Applying modern data science techniques to improve global mental well-being, with the collaboration of Big Tech firms

Note: I am not a medical, or mental health professional. This article is the exploration of an idea without the assessment from professionals in the health space. This can be seen as a thought experiment.

Introduction

This application of a systems approach focuses on one possible solution to a bigger problem (and the obstacles preventing it from being a solution) rather than the core problem. The problem is a social and medical one: mental health.?

There is a global army of counselors, therapists, psychologists, psychiatrists, academics, and scholars around the globe who work towards healing the world on a daily basis. Estimates place the share of the global population with a disorder at around 10% (Ritchie & Roser 2018), and that’s only for official disorders and disregarding general unhappiness or well-being. To help the mind of anyone, the best source of information is directly from the patient’s mind, and so any care person has to ask questions, listen to, and interpret the answers.

The objective here is to determine where business and technology can be inserted into the system, how it could help, to what degree it can help, and finally what needs to be done to make it a reality. This would involve determining who needs to work together (who has the resources: data and the ability to interpret data), how they need to work together (shared channels for communication), and what obstacles are in the way of them working together (laws and lack of incentives).?

Let us begin by mapping out the Challenge Landscape for the core problem, before moving onto the “new treatment” sub-problem.

The Challenge Landscape

As I mentioned, the objective is to find a business solution to this problem; how business can help, and which specific organizations are able to do so. The challenge lies in the domain of one of the solutions. Before we get there, however, we need to see the broader problem at the highest level first:

At the heart of our map is our problem, and on either vertical side of the problem lies the causes (above), and the solutions (below). The problem at the heart of this problem canvas is what I have already referred to as the “core problem”: the spectrum of mental health issues.

The top side of the canvas speaks to the root causes of the problem; physiological & environmental factors which lead to unbalanced and/or unhealthy minds. This side of the canvas also hints at what states of mind are captured in the “problem” section of the canvas. Mental health issues are not limited to depression and other physiological conditions. I am defining it here to also include overall happiness, mental wellbeing, or even discriminatory thoughts/behaviors. Discrimination and prejudice are signs of an unhealthy mind and also manifest in behavior that hurts others and is toxic to society. Studies have suggested that extreme racism, for example, can be a symptom of a psychotic disorder (Poussaint, 2002).?

Drilling down into our causes further we find deeper root causes. Some of these causes are fixed (e.g., genetic predisposition), and some can theoretically be changed (e.g., an abusive partner could have an even deeper root cause for their unhealthy outlets, which could be resolved with the right help).

Now that we have understood the problem and the root causes, we can investigate the solution (or “treatment”) side of the canvas. Treatments are broadly deployed in the form of medication, and/or therapy. Medication is distributed based on the type of physiological condition the patient is facing and is largely broken down into 5 main groups of psychiatric medications. As for therapy, different types of therapy are used on a case-by-case basis for each patient, these typically involve either external changes (one’s environment often can’t be changed and it needs to be escaped – as with an abuser who won’t change their habits) or an internal one like changing one’s outlook using psychotherapy techniques (e.g., Cognitive Behavior Therapy (CBT), Dialectical Behavioral Therapy (DBT), Rational Emotive Behavior Therapy (REBT)). How are these evaluations conducted? Just like any medical evaluation, they are done by looking at the symptoms.

Prescribing the right sort of treatment requires first an evaluation from which an expert can tailor the right treatment for the patient, and it is here we start to arrive at our subproblem. Applying either of these two forms of treatment requires evaluation, performed by other humans. The same way that trained AI algorithms are being used to help physicians detect medical conditions in lung x-rays, this is where we have room to make massive improvements in the way we evaluate, and to some degree treat mental health issues.

Evaluation is the first step in the treatment process, from which a trained medical expert will evaluate the symptoms presented by a patient. This presents a massive bottleneck and problem in the treatment process. Millions of people worldwide are suffering every day, and either do not have access to treatment because they cannot afford it, because of the stigma against getting help, or because the problem is just not taken seriously. The lack of relevant government action in the past is a clear indicator of this with many countries having no national mental health policy (Thornicroft & Maingay, 2002). There are also not enough practitioners to go around; relative to the amount of help that is needed there is just not enough. This bottleneck in evaluation is our subproblem and where I believe we can start to leverage technology to make a significant difference in the world.

