The CAARS 2–ADHD Index was developed with the use of machine learning algorithms, which identified the optimal set of 12 items for the Self-Report and 12 items for the Observer rater forms that would distinguish between general population and ADHD cases. Responses to the set of items are summed into a raw score and then converted to a probability score. Results of the CAARS 2–ADHD Index are reported as a probability score that communicates the degree to which an individual’s raw score resembles a score from the ADHD Reference Sample rather than the General Population. There is strong evidence for reliability, based on analyses regarding internal consistency, precision of measurement, test-retest reliability, and inter-rater reliability. Additionally, evidence supports the validity of the CAARS 2–ADHD Index probability score with respect to its ability to correctly classify individuals with and without an ADHD diagnosis. Finally, there is substantial support for the fairness and generalizability of the CAARS 2–ADHD Index, as there is a lack of evidence of measurement bias for demographic subgroups. The CAARS 2–ADHD Index meets reliability, validity, and fairness standards and guidelines for psychometric tests (AERA, APA, & NCME, 2014).