PT - JOURNAL ARTICLE AU - Michal Svoboda AU - Jan Svancara AU - Jiri Jarkovsky AU - Barbora Novotna AU - Patrice Popelkova AU - Marek Plutinsky AU - Kristian Brat AU - Jaromir Zatloukal AU - Eva Volakova AU - Karel Hejduk AU - Eva Kocova AU - Michal Kopecky AU - Katerina Neumannova AU - Miroslava Fecaninova AU - Petr Safranek AU - Tomas Dvorak AU - Ondrej Sobotik AU - Zuzana Liptakova AU - Lucie Heribanova AU - Blanka Snelerova AU - Marie Majerciakova AU - Pavlina Musilova AU - Vladimir Koblizek TI - Risk scores for predicting death in COPD patients AID - 10.1183/13993003.congress-2016.OA1508 DP - 2016 Sep 01 TA - European Respiratory Journal PG - OA1508 VI - 48 IP - suppl 60 4099 - http://erj.ersjournals.com/content/48/suppl_60/OA1508.short 4100 - http://erj.ersjournals.com/content/48/suppl_60/OA1508.full SO - Eur Respir J2016 Sep 01; 48 AB - Introduction and objectives: COPD is a heterogeneous disorder associated with significant mortality. Prognostic markers or indices are needed especially in populations with severe COPD. The main objective was to develop a risk score for predicting death, and to compare this score with already existing indices (BODE, ADO, Yokohama).Methods: A total of 637 patients were involved in the analysis. The patient cohort was divided into a training set (N=453, 71%) and a testing set (N=184, 29%). The best cut-off values for predicting death were determined for all of the aforementioned variables by applying ROC analysis on the training set, and two risk scores were subsequently developed. The first index corresponded to the sum of statistically significant binary parameters based on the Cox proportional hazards model. In the second index, individual factors were HR-weighted. Both new indices were tested on the testing set. Finally, long-term survival and the abilities of new and existing indices (BODE, ADO, Yokohama) to predict deaths were compared; the existing indices had also been categorised by the ROC analysis.Results: The abilities of new and existing indices to predict death were compared: while the BODE index (with cut-off value ≥ 4) showed the highest sensitivity (0.870), the newly developed indices showed the highest specificity (0.702 and 0.702 for the first and second index, respectively) and the highest overall accuracy (0.693 and 0.710 for the first and second index, respectively).Conclusion: It was confirmed that indices represent a useful tool for predicting death in COPD patients. The BODE index has the highest sensitivity, whereas newly developed indices display higher specificity and overall accuracy.