Feasibility of a Machine Learning-Based Smartphone Application in Detecting Depression and Anxiety in a Generally Senior Population

• DOI: 10.3389/fpsyg.2022.811517
• Publication Date: 2022-04-08T14:29:31Z
• Authors: David Lin; Tahmida Nazreen; Tomasz Rutowski; Yang Lu; Amir Harati; Elizabeth Shriberg; Piotr Chlebek; Michael Aratow
• Keywords: Depression; Patient Health Questionnaire

Depression and anxiety create a large health burden increase the risk of premature mortality. Mental screening is vital, but more sophisticated monitoring methods are needed. The Ellipsis Health App addresses this need by using semantic information from recorded speech to screen for depression anxiety.The primary aim study determine feasibility collecting weekly voice samples mental screening. Additionally, we demonstrate portability improved performance Ellipsis' machine learning models patients various ages.Study participants were current at Desert Oasis Healthcare, mean age 63 years (SD = 10.3). Two non-randomized cohorts participated: one with documented history within 24 months prior (Group Positive), other without Negative). Participants 5-min 6 weeks via App. They also completed PHQ-8 GAD-7 questionnaires assess anxiety, respectively.Protocol completion rate was 61% both groups. Use beyond protocol 27% Group Positive 9% Negative. showed an AUC 0.82 combined groups when compared threshold score 10. Performance high senior as well younger ranges. many spoke longer than required 5 min.The demonstrated in recordings among Transformer methodology maintain improve over LSTM applied population.