Questionnaire-based attention deficit disorder with hyperactivity (ADHD) screening tests may not always be objective or accurate, owing to both subjectivity and prejudice. Despite attempts to develop objective measures to characterize ADHD, no widely applicable index currently exists. The principal aim of this study was to determine whether high-resolution activity features could provide a sufficient analytical foundation for determining the activity transitions of children with ADHD and to develop a decision support model for ADHD screening by monitoring children’s school activities using a 3-axial actigraph. Actigraphs were placed on the non-dominant wrists of 153 children for 3 hours, while they were at school. Children who scored high on the questionnaires were clinically examined by child psychiatrists, who then confirmed ADHD. Mean, variance, and ratios of activity within partitioned activity regions (a unit of 0.1G) were extracted as activity features. As a primary research on the effect by contexts of school life, activity features were extracted for three courses, including art, language and math. And, they were compared between the ADHD and non-ADHD groups via principal component analysis (PCA) and other statistical analyses. For, screening model construction, two decision tree models were constructed using the C5.0 algorithm after feature selection step: [A] from whole hours (class + playtime) and [B] during classes. Accuracy, sensitivity, and specificity were evaluated. Positive/negative predictive value, odds ratio, relative risk, likelihood ratio and area under ROC curve (AUC) were also calculated for model evaluation.
In the comparison analysis on course contents, especially in the art course, the two groups of children were almost completely separable by the new features, and the activity distributions of the groups differed significantly over a broad range. Those findings also showed that the course contents appeared to influence the activity patterns of children with ADHD. Monitoring the actual magnitude and counts of activity over a broad range could facilitate deeper investigations into the distributions or patterns of activities. And, two 5-depth decision trees were constructed by C5.0 algorithms. [Model A] three non-ADHD children were misclassified, resulting in an accuracy score of 97.89%. Sensitivity and NPV were 1.00. Specificity and PPV were 0.98 and 0.58-0.77, respectively. [Model B] 11 non-ADHD children were misclassified, resulting in an accuracy score of 92.23%. Specificity and PPV were scored at 0.92 and 0.27-0.47, respectively. Objective screening of latent ADHD patients can be accomplished with a simple watch-like sensor, which is worn for just a few hours while the child attends school. The model proposed herein can be applied to a great many children without heavy cost in time and manpower cost, and would generate valuable results from a public health perspective.
