Computer scientists at Johns Hopkins have built an algorithm that uses Twitter to gather information about mental illness trends.
Researchers combed 8 billion public tweets for users who shared their diagnoses of depression, post-traumatic stress disorder, bipolar disorder and seasonal affective disorder. They further looked for commonly used words among these populations in an effort to identify language cues that might be related to the illnesses.
The data analysis revealed two key findings: Signs of depression were more common in U.S. regions with higher unemployment rates and PTSD was more common at military bases that frequently deployed soldiers to Iraq and Afghanistan.
While neither of these discoveries are particularly surprising, they do suggest that Twitter could be an effective platform for collecting location-based mental health data. This type of data collection could also help psychologists investigate the language that tends to accompany various mental health problems.
“One of the most exciting possibilities that this enables is to examine the language of many mental health conditions simultaneously,” Glen Coppersmith, a senior research scientist at Johns Hopkins, said in an email to The Huffington Post. “And perhaps find as-of-yet undiscovered connections across disorders.”
The researchers hope the data, the reach of which could be difficult to obtain using other methods, will be useful for treatment providers and public health officials. While Twitter analysis will never replace traditional methods of collecting data on mental health, the new technologies could be a relatively low-cost and efficient supplement to conventional methods of data collection.
Previously, the same research team used Twitter to track flu cases, in order to generate a more complete picture of how flu cases were dispersed geographically. In this case, the researchers’ goal is also to promote discussion about mental illness and help reduce the stigma around these issues.
“With many physical illnesses, including the flu, there are lots of quantifiable facts and figures that can be used to study things like how often and where the disease is occurring, which people are most vulnerable, and what treatments are most successful,” Coppersmith said in a statement. “But it’s much tougher and more time-consuming to collect this kind of data about mental illnesses because the underlying causes are so complex and because there is a long-standing stigma that makes even talking about the subject all but taboo.”
Of course, there are some significant limitations to this manner of data collection. For one, the subset of the population willing to publicly disclose a mental health diagnosis is very small, and may not be an accurate representation of the overall demographic of people who have been diagnosed with a mental illness. The method may also not be effective for less-common mental illnesses or illness that are less likely to be discussed on Twitter, based on the demographic of the platform’s users (Alzheimer’s, for instance).
“This is one place we can look for quantifiable information relevant to mental health,” Coppersmith said. “We see this a small step towards an inexpensive way to measure mental health, which might allow us to measure and improve treatments and outcomes.”