Sample population range
There were two types of studies based on the sample population type: studies with human subjects as sample and studies with data extracted online as the sample population. While the range of the sample size for human subjects varied from 20 adolescents to 526 adults, the sample size for posts differed from 132 tweets to 620,000 posts from various social media platforms. Studies with posts or tweets as the sample population collected the data from either one single source or made use of multiple sources.
Mental health domain
The included studies either concentrated only on a particular illness like depression or studied multiple illnesses. Fourteen studies focused on general mental health disorders without specifying one particular disorder11,12,14,15,16,17,18,19,21,23,25,28,32,35. Twelve studies were on depression8,9,13,20,22,24,26,27,29,31,33,36, two on schizophrenia10,36 and four discussed suicide29,30,34,36. Individual studies also discussed other illnesses such as anxiety36, stress8, bipolar disorder20 and post-traumatic stress disorder20.
3.1 Characteristics of studies
A summary of the selected articles is presented in Table 1. Of the thirty studies, ten studies discussed interventions that used social media, five were observational studies, fourteen studies employed machine learning methods and one randomized control trial. The intervention based studies included three qualitative analyses studies that respectively examined the comments under YouTube videos related to schizoaffective disorders8, the posts from various social media platforms to understand what users expected in this exchange28, and the tweets to see why users use social media for mental health services9. Intervention based studies also included three thematic analyses14,16,26 with various purposes such as assessing the power of Twitter for mental health and studying experiences of youth using social media for the same purpose. One longitudinal text analysis study32 was conducted to explore timelines of users that committed suicide to identify patterns and another text outlined a framework that proposed an e-psychology platform13. In addition, there was a conceptual model21 and an empirical model27 to promote the social media platform for mental health services.
One randomized control trial11 was carried out to develop and test a tool that helped prevent depression. Two surveys10,19 were carried out to see the interest of psychiatric patients in using technology for treatment and how social media can help manage them. Another study conducted a survey to study exclusively about the ‘Historically black colleges and universities’ (HBCU) population. One cross sectional study6 to investigate the popularity of mobile apps for mental health and an online questionnaire33 to study the effectiveness of social media interventions were included in the thirty articles.
The fourteen machine learning based studies7,1,15,17,18,20,22,24,25,29,30,31,34,35 included ten studies that utilized machine learning techniques to predict or classify users with depression12,20,22,24,25,30,31, anxiety6, suicide ideation29 or any cognitive distortions35. Two studies23,38 used Natural Language Processing (NLP) to extract and transform information in a post or comment to predict possible post-traumatic stress disorder (PTSD), depression, bipolar disorder, and seasonal affective disorder (SAD) and two studies18,31 described about tools with machine learning that helped in the collection and exploration of big data.
Ten articles out of the thirty examined accounts or collected posts from different social media platforms rather than focusing on one of them15,16,19,20,21,24,28,29,32,33. Examinations and findings of seven articles were based solely on Twitter9,12,13,14,17,18,27. One study investigated comments under videos in YouTube8 and four study concentrated on Facebook22,25,30,31. Six studies analyzed the usage and benefits of different mobile applications related to mental health6,10,11,17,23,26 while three studies collected posts from online websites that give access to posts from social media accounts and personal blogs freely7,34,35.
3.2 Study outcomes
All the studies involved arrived at the conclusion that social media has the potential to be used as a mental health service provider. Eight studies particularly illustrated how social networks can act as a mental health consultant and improve the wellbeing of people suffering from mental illnesses5,18,21,22,24,29,30,31. One study reported that 16% of the people they interviewed believed that the usage of apps related to mental health can bring positive results and help people understand more about their mental issues4. Four studies identified important themes that proved the use of social media to be convenient as they provide immediate response and continuous interaction and care7,14,27,30. The presence of peer support was an important factor for people approaching social media over traditional consultation methods. In a study where YouTube comments were examined, people with schizoaffective disorders posted comments under videos so that they could identify themselves with others and have support from fellow schizoaffective users6. Subjects of five studies confirmed they were more comfortable and less conscious using social media as their service providers for mental health36,19,22,37,38.
Fourteen papers used machine learning algorithms to analyze the data extracted online. The results of the studies that used classification models to predict depression, yielded above 70% accuracy scores in all three cases25,26,37. One study categorized depression into four levels and through their model, classified a user with depression into any of the four categories25. The studies that employed Natural Language Processing techniques were able to extract information from the posts or comments to predict possible mental illnesses18,23,38. Another study that made use of a multilevel predictive model to detect depression with respect to another variable concluded that a multilevel predictor had more performance capability than a model that predicted depression single-handedly34. Three studies that analyzed posts related to suicide found commonly that there is always an increase in the occurrence of posts or tweets nearing the time of suicide28,31,35 and it is viable to detect and track suicide behaviors through social media.
Other studies focused their work on two particular populations:- Sexual and gender minorities (SGM) and Historically black colleges and universities (HBCU). The study that took ‘sexual and gender minorities’ (SGM) as their population described that SGM exhibited more negative feelings through Twitter compared to others12. The study that tested the use of an app specially designed for students in ‘Historically black colleges and universities’ (HBCU) called ‘Soul Glow’ reported that the app provided a platform for students with mental health issues to share their stories and interact openly24.
3.3 Patient reported experiences
One of the most prominent advantages of social media is its reach and accessibility. McClellan, C.28 states the fact that mobile phones and social media are ubiquitous and personal in nature, thus helping the user keep their health status private. Van Rensburg, S. H et al.30 highlights the rapid response and better communication ability of social media to be outstanding. Patients felt like they were heard and got immediate responses6. Since social media is available all over the world, people from different countries, backgrounds and cultures were able to communicate and connect to each other based on their common mental health disorders18. This allowed people to talk about a stigmatized topic without the fear of being judged or labeled. Berry N et al. in their thematic analysis found that users prefer using Twitter as it provides a sense of belongingness and community and thus could be identified with others. They also found that tweeting about mental health issues aids in raising awareness and reduce the stigma attached with it all the while helping the users with such illnesses cope with their state of mind. Zhang Q et al. in their research states that with considering social media as a mental health provider users or patients can get treatment any time regardless of place or hours and can use them according the desires of the patients. This method also helps reducing the dependency on therapist and makes the work of both therapist and patient easier.
Along with the advantages certain disadvantages or challenges were also emphasized. Privacy is the most notorious among them. People are taking a very high risk of losing their confidentiality while discussing their mental illnesses online and public6,4,25. Reliability of information available online is another risk. Patients should take care of the facts and figures they get online and make sure they are accurate and reliable. Also it is not necessary that a particular way of treatment can be applied or effective on everyone. Fergie G et al. details in their study about the concerns of validity and reliability of information online and distinguishes between content produced professionally that is evidence-based and social media content.