Abstract 12602: Predictive Modeling for ADHD in Adolescents With Critical Congenital Heart Disease

IntroductionADHD is the most common behavioral diagnosis reported in patients (pts) with critical congenital heart disease (CHD), with lifetime prevalence 3-5 times greater than in the typically-developing population. We sought to identify predictors of ADHD symptoms in teens after CHD surgery using...

Full description

Saved in:
Bibliographic Details
Published inCirculation (New York, N.Y.) Vol. 140; no. Suppl_1 Suppl 1; p. A12602
Main Authors Majeed, Amara, Wypij, David, Rofeberg, Valerie, Bellinger, David, Newburger, Jane
Format Journal Article
LanguageEnglish
Published by the American College of Cardiology Foundation and the American Heart Association, Inc 19.11.2019
Online AccessGet full text

Cover

Loading…
More Information
Summary:IntroductionADHD is the most common behavioral diagnosis reported in patients (pts) with critical congenital heart disease (CHD), with lifetime prevalence 3-5 times greater than in the typically-developing population. We sought to identify predictors of ADHD symptoms in teens after CHD surgery using machine learning.MethodsWe analyzed data from three single-center studies of neurodevelopment in adolescents with d-TGA, TOF, and Fontan; each dataset included preoperative, operative, and post-operative variables. We developed a classification tree model to predict our primary outcome, a score of ≥65 on the parent-reported Conners ADHD Index T score, the threshold for clinical concern, and a regression tree model to predict the continuous ADHD Index T score.ResultsIn 380 pts (137 d-TGA, 90 TOF, 153 Fontan), mean (±SD) ADHD Index T score was 57.5±13.6 and 28% had scores ≥65. The classification model in the combined cohort (Figure, sensitivity 95%, specificity 36%) revealed highest ADHD concern in two pathways1) ≥3 caths and ≥4 operative complications (72% of clinical concern), and 2) ≥3 caths and <4 operative complications when Hollingshead social class <58, total support time ≥82 min, and DHCA time <45 minutes (52%), and even higher if also ≥1 neurologic event (83%). In regression tree modeling (R=11%), the highest (worst) scores were seen in pts with diagnoses other than TGA, family class ≤51, and seizures (mean=73); lowest (best) scores were in d-TGA pts (mean=54). Limitations include variation in risk factors in analyses within diagnostic group, and high intercorrelation of some medical risk factors.ConclusionsADHD symptoms by parent report are more common in teens with lower social class, more catheterizations and operative complications, longer support times, and postoperative neurologic events. Future studies should test whether early identification of risk factors and institution of targeted intervention can reduce later symptoms of ADHD.
ISSN:0009-7322
1524-4539
DOI:10.1161/circ.140.suppl_1.12602