The Sound of Mental Health: Audio Features as Indicators of Depression and Anxiety Symptoms in Behavioral Treatment

Measurement-based care (MBC) is vital in psychotherapy for evaluating client progress and informing treatment decisions. Despite its importance, self-report assessments frequently encounter low completion rates, diminishing their efficacy. Therefore, there is a pressing need to develop non-intrusive, objective methods to monitor mental health symptoms. This study aimed to investigate the potential of audio features extracted from therapy sessions to predict depression and anxiety symptoms in community treatment settings. Utilizing machine learning techniques, we aim to bridge the gaps in current MBC practices, thereby enhancing the monitoring and outcomes of behavioral health treatments. Machine learning models analyzed audio features from 2,348 therapy sessions across varied behavioral health programs, aiming to predict mild versus moderate-severe symptoms of depression and anxiety. The models demonstrated high efficacy, with accuracy rates of 75.1% for anxiety and 73.2% for depression. Precision and recall were also strong, particularly for anxiety (81.3% precision, 61.3% recall) and depression (71.4% precision, 68.1% recall). This study highlights the potential of audio analysis as a cost-effective, non-intrusive approach for enhancing MBC. Despite limitations, such as limited demographic data, the findings suggest significant implications for improving therapy monitoring and patient outcomes.