/2011; 26(3):622-35. · 1.31 Impact Factor