The status quo in this situation is to not bring together the available resources (data, algorithms, developers, medical experts) to create a symptom analysis tool.

What is holding the status quo in place?

Several factors are holding the status quo in place, mostly revolving around trust, business incentive, and system complexity.

Trust

The information we’d need to divulge, our “symptoms”, are sensitive data. The vulnerability involved in sharing them implies a significant need for trust. A strong foundation of trust does not exist, given the track record of data use and the relative infancy of this technology. Three reasons for a lack of trust:

·?????Fear of technology – technology is complex and difficult to understand

·?????Data security – past data breaches have changed public sentiment

·?????Data privacy/data use – the privacy, use, and handling of data by businesses is undermined by past events and their expected profit motives

?All of the above has led to data protection legislation (e.g., the EU’s GDPR, South Africa’s POPIA) to protect the consumer, while limiting potential uses of big data – another thing holding the status quo in place. Legislation needs to be applied with more surgical precision.

Business incentive

If any company were to try and implement a solution like what is proposed, there would be a lack of incentive for them to do so. Implementation would require resources and bear a cost. The incentive would be the image and publicity benefits that would have an indirect effect on company revenues. A company could theoretically charge for such a service/product, but no single company has enough user data (except maybe social networking giant Facebook) to make a very useful implementation.

?This leads us to the second issue with an incentive: any proposed solution’s success rate and efficacy are positively correlated with the number of companies involved and the degree to which they cooperate and share data. The relative contribution-to-benefit ratio will differ from company to company, and this imbalance could lead companies with larger data pools to be hesitant with sharing customer data.?

Complexity

The mental healthcare system is an incredibly complex system, which even requires a thorough systems approach analysis to understand well (Furst et al., 2020). Change in this system takes time and has to go make its way through various levels of academia, government, and other regulatory bodies before being ready to enact. This system is of course necessary to safeguard society’s wellbeing but can also hold the status quo in place.

It is important to note that a proposed solution can find ways to overcome this if executed and prioritized correctly. It would need to go through equally strict regulatory assessments as part of its development and testing.

Stakeholder Map

Briefly mentioned above, the stakeholders are largely grouped into businesses, “treatment tech”, and groups representing public interest (experts and government). A well-implemented bridge between businesses (who hold the data) and treatment tech would be the solution to our problem if it were not for the factors mentioned above that are holding the status quo in place. See the Stakeholder Map online here.

Fears

At first thought, two fears jump out around this solution: privacy, and due diligence. With regards to the former: this system would only be implemented with the strictest of data protection standards (the same already used for non-anonymized customer data). The latter concern arises due to the thought of machines making important decisions in a rule-based manner, but this would not be fully implemented until the statistics begin to show promising results. Until then, any final decision will need to be reviewed and approved by trained medical experts.

Understanding the Existing Solution Efforts

When we talk about existing solution efforts, it’s difficult to avoid broadly defining these and listing a few dozen apps assisting with mental health and wellbeing. The truth is that these are all to some degree furthering the cause. We can look at some of the most popular to help us understand the existing solution efforts.

Wellbeing apps

These are the most popular, and the most successful apps today. Wide adoption is relatively safe and easy since the advice being given is not officially medical advice. Even so, when put into practice the advice Headspace provides leads to benefits such as:

·?????Improved focus (Bennike et al., 2017)

·?????Reduced stress (Economides et al., 2018)

·?????Increased compassion (Lim et al., 2015)

·?????Reduced aggression (DeSteno et al., 2017)

·?????Increased well-being (Howells et al., 2016)

Similar apps like Calm, and Insight Timer provide similar services and are growing in popularity. They are however hindered in the depth and complexity of the advice they can give, which focuses on well-being but never touches on mental health.

Mental Health apps

There is a small variety of apps that try to go a step further on the mental health side of things. There is MY3, which helps suicidal individuals stay connected to someone in difficult times, and there are apps like CompanionMx and What’s Up which are the closest thing to what I am proposing. By description, they provide tailored individual or group (for organizations) advice based on big data analysis. The issue is that after approximately four years on the iOS App Store, both have little traction. For comparison, the What’s Up Store page has 9 ratings with an average score of 2.8/5, the Companion page has 4 ratings with an average score of 3.5/5, and the Headspace app has 750,000 ratings with a 4.9/5 average score. An app like the one I’m proposing needs more data, and it needs data from the things we consume in our everyday lives (media, shopping, location habits, etc.). To do this requires consumer consent and Big Tech support.?

