Prediction Technology: The community feedback period for this project began on 9/12/18 and will end on 11/11/2018

What is the problem that needs to be addressed? Please describe how it is related to mental health.

Anxiety and depression are the top causes of disease burden in the United States and cost tens of billions of dollars a year. Although current treatment strategies such as cognitive behavior therapy and community-based peer support groups have been successfully used to improve symptoms of anxiety and depression, these treatments require a great deal of time and economic resources. Cost-effective treatment for anxiety and depression can be especially difficult in large institutions where psychologists receive an overwhelming number of patients, reducing treatment effectiveness and leading patients to feel neglected.

To learn how to improve the effectiveness of mental health interventions, researchers have attempted to collect data on the needs of patients with anxiety and depression, such as through interviews, surveys, and tracking patient cases/reports. However, these methods suffer from a number of limitations including a long lag-time in reporting, as it often takes years for government agencies and researchers to collect interview and survey data. New methods are needed to address the lack of data, and the need for higher quality data to improve the lives of people suffering from these mental health issues.

 

Why is this a concern for Orange County? What can Orange County and other counties learn from this project?

Orange County has a large number of individuals with anxiety, depression, and related mental health disorders, including a growing number of veterans suffering from anxiety, depression, and post-traumatic stress (PTSD), a growing homeless population suffering from similar mental and behavioral health issues, and a growing number of youth suffering from these mental health issues and at high risk for suicide. New methods are needed to prevent these growing risks for Orange County residents and these important populations.

 

What is currently being done to resolve this problem in our county and throughout the United States? If applicable: Is it working; why or why not?

A number of methods are being used to both collect data as well as intervene to reduce anxiety and depression, both broadly, as well as more focused among veterans, homeless populations, and youth. For example, a large number of social services already exist that can provide therapy (e.g., CBT), outreach, skills training, and class such as yoga/meditation. Many of these are also available online, such as through mobile apps for CBT. Data collection methods can help inform these interventions. For example, the typical methods used to better understand the needs of people with anxiety and depression include interviews, surveys, and patient case reports. Although these methods are working to some extent, they also suffer from limitations. For example, the methods used to collect data on the needs of patients with anxiety and depression are not real-time or near real-time. Collecting interview and survey data, aggregating it, and analyzing it can take years. It is difficult to understand the needs of people with anxiety and depression, and to tailor interventions toward them, if the data are based on reports from years ago. New methods are needed to better monitor in real-time, or better yet predict, the people at risk for anxiety and depression as well as their mental health-related needs.

 

What is new or different about this project idea? Please describe how this differs from what is already being done (Question 6). Please list any research that was done on this topic.

No known mental health research/interventions are incorporating the proposed methods.

 

What is the project idea? Please describe how this project will operate.

This project will leverage the growing amount of data from people’s social media and smartphones and use it to monitor and predict people’s anxiety and depression, and related mental health needs. The ultimate goal of the project will be to collect data in real-time, in an organic comfortable environment, that can help inform tailored interventions for OC residents suffering from and/or at risk for anxiety and depression. We will invite OC participants from a variety of high-risk populations (e.g., veterans, college students, homeless populations) to complete depression and anxiety screening assessments to identify those with anxiety and/or depression. These individuals will be invited to participate in a study where they can choose to provide access to their social media data (e.g., Facebook, Twitter, etc). They will read about their risks of participation (e.g., informed consent; this will be approved by our university ethical review board). Among those who allow access, we will collect and analyze their real-time social media/smartphone data while having them complete periodic surveys to assess their mental health-related behaviors and outcomes. We will use advanced computer science/artificial intelligence methods to “train” machines to learn how to identify patterns from their social media data that can be used to predict their anxiety and depression risk and related mental health outcomes. We will also study their comfort in us using these methods to track their mental health to better understand how to safely and ethically collect real-time social media data and tailor it for personalized interventions.

