Mental Health, AI

What role can AI play in diagnosing mental health patients?

Diagnosing mental health disorders involves numerous steps and isn’t always easy. Doctors may perform a physical exam or order lab tests to determine or rule out physical conditions that may be causing a patient’s symptoms. However, there is no specific medical test to directly diagnose a mental health condition. A psychological evaluation is also performed, during which patients are asked to talk about their emotions, thoughts, behavior, and symptoms. Different symptoms are associated with different mental illnesses, so doctors can use the patient’s responses as well as reference the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) to help determine a diagnosis.

There’s another technology that’s emerging as an additional resource when diagnosing mental health patients: artificial intelligence (AI). Even though using AI for this purpose is relatively new, it’s already showing great potential. Various AI tools are allowing mental health professionals to predict mental health conditions by analyzing medical data and written and verbal communication. Here’s a closer look at the role AI can play in diagnosing mental health patients.  

Mental Health, AI

Predicting the onset of mental health issues using AI and medical data

Data is a valuable resource, and AI makes it even more valuable by using it to generate actionable insights. With the right AI tools, mental health care professionals can identify signs of mental health issues in patients and even predict the onset of mental health conditions. 

Lumiata, for example, uses AI and machine learning (ML) to predict the onset of depression, dementia, and bipolar disorder, among other acute and chronic diseases. We start by bringing a client’s data into our ecosystem, then optimizing it for machine learning by validating, normalizing, and cleansing it. We also add our proprietary disease codes, which help identify risk factors regarding the onset or progression of different diseases, and create Person360 records, longitudinal records of each person in the dataset. The data can then be used with the clinical ML models in our Spectrum AI Model Catalog, including our disease onset model, which can predict the onset of more than 20 diseases within the next 12 months.  

Lumiata can identify markers for depression by looking at multiple complex conditions. For example, if obesity or malnourishment is predicted in a patient along with a history of or predictions for deficient thyroid activity, this is a strong indicator for Major Depressive Disorder (MDD). A history of the cardiac bypass with no history of completed physical rehabilitation is another strong marker for MDD. 

We combine our client’s optimized data with our AI technology to predict the onset of mental health conditions, giving mental health care professionals another tool in their tool belt to use when diagnosing mental health patients.

Predicting mental health issues using AI and written language

AI can be used to assess written messages and identify signs of mental health issues. Some apps, for example, have patients “talk” to chatbots via text-based messages, which can then be analyzed for signs of mental illness. In December 2020, researchers announced they had used data from Facebook messages to predict psychiatric diagnoses up to 18 months before the user was officially diagnosed. [1]

Michael Birnbaum, an assistant professor at the Feinstein Institutes for Medical Research, led the study. The team collected 3,404,959 Facebook messages and 142,390 images across 223 participants, some of who had diagnosed mental health disorders, including mood disorders and schizophrenia spectrum disorders, and some who were considered healthy. The researchers looked at features that had been uploaded prior to the participant’s first hospitalization, some up to 18 months prior. [2]

Since the researchers knew when the participants had been diagnosed, they could work backward and see how far prior to the diagnosis a prediction could be made about that person’s mental health. 

The team found that the use of swear words was indicative of mental illness in general, and negative emotions as well as perception words like “see,” “hear,” and “feel” were indicative of schizophrenia. Photos with blue tones were associated with mood disorders. [3] The ability to predict mental health issues by using AI and Facebook posts or other text-based messages could allow professionals to diagnose patients months before they would otherwise receive a diagnosis from traditional methods.

Predicting mental health issues using AI and verbal language

Apps can ask people questions and have people answer by speaking their responses into their phones. Apps can also have people respond via text-based responses. The apps then assess the responses. Changes in pace, tone, etc. may signify a mental health issue. If the app identifies areas of concern, it can alert the person and their healthcare provider.

In a similar way that AI can be used to assess written text and identify signs of mental health disorders, AI can also be used to assess verbal language. Speech and mental health are closely linked, according to Dr. Henry Nasrallah, a psychiatrist at the University of Cincinnati Medical Center. He says that fast speech can be indicative of mania, monotone can point to depression, and disjointed word choice can be connected to schizophrenia. While these traits must reach a certain level before a human clinician will notice them, AI could be trained to identify the traits when they’re more subtle. [4]

Peter Foltz, a research professor at the University of Colorado Boulder, and his team designed a mobile app that guides users through verbal exercises like answering questions about how they’re feeling or telling a story. AI then analyzes the audio content and looks for signs of mental distress by comparing the content to the user’s previous responses as well as responses from a larger patient population. In a study involving 225 participants, the app was able to identify speech-based signs of mental distress at least as well as clinicians. [5]

What’s the impact of using AI when diagnosing mental health patients?

AI enables mental health care professionals to better identify a patient’s risk factors for mental health conditions and predict mental health conditions sooner than traditional diagnostic methods allow.

These predictions can result in earlier diagnosis, which enables mental health care providers to address the patient’s risk factors sooner, provide better care in a timely manner, and achieve better outcomes for their patients.
To see how Lumiata can support your organization in diagnosing mental health patients, click here to request a demo.