Virtual Summit Shorts: Using artificial intelligence to improve mental health

To commence the discussion, session Chair Lucy Hartley (Claims Service Manager, Swiss Re) introduced Matt, Brendan and Joseph, explaining that their mental health research project investigates the intersect of mental health and insurance using artificial intelligence (AI) methods and data.

“So far, we have learnt that mental health is such a huge and complex subject,” Matt explained.

Yet through a scoping literature review which included, “an application of machine learning techniques to distil the very large number of research papers down to more relevant ones,” Matt explained that the preliminary learnings extracted so far, have provided three real-life examples of the application of AI and machine learning. These findings support the notion that AI can improve mental health detection and services.

With statistics sourced from Beyond Blue and The Black Dog Institute, Matt noted that, “While there have been positive outcomes in reducing the stigma associated with mental health illness, there’s still a real lack of understanding in this space – particularly, the key drivers that are influencing the increases in incidence and prevalence that we’ve seen.”

This observation led to Matt asking two questions – a) is it possible to identify a deterioration in people’s mental health early enough to make a difference; and b) what makes treatments work for some people and not others?

“In terms of doctors in Australia we know that the number of mental health plans are increasing yet we hear that doctors often don’t have enough time to deal with mental health cases, which may be more complex than other health conditions.”

With this knowledge, Matt reflected on the capability of the insurance sector in providing sustainable insurance cover for people who are suffering with their mental health.

Brendan reflected on the scoping review portion of the research project and notes that, “Scoping literature reviews are essential attempts to assess the breadth of the scientific body of the research.”

Brendan categorised the literature of the scoping review into four focus areas:

• Developing AI predictive models using mental health and insurance data;

• Observing factors in mental health and insurance data using AI techniques;

• Industry and commercial research of AI using mental health and insurance data; and

• Methodological and ethical consideration of AI using mental health and insurance data.

In doing so, Brendan noted that “Our goal wasn’t to develop a generalised model but to screen thousands and thousands of papers to narrow down to a group of papers that we can analyse,” to answer the two questions posed above.

Brendan intertwined the case study, “Deep Neural Networks Detect Suicide Risk from Textual Facebook Posts”, to highlight the key benefit of using Facebook or social media data as a source of AI.

Brendan introduced another case study, “Medical Comorbid Diagnoses Among Adult Psychiatric Inpatients”, that used AI-based methods and techniques to examine medical care mobility available to psychiatric patients. Linking claims data to electronic health records, this case study describes the prevalence of medical conditions among individuals diagnosed with psychiatric disorders.

The third case study that Brendan introduced, “The Importance of Unresolved Fatigue in Depression: Costs and Comorbidities”, aimed to observe the costs associated with patients experiencing significant fatigue and depression, to gather feedback from patients’ and customers’ surveys and insurance.

To understand the preliminary findings from these data sets, Brendan explained that the research project undertook an environmental scan of the public mental health datasets to “understand those different indicators and measures of mental wellbeing across different research datasets.” They found common demographic factors included age, gender, education, ethnicity, relationship/families, employment, and socioeconomic status. These common indicators outline the importance of adjusting for multiple risk factors.

To ensure that the research project was aligned with the needs of the organisations across the different sectors, Joseph reflected on the stakeholder engagement sessions that MNC held with mental health organisations, superannuation funds and insurance companies.

“When speaking to the mental health organisations, the main issues and needs were around directing government funding to the most needed areas, accessibility of mental health services and measuring the effectiveness and impact of these services,” Joseph recalled.

In contrast to mental health organisations, Joseph noted that for superfunds, “the focus is the members’ best interests. How do we help members achieve the best retirement outcome?”

Joseph noted that, “Even though the sectors seem to be facing different issues and needs, there are several common themes that emerge from our discussions which can be summarised into five main areas:

• Health System: Lack of coordination and poor integration of care for people with mental illness;

• Risk Factors and Models: A need for to understand why some people claim, but not others with the same mental health condition;

• Diagnosis: Diagnosis and criteria of mental health condition is subjective;

• Treatment and Outcome: Mental health is unavoidable. Should we focus on recovery and resilience instead? What does the community need?; and

• Longitudinal Studies: How do people develop from mild mental health illness symptoms to more severe conditions?

Concluding the discussion Matt reflected, “that there is so much we don’t know about mental health and it’s essential that we drill down into the areas in which we can have an impact.

”To conclude the session, Lucy explained that the MNC team are re-engaging with interested stakeholders and formally requesting data assets that can assist in their research. MNC are also reaching out to other organisations that may be interested.

Whilst the road ahead to solving the impacts of mental health problems in society is a complex one, working together with the insurance and other sectors using different skills and resources is essential to developing valuable data insights. Only by doing this can we ensure that all insurance products and services are effective, sustainable, and equable for people with lived experiences of mental health problems.

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.