Abstract 20169: A Novel Early Warning System to Predict Critical Events in Children With Congenital Heart Disease
| Autor: | Andrew Y Shin, Rui Liu, Vamsi Yarlagadda, Doff McElhinney, Christopher Almond, Xuefeng Ling |
|---|---|
| Rok vydání: | 2015 |
| Předmět: | |
| Zdroj: | Circulation. 132 |
| ISSN: | 1524-4539 0009-7322 |
| DOI: | 10.1161/circ.132.suppl_3.20169 |
| Popis: | Background: The outcome of hospitalized patients suffering clinical decline is contingent upon timely recognition. Current algorithms utilize select clinical variables for predictive modeling but have generally excluded congenital heart disease due to heterogeneity of patient and disease. The purpose of this study is to develop a model to predict decline in congenital heart disease patients. Methods: We analyzed all patients admitted to the Heart Center at Stanford Children’s Health between 9/1/2009 and 12/31/2012. We profiled the phenotypic changes in the EMR and measured entropy defined as the degree of fluctuation in all EMR values at 3-hour intervals. Predictive modeling was performed to identify temporal and directional relationships between entropy and the occurrence of a critical event, compared to the entropy of matched control patients. The entropy index was derived retrospectively in the first two-thirds of patients, and applied prospectively in the final third to determine predictability. Results: A total of 2,658 encounters occurred. A total of 71 (2.6%) critical events were observed as defined by calls for rapid response team, code blue, and ECMO. Peak entropy within the EMR was observed at mean of 15 (±5.4) hours prior to critical events (Figure). In the prospective arm, the area under the ROC curve for predicting critical events was 0.80 (95% confidence interval [CI]: 0.75-0.83). The sensitivity and specificity derived from the ROC curve was 68.4% and 74.6%, respectively. Only 2% of EMR variables were found in all cases and 14% of all critical events shared 3 or more common variables. Conclusion: Common EMR variables that cluster uniquely in relation to critical events among children hospitalized for cardiac disease are rare, which poses a challenge to early warning systems that are based on specific variables. Leveraging the EMR to detect increased entropy as a novel predictor of crisis event in this population merits further investigation. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|