MQ Ambassador Dr Esther Beierl is a data scientist, trial statistician, and psychometrician in psychology and mental health research (currently the University of Cambridge, previously the University of Oxford), yoga teacher, and personal trainer. She also has personal lived experience of mental health conditions. For Research Appreciation Day, Esther explains why data science is integral for mental health research.

Individual needs

75% of mental health problems start before adulthood, and 50% of all lifetime mental health problems begin before the age of 14 years (MQ Mental Health Research, 2017). Mine did, too, nearly 30 years ago.

As a child, I had unique needs (and still have), which come with advantages and challenges for myself and others. My nervous system functions differently than most people’s. I am highly intelligent and extremely sensitive, particularly regarding auditory but also olfactory, tactile, and visual stimuli. I am also very susceptible to social dynamics and highly empathetic. I absorb and process much more information and faster than others, which is both a blessing and a curse. I get overstimulated and overwhelmed sooner than others, requiring more downtime in peace and quiet to recharge.

The invalidation and disregard of my specific needs contributed to my mental health struggles in the first place and also made it nearly impossible for me to access appropriate treatments at that time.

My vision is that no child, teenager, or young adult goes through the suffering I experienced.

As a data scientist at the forefront of psychological and mental health research, I know what is possible now and am confident we can improve.

I see great potential in developing data-informed therapies in these technologically innovative times. The prospect of customising therapies to suit individual needs using data science promises to improve treatment outcomes.

What can data science achieve that was not possible before?

Over 90% of the world’s data has been generated in the last two years, and data will continue to grow exponentially.

The amount of data related to human nature and mental illness has increased, and it is now easier to access large datasets. We can use multidimensional approaches to gather a wide range of information (such as working with behavioural data online and offline, self-reports and expert ratings, cognitive and emotional measures, text narratives, physiological and biological data, etc.).

Meanwhile, statistical approaches for working with these large datasets are being developed and improved.

By leveraging large datasets and sophisticated statistical tools, researchers can ask more challenging research questions and more accurately capture the complexity of mental illness in their studies. Compared to traditional approaches and building upon previous theory-driven research and population averages, a research approach based on data science can complement and enhance current clinical practice.

Why is it essential to achieve this? Why does it matter?

I firmly believe that research based on data science can significantly enhance our understanding, prediction, treatment, and prevention of mental illness.

Understanding the interplay of risk factors and predictors with their impact on mental illness in detail can enhance the accuracy of diagnosis. This improved understanding also enables us to make more precise predictions about how these factors influence an individual’s mental health outcome, as well as how they can exacerbate or perpetuate a mental health issue. Additionally, gaining more knowledge about risk factors and predictors offers research opportunities aimed at preventing mental illness. Further research on specific patterns in individuals who have not responded to traditional treatment approaches (such as I) or in subgroups (e.g., people from minority groups) presents opportunities to adapt compounds of existing treatments or develop entirely new therapies that could benefit individuals like us.

I would even argue that data science for mental health research can ultimately contribute to a more just healthcare system and society.

How is this achieved?

Research examples to illustrate how mental health research based on complex computational approaches works and how the healthcare system can practically use its results include my own research project or the MQ-funded research by Zac Cohen and Rob DeRubeis.

During my postdoctoral research at the University of Oxford, I developed a data science algorithm to predict the onset of post-traumatic stress disorder (PTSD) 1 month after a traumatic event by predictors and risk factors assessed shortly after the trauma in a prospective study of survivors of assault or road traffic accidents (Beierl et al. 2024). Our research can be used to identify individuals at risk for PTSD.

The Stratified Medicine Approaches for Treatment Selection (SMART) Mental Health Prediction Tournament by Zac Cohen and Rob DeRubeis used data from the NHS Talking Therapies program (Clark, 2018) to predict whether individuals would benefit most from a low- or high-intensity treatment and how effective the algorithms would be at identifying the suitable therapies.

MQ’s Mental Health Data Science Group continues to advance research and policy (McIntosh et al., 2016; Russ et al., 2019).

What are the challenges?

Despite the potential of data science for precision medicine, contextual, statistical, technical, and clinical/ practical challenges must be addressed. Contextual challenges include societal, financial, and ethical considerations. Statistical challenges refer to model evaluation (e.g., which statistical model best represents the data?), interpretation and drawing sensible conclusions from these complex algorithms, generalisability to real-world settings, and clinical utility. Clinical and practical challenges apply to clinicians’ needs and concerns (e.g., how these algorithms can best aid their clinical practice).