Artificial intelligence and machine learning

Use cases do exist where technology is pushing in the proposed direction, albeit mostly taking place in the sphere of academia. A great example of this is the use of Natural Language Processing (NLP) and Machine Learning (ML) technologies to analyze and classify text in an internet support group for those with mental health issues. Such technology is analyzing systems to triage individuals based on the degree of severity. It is a simple application of what I am proposing but with plain text data, such a system was already able to build a “competitive” triage classifier that would assist human moderators (Ferraro et al., 2020).

Further evidence of successful ML applications in the mental healthcare space is in the managing of mental health services and systems. In one instance, causal modeling (more advanced than typical predictive modeling) was used for a better understanding of mental health services and systems behavior (Almeda et al., 2019). A similar application has been performed using simulations and more basic statistical methods, but still for mental health systems management, yielding positive, real-world results (García-Alonso, 2019). While different from individual patient treatment, it shows that academia is getting used to using ML techniques and there is a place for it.?

Data from just last year also suggests that cases can also be handled via teleconsultation services with promising results (López Seguí et al., 2020). This would prove to create a more efficient allocation of care resources that potentially would not be bound by borders.

Finally, a paper reviewing 28 different studies applying AI to mental health problems found that “Collectively, these studies revealed high accuracies and provided excellent examples of AI’s potential in mental healthcare” (Graham et al., 2019). Those studies are still academic proof-of-concept works, but it’s time for this concept to leave academia and get the data it needs.

Identification of Gaps and Levers of Change

Tying everything together, there can be a clear and modular system whereby ML algorithms act as symptom analysis tools, humans review this data and consult patients via teleconsultation and prescribe treatment with the aid of an app or other technological platform – drastically increasing the effectiveness of care experts.

Big Tech collectively has enough of our behavioral data, that pairing it with our mental state, either as a physiologically healthy individual, or not, could lead to profound insights and progress on moving forward human wellbeing. It’s still not being done, and such developments will not take place until certain gaps are closed.

Key Insights and Lessons Learned

Mapping out the challenge landscape is an insightful experience. A mentally healthier world is a more compassionate and caring one. It is one with less hatred, less misogyny, prejudice, and even respect for the environment. The world’s mental health and wellbeing is, therefore, a cause that should be given high importance among society’s other very pressing social and environmental issues. Mapping out the challenge landscape for the first time allows one to envision exactly where technology would fit and be part of the solution.

It is also now clear just how big of a challenge this would be. Without knowing the exact technicalities of it, I know enough to confidently say that this system can certainly be built, and it’s not as difficult a task as one would think at first thought. It’s essentially just applying regression algorithms to vast amounts of consumer usage data. This is not very different from how many psychology studies are performed today – it would just have a lot more data to work with. There are of course ethical concerns that jump out, all of which should not be brushed aside, but rather used as guides for implementation in the right way. There is a fine, but clearly defined line between surveillance and surveying. Staying on the right side, utilizing the most powerful algorithmic tools in human history while practicing consent and transparency could however lead to very worthwhile degrees of human well-being.

Before researching the problem, it was not easy to identify and isolate the gaps keeping the status quo in place. It is now clear that data protection laws, lack of a financial incentive, and fear of losing competitive advantage are the key obstacles to making this a reality.?

In the face of how difficult a task this would be, it would be an immensely worthwhile endeavor for technology companies to engage in and one which government should pave the way for. This analysis has also made execution steps clear: get Big Tech leadership figures around a table and see if this is feasible. Using the gaps and levers of change identified, it would be very easy for a relatively small committee from the most relevant and powerful firms in the technology space to determine feasibility and to create an execution plan. It will be a complicated project to finish, but not difficult to start.

Big Tech is often blamed for causing or exacerbating problems, especially in the social media space. They could let this be an opportunity to do good with the powerful tools they have – and they are encouraged to.