 

Additional Information:

For more information on these types of research methods: predictiontechnology.ucla.edu

 

August 28, 2018

11 comments

Below are two articles that mention companies that are capturing data in a similar way. How would this project be different?

https://moodsurfing.com/data-mining-mental-health-digital-phenotyping/
https://www.techemergence.com/diagnosing-and-treating-depression-with-ai-ml/

Great find and cool articles! This project would be different than the products/companies in the links because this project is focused on identifying the efficacy and potential effectiveness of a new process, or method, rather than a product. Those links showed products that are doing similar type of work, but one problem is that they didn’t study or report findings on the process and methods behind the product, just the results of the product itself. If the product works, or doesn’t work, we don’t know why. That makes it difficult to apply the insights into new technologies and approaches. Almost every specific product has a life or cycle, and people eventually lose interest. It might not be because of a problem with the product, but just might be that the product has become outdated. But if we study the process or method behind making something work, then we can create new products and tools that use what we’ve learned. That’s what this project tries to do. It attempts to study how (and through what means) social data can monitor and predict mental health outcomes. For example, it might find that people’s personality characteristics can be identified by their social media data, and that this can be used to predict people’s mental health outcomes. That would be an insight that could be incorporated into a lot of new technologies or projects in the future, rather than just into one.
BTW, I served on the ethics review board for one of the companies listed in the link for an internal study they did on depression– they are doing some cool things!

What would be the specific learning objectives?

we would want to know 1) whether it’s possible to use social media and smartphone data to predict people’s mental health states (e.g., do people post enough information about their mental health states on social media to be able to analyze it and sufficiently model/use it to predict their mental health states, 2) learn what they talk about on social media and through their smartphone interactions that is related to their mental health, and 3) see whether we can use their information from social media and smartphone data to predict their real-time/future mental health outcomes such as anxiety and/or depression.

How would the success of this project be measured?

We typically measure success of these types of projects in 3 ways: 1) is this a feasible project idea to carry out/it possible for us to get the data we need (e.g., will people post at least 500 words on social media (this is the amount of text we’ve found is the minimum needed to come up with adequate models), 2) will people express information about their mental health while we’re monitoring them (e.g., do they talk about mental health information and how they’re feeling, and if so, what do they discuss?) and 3) can we identify some associations between their social media data and actual mental health states (e.g., does the information we learn in #2 from their social media seem to “statistically” predict their mental health)?

Will the tracking of these individuals include therapeutic services if their profile shows they are in need of help?

That’s a great idea. We thought it would just study the process and methods behind using social media data to predict mental health outcomes, but the insights gained could also be used to route people to therapeutic services. For example, we could in a first phase identify algorithms that can use social media data to predict people’s mental health status. Once we’ve done that we could use the algorithms/models for detecting people at risk or in need to route them to mental health services. I think it would be best to do this in two phases though as they are slightly different projects with different methods of evaluation: one is an analysis of data to monitor and predict mental health status, the other would be an analysis of an intervention that uses the models that predict mental health status. We’ve done a lot of work with interventions too (for example, see digitalbehavior.ucla.edu) and integrate one of our interventions or just route people at high risk to specific clinical settings. Anyone else have thoughts on whether to do this in one phase or two?

It seems as though there would need to be information in the informed consent documents that participants would sign that outlines protocols for participants who may be at risk of self harm or in need of ongoing mental health support. I feel having a mental health counselor on staff for this project would be the most appropriate option and this person could provide referrals or crisis intervention.

Valera Health (https://valerahealth.com/) has implemented similar technology, originally intended to detect psychosis. Valera required at least a $3.4 million investment and is staffed/supported by eight MDs (many with dual degrees such as MBA, MPH, etc) and a technical development team of about 10 people (https://valerahealth.com/team/). The expense of this team is very likely cost-shared with clinical/academic income.

Since the original pitch of using Machine Learning Artificial Intelligence, Valera Health has pivoted to providing a communications platform to connect patients with providers. Valera has since downplayed its Machine Learning Artificial Intelligence technology and I am not aware of any peer reviewed publications describing meaningful outcomes of the original AI effort that cite reduced morbidity/mortality, reduced healthcare expenditure, improved patient income, improved patient quality of life, etc..

The more traditional screening methods for depression (SIG-E_CAPS) don’t require such a hefty investment. While the merits of this proposal are interesting, the investment required is more appropriate for a NIH or NSF grant.

It sounds like you have a couple of questions
Is this novel if other companies are doing things slightly related?
You mentioned companies working on machine learning, but that’s a huge field and those companies aren’t specifically doing what was pitched here by the team.

Is this too expensive/Isn’t the NIH or NSF the correct funder for bigger grants?
How do we know what the team needs to complete or what the funders have to offer? And NSF and NIH grants have lots of different sizes, not just big grants, so that wouldn’t be what determines where a grant goes.

Leave a Reply

Translate »
X