Machine Learning Differentiates ADHD From Disorders With Similar Symptoms

A machine-learning analysis using the short version of the Conners’ Adult ADHD Rating Scales (CAARS) self-report misclassified only 5% of adults with attention-deficit/hyperactivity disorder (ADHD) in a study that also included adults with obesity or problematic gambling and control subjects. Researchers published their findings in Scientific Reports.

“This is an excellent rate and highly important,” researchers wrote, “as the CAARS is known to have successfully differentiated patients from healthy controls, but differential diagnoses proved much more difficult, especially for females or other rating scales.”

Because symptoms of adult ADHD resemble symptoms of other disorders, and because ADHD, gambling disorder, and obesity diagnoses have an overlap of about 20%, adult ADHD can be challenging to diagnose, the study explained. Consequently, using established diagnostic instruments to differentiate among conditions is important.

Strategies for Treating Patients With Both ADHD and SUD

The study included 1629 adults who completed the short form of the CAARS self-report, which consists of 26 items. Among participants, 385 had newly diagnosed adult ADHD and were medication-naïve, 135 had morbid obesity, 517 had problematic gambling, and 592 had no diagnosed disorder and were considered control subjects. Researchers used machine learning and all 26 items of the short version of the CAARS, age, and gender for analyses. Some 70% of data were used for training, and 30% for testing.

In the first test data analysis, just 13 patients with adult ADHD were misclassified, according to the study. In differentiating participants with adult ADHD, obesity, problematic gambling, and a control group, machine learning analysis had a global accuracy of .80 and precision ranges between .78 (gambling) and .92 (obesity).

“Our results demonstrate that with machine learning analyses, the various groups of patients we assumed to be similar in terms of impulse control and emotion regulation were very well distinguishable from each other as well as from the control group,” researchers wrote.

The short version of the CAARS self-report, they concluded, appears a promising instrument for classifying disorders with ADHD-like symptoms. 

—Jolynn Tumolo

Reference

Christiansen H, Chavanon ML, Hirsch O, et al. Use of machine learning to classify adult ADHD and other conditions based on the Conners' Adult ADHD Rating Scales. Scientific Reports. 2020;10(1):18871.