Questionnaire-based computational screening of adult ADHD

Use of Machine Learning Models to Differentiate Neurodevelopment Conditions Through Digitally Collected Data: Cross-Sectional Questionnaire Study

BACKGROUND

Diagnosis of child and adolescent psychopathologies involves a multifaceted approach, integrating clinical observations, behavioral assessments, medical history, cognitive testing, and familial context information. Digital technologies, especially internet-based platforms for administering caregiver-rated questionnaires, are increasingly used in this field, particularly during the screening phase. The ascent of digital platforms for data collection has propelled advanced psychopathology classification methods such as supervised machine learning (ML) into the forefront of both research and clinical environments. This shift, recently called psycho-informatics, has been facilitated by gradually incorporating computational devices into clinical workflows. However, an actual integration between telemedicine and the ML approach has yet to be fulfilled.

OBJECTIVE

Under these premises, exploring the potential of ML applications for analyzing digitally collected data may have significant implications for supporting the clinical practice of diagnosing early psychopathology. The purpose of this study was, therefore, to exploit ML models for the classification of attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) using internet-based parent-reported socio-anamnestic data, aiming at obtaining accurate predictive models for new help-seeking families.

METHODS

In this retrospective, single-center observational study, socio-anamnestic data were collected from 1688 children and adolescents referred for suspected neurodevelopmental conditions. The data included sociodemographic, clinical, environmental, and developmental factors, collected remotely through the first Italian internet-based screening tool for neurodevelopmental disorders, the Medea Information and Clinical Assessment On-Line (MedicalBIT). Random forest (RF), decision tree, and logistic regression models were developed and evaluated using classification accuracy, sensitivity, specificity, and importance of independent variables.

RESULTS

The RF model demonstrated robust accuracy, achieving 84% (95% CI 82-85; P<.001) for ADHD and 86% (95% CI 84-87; P<.001) for ASD classifications. Sensitivities were also high, with 93% for ADHD and 95% for ASD. In contrast, the DT and LR models exhibited lower accuracy (DT 74%, 95% CI 71-77; P<.001 for ADHD; DT 79%, 95% CI 77-82; P<.001 for ASD; LR 61%, 95% CI 57-64; P<.001 for ADHD; LR 63%, 95% CI 60-67; P<.001 for ASD) and sensitivities (DT: 82% for ADHD and 88% for ASD; LR: 62% for ADHD and 68% for ASD). The independent variables considered for classification differed in importance between the 2 models, reflecting the distinct characteristics of the 3 ML approaches.

CONCLUSIONS

This study highlights the potential of ML models, particularly RF, in enhancing the diagnostic process of child and adolescent psychopathology. Altogether, the current findings underscore the significance of leveraging digital platforms and computational techniques in the diagnostic process. While interpretability remains crucial, the developed approach might provide valuable screening tools for clinicians, highlighting the significance of embedding computational techniques in the diagnostic process.

Psychometric properties of the Chinese version of the Quick Delay Questionnaire (C-QDQ) and ecological characteristics of reward-delay impulsivity of adults with ADHD

BACKGROUND

The Quick Delay Questionnaire (QDQ) is a short questionnaire designed to assess delay-related difficulties in adults. This study aimed to examine the reliability and validity of the Chinese version of the QDQ (C-QDQ) in Chinese adults, and explore the ecological characteristics of delay-related impulsivity in Chinese adults with attention-deficit/hyperactivity disorder (ADHD).

METHODS

Data was collected from 302 adults, including ADHD (n = 209) and healthy controls (HCs) (n = 93). All participants completed the C-QDQ. The convergent validity, internal consistency, retest reliability and confirmatory factor analysis (CFA) of the C-QDQ were analyzed. The correlations between C-QDQ and two laboratory measures of delay-related difficulties and Barratt Impulsiveness Scale-11 (BIS-11), the comparison of C-QDQ scores between ADHD subgroups and HCs were also analyzed.

RESULTS

The Cronbach's α of C-QDQ was between 0.83 and 0.89. The intraclass correlation coefficient of C-QDQ was between 0.80 and 0.83. The results of CFA of C-QDQ favoured the original two-factor model (delay aversion and delay discounting). Significant positive associations were found between C-QDQ scores and BIS-11 total score and performance on the laboratory measure of delay-related difficulties. Participants with ADHD had higher C-QDQ scores than HCs, and female ADHD reported higher scores on delay discounting subscale than male. ADHD-combined type (ADHD-C) reported higher scores on delay aversion subscale than ADHD-inattention type (ADHD-I).

CONCLUSION

The C-QDQ is a valid and reliable tool to measure delay-related responses that appears to have clinical utility. It can present the delay-related impulsivity of patients with ADHD. Compared to HCs, the level of reward-delay impulsivity was higher in ADHD.

Exploring telediagnostic procedures in child neuropsychiatry: addressing ADHD diagnosis and autism symptoms through supervised machine learning

Recently, there has been an increase in telemedicine applied to child neuropsychiatry, such as the use of online platforms to collect remotely case histories and demographic and behavioral information. In the present proof-of-concept study, we aimed to understand to what extent information parents and teachers provide through online questionnaires overlaps with clinicians' diagnostic conclusions on attention-deficit/hyperactivity disorder (ADHD). Moreover, we intended to explore a possible role that autism spectrum disorders (ASD) symptoms played in this process. We examined parent- and teacher-rated questionnaires collected remotely and an on-site evaluation of intelligence quotients from 342 subjects (18% females), aged 3-16 years, and referred for suspected ADHD. An easily interpretable machine learning model-decision tree (DT)-was built to simulate the clinical process of classifying ADHD/non-ADHD based on collected data. Then, we tested the DT model's predictive accuracy through a cross-validation approach. The DT classifier's performance was compared with those that other machine learning models achieved, such as random forest and support vector machines. Differences in ASD symptoms in the DT-identified classes were tested to address their role in performing a diagnostic error using the DT model. The DT identified the decision rules clinicians adopt to classify an ADHD diagnosis with an 82% accuracy rate. Regarding the cross-validation experiment, our DT model reached a predictive accuracy of 74% that was similar to those of other classification algorithms. The caregiver-reported ADHD core symptom severity proved the most discriminative information for clinicians during the diagnostic decision process. However, ASD symptoms were a confounding factor when ADHD severity had to be established. Telehealth procedures proved effective in obtaining an automated output regarding a diagnostic risk, reducing the time delay between symptom detection and diagnosis. However, this should not be considered an alternative to on-site procedures but rather as automated support for clinical practice, enabling clinicians to allocate further resources to the most complex cases.