How to Predict Mental Illnesses: The Digital Future of Mental Healthcare

Before patients receive mental healthcare, they often must overcome several barriers. They can be tangible, such as a lack of financial resources or limited professionals in the area, or intangible, such as stigma or implicit racial bias.³ 

We tend to avoid engaging in the mental effort to challenge our status quo - even if our status quo has a negative effect on our wellbeing. We often rationalize our declining mental health, reasoning “it’s not that bad” or “everybody gets stressed”.

Due to our aversion to change, people tend to only seek out help once mental illness has become so overwhelmingly painful that the status quo is more painful than taking action. The increased severity of the problem before seeking help makes it more difficult for practitioners to address mental health issues.

Understanding mental health from this perspective, we can say that mental health treatment is fundamentally reactive: we tend to wait for mental health to decline to a dangerous state before we get help. Just as we don’t wait for a heart attack before we take care of our bodies, we shouldn’t wait for serious mental illnesses to take care of our minds. 

With the rise of smartphones and smartwatches, we’ve seen the increased power and ability of health-tracking apps. These apps take countless health data points (physical activity, resting heart rate, blood pressure, etc.) and organize them into an informative picture of overall health.

These apps empower users to fully understand the oscillating trends in their physical well-being. When there is a downturn in one category, users - or increasingly, their healthcare practitioners - can implement techniques to address their health problems, preventing more serious illness.  

While these apps already track physical health indicators that can help us improve mental health (i.e. sleep, exercise), there is an emerging field of data science that suggests that deeper indicators of mental health could soon be available to us.

According to this research, just like how our physical data goes into the health app, our online language and behavior could be highly predictive of our broader mental state.⁴ 

Absolute(ly) dangerous language: 

Studies have shown that the use of absolutist language (never, always, completely, nothing), negative emotion words, and first-person pronouns often appear in the online language of those with eating disorders, anxiety, and affective disorders.⁵ While this is largely correlation, there is still plenty of data that suggests that language could be predictive as well. 

AI can detect hidden language indicators: 

Advanced natural language processing and machine learning have been showed to predict depression severity and administer optimal treatments when given small pieces of narrative text.⁶ 

Given writing samples, natural language processing has been able to assess suicide risk in pediatric populations and identify linguistic features characteristic of early stage dementia.⁶  In one case, researchers had a 71% accuracy at predicting postpartum depression by analyzing birth-related phrases and pre-birth language on the new mothers’ Twitter accounts.⁶ 

While there is no substitute for consistent, in-person counseling, these computerized language analyses have been shown to be highly effective at uncovering a range of latent psychological conditions through conversational chatbots, artificial intelligence, and natural language processing.

If we can detect upticks in negative and absolutist language, we’re then able to administer personalized nudges that can transform these self-defeating, negative thought patterns into positive and healthy ones. A standout technique for this is cognitive behavioral therapy, which focuses on identifying, challenging, and replacing negative thought patterns. 

Scenario: Predictive analytics in your pocket 

For example, if your phone detects that you have been using more negative or absolutist language on social media, your phone sends a push notification that provides you with self talk strategies, daily habits, or words of encouragement that help you correct that behavior before it begins to spiral out of control.

If the data is suggesting your situation could be serious, the app could administer nudges that increase the salience of available options for help. If you allow it, it could also contact your medical provider or family member if needed, just like physical health applications. 

Paired with effective nudges towards healthy behaviors, applying data-driven cognitive behavioral therapy could break the pattern of reactive mental healthcare, as we can tackle the negative thoughts as soon as we see a downturn in language data.

While research on predictive mental health analytics is still emerging, practitioners have already begun to apply techniques from it successfully to alleviate a bevy of mental health issues. Here at The Decision Lab we collaborated to develop Hikai, an AI-powered mental health platform. Using a conversational chatbot, Hikai leveraged cognitive behavioral therapy to improve workplace mental health. Within a pilot group, 71% felt more engaged by their work, while 82% found that it helped reduce their stress levels significantly. You can find out more in our case study here.

The Decision Lab is a behavioral consultancy that uses science to advance social good. The mental health crisis is one of the world’s greatest challenges, and combining smart data with behavioral science could be a big step in the right direction. We have long worked with leading mental health organizations to develop innovative, empathic, and scalable mental health solutions. If you’d be interested in tackling the mental health crisis together, contact us.