Please activate JavaScript!
Please install Adobe Flash Player, click here for download


ISCB 2014 Vienna, Austria • Abstracts - Poster Presentations 123Wednesday, 27th August 2014 • 11:00-11:30 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust The aim of the present study was to test a hypothetical model to evalu- ate: 1) optimism, stigma perception and social support, have an impact on QoL, controlling for socio-demographic and clinical variables; 2) optimism exerts a mediator effect between stigma perception and QoL. Study comprises a sequential sample of 215 volunteer obese patients, ap- proached by their physicians, in outpatient departments of principal hos- pitals in Portugal and completed self-report questionnaires to assess so- cio-demographic and clinical, psychosocial and QoL variables. Structural Equation Modeling (SEM) was used to test the quality of the hypothesized model. Results showed that the hypothesized model fitted the data reasonably well, CFI=0.9, RMSEA=0.06, X(276)2 =525.75, p<0.001 (sensible to sample size). Controlling for socio-demographic and clinical variables, all factors had a simultaneous independent statistically significant impact in QoL, demonstrating that a more optimistic attitude, a lower stigma perception and more social support contribute to a better general well-being, a better physical health and a better mental health. Results also showed a partial mediation effect of optimism between stigma perception and general well-being/mental health.   P3.5 Development and validation of clinical prediction models P3.5.3 Logistic regression and linear discrimant analysis for assessing factors related to genetic anemia: a comparison of both approaches U Aguirre1,2 , E Urrechaga3 1 Research Unit, Hospital Galdakao-Usansolo, Galdakao- Usansolo, Spain, 2 REDISSEC Health Services Research on Chronic PatientsNetwork, Bilbao, Spain, 3 Laboratory, Hospital Galdakao- Usansolo, Galdakao-Usansolo, Spain   Aims: Logistic regression (LR) and linear discriminant analyses (LDA) are statistical methods which can be used for the evaluation of the associa- tions between various covariates and a categorical outcome. Both meth- odologies have been extensively applied in research, especially in medical and sociological sciences. Although the theoretical properties have been studied extensively throughout the literature, the choice of the proper method in data analysis is still a question for the researcher.The aim of this work is to explore the performance of the two analytical methods to the detection of genetic anemia. Methods: A set of 1108 blood samples was divided into the training (60%) and test groups (40%). Red Blood Cells (RBC), hemoglobin (Hb), mean cell volume (MCV), mean cell hemoglobin (MCH) and RBC distribution width (RDW) were considered as independent variables whereas the ge- netic anemia as outcome. LR and LDA methods were applied to both data sets. Sensitivity, Specificity, Negative and Positive Predictive Values and Accuracy have been evaluated. Results: When trying to classify genetic anemia, 82.5% and 78.21 % of the genetic anemias were correctly classified by LDA and LR, respectively. As for the sensitivity, LR showed higher value than LDA approach. Overall ac- curacy was higher (81.37%) using LDA than LR (79.30%). Conclusions: LDA presents the advantage over classical analysis (LR) that it can be applied to discriminate two groups: acquired anemia and genetic anemia, independently of the clinical state of the carrier at the moment of the analysis.   P3.5.8 An investigation of performance measures developed to validate risk models for survival data G Ambler1 , S Rahman2 , B Choodari-Oskooei1 , RZ Omar1 1 UCL, London, United Kingdom, 2 Institute of Statistical Research and Training, Dhaka, Bangladesh   When developing a risk prediction model for survival data it is essential that the performance of the model is evaluated in validation data using appropriate performance measures. Although a number of measures have been proposed, there is only limited guidance regarding their use in prac- tice. A simulation study based on two clinical datasets was conducted to in- vestigate a wide range of performance measures. Measures were selected from categories that assess overall performance (Graf’s Brier score and IBS, Schemper’s V and measures from Kent and O’Quigley, and Schmid), dis- crimination (Harrell, Uno and Gonen’s concordance indices and Royston’s D) and calibration (calibration slope) of a model and were evaluated with respect to their robustness to censoring and ease of interpretation. Some of the measures needed to be modified for use in validation data. The overall performance measures were all reasonably robust to moder- ate levels of censoring. The most commonly used discrimination measure, Harrell’s C, was considerably affected by censoring and tended to increase as censoring increased. In contrast, Uno and Gonen’s C indices were rea- sonably stable in the presence of censoring, as was Royston’s D. The cali- bration slope was not affected by censoring. We recommend that Uno’s C is used in practice to quantify concordance and that D is reported alongside since it has an appealing interpretation. Any of the overall performance measures could be recommended but we prefer Graf’s measures as they are robust to high levels of censoring. The calibration slope can also be recommended. P3.5.10 PREVEXEPOC: a computer tool for risk stratification of patients with exacerbated COPD based on a predictive severity scoring system I Arostegui1,2 , MJ Legarreta1 , I Barrio1 , A Unzurrunzaga3 , JM Quintana3 1 University of the Basque Country UPV/EHU, Leioa, Spain, 2 REDISSEC, Galdakao, Spain, 3 Hospital Galdakao-Usansolo, REDISSEC, Galdakao, Spain   Limited information is available about predictors of short-term mortality in patients with exacerbated COPD (eCOPD). The goal of the study was to propose a method for the development of prognostic severity scores for risk stratification of patients with eCOPD and to make them available as easy to use tools for clinical decision-making process. The method we propose started with the development of a prediction model for short-term mortality in patients with eCOPD internally and externally validated. The next step consisted on creating a prognostic severity score that predicted the risk of mortality based on the previous statistical model. The predictive accuracy of the severity score was inter- nally and externally validated and comparison with the original model was performed. A final step consisted on categorizing the severity score into 4 levels that defined risk groups of mortality using a novel approach to categorizing continuous variables in prediction models. Discrimination ability was tested using the area under the receiver operating character- istic (ROC) curve (AUC) and comparison of continuous and categorical versions of the score was performed with the integrated discrimination improvement (IDI). The methodology was applied to a cohort study of patients with an eCOPD. A severity score was created without significant loss of predictive accuracy compared to the predictive model. The score was categorized into risk categories, without significant loss in discrimination ability, IDI

Pages Overview