Techniques of Explainable Artificial Intelligence and Machine Learning in Digital Mental Health Intervention

Abstract

The objective of this study is to examine the use of explainable artificial intelligence (XAI) for gathering and analyzing health-related data in medical services applications, particularly in predicting outcomes in mental health care. Artificial Intelligence (Al) is increasingly utilized in medical services and wearable technologies, such as Fitbits. Strategies in the field of explainable Al (XAI) aim to clarify the predictions made by Al systems, enhancing their application in health-related contexts. Data was collected from 970 individuals, incorporating clinical and sociodemographic information. The analysis focused on factors such as self-reported motivation, the type of reference (self vs. healthcare provider), and the Work Productivity and Activity Impairment Questionnaire. Additionally, pre-treatment scores from the Patient Health Questionnaire-9 and General Anxiety Disorder Screener-7 were evaluated. The study identified that self-reported motivation and the specified reference type significantly contributed to predictive outcomes. The irregular forest model achieved an accuracy of 0.71 for the test set, with a base rate of 0.67, an AUC of 0.60, and a p-value of 0.001. Adjusted accuracy was noted at 0.60, indicating a significant ability to predict reductions in anxiety and depressive symptoms. To improve the personalization and effectiveness of mental health care, it is essential to advance predictive models that accurately evaluate individual responses to therapeutic interventions.