Depression Screening Using Smartphone Technologies: A Data Driven Approach

BACKGROUND: Depression is a common mental health disorder that frequently goes undetected and untreated. Early detection of depression symptoms can positively impact patient outcomes, and smartphone apps hold significant promise as effective tools for depression screening. As the number of smartphone apps developed for depression screening increase, many offer the hope that depression can be detected using passively collected data from the phone sensors. Furthermore, the usability of such applications can influence their acceptance and adoption among at-risk populations. OBJECTIVES: This dissertation aims to: 1) identify and evaluate depression screening apps on their usefulness, usability, and integration and infrastructure (Paper 1); 2) explore the use of passively collected behavioral markers measured by geographic location (GPS) sensors to identify depressive symptom severity (Paper 2); and 3) formatively evaluate the usability a new depression screening mobile application using a clickable prototype approach (Paper 3). METHODS: The apps included in the review for Paper 1 were identified by searching both the scientific literature and commercial marketplace. For Paper 2, the StudentLife dataset was used to evaluate depressive symptom severity. This dataset was collected from 48 college students over a 10-week period, which included GPS phone sensor data and the Patient Health Questionnaire 9-item (PHQ-9). For Paper 3, two usability evaluation methods were involved: (1) a heuristics evaluation and (2) end-user testing with a think-aloud testing method. RESULTS: A total of 49 apps were identified after searching both the scientific literature and the commercial market (Paper 1). The results of Paper 2 revealed that a number of GPS features, including location variance and percent time at home were correlated with PHQ-9 scores. For Paper 3, participants were able to perform the predesigned tasks with the prototype and expressed mostly positive responses about the perceived usability measures regarding the interface. CONCLUSION: In response to the growing development of depression screening apps, there is a need to ensure that the apps provided use validated and accurate screening tools, have a high level of usability, and support users with a robust privacy and safety infrastructure. Our findings were also consistent with past research demonstrating that GPS features may be an important and reliable predictor of depressive symptom severity. Lastly, we demonstrated that our usability evaluation approach could be used to quickly and effectively identify usability problems in a health care application at an early stage of the development process using a clickable prototype