Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Jun;58(6):893-905.
doi: 10.1007/s00127-022-02415-7. Epub 2023 Feb 28.

Predictive modelling of deliberate self-harm and suicide attempts in young people accessing primary care: a machine learning analysis of a longitudinal study

Catherine M McHugh  1   2 , Nicholas Ho  3 , Frank Iorfino  3 , Jacob J Crouse  3 , Alissa Nichles  3 , Natalia Zmicerevska  3 , Elizabeth Scott  3   4   5 , Nick Glozier  3   6 , Ian B Hickie  3
Affiliations

Predictive modelling of deliberate self-harm and suicide attempts in young people accessing primary care: a machine learning analysis of a longitudinal study

Catherine M McHugh et al. Soc Psychiatry Psychiatr Epidemiol. 2023 Jun.

Abstract

Purpose: Machine learning (ML) has shown promise in modelling future self-harm but is yet to be applied to key questions facing clinical services. In a cohort of young people accessing primary mental health care, this study aimed to establish (1) the performance of models predicting deliberate self-harm (DSH) compared to suicide attempt (SA), (2) the performance of models predicting new-onset or repeat behaviour, and (3) the relative importance of factors predicting these outcomes.

Methods: 802 young people aged 12-25 years attending primary mental health services had detailed social and clinical assessments at baseline and 509 completed 12-month follow-up. Four ML algorithms, as well as logistic regression, were applied to build four distinct models.

Results: The mean performance of models predicting SA (AUC: 0.82) performed better than the models predicting DSH (AUC: 0.72), with mean positive predictive values (PPV) approximately twice that of the prevalence (SA prevalence 14%, PPV: 0.32, DSH prevalence 22%, PPV: 0.40). All ML models outperformed standard logistic regression. The most frequently selected variable in both models was a history of DSH via cutting.

Conclusion: History of DSH and clinical symptoms of common mental disorders, rather than social and demographic factors, were the most important variables in modelling future behaviour. The performance of models predicting outcomes in key sub-cohorts, those with new-onset or repetition of DSH or SA during follow-up, was poor. These findings may indicate that the performance of models of future DSH or SA may depend on knowledge of the individual's recent history of either behaviour.

Keywords: Adolescence; Deliberate self-harm; Machine learning; Suicidal behaviour; Suicide attempt; Youth.

PubMed Disclaimer

Conflict of interest statement

Catherine M. McHugh, Frank Iorfino, Jacob J Crouse, Alissa Nichles, Natalia Zmicerevska, Nicholas Ho, and Nick Glozier report no conflicts of interest. Elizabeth Scott is the clinical director of the St Vincent’s Youth Mental health programme. She has received honoraria for educational seminars related to the clinical management of depressive disorders supported by Servier and Eli-Lilly pharmaceuticals. She has participated in a national advisory board for the antidepressant compound Pristiq, manufactured by Pfizer. She was the National Coordinator of an antidepressant trial sponsored by Servier. Ian Hickie was an inaugural Commissioner on Australia’s National Mental Health Commission (2012–18). He is the Co-Director, Health and Policy at the Brain and Mind Centre (BMC) University of Sydney. The BMC operates an early-intervention youth services at Camperdown under contract to headspace. Professor Hickie has previously led community-based and pharmaceutical industry-supported (Wyeth, Eli Lily, Servier, Pfizer, AstraZeneca) projects focused on the identification and better management of anxiety and depression. He was a member of the Medical Advisory Panel for Medibank Private until October 2017, a Board Member of Psychosis Australia Trust and a member of Veterans Mental Health Clinical Reference group. He is the Chief Scientific Advisor to, and an equity shareholder in, Innowell. Innowell has been formed by the University of Sydney and PwC to deliver the $30 m Australian Government-funded ‘Project Synergy’. Project Synergy is a three-year programme for the transformation of mental health services through the use of innovative technologies.

Figures

Fig. 1
Fig. 1
Machine learning model building. Model 1 predicts all participants who report DSH at follow-up (n = 100, yellow group) relative to those with no self-harm at follow-up (n = 352, blue group). Model 2 predicts participants who report SA at follow-up (n = 57, red group) relative to those with no self-harm at follow-up (n = 352) (blue group). Model 3 samples only those without self-harm at baseline (n = 292). Model 4 samples only those with either DSH or SA at baseline (n = 217)
Fig. 2
Fig. 2
Method of behaviour at baseline and likelihood of further DSH or SA. Samples only those young people who reported DSH or SA at baseline (n = 342) and includes those lost to follow-up. A presents the proportion of young people in each outcome group by method of DSH or SA at baseline, and B presents the odds ratio of either outcome, DSH or SA, according to baseline behaviour and method. Reference group for odds ratio calculations was those who completed follow-up and reported no further self-harm
Fig. 3
Fig. 3
Heatmap displaying frequency of variable selection by each algorithm in ML Model 1 and Model 2. Frequency of variable selection represents the sum of 100 trials of each algorithm (AUCRF, Boruta, LASSO, Elastic-net and logistic regression), there for 500 trials per Model in total. Variables ranked 1–10 by frequency of selection across the 500 trials are labelled. A presents variable selection in Model 1 (deliberate self-harm) and B presents variable selection in Model 2 (suicide attempts)
Fig. 4
Fig. 4
Variables important to the prediction of DSH and/or SA. Important variables were those that were selected in at least 70% of algorithm trials
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Moran P, Coffey C, Romaniuk H, Olsson C, Borschmann R, Carlin JB, et al. The natural history of self-harm from adolescence to young adulthood: a population-based cohort study. Lancet. 2012;379(9812):236–243. doi: 10.1016/S0140-6736(11)61141-0. - DOI - PubMed
    1. Iorfino F, Hermens DF, Cross SP, Zmicerevska N, Nichles A, Groot J, et al. Prior suicide attempts predict worse clinical and functional outcomes in young people attending a mental health service. J Affect Disord. 2018;238:563–569. doi: 10.1016/j.jad.2018.06.032. - DOI - PubMed
    1. Chung DT, Ryan CJ, Hadzi-Pavlovic D, Singh SP, Stanton C, Large MM. Suicide rates after discharge from psychiatric facilities: a systematic review and meta-analysis. JAMA Psychiat. 2017;74(7):694–702. doi: 10.1001/jamapsychiatry.2017.1044. - DOI - PMC - PubMed
    1. Kapur N, Cooper J, O’Connor RC, Hawton K. Non-suicidal self-injury v. attempted suicide: new diagnosis or false dichotomy? Br J Psychiatry. 2013;202(5):326–328. doi: 10.1192/bjp.bp.112.116111. - DOI - PubMed
    1. Franklin JC, Nock MK (2016) Nonsuicidal self-injury and its relation to suicidal behavior. In: The Oxford handbook of behavioral emergencies and crises, pp 401–16