Twitter  Linkedin
 

Back to 2016 Annual Meeting


The Pediatric Trauma Assessment and Management Database: Leveraging Existing Data Systems to Predict Functional Status After Pediatric Injury
Katherine Flynn-O’Brien, MD, MPH1, Mary E. Fallat, MD2, Thomas B. Rice, MD3, Christine M. Gall, DrPh, MS, RN4, Michael L. Nance, MD5, Jeffery P. Upperman, MD6, David M. Gourlay, MD7, John P. Crow, MD8, Fredrick P. Rivara, MD MPH9, 1Department of Surgery & Harborview Injury Prevention and Research Center, University of Washington, Seattle, WA, USA, 2Department of Surgery, University of Louisville and Kosair Children’s Hospital, Louisville, KY, USA, 3Department of Pediatrics, Medical College of Wisconsin, Milwaukee, WI, USA, 3Virtual Pediatric Systems, LLC., Los Angeles, CA, USA, 4SCL Health, Broomfield, CO, USA, 5Department of Surgery, Division of Pediatric General, Thoracic, and Fetal Surgery Children’s Hospital of Philadelphia, Philadelphia, PA, USA, 6Department of Surgery, Division of General Pediatric Surgery, Children’s Hospital of Los Angeles, Los Angeles, CA, USA, 7Department of Surgery, Division of General Pediatric Surgery, Children’s Hospital of Wisconsin, Milwaukee, WI, USA, 8Department of Surgery, Division of General Pediatric Surgery, Akron Children’s Hospital, Akron, OH, USA, 9Department of Pediatrics & Harborview Injury Prevention and Research Center, University of Washington, Seattle, WA, USA.

Background: Non-mortality outcomes are essential to evaluate after pediatric trauma, however they are rarely included in large clinical databases, limiting investigation and efforts to improve care on a national level. We merged two national data systems to create a prediction model for functional status after injury in critically injured children.

Methods: Trauma Registry (TR) and Virtual Pediatric Systems (VPS) from five Pediatric Trauma Centers were merged for children discharged from a PICU after traumatic injury in 2013. Through purposeful covariate selection testing over 100 variables, we created de novo a prediction model for functional status using dichotomized Pediatric Overall Performance Category scale (POPC). All univariate and multivariable regressions controlled for clustering by site.

Results: Of the 688 children included; 5.1% died from their injuries, 77.3% were discharged from the PICU with good performance or mild disability and 17.6% with moderate/severe disability or coma. The final POPC model included 14 variables demonstrated adequate discrimination (C-statistic = 0.896) and calibration (Figure). The probability of poor outcome varied significantly by site (p <0.0001).

Conclusions: Merging two data systems allowed for risk-adjusted modeling for functional status on a multi-institutional level. Merging existing data is feasible, innovative, and has potential to impact care with minimal new resources.


Back to 2016 Annual Meeting