Big data mining and new hypotheses in mental health research

BigdataThis is a guest post by the computational epidemiologist Dr. John Ayers:

Most of us are aware of the “big data” revolution fueled by electronic information. It has been suggested that big data, along with hypothesis-free methods popularized by films such as Moneyball, will allow for an unprecedented growth of knowledge across disciplines, including epidemiology and preventive medicine. While I am a bit more circumspect in expectations (there is no substitute for survey data in many cases), I do believe that electronic data collected for a fraction of the cost of survey data can work hand-in-hand with research derived from more traditional sources.

Our study, published this month in the American Journal of Preventive Medicine, is a great example of the complementarity of big data approaches to mental health research. Previous work had identified mood symptoms as varying in many individuals suffering from depression. Using Google search as a proxy for changing patterns in mental illness, we sought to better understand seasonality in mental health.

For the analysis, all Google mental health queries were monitored in the U.S. and Australia from 2006 to 2010. Additionally, queries were subdivided among those including the terms ADHD (attention deficit-hyperactivity disorder); anxiety; bipolar; depression; anorexia or bulimia (eating disorders); OCD (obsessive-compulsive disorder); schizophrenia; and suicide. A wavelet phase analysis was used to isolate seasonal components in the trends, and based on this model, the mean search volume in winter was compared with that in summer.



While some conditions, such as seasonal affective disorder are known to be associated with seasonal weather patterns, the connections between seasons and a number of other major disorders was surprising. We found eating disorder searches were down 37 percent in summer versus winter in the U.S., and 42 percent in Australia. Schizophrenia searches decreased 37 percent during U.S. summers and by 36 percent in Australia. Bipolar searches were down 16 percent during U.S. summers and 17 percent during Australian summers; ADHD searches decreased by 28 percent in the U.S and 31 percent in Australia during summertime. OCD searches were down 18 percent and 15 percent, and bipolar searches decreased by 18 percent and 16 percent, in the U.S. and Australia respectively. Searches for suicide declined 24 and 29 percent during U.S. and Australian summers and anxiety searches had the smallest seasonal change — down 7 percent during U.S. summers and 15 percent during Australian summers.

Typically, telephone surveys are used to assess population mental health, but this approach has a large margin of error because respondents may be reluctant to give honest answers about their mental health. This approach also has high material costs and as a result, investigators are not able to collect as much data as they need to assess seasonal patterns, especially for rare mental illness. Data availability (or the lack there of) has tremendous consequences on theoretical and subsequently clinical developments in mental health. For example, we saw strong seasonal patterns for schizophrenia, a disease for which symptom severity had not been closely associated with seasonal patterns. In contrast, tremendous attention had been given to seasonal birth patterns in schizophrenia. Why? Population surveys readily collect birth date without any added planning, concerns with sensitivity/reliability/validity, or additional budgeting. Since all theories are based on some data, our approach can provide the beginning data stream for theoretical development in global mental health seasonality.



Clearly, these results are not intended to be definitive. Further research is needed, especially for understanding the link between search patterns and symptomatology. However, intuition suggests that these results are reflective of an important link between the seasons and mental health that goes beyond our previous understanding of these conditions. This kind of work can continue to cost effectively inform the field on a variety of vital health topics, and ours are just the beginning steps.

Disclosure: Dr. Ayers holds an equity stake in Directing Medicine LLC that advises hospitals, allied health groups and industry on data mining strategies.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s