Improving Clinical Risk Stratification for Patients with Suicidal or Self Harm Behaviours

Research Report Précis

Improving Clinical Risk Stratification for Patients with Suicidal or Self Harm Behaviours by Application of Machine Learning and Nonlinear Methods to Analyse Dynamic Risk Factors

Dr Michael McCullough1, Professor Andrew Page2, Professor Michael Small1, and Associate Professor Geoff Hooke3.

1School of Physics, Mathematics and Computing, UWA

2School of Psychological Science, UWA

3Perth Clinic

October 26, 2018

Young Lives Matter – UWA Foundation

The Young Lives Matter Foundation is a large-scale, cross-disciplinary research initiative established by The University of Western Australia focussed on combating issues of suicide that continue to challenge 15-24 year olds.

UWA has assembled a team of world leading experts in mental health and allied fields that has the capacity to leverage a diversity of expertise across relevant disciplines including clinical, medical, psychology, social science, mathematics and research institutes such as the Telethon Kids Institute.

Research Report Précis

Existing clinical tools and statistical models for suicide and self-harm risk stratification are generally considered to be inaccurate and ineffective. Research in this field is now investigating machine learning and artificial intelligence as a means to model the vast and complex collection of suspected risk factors. It is also now clear that many of the typical psychological risk factors which are often treated as static variables will in fact vary signicantly over as little as several hours and should be investigated as dynamic interdependent variables.

Perth Clinic is a private psychiatric clinic located in West Perth with an outstanding database of high quality psychological measures. In addition, they have established sophisticated monitoring and feedback systems. These features set Perth Clinic apart and presents a unique opportunity to investigate psychological risk factors as dynamic interdependent variables.

Researchers from The University of Western Australia (UWA), in collaboration with Perth Clinic, have already produced several pioneering investigations into the dynamic trajectories and interdependent relationships between variables in this data using latent growth class analysis and cross-lagged panel analysis.

This current study will comprehensively reanalyse existing data using high dimensional machine learning methods while adopting a networked dynamical systems approach. Given the magnitude and quality of the data, a comprehensive re-analysis is proposed using multivariate methods from the field of machine learning to investigate the possibility of improved risk stratification.

Perth Clinic collaboration

As part of Perth Clinic’s commitment to Mental Health Recovery, it values a long-standing collaboration with The University of Western Australia.

One of many shared interests between Perth Clinic and UWA is understanding and preventing adverse outcomes in mental health treatment, such as self-harm and suicide.

This research with Psychology has been challenging and rewarding. The benefits have translated into innovative care for patients, with improved outcomes, that is central to a long-held philosophy at Perth Clinic.