Preventing suicide before it happens, is one of the most challenging tasks for the police force as well as family members. Observed main reasons behind incidents of suicide cases are daily personal lifestyle, working condition, social life behavior and person’s depression condition. According to recent studies, examining the effects of social interactions and context practices on various modes of expression, such as visual, textual, and social activity using social media is possible to be used to predict depression signs. The paper has put forward elements of reviewing initial studies on social media depression detection. Four digital libraries were searched for primary studies: ScienceDirect, SpringerLink, IEEE Xplore Digital Library and Association for Computing Machinery (ACM) Digital Library to broaden the results. The technique of this study is to review each article. Twenty-eight initial studies were examined. In the conclusion of this study, geotagging is the most analyzed social media platform technique to find locations shared by people. Hashtags were the most applied for depression detection. Rule-based sentiment analysis can grasp phrases that contain sarcasm to detect depression. Multimodal features such as users’ comments and image posts with machine learning or deep learning acquire the best output results to detect depression.