References

Thornicroft, G., & Maingay, S. 2002. “The global response to mental illness.”?BMJ (Clinical research ed.),?325(7365), 608–609. https://doi.org/10.1136/bmj.325.7365.608?

Furst, Mary Anne, Nasser Bagheri, & Luis Salvador-Carulla. 2021. “An Ecosystems Approach to Mental Health Services Research.”?BJPsych International?18 (1). Cambridge University Press: 23–25. https://doi.org/10.1192/bji.2020.24

Bennike, I.H., Wieghorst, A. & Kirk, U. 2017. “Online-based Mindfulness Training Reduces Behavioral Markers of Mind Wandering.”?J Cogn Enhanc?1,?172–181. https://doi.org/10.1007/s41465-017-0020-9

Economides, M., Martman, J., Bell, M.J.?et al.?2018. “Improvements in Stress, Affect, and Irritability Following Brief Use of a Mindfulness-based Smartphone App: A Randomized Controlled Trial.”?Mindfulness?9,?1584–1593. https://doi.org/10.1007/s12671-018-0905-4

Lim, D., Condon, P., & DeSteno D. 2015. “Mindfulness and Compassion: An Examination of Mechanism and Scalability.” PLoS ONE 10(2): e0118221. https://doi.org/10.1371/journal.pone.0118221

DeSteno, D., Lim, D., Duong, F., & Condon, P. 2017. “Meditation Inhibits Aggressive Responses to Provocations” Mindfulness. https://doi.org/10.1007/s12671-017-0847-2

Howells, A., Ivtzan, I. & Eiroa-Orosa, F.J. 2016. “Putting the ‘app’ in Happiness: A Randomised Controlled Trial of a Smartphone-Based Mindfulness Intervention to Enhance Wellbeing.”?J Happiness Stud?17,?163–185. https://doi.org/10.1007/s10902-014-9589-1

Ferraro, G., Loo Gee, B., Ji, S.?et al.?2020. “Lightme: analysing language in internet support groups for mental health.” Health Inf Sci Syst?8,?34. https://doi.org/10.1007/s13755-020-00115-7

López Seguí F,?Walsh S,?Solans O,?Adroher Mas C,?Ferraro G,?García-Altés A,?García Cuyàs F,?Salvador Carulla L,?Sagarra Castro M,?Vidal-Alaball J 2020. “Teleconsultation Between Patients and Health Care Professionals in the Catalan Primary Care Service: Message Annotation Analysis in a Retrospective Cross-Sectional Study” J Med Internet Res ;22(9):e19149. https://doi.org/10.2196/19149

Almeda, Nerea; García-Alonso, Carlos R.; Salinas-Pérez, José A.; Gutiérrez-Colosía, Mencía R.; Salvador-Carulla, Luis. 2019. "Causal Modelling for Supporting Planning and Management of Mental Health Services and Systems: A Systematic Review"?Int. J. Environ. Res. Public Health?16, no. 3: 332. https://doi.org/10.3390/ijerph16030332

García-Alonso CR, Almeda N, Salinas-Pérez JA, Gutiérrez-Colosía MR, Uriarte-Uriarte JJ, Salvador-Carulla L 2019. “A decision support system for assessing management interventions in a mental health ecosystem: The case of Bizkaia (Basque Country, Spain).” PLoS ONE 14(2): e0212179. https://doi.org/10.1371/journal.pone.0212179

Chung, Y., Salvador-Carulla, L., Salinas-Pérez, J.A.?et al.?2018. “Use of the self-organising map network (SOMNet) as a decision support system for regional mental health planning.”?Health Res Policy Sys?16,?35. https://doi.org/10.1186/s12961-018-0308-y

Ritchie, H. & Roser, M. 2018. “Mental Health” Our World in Data. https://ourworldindata.org/mental-health

Poussaint, Alvin F. 2002. “Yes: it can be a delusional symptom of psychotic disorders.”?The Western journal of medicine?vol. 176,1: 4. https://doi:10.1136/ewjm.176.1.4

Graham, S., Depp, C., Lee, E.E.?et al.?2019. ?Artificial Intelligence for Mental Health and Mental Illnesses: an Overview.”?Curr Psychiatry Rep?21,?116. https://doi.org/10.1007/s11920-019-1094-0