Predictive Neurocognitive Model of Attention Deficit Hyperactivity Disorder Diagnosis

Attention Deficit Hyperactivity Disorder (ADHD) is a neurodevelopmental clinical entity associated with a reduction in brain maturation mainly at the frontotemporal level, generating neurocognitive deficits. This disorder usually presents in a comorbid form. Our research aimed to identify the neurocognitive characteristics of ADHD (combined, hyperactive/impulsive and inattentive) of pure presentation or with comorbidity with oppositional defiant disorder, specific learning disorder or autism spectrum disorder. We used a sample of N=712 school children with a diagnosis of ADHD, aged between 6 and 13 years, using non-probability sampling. We built predictive models to establish probabilistic rules with different types of variables (total IQ, working memory index, perceptual reasoning index, processing speed index, phonological/semantic fluency, attention, visuoverbal memory, verbal memory, working memory, visual memory, visual perception, constructional praxias), we used the automatic learning technique of Decision Trees (DTCM) in Rcran 4.2.1 software, which allowed us to establish a clinical hierarchy. The data for the analysis were obtained from the results of the psychometric tests provided to the sample (WISC-IV, Verbal Fluency, TMT, Visuoverbal Memory Scale, Verbal Memory Curve, Wechsler Memory Scale, Rey/Osterrieth Complex Figure). We conclude that children with pure ADHD present poor performance on tasks assessing the working memory index and perceptual reasoning that are not explained by deficits in IQ. Deficits in working memory are generalizable to all presentations and comorbidities of ADHD. One of the main advantages of DTCM over other predictive machine learning techniques is the possibility of differentiating the hierarchy of importance of the dependent variables, in this case, allowing the identification of the most important variables in four different populations of children diagnosed with ADHD.