Very carefully. Prediction models can save time and resources, enabling clinicians and nurses to enhance clinical care. The performance of linear and nonlinear assistance vector machines (SVM) as prediction models for the tacrolimus blood concentration in liver transplantation individuals is compared with linear regression evaluation. Procedures 5 hundred and twenty-three tacrolimus blood concentration levels, with each other with 35 other relevant variables from 56 liver transplantation patients in between 2002 and 2006, were extracted from Ghent University Hospital database (ICU Information and facts Program IZIS) (Centricity Essential Care Clinisoft; GE Healthcare). Multiple linear regression, and assistance vector regression with linear and nonlinear (RBF) kernel functions have been performed, just after selection of relevant information elements and model parameters. Performances on the prediction models on unseen datasets were analyzed with fivefold cross-validation. Wilcoxon signed-rank evaluation was performed to examine differences in performances in between prediction models and to analyze variations among actual and predicted tacrolimus blood concentrations. Outcomes The imply absolute difference with all the measured tacrolimus blood concentration inside the predicted regression model was two.34 ng/ml (SD 2.51). Linear SVM and RBF SVM prediction models had imply absolute differences together with the measured tacrolimus blood concentration of, respectively, 2.20 ng/ml (SD 2.55) and 2.07 ng/ml (SD 2.16). These variations have been within an acceptable clinical range. Statistical analysis demonstrated considerable greater performance of linear (P < 0.001) and nonlinear (P = 0.002) SVM (Figure 1) in comparison with linear regression. Moreover, the nonlinear RBF SVM required only seven data components to perform this prediction, compared with 10 andFigure 1 (abstract P471)P470 Comparison of intensive care unit mortality performances: standardized mortality ratio vs absolute risk reductionB Afessa, M Keegan, J Naessens, O Gajic Mayo Clinic College of Medicine, Rochester, MN, USA Critical Care 2007, 11(Suppl 2):P470 (doi: 10.1186/cc5630) Introduction The aim of this study was to assess the role of absolute risk reduction (ARR) to PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/20800409 measure ICU performance as an option to the standardized mortality ratio (SMR). Techniques This retrospective study entails patients admitted to 3 ICUs of a single tertiary medical center from January 2003 through December 2005. Only the first ICU admission of each patient was integrated inside the study. The ICUs have been staffed similarly. We abstracted data in the APACHE III database. For each and every ICU, the SMR and ARR with their 95 Prostaglandin E2 chemical information self-assurance intervals (CI) were calculated. ICU performance was categorized as shown in Table 1. When comparing ICUs, in the event the 95 CI in the SMR or the ARR overlap amongst the units, the performances had been regarded as related. If there was no overlap, the variations in overall performance have been thought of statistically substantial. Results Through the study period, 12,447 patients had been admitted towards the three ICUs: 4,334 towards the medical ICU, 3,275 to the mixed ICU and four,838 for the surgical ICU. The predicted mortality rates have been 19.5 , 16.0 and 9.0 plus the observed mortality rates 14.eight , 9.7 and 4.3 for the medical, mixed and surgical ICUs, respectively. The SMR and ARR in mortality for every ICU are presented in Table 2. Conclusions ICU mortality performances assessed by SMR and ARR give diverse benefits. The ARR might be a far better metric when comparing ICUs having a diverse.