Development of a risk calculator to predict attention-deficit/hyperactivity disorder in very preterm/very low birth weight newborns.

Abstract

Very preterm/very low birth weight (VP/VLBW) newborns can have lifelong morbidities, as attention-deficit/hyperactivity disorder (ADHD). Clinicians have no markers to discriminate which among those individuals will develop later ADHD, based only on the clinical presentation at birth. Our aim was to develop an individualized risk calculator for ADHD in VP/VLBW newborns. This retrospective prognostic study included a consecutive sample of all VP/VLBW children (gestational age <32 weeks and/or birth weight <1.5 kg) born between 2010 and 2012 from a clinical cohort in a Brazilian tertiary care hospital. Children were clinically assessed at 6 years of age for ADHD using the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS). The least absolute shrinkage and selection operator (LASSO) method was used for model-building. Ninety-six VP/VLBW children were assessed at 6 years of age (92% follow-up), of whom 32 (33%) were diagnosed with ADHD. The area under the ROC curve (AUC) for ADHD prediction based on seven parameters (late-onset sepsis confirmed by blood culture, necrotizing enterocolitis, neonatal seizures, periventricular leukomalacia, respiratory distress syndrome, length of hospital stay, and number of maternal ADHD symptoms) was .875 (CI, 0.800-0.942, p < .001; AUC corrected for optimism with bootstrapping: .806), a performance that is comparable to other medical risk calculators. Compared to approaches that would offer early intervention to all, or intervention to none, the risk calculator will be more useful in selecting VP/VLBW newborns, with statistically significant net benefits at cost:benefits of around 1:2 to around 10:6 (range of ADHD risk thresholds of 32%-62%, respectively). It also showed specificity for ADHD compared to other prevalent child psychopathologies. The risk calculator showed good performance for early identification of VP/VLBW newborns at high risk of future ADHD diagnosis. External validity in population-based samples is needed to extend clinical usefulness.