12-18 Yaş Aralığında Dikkat Eksikliği ve Hiperaktivite Bozukluğu Tanısının ve İlişkili Faktörlerin Makine Öğrenmesi ile İncelenmesi

Abstract

This study examined the clinical profile of ADHD in adolescents who were referred to the psychiatry clinic for the first time from the pediatric outpatient clinic with ADHD characteristics and who did not receive any treatment, by comparing the cases diagnosed with ADHD and not diagnosed with ADHD in terms of sociodemographic, developmental and clinical characteristics, comorbid psychiatric disorders, emotional, behavioral and cognitive problems, executive functioning, functionality, perceived stress level, anxiety and depressive symptoms; It is a study in which childhood and late-onset ADHD were compared based on the “age of onset” of adolescents diagnosed with ADHD. With the data obtained, a model was created using machine learning algorithms for the prediction of ADHD diagnosis in adolescence; the first phase of the study was conducted as cross- sectional descriptive and the second phase as metadological. The study included 203 adolescents (101 girls, 102 boys) between the ages of 12-18, and adolescents were administered the Schedule for Affective Disorders and Schizophrenia for School Age Children Present and Life-time DSM-5 (K-SADS-PL) psychiatric diagnostic interview, Conners-Wells Adolescent Self-Report Scale-Revised Long Form (CASS-L), Weiss Functional Impairment Rating Scale- Self Report Form (WFIRS- S), Perceived Stress Scale (PSS) and The Revised Child Anxiety And Depression Scale-Child Version (RCADS); Conner's Parent Rating Scale (CPRS), Weiss Functional Impairment Rating Scale-Parent (WFIRS-P) and Behavioral Rating Inventory of Executive Functions (BRIEF). In the evaluation made with K-SADS- PL, 21.2% of the adolescents were diagnosed with ADHD, 53.5% of those diagnosed with ADHD had childhood onset and 46.5% had late onset. In adolescents diagnosed with ADHD, maternal age at birth was younger, postnatal complications were higher, paternal education level was lower, family structure was single parent/extended/dispersed family, and family history of ADHD was more frequent. It was determined that adolescents diagnosed with ADHD were more often comorbidity with Disruptive Behavioral Disorders, emotional, behavioral and cognitive problems, anger control difficulties, problems in family and peer relationships, and that functionality was negatively affected in many areas and impairment in executive functions was more severe. While adolescents who were not diagnosed with ADHD were diagnosed with Social Anxiety Disorder at a higher rate, there was no difference between the groups in terms of perceived stress level, anxiety and depression symptoms. Among the adolescents diagnosed with ADHD, it was determined that in the late-onset group, female gender and inattention type were predominant in clinical appearance, family history of ADHD was less, Disruptive Conduct Disorders accompanied at a lower rate, and impairment in the executive function indicator was less severe. Among the models created for the prediction of ADHD diagnosis in adolescence, the Random Forest (RF) algorithm was the model with the highest prediction performance by reaching a high accuracy value of 88.7%, while the variables that optimized the performance of the model in the prediction of ADHD diagnosis were executive function functions, Conners total score based on adolescent self-report, family history of ADHD, and CPRS total score based on parental report, in order of importance. In this study, the effectiveness of machine learning algorithms in predicting ADHD diagnosis in adolescence was emphasized by using a random forest tree algorithm. It was determined that the assessment of executive function in adolescence was the most important attribute in predicting the clinical diagnosis.