Mental health support is in short supply. While one in five U.S. adults live with a mental illness, as of March 2023, 160 million Americans live in areas with mental health professional shortages.
At the same time, a limited supply of psychiatrists and other healthcare professionals leaves many service providers overburdened, with a limited number of staff under pressure to diagnose and treat patients in the shortest time possible.
Artificial intelligence (AI) could be a game-changer for the mental health industry not only because it helps study the symptoms, predictors, and signals that can be used to diagnose conditions like Alzheimer’s, depression, and schizophrenia but because it can also be used to discover new treatments to address them.
How AI Can Help Mental Healthcare Professionals
In recent years, a number of researchers have been experimenting with AI, machine learning (ML), and deep learning to see if they can be used to diagnose and treat mental conditions.
For instance, in 2022, Stanford researchers used machine learning to predict opioid use in patients. As part of the study, the researchers processed the deidentified data of 180,000 Medicaid employees to scan for the key indicators of chronic opioid use.
This approach led to new insights into the treatment of opioid addiction. More specifically, the study showed that the popularly prescribed opioid pain treatment Tramadol could be used to predict long-term opioid use.
It also found that 29.9% of first-timers or those who’d taken opioids for less than 2 months were at risk of opioid dependency.
In another study, researchers from Queen’s University Canada demonstrated how AI and deep learning could be used to process clinical interview transcriptions and automatically assess the severity of a patient’s depression. Here, AI helped to standardize and accelerate the assessment process.
It’s important to note that AI’s diagnostic capabilities aren’t meant to replace a mental professional’s judgment but to give them access to more insights that they can use to inform their decisions about how to treat and support their patients.
Is AI Accurate Enough to Predict Mental Health Conditions?
When using AI in a healthcare setting, it’s paramount that the insights generated from a dataset are as accurate as possible, as real lives are at stake. The wrong decision or diagnosis can lead to vulnerable individuals going without the support they need.
However, one of the key challenges in using AI to predict mental health conditions is that the accuracy of their diagnosis or predictions depends on the quality and accuracy of the data they’re trained on.
Clinical researcher Sarah Graham et al. reviewed 28 studies that used AI to predict, classify, or subgroup mental health illnesses such as depression, schizophrenia, and other psychiatric illnesses and found that their overall accuracy ranged between 63%-92%.
While 63% is relatively low, 92% is a lot more promising. It suggests that if researchers take the time to feed AI systems the right data signals, they can drastically increase the accuracy of their results.
For example, a study released earlier this year found that AI (CRANK-MS) could be used to predict Parkinson’s disease with 96% accuracy.
This was possible due to the detailed data available on the test subjects, with the study following 78 individuals from Spain, who provided a blood sample between 1993 to 1996 and were followed for 15 years.
The key take-home here is that AI systems can make much more reliable inferences when they’re provided with rich signals to derive insights.
Ethical Concerns
Of course, while AI has a useful role to play in supporting mental health patients, it must be applied carefully. In practice, this means healthcare professionals can’t rely on AI to diagnose mental health conditions but instead to better inform their own diagnosis and understanding of a patient’s condition.
For example, if a psychologist or a psychologist assesses an individual for depression, they could use their own knowledge to assess the severity of their condition and use additional insights generated by AI to increase confidence in their diagnosis or crosscheck potential treatment options.
It is also important to note that organizations that want to apply AI to process patient data need to collect permission from those subjects, or de-identify their data, to protect their data privacy rights.
Failure to do so could result in significant legal liabilities under data protection frameworks such as the GDPR and HIPAA or other types of reputational risk.
Just recently, the U.S. Senate has called for scrutiny of mental health apps like BetterHelp and Talkspace over concerns that vendors could be collecting and sharing private information with third parties.
Augmenting Healthcare Providers
AI isn’t a silver bullet for the mental health crisis, but it does provide opportunities for practitioners and clinical researchers to augment their capabilities to diagnose and treat patients.
Fundamentally, the more data there is to support a diagnosis or a treatment, the better.