A Sentiment Analysis of User Reviews of Depression Apps Features (Preprint)
BACKGROUND Mhealth apps are promising to overcome barriers to access mental health care. Adoption and continuous use, however, depends on users’ decisions. App reviews both reflect and influence users’ attitude and experience towards apps and influence their propensity to use mhealth apps. OBJECTIVE We investigate user app reviews on specific features in depression apps (psychoeducation, medical assessment, therapeutic treatment, supportive resources and entertainment). METHODS We extracted 3,261 user reviews of depression apps, isolated reviews associated with single feature apps. We then analyzed reviews using LIWC, a natural language analytical tool and contrasted language patterns associated with different features. RESULTS Medical Assessment features stand out for the strong negative emotions and negative ratings they generate, as users receive potentially disturbing feedback on their condition. Symptom Management and Entertainment features generate less negative emotions and anxiety. Therapeutic Treatment features also generate more positive and fewer negative emotions, even though user experience is less authentic (i.e., reflecting a personal experience). CONCLUSIONS Developers should be cautious in their choice of features when they are targeting potentially vulnerable users. Medical assessment feedback being riskier while offering information, contacts or even games may be a safer starting point to engage people with depression. App features emerged as a key dimension to consider when investigating user experience with mhealth apps. Methodologically, app reviews can be leveraged to investigate specific app features at the level of a family of apps. Specifically, Natural Language Analysis proved to be a responsive tool to investigate behaviors related to a quickly changing app environment.