Mental healthcare is difficult for many people to access, whether because of cost, stigma or waiting times. According to a recent pre-covid study in England [1] 18% of people are experiencing a mental health issue, yet only 5% actually access help through an employer’s assistance plan (if one even exists) [2] . The majority suffer in silence. In addition to the human cost, there are significant economic costs. In the U.S., mental health issues cost employers thousands of dollars per employee per year in medical costs and lost productivity [3].

At our previous company, we lost a colleague to suicide. Since then we’ve been on a mission to bring evidence-based mental healthcare to as many people as possible. We were surprised to find that Cognitive Behavioural Therapy is backed by over 30 years of research, and over 250 meta-analyses [4], yet remains difficult for people to access. CBT has been shown to be effective for a wide range of mental health conditions, even when administered digitally [5] or in books (so called bibliotherapy) [6]. While we do not believe that self or app administered CBT can replace the human touch, it does provide a convenient, affordable, always accessible and stigma-free way for many users to access support, when they may not have been able or ready to seek it in person. In the UK, the NHS presents a progressive “stepped model” of mental healthcare, depending upon severity from step 1 to step 4. We see apps playing a part in step 1, step 1a and possibly step 2 of this model [7].

While there are a number of CBT apps available, we wanted to base ours around Low-Intensity Cognitive Behavioural Therapy, a relatively new variant which has been used extensively in England to treat people with mild to moderate conditions at scale. In low-intensity CBT, the patient more or less treats themselves by doing exercises, and the therapist supports the process. This approach to CBT is especially suited to delivery through a digital form. Our clinical director, Professor Paul Farrand, was the architect of using Low-Intensity Cognitive Behavioural therapy at scale in the UK’s National Health Service (NHS), and the editor of one of the first international textbooks on the subject [8].

Our app presents the user with a chatbot which guides them through CBT techniques and exercises. We went for a chat-style UI to make the software more engaging—it’s a challenge to make an app of this kind be more than just a “content board”. Our chatbot engine is graph-based. At some nodes in the graph, classifiers applied to the user’s free text entry determine the path through this graph, along with the user’s state and past inputs. We want the paths within the app to be predictable and ensure that the chatbot doesn’t do anything unexpected. We’d seen more freeform generative seq2seq style chatbots produce potentially dangerous output at times in other apps and wanted to avoid this. Also, CBT has been shown to often work better when “therapist drift” is minimized—this is where the therapist drifts away from the prescribed structure of the exercises in CBT [9]. A graph-based conversation is a good way to keep things on track.

We offer the app as an employee wellness benefit to organizations. Since only 5% of employees tend to use the Employee Assistance Programme (EAP) at their employer, the app offers a way to support the “missing middle” for whom mindfulness/meditation apps are not sufficient, but who aren’t ready to access their employer’s EAP. We offer use of the app completely anonymously if the user wishes. It’s in the Apple App Store at https://apps.apple.com/app/iona-anxiety-mood-support/id14913... and Google Play at https://play.google.com/store/apps/details?id=com.iona.menta....

We’d be really interested to hear from the community about any ideas or experiences you've had when it comes to delivering mental health support digitally. We appreciate this is a sensitive topic so if you’d rather email us we’d love to hear from you at [email protected]

[1] https://assets.publishing.service.gov.uk/government/uploads/...

[2] https://www.eapa.org.uk/wp-content/uploads/2014/02/UK-EAPA-R...

[3] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1924724/ indicates major depressive disorder alone costs $4426 per ill worker per annum

[4] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3584580/

[5] There are a wealth of papers on so-called “iCBT” or internet delivered Cognitive Behaviour Therapy, covering both fully digital and human-in-the-loop variations. The approach dates back to the days when iCBT was actually just static content delivered on CD-ROM. I attach a couple below https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5421393/ https://pubmed.ncbi.nlm.nih.gov/23419552/

[6] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5788928/

[7] http://www.swscn.org.uk/wp/wp-content/uploads/2015/03/non-gu...

[8] https://us.sagepub.com/en-us/nam/low-intensity-cbt-skills-an...

[9] https://pubmed.ncbi.nlm.nih.gov/26752326/ https://pubmed.ncbi.nlm.nih.gov/19036354/ https://www.vgct.nl/stream/vr-keynote-9.15-waller.pdf