182 Views

Oral Communication Session

Pediatric Hemostasis and Thrombosis

Session: Diagnostic tools to improve the care of pediatric hemostasis disorders

OC 72.2 - Computable phENotype To Identify Pulmonary Embolism in chilDrEn: The CENTIPEDE study

Wednesday, June 26, 2024

10:30 – 10:45 ICT

Room: Ballroom B3

Presenting Author(s)

Hilary Whitworth, MD, MSCE

Instructor
Division of Hematology, Children's Hospital of Philadelphia
Philadelphia, Pennsylvania, United States

Co-Author(s)

KW

Kristina Wagner

pediatric hospitalist
Children's Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
AZ

Allison Zelinski

Biostatistician
Applied Clinical Research Center; Children’s Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
ED

Elana DiCocco

Senior Clinical Research Coordinator
Applied Clinical Research Center; Children’s Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
AG

Amy Goodwin Davies

Data Analytics Lead
Applied Clinical Research Center; Children’s Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
RJ

Rochelle Jordan

Clinical Research Coordinator
Applied Clinical Research Center; Children’s Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
MH

Miranda Higginbotham

Clinical Research Project Manager
Applied Clinical Research Center; Children’s Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
LB

L. Charles Bailey

Associate Professor
Division of Hematology, Children's Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
LR

Leslie Raffini

Professor
Division of Hematology, Children's Hospital of Philadelphia
Philadelphia, Pennsylvania, United States
Madhvi Rajpurkar

Professor
Division of Pediatric Hematology Oncology, Carman and Ann Adams Department of Pediatrics, Wayne State University School of Medicine, Children's Hospital of Michigan, USA
Detroit, Michigan, United States

Background: Pediatric pulmonary embolism (PPE) is rare at individual centers despite an overall rising incidence. Multicenter studies are necessary but challenging due to inaccurate identification of PPE in administrative data. ASPECT is a validated computable phenotype for PE in adults that has not been evaluated in pediatrics (Klil-Drori et al., 2019).

Aims: We aimed to determine the positive predictive value (PPV), negative predictive value (NPV), sensitivity, and specificity of ASPECT for PPE.

Methods: ASPECT was evaluated in the PEDSnet database, a multicenter learning health system network, using Children’s Hospital of Philadelphia electronic medical record data from patients ≤21 years with an inpatient or emergency department (ED) visit. Predicted cases were identified by a SNOMED code for PE, anticoagulant prescription, and PE diagnostic imaging study. Predicted non-cases had an anticoagulant prescription in an inpatient or ED visit and were not identified as cases. To improve identification of false negatives (FN), predicted non-cases were enriched 20% with patients who had both an anticoagulant prescription and a PE SNOMED code. A blinded chart reviewer assessed all predicted cases and an equal number of predicted non-cases. Two adjudicators confirmed chart review and identified reasons for FN and false positives (FP). The study was approved by the Institutional Review Board and funded by The Children’s Foundation.

Results: Demographics of true positives and true negatives are summarized in Table 1. Of the 139 predicted cases and 139 predicted non-cases, 151 were PE positive and 127 were PE negative on chart review. The PPV was 92%, NPV was 83%, sensitivity was 85%, and specificity was 91% (Table 2).

Conclusion(s): ASPECT performed well in a pediatric population however was inferior to the adult validation. Sequential evaluation of different algorithms to improve the computable phenotype is ongoing. Future work also includes validation of PE outcomes.