Can AI help in navigating mental health?
How can natural language processing programmes offer personalised and immediate care? How can bias be mitigated in these chatbots? Do they help clinicians as well?
The story so far:
We live in a world where therapy is a text away. Natural language processing (NLP), a branch of Artificial Intelligence (AI), enables computers to understand and interpret human language that mirrors human comprehension. In mental healthcare, we are already seeing a rapid evolution of use cases for AI with affordable access to therapy and better support for clinicians.
How does it help patients?
External and internalised stigma persists across demographics and countries.
Through textbased platforms and virtual mental health assistants, NLP programs provide privacy and anonymity that can improve helpseeking behaviour. For users, the chatbot can support them in reframing thoughts, validating emotions and providing personalised care, especially in the absence of human support. Not only is this beneficial when a therapist is not accessible, but it also helps improve patient health outcomes just as well as inperson care. Mental health treatment requires continuity of care to take a more holistic approach and reduce instances of relapse. For example, digital therapy assistants can help point you to resources for healthier coping in instances of distress, grief, and anxiety. Since these chatbots are scalable, costeffective, and available 24x7, they could therefore be integrated into existing health programs. Additionally, companies building chatbots must proactively expand the scope of service delivery through partnerships and collaborations for followup services such as referrals, inperson treatment, or hospital care, where needed.
How does it help clinicians?
Mental health illnesses have complex causes of origin, making it difficult to design a straightforward protocol or make a quick and accurate diagnosis. By using vast datasets, AI tools can help summarise information including clinical notes, patient conversations, neuroimages, and genetic information. This can help clinicians get up to speed with the entire patient history, saving valuable time during sessions.
Recent advancements in NLP programs have demonstrated the ability to forecast responses to antidepressants and antipsychotic drugs by analysing brain electrical activity, neuroimages, and clinical surveys. Such predictive capability can streamline treatment decisions and minimise the risk of ineffective interventions. Some chatbots are also creating etriaging systems that can significantly reduce wait time and free up valuable clinical personhours. With improving bandwidth, mental health providers can devote a higher proportion of time to severe mental illnesses, such as bipolar disorder and schizophrenia, where specialised care is required.
What’s next?
There is immense potential and promise in these applications and we expect to see a growing adoption. Going forward, companies must refine these applications by using more diverse populationwide datasets to minimise bias. These programs can also incorporate a wider set of health indicators for a comprehensive patient care experience. We expect greater success of these programs if they are guided by a conceptual framework for improving health outcomes and rigorously and continuously tested.
In the pursuit of innovation, governments and institutions need to prioritise user safety and wellbeing by ensuring adherence to global compliance standards. As these applications evolve, we must persist in updating our beliefs, governing laws and regulations, and demanding better standards of care.
Iti Bhargava and Namrata Rao are researchers in mental health in India, and Manmath Goel is a healthcare investor.