If you’d like to delve deeper into the topic, I recommend reading our paper (Deisenhofer, Barkham, Beierl, et al., 2023). We have thoroughly discussed and addressed those challenges in our ‘Implementing Precision Medicine’ framework.

My vision can become a reality, but we have yet to achieve it.

Eventually, with the invaluable support of exceptionally open-minded and competent healthcare professionals, I was able to formulate my own personalised therapy plan tailored to my needs. I am grateful for where I am, but I continue to live with several chronic diagnoses, which lead to significant limitations in my daily life.

In this era of data science, the focus should be on funding cutting-edge research based on quantitative data and including lived experience in qualitative data to prevent and specifically target individual trajectories of mental illness like mine. Otherwise, selecting an appropriate therapy becomes a cumbersome process of trial and error.

I aim to contribute to my vision through advocacy for mental health, reshaping personal experiences and conducting further research

You can find Esther on social media in the following handles: X: @EBeierl, Instagram: @estherbeierl, Substack: @estherbeierl

References

Beierl, E. T., Böllinghaus, I., Clark, D. M., Glucksman, E., & Ehlers, A. (2024). Data science for mental health: Development of a predictive algorithm to identify individuals at risk for PTSD 1 month after trauma within hours to days after trauma [Manuscript in preparation]. Department of Experimental Psychology, University of Oxford, UK.

Clark, D. M. (2018). Realizing the mass public benefit of evidence-based psychological therapies: The IAPT Program. Annual Review of Clinical Psychology, 14, 159-183, https://doi.org/10.1146/annurev-clinpsy-050817-084833

Deisenhofer, A. K., Barkham, M., Beierl, E. T., Schwartz, B., Aafjes-van Doorn, K., Beevers, C. G., Berwian, I. M., Blackwell, S. E., Bockting, C. L., Brakemeier, E. L., Brown, G., Buckman, J. E. J., Castonguay, L. G., Cusack, C. E., Dalgleish, T., de Jong, K., Delgadillo, J., DeRubeis, R. J., Driessen, E., Ehrenreich-May, J., Fisher, A. J., Fried, E. I., Fritz, J., Furukawa, T. A., Gillan, C. M., Gómez Penedo, J. M., Hitchcock, P. F., Hofmann, S. G., Hollon, S. D., Jacobson, N. C., Karlin, D. R., Lee, C. T., Levinson, C. A., Lorenzo-Luaces, L., McDanal, R., Moggia, D., Ng, M. Y., Norris, L. A., Patel, V., Piccirillo, M. L., Pilling, S., Rubel, J. A., Salazar-de-Pablo, G., Schleider, J. L., Schnurr, P. P., Schueller, S. M., Siegle, G. J., Saunders, R., Uher, R., Watkins, E., Webb, C. A., Wiltsey Stirman, S., Wynants, L., Youn, S. J., Zilcha-Mano, S., Lutz, W., and Cohen, Z. D. (2024). Implementing precision methods in personalizing psychological therapies: Barriers and possible ways forward. Behaviour Research and Therapy, 172(9), 104443. doi: https://doi.org/10.1016/j.brat.2023.104443

McIntosh, A. M, Stewart, R., John, A., Smith, D. J., Davis, K., Sudlow, C., Corvin, A., Nicodemus, K., Kingdon, D., Hassan, L., Hotopf, M., Lawrie, S. M., Russ, T., C., Geddes, J. R., Wolpert, M., Wölbert, E., Porteous, D. J., & the MQ Data Science Group (2016). Data science for mental health: a UK perspective on a global challenge. The Lancet Psychiatry, 3(10), 993-998. https://doi.org/10.1016/S2215-0366(16)30089-X

MQ Mental Health Research (2017). MQ’s manifesto for young people’s mental health. https://www.mqmentalhealth.org/wp-content/uploads/MQManifestoforyoungpeoplesmentalhealth2017.pdf

MQ Mental Health Research (n.d.). The Stratified Medicine Approaches for Treatment Selection (SMART) Mental Health Prediction Tournament. https://www.mqmentalhealth.org/research/the-stratified-medicine-approaches-for-treatment-selection-smart-mental-health-prediction-tournament/

Russ, T. C., Wölbert, E., Davis, K. A. S., Hafferty, J. D., Ibrahim, Z., Inkster, B., John, A., Lee, W., Maxwell, M., McIntosh, A., Stewart, R., & the MQ Data science group (2019). How data science can advance mental health research. Nature Human Behaviour, 3, 24-32.  https://doi.org/10.1038/s41562-018-0470-9