BACKGROUND
Attention-deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder characterised by a persistent pattern of inattention, hyperactivity, and/or impulsivity that causes significant impairment across various domains. Delayed diagnosis and treatment for ADHD can be harmful to people, leading to broader mental health and social care difficulties. A diagnostic decision tool was developed using Artificial Intelligence (AI) to assist clinicians with the diagnosis of ADHD in adults. This tool is built by incorporating both expert clinician knowledge using a knowledge-based approach and historical clinical data using machine learning leading to an efficient hybrid approach.
OBJECTIVE
To evaluate the efficacy of the AI tool in clinical practice in the diagnosis of adult ADHD.
METHODS
Routine clinical data were collected by a specialist adult ADHD Clinic from people undergoing an assessment. A hybrid diagnostic tool for adult ADHD is proposed that combines a knowledge representation (KR) and machine learning (ML) model. The former encodes expert medical knowledge about ADHD diagnosis into a rule-based expert system; the latter is the optimal model after training, fine-tuning, and testing an ensemble of classification models on the acquired clinical data using 285 patient records. Each model predicts the diagnostic outcome by suggesting a positive or negative ADHD case, whereas in the case of rule-based and hybrid models it is also possible to yield a “expert” result.
RESULTS
Accuracy was expressed using the ratio of true/false positive/negatives (f1 score) and is calculated as 90.18%, 83.67% and 94.39% for the KR, ML and hybrid model, respectively. The proposed model, outperforms our previous work, showing that additional clinical data improved the accuracy and reduced the number of “expert” type diagnostic outputs.
CONCLUSIONS
The tool using AI is safe and effective in identifying people who may have ADHD using routine clinical data hence speeding up the diagnostic process. Further research can focus on the acceptability of using this technology.