Investigators analyzed 91 studies to develop a framework for future research.
The potential is clear. Oozing with photos, videos, and text, social media platforms offer untold data to researchers studying public health issues. Since some people assume a sense of anonymity exists on these websites, they’re especially useful for exploring topics that otherwise might be cloaked in secrecy.
Take drug use, for example.
A new study published in the Journal of Medical Internet Research examined 91 papers that used analytics to mine such data from social media sites like Twitter, Instagram, and Facebook, confirming their power in this area. Led by Sunny Jung Kim, PhD, a biomedical data science expert in the Geisel School of Medicine at Dartmouth College, the researchers developed a 4-part framework to guide future inquiries.
The “large scale” of social media posting made by people who use or have used prescription drugs like opioids for nonmedical purposes will “provide insight into important novel discoveries of collective public health risk behavior,” Kim and her colleagues wrote.
“More specifically, social media communication data aggregated by drug use-related search keywords,” they went on, “can indicate the level and stage of drug dependence, the actions of patients engaging in addiction recovery support groups, former users with or without relapse episodes, or current users with or without dependence.”
What does that mean? Well, according to the review, researchers can use social media drug data to develop insights for the individual and society at large. What’s more, those takeaways may relate to almost any aspect of problematic drug use, from coordinating just-in-time interventions and preventative campaigns to taking a bird’s eye view of changes in associated black markets or widespread use habits.
Kim’s team broke down their framework into 4 sections describing what sorts of information can come from these social media probes and why they’re important.
1. User characteristics. Most studies reported on their subjects using general groups, like college students, adolescents, Twitter users, or more. They often found that the type of drug use varied by demographic. Gaining a greater understanding here, Kim and her team said, can inform interventions and other campaigns. So far, though, research has paid less attention to analyzing this in conjunction with the second spoke in the wheel of the framework.
2. Communication characteristics. Every study in the review analyzed this component, which focused on how people discuss drug use, past or present. For example, one report detailed how some people spoke of prescription stimulants as a so-called “study aid,” examining how social network factors explained nonmedical use. This framework deals with how people portray their feelings about drugs and risky behaviors, according to the authors.
3. Mechanisms and predictors. This area remains “largely underexplored,” the researchers noted. But social media data on attitudes and risk perceptions can shed light on the motivation for sharing potentially damaging information and even produce clinical insights. Reaching out to other people with substance use disorders, for instance, could signal a need for social support.
4. Psychological and behavioral outcomes. Although studies have touted the significance of this aspect, it remains understudied. Future research into this might require population-level-based surveys and longitudinal follow-up interviews, the authors wrote. It could yield models and insights regarding the clinical implications of social media as behavioral intervention platforms, they noted.
The study advocated future research mix methods, incorporating survey, recruitment, and more. It also pushed for a multidisciplinary approach, bringing together data scientists, social scientists, and clinicians. All the while, they must consider the precision and sensitivity of social media big data, the authors said.
Going forward, investigators may use this framework to execute various computational linguistic, social network, and machine learning analyses.
But, the researchers added, they must keep in mind the ethics of data mining, interviewing, and disseminating results gathered on a platform that’s public but in some cases incorrectly thought to be private or anonymous.
Just 4 studies claimed to have gained Institutional Review Board approval, while 2 said the requirement was waived. “In most studies, potential ethical issues and practices were not discussed in detail,” the authors wrote. “This might be, in part, because the social media data in their studies was considered publicly open or because discussing ethical aspects was not directly within the scope of their study.”
In the future, however, the researchers urged their peers to strike a balance between ethical principles and potential scientific discoveries. This, they noted, should begin prior to launching a study.