About Nesta

Nesta is an innovation foundation. For us, innovation means turning bold ideas into reality and changing lives for the better. We use our expertise, skills and funding in areas where there are big challenges facing society.

Precision, precision: how will predictive analytics change mental health care?

Precision medicine has been championed for its potential to transform the way we diagnose and treat people. The aim is to move towards a highly individualised approach using new forms of information, from genetics to data sourced from digital health trackers, to more accurately diagnose and select treatment options. But while precision medicine is attracting the bulk of attention and investment, the revolutionary possibilities presented by ‘precision psychiatry’ remain comparatively under-explored.

The challenges of diagnosis and treatment in a complex field

Compared to other areas of health, it is far more difficult to base decision on the underlying causes or markers (measurable indicators) of mental health conditions. Only a third of people suffering from depression benefit from the first antidepressant they try and it generally takes several months to determine whether an antidepressant is proving effective. If it proves ineffective, clinicians are back at square one: is it worth upping the dosage, or prescribing another antidepressant? And medication is only one treatment option (talking therapy or cognitive behavioural therapy could prove more effective). At present we lack reliable approaches for predicting how an individual might respond to any combination of treatment options.

When treating and diagnosing physical health conditions, doctors have access to a range of tests for important biomarkers of disease (such as glucose levels from blood tests), whereas there are few equivalents in psychiatry that are able to supply the same quantifiable information as a basis for diagnosis. The challenge is further complicated by our limited understanding of the brain and the fact that psychiatric conditions are often the result of a combination of interdependent social, economic, neurological and genetic factors - the interplay between which is yet to be fully understood.

At present we lack reliable approaches for predicting how an individual might respond to any combination of treatment options.

These issues are only compounded by the reliance on a contentious system for diagnosis- the Diagnostic and Statistical Manual of Mental Disorders (DSM). Since 1980 it has defined the way the Western world talks about, researches, and treats mental illness. As Thomas Insel, former director of the US National Institute for Mental Health, observed, the DSM’s “...weakness is its lack of validity. Unlike our definitions of ischemic heart disease, lymphoma, or AIDS, the DSM diagnoses are based on a consensus about clusters of clinical symptoms, not any objective laboratory measure." The British Psychological Society has criticised the system because proposed diagnoses were “clearly based largely on social norms, with 'symptoms' that all rely on subjective judgements.”

The promise of precision psychiatry

An ambitious goal for precision psychiatry is to identify biomarkers which could, in the future, be used to tailor treatment options to a specific individual. The most obvious place to look for these biomarkers is in the brain.

One emerging research approach is to place participants in brain scanners before they are given a course of treatment. Next, machine learning algorithms (an automated way of analysing large data sets to find correlations in the data) are trained to look for differences in the brain scans of participants who responded successfully to a treatment compared to those who didn’t. These algorithms have then been tested using brain scans from a separate group of patients to see if they could predict whether they would respond successfully to treatment or not. Early results suggest this approach may prove successful in predicting the response of people to treatment options for a range of different conditions including lithium for bipolar disorder, CBT for social anxiety, and antidepressants for depression.

The next step for the field is to test these neural biomarkers prospectively: is it possible to use brain scans to predict the response of an individual to a range of treatment options, and then use this information to guide therapeutic decisions? This possibility may not be that remote. One study has already used brain scans to place people with depression into one of four categories based on their neural activity, finding one group in particular to respond well to Transcranial Magnetic Stimulation (TMS), while two other groups saw little effect. In principle, this approach could then be used to test if someone considering TMS fell into the responsive category before they even started the therapy.

But while these neural markers could have considerable potential, expensive brain scans are unlikely to be a panacea for the challenges associated with diagnosing and treating mental health problems. For a start, the cost of brain scans could easily prove prohibitive when we are already seeing an acute squeeze on budgets for mental health services. In addition, such highly technical experimental approaches are time consuming and only constitute a single observation, therefore are less reliable than multiple measurements. So if the potential of precision psychiatry is to be realised, it would be necessary to find biomarkers capable of guiding diagnosis and treatment, but which are easier and cheaper to collect than those obtained by brain scans.

