Bayesian probability approach to ADHD appraisal

Accurate diagnosis of attentional disorders such as attention-deficit hyperactivity disorder (ADHD) is imperative because there are multiple negative psychosocial sequelae related to undiagnosed and untreated ADHD. Early and accurate detection can lead to effective intervention and prevention of negative sequelae. Unfortunately, diagnosing ADHD presents a challenge to traditional assessment paradigms because there is no single test that definitively establishes its presence. Even though ADHD is a physiologically based disorder with a multifactorial etiology, the diagnosis has been traditionally based on a subjective history of symptoms. In this chapter we outline a stochastic method that utilizes a Bayesian interface for quantifying and assessing ADHD. It can be used to combine of a variety of psychometric tests and physiological markers into a single standardized instrument that, on each step, refines a probability for ADHD for each individual based on information provided by the individual assessments. The method is illustrated with data from a small study of six college female students with ADHD and six matched controls in which the method achieves correct classification for all participants, where none of the individual assessments was capable of achieving perfect classification. Further, we provide a framework for applying this Bayesian method for performing meta-analysis of data obtained from disparate studies and using disparate tests for ADHD based on calibration of the data into a unified probability scale. We use this method to combine data from five studies that examine the diagnostic abilities of different behavioral rating scales and EEG assessments of ADHD, enrolling a total of 56 ADHD and 55 control subjects of different age groups and gender.