Screening for Adulthood ADHD and Comorbidities in a Tertiary Mental Health Center Using EarlyDetect: A Machine Learning-Based Pilot Study

Abstract

Screening for adult Attention-Deficit/Hyperactivity Disorder (ADHD) and differentiating ADHD from comorbid mental health disorders remains to be clinically challenging. A screening tool for ADHD and comorbid mental health disorders is essential, as most adult ADHD is comorbid with several mental health disorders. The current pilot study enrolled 955 consecutive patients attending a tertiary mental health center in Canada and who completed EarlyDetect assessment, with 45.2% of patients diagnosed with ADHD. The best ADHD classification model using composite scoring achieved a balanced accuracy of 0.788, showing a 2.1% increase compared to standalone ADHD screening, detecting four more patients with ADHD per 100 patients. The classification model including ADHD with comorbidity was also successful (balanced accuracy = 0.712). The results suggest the novel screening method can improve ADHD detection accuracy and inform the risk of ADHD with comorbidity, and may further inform specific comorbidity including MDD and BD.

Keywords: ADHD comorbidity; ADHD screening; adult ADHD; machine-learning; mental health.

Figure 1.

Model performance on ADHD classification. Note. Panel A presents the confusion matrix of the LOOCV performance for the best-performing model. Panel B presents the Receiver Operating Characteristic Curve. AUC stands for the area under the curve.

Figure 2.

Top 10 features on ADHD classification. Note. The average coefficient values were calculated from all individual Elastic Net models in LOOCV. ¹“As a child, were you often fidgety, restless, unable to concentrate or remember things, disorganized, or impulsive and did these symptoms make it difficult to complete tasks such as homework or get along with others?” ²“Do you still experience some of these symptoms as an adult?” ³“The symptoms have disrupted your social life/leisure activities.” ⁴“How often do you feel overly active and compelled to do things, like you were driven by a motor?” ⁵“Have you ever had a period of time when you were feeling ‘up’ or ‘high’ or ‘hyper’ or so full of energy or full of yourself that you got into trouble, or that other people thought you were not your usual self? (Do not consider times when you were intoxicated on drugs or alcohol.).”

Figure 3.

LOOCV results on the classification of ADHD with comorbidity table. Note. Panel A presents the confusion matrix of the LOOCV performance for the model using individual questions from ASRS. Panel B presents the confusion matrix of the LOOCV performance for the model using all individual questions. Individual cell presents frequency counts followed by row-wise proportions.