From diagnose-and-treat to predict-and-prevent?

Could the answer be found in the data captured by devices we use every day, such as mobile phones or health trackers? Mindstrong, a new Silicon Valley startup, is taking exactly this approach.

Mindstrong (which boasts the aforementioned Thomas Insel on its team) has created an app which continuously and passively monitors patterns of behaviour on smartphones such as the timing of swipes, key presses, and spacebar taps. Using large patient datasets and machine learning, it has been able to replicate results from a range of neurobiological test scores, including scales used to measure depression and anxiety, simply by drawing on information about how someone uses their phone. Mindstrong believes that the app could be used to provide an objective measure of someone’s mental health, one which could monitor how wellbeing fluctuates over time. Data collection from the app works on an ‘opt-in’ basis, with Mindstrong maintaining that strict informed consent needs to underpin an individual’s interaction with the app and any subsequent sharing of data with a doctor or caregiver.

It is already working in partnership with BlackThorn therapeutics, a biopharmaceutical company, to use these digital biomarkers to identify which patients out of a large cohort will be most responsive to its treatments. It is also now being used as part of the ambitious Aurora study, which is looking for early signals of PTSD among civilian survivors of trauma.

Far more still needs to be understood about the brain and the physiological basis of mental health before precision psychiatry can start to deliver on the promise.

Mindstrong believes that a cheap, widely accessible way of measuring markers of mental health will enable clinicians to move from a model of diagnose-and-treat, to one of predict-and-prevent. The next step will be to determine how applicable this approach could be for a wider range of conditions, and how well it could act as a proxy for the neural markers already identified from brain scans.

Where next?

The combination of brain imaging, big data, machine learning capabilities and digital health trackers have made it possible to test the theory of precision psychiatry. However, it important to note that these developments are not set to overturn existing approaches to diagnosis and treatment in the foreseeable future. Far more still needs to be understood about the brain and the physiological basis of mental health before precision psychiatry can start to deliver on the promise. There are also a set of more ‘technical’ challenges to be overcome:

  1. Ensuring the underlying data is representative

Training datasets for predictive models in precision psychiatry must be built from diverse population samples. Algorithms built using narrow data sets in terms of age or ethnicity will have little generalisable predictive power for the wider population. As the field of precision psychiatry evolves, it will need to learn from mistakes made in other fields such as genomics and psychology, and ensure the evidence base is built upon diverse foundations.

  1. Designing-in data privacy and informed consent

Precision psychiatry is founded on deriving insights based on data collection from individuals - the management and sharing of this data raises complex issues around privacy and security. These problems are far from unique to this field, they are being sharply felt across the health system by both private and public organisations. However, within the field of mental health, the ethical issues are arguably even more sensitive, given that some of the most vulnerable groups in society are affected. There are ways to mitigate these problems. Carefully designed consent agreements which clearly outline how the data will be used and shared as well as greater transparency around how the data is stored, used and analysed will help people make more informed decisions about services. And, going further, the success of the network PatientsLikeMe shows the benefit of people being able to collect and manage their own health data.

  1. Accounting for a broader set of factors

In order to improve predictive treatment models that use brain imaging, it is likely that researchers will need to build in other kinds data about an individual’s life experiences. Research by Leanne Williams’ lab at Stanford found that including questionnaire data from participants assessing their early life stressors improved the accuracy of their models (which were previously just based on neural biomarkers) when it came to anticipating how someone with depression would respond to antidepressants. This raises the possibility that the inclusion of other psychological, genetic, and biological factors may further improve the predictive strength of the computational models.

Moves to innovate and develop new approaches in psychiatry will always be contentious, and precision psychiatry will no doubt polarise opinion - but the next chapter in its evolution will be well worth watching closely.

Author

Jack Pilkington

Jack Pilkington

Jack Pilkington

Explorations Intern

Jack worked in our Explorations team.

View profile