FASDetect as a machine learning-based screening app for FASD in youth with ADHD

www.nature.com/npjdigitalmed ARTICLE OPEN FASDetect as a machine learning-based screening app for FASD in youth with ADHD Lukas Ehrig1,2,7, Ann-Christin Wagner2,7, Heike Wolter2, Christoph U. Correll2,3,4, Olga Geisel2,8 and Stefan Konigorski 1,2,5,6,8 ✉ Fetal alcohol-spectrum disorder (FASD) is underdiagnosed and often misdiagnosed as attention-deficit/hyperactivity disorder (ADHD). Here, we develop a screening tool for FASD in youth with ADHD symptoms. To develop the prediction model, medical record data from a German University outpatient unit are assessed including 275 patients aged 0–19 years old with FASD with or without ADHD and 170 patients with ADHD without FASD aged 0–19 years old. We train 6 machine learning models based on 13 selected variables and evaluate their performance. Random forest models yield the best prediction models with a cross-validated AUC of 0.92 (95% confidence interval [0.84, 0.99]). Follow-up analyses indicate that a random forest model with 6 variables – body length and head circumference at birth, IQ, socially intrusive behaviour, poor memory and sleep disturbance – yields equivalent predictive accuracy. We implement the prediction model in a web-based app called FASDetect – a user-friendly, clinically scalable FASD risk calculator that is freely available at https://fasdetect.dhc-lab.hpi.de. 1234567890():,; npj Digital Medicine (2023)6:130 ; https://doi.org/10.1038/s41746-023-00864-1 INTRODUCTION Fetal alcohol-spectrum disorder (FASD) is an umbrella term for medical conditions caused by prenatal alcohol exposure, including fetal alcohol syndrome (FAS), partial fetal alcohol syndrome (pFAS), alcohol related birth defects (ARBD), and alcohol-related neurodevelopmental disorder (ARND). The global prevalence of FASD is estimated to be between 2–5% of the Western world’s population1. Despite the prevalence rate, FASD is highly underdiagnosed and many patients miss out on the beneficial effects of an early childhood diagnosis and subsequent early intervention2–5. Established diagnostic systems for FASD are based on the manifestation of growth deficiencies, craniofacial dysmorphia, central nervous system damage/dysfunction, and gestational alcohol exposure6,7. These neuropsychological impairments can manifest as deficits in intelligence, learning, memory, executive function and academic achievements, language and motor development and attention8. People with FASD have a higher risk to develop secondary psychiatric conditions, like conduct disorder, attention-deficit/hyperactivity disorder (ADHD) and sleep disorders, as well as to experience adverse life events8–11. Hyperactivity, inattention and impulsivity are characteristically seen both in patients with ADHD and FASD. More than half of FASD patients suffer from comorbid ADHD11. These overlapping symptoms of FASD and ADHD complicate the diagnostic process and can lead to misdiagnosis as well as delayed intervention for FASD. In a study conducted in 547 children and adolescent who were adopted or in foster care and who underwent a comprehensive multidisciplinary diagnostic evaluation to identify FASD, 156 youth met criteria for FASD, but in as many as 80% the FASD diagnosis had been missed and 6% were misdiagnosed within the FASD spectrum. The mental health diagnosis most commonly given to those children upon referral was ADHD12. The very high proportion of missed FASD diagnosis and youth receiving a misdiagnosis underscore the importance of evaluating youth diagnosed with ADHD in order to detect any missed FASD diagnosis. The purpose of the present study is to (i) develop a machine learning algorithm for detection of FASD in patients with ADHD symptoms based on retrospectively gathered out-patient data, and (ii) subsequently use this algorithm to create an easy and fast as well as clinically scalable online screening tool. Based on the analysis of medical record data from a German University outpatient department including 275 patients with FASD with or without ADHD and 170 patients with ADHD without FASD, we identify a random forest model based on 6 variables – body length and head circumference at birth, IQ, socially intrusive behaviour, poor memory and sleep disturbance –that yields sufficient accuracy to differentiate youth with versus without FASD. We implement this algorithm in a screening tool called FASDetect which is easy to use and yields a quick screening result. RESULTS Study sample This study was conducted at the outpatient unit of the department of child and adolescent psychiatry at the Campus Charité Virchow of the Charité Universitätsmedizin Berlin, Germany. The sample for the analysis was selected to allow a comparison of patients with a diagnosis of ADHD with patients with a diagnosis of FASD. More specifically, a group of consecutively assessed patients with a clinical diagnosis of ADHD without FASD and a group of patients with an expert diagnosis of FASD (with or without comorbid ADHD) was compared. Altogether, 694 patients with ADHD symptoms were identified consecutively from the general patient pool being potentially 1 Digital Health Center, Hasso Plattner Institute for Digital Engineering, University of Potsdam, Potsdam, Germany. 2Department of Child and Adolescent Psychiatry, Charité Universitätsmedizin Berlin, Berlin, Germany. 3The Zucker Hillside Hospital, Department of Psychiatry, Northwell Health, Glen Oaks, NY, USA. 4Donald and Barbara Zucker School of Medicine at Hofstra/Northwell, Department of Psychiatry and Molecular Medicine, Hempstead, NY, USA. 5Hasso Plattner Institute for Digital Health at Mount Sinai, Icahn School of Medicine at Mount Sinai, New York, NY, USA. 6Department of Statistics, Harvard University, Cambridge, MA, USA. 7These authors contributed equally: Lukas Ehrig, Ann-Christin Wagner. 8These authors jointly supervised this work: Olga Geisel, Stefan Konigorski. ✉email: Published in partnership with Seoul National University Bundang Hospital L. Ehrig et al. 1234567890():,; 2 Fig. 1 Flow chart of the ADHD and FASD patient groups included in the study. eligible for the study. 256 of the 694 ADHD patients had a confirmed FASD diagnosis and therefore were excluded from the ADHD pool. Further, 141 patients were excluded from the ADHD group due to an unconfirmed ADHD diagnosis; 58 because they had a suspected but not confirmed FASD diagnosis; 37 due to other severe medical, psychiatric, or neurological conditions; and 32 patients were excluded because patient records were unavailable. This yielded in total 170 patients in the ADHD group. The consecutively enrolled FASD group was recruited from the specialist center and consisted of 275 youth, including 129 FASD patients with comorbid ADHD and 146 patients without comorbid ADHD diagnosis. These 275 patients included most of the 256 FASD patients from the general patient pool. See also Fig. 1 for an illustration of the two study groups. Table 1. Patient characteristics. Variable All patien (...truncated)