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

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.

Female cancer survivors are more likely to be at high risk of malnutrition and meet the threshold for clinical importance for a number of quality of life subscales.

BACKGROUND: The present study aimed to explore malnutrition risk, handgrip strength and quality of life (QOL) in cancer survivors. METHODS: In total, 232 individuals completed a demographic questionnaire, Patient-Generated Subjective Global Assessment Short Form and the European Organization for Research and Treatment of Cancer QOL Questionnaire (EORTC QLQ-C30). Handgrip strength was determined using a spring-loaded handgrip dynamometer and anthropometric measurements were taken by an oncology nurse. Frequencies and distribution data, analysis of variance and chi-squared tests were then conducted. RESULTS: The majority of the cohort were female (n = 141; 60.8%) had breast cancer (n = 62; 26.7%) and the mean ± SD body mass index (BMI) was 26.6 ± 6.2 kg m-2 . Less than a one-third reported seeing a dietitian (n = 68; 29.3%). Over one-third reported recent weight loss (n = 88; 37.3%). Some 40.9% (n = 95) were at moderate to high risk of malnutrition, with women more likely than men to be classified as high risk (p < 0.05). Mean ± SD handgrip strength was 25 ± 15 kg and this differed significantly by gender (p = 0.00), cancer type (p = 0.01) and BMI classification (p = 0.01). One-fifth of individuals were classified as having dynapenia (n = 48; 21.1%). Median (interquartile range) QOL score was 66.7 (33.3). The proportion of individuals meeting the threshold for clinical importance for QOL subscales ranged from 12.5% (constipation) to 42.7% (physical functioning). Females were more likely than males to meet the threshold for physical functioning (p = 0.00), fatigue (p = 0.02) and pain (p = 0.01). CONCLUSIONS: Females are more likely than males to be at high risk of malnutrition and meet the threshold for clinical significance for several QOL subscales.

Premature mortality in young people accessing early intervention youth mental healthcare: data-linkage cohort study.

BACKGROUND: Understanding premature mortality risk from suicide and other causes in youth mental health cohorts is essential for delivering effective clinical interventions and secondary prevention strategies. AIMS: To establish premature mortality risk in young people accessing early intervention mental health services and identify predictors of mortality. METHOD: State-wide data registers of emergency departments, hospital admissions and mortality were linked to the Brain and Mind Research Register, a longitudinal cohort of 7081 young people accessing early intervention care, between 2008 and 2020. Outcomes were mortality rates and age-standardised mortality ratios (SMR). Cox regression was used to identify predictors of all-cause mortality and deaths due to suicide or accident. RESULTS: There were 60 deaths (male 63.3%) during the study period, 25 (42%) due to suicide, 19 (32%) from accident or injury and eight (13.3%) where cause was under investigation. All-cause SMR was 2.0 (95% CI 1.6-2.6) but higher for males (5.3, 95% CI 3.8-7.0). The mortality rate from suicide and accidental deaths was 101.56 per 100 000 person-years. Poisoning, whether intentional or accidental, was the single greatest primary cause of death (26.7%). Prior emergency department presentation for poisoning (hazard ratio (HR) 4.40, 95% CI 2.13-9.09) and psychiatric admission (HR 4.01, 95% CI 1.81-8.88) were the strongest predictors of mortality. CONCLUSION: Premature mortality in young people accessing early intervention mental health services is greatly increased relative to population. Prior health service use and method of self-harm are useful predictors of future mortality. Enhanced care pathways following emergency department presentations should not be limited to those reporting suicidal ideation or intent.

Neurocognitive functioning predicts suicidal behaviour in young people with affective disorders.

BACKGROUND: Neurocognitive impairment is recognised as a risk factor for suicidal behaviour in adults. The current study aims to determine whether neurocognitive deficits also predict ongoing or emergent suicidal behaviour in young people with affective disorders. METHODS: Participants were aged 12-30 years and presented to early intervention youth mental health clinics between 2008 and 2018. In addition to clinical assessment a standardised neurocognitive assessment was conducted at baseline. Clinical data was extracted from subsequent visits using a standardised proforma. RESULTS: Of the 635 participants who met inclusion criteria (mean age 19.6 years, 59% female, average follow up 476 days) 104 (16%) reported suicidal behaviour during care. In 5 of the 10 neurocognitive domains tested (cognitive flexibility, processing speed, working memory, verbal memory and visuospatial memory) those with suicidal behaviour during care were superior to clinical controls. Better general neurocognitive function remained a significant predictor (OR=1.94, 95% CI 1.29- 2.94) of suicidal behaviour in care after controlling for other risk factors. LIMITATIONS: The neurocognitive battery used was designed for use with affective and psychotic disorders and may not have detected some deficits more specific to suicidal behaviour. CONCLUSION: Contrary to expectations, better neurocognitive functioning predicts suicidal behaviour during care in young people with affective disorders. While other populations with suicidal behaviour, such as adults with affective disorders or young people with psychotic disorders, tend to experience neurocognitive deficits which may limit their capacity to engage in some interventions, this does not appear to be the case for young people with affective disorders.

Can machine-learning methods really help predict suicide?

PURPOSE OF REVIEW: In recent years there has been interest in the use of machine learning in suicide research in reaction to the failure of traditional statistical methods to produce clinically useful models of future suicide. The current review summarizes recent prediction studies in the suicide literature including those using machine learning approaches to understand what value these novel approaches add. RECENT FINDINGS: Studies using machine learning to predict suicide deaths report area under the curve that are only modestly greater than, and sensitivities that are equal to, those reported in studies using more conventional predictive methods. Positive predictive value remains around 1% among the cohort studies with a base rate that was not inflated by case-control methodology. SUMMARY: Machine learning or artificial intelligence may afford opportunities in mental health research and in the clinical care of suicidal patients. However, application of such techniques should be carefully considered to avoid repeating the mistakes of existing methodologies. Prediction studies using machine-learning methods have yet to make a major contribution to our understanding of the field and are unproven as clinically useful tools.

Association between suicidal ideation and suicide: meta-analyses of odds ratios, sensitivity, specificity and positive predictive value.

BACKGROUND: The expression of suicidal ideation is considered to be an important warning sign for suicide. However, the predictive properties of suicidal ideation as a test of later suicide are unclear.AimsTo assess the strength of the association between suicidal ideation and later suicide measured by odds ratio (OR), sensitivity, specificity and positive predictive value (PPV). METHOD: We located English-language studies indexed in PubMed that reported the expression or non-expression of suicidal ideation among people who later died by suicide or did not. A random effects meta-analysis was used to assess the pooled OR, sensitivity, specificity and PPV of suicidal ideation for later suicide among groups of people from psychiatric and non-psychiatric settings. RESULTS: There was a moderately strong but highly heterogeneous association between suicidal ideation and later suicide (n = 71, OR = 3.41, 95% CI 2.59-4.49, 95% prediction interval 0.42-28.1, I2 = 89.4, Q-value = 661, d.f.(Q) = 70, P ≤0.001). Studies conducted in primary care and other non-psychiatric settings had similar pooled odds to studies of current and former psychiatric patients (OR = 3.86 v. OR = 3.23, P = 0.7). The pooled sensitivity of suicidal ideation for later suicide was 41% (95% CI 35-48) and the pooled specificity was 86% (95% CI 76-92), with high between-study heterogeneity. Studies of suicidal ideation expressed by current and former psychiatric patients had a significantly higher pooled sensitivity (46% v. 22%) and lower pooled specificity (81% v. 96%) than studies conducted in non-psychiatric settings. The PPV among non-psychiatric cohorts (0.3%, 95% CI 0.1%-0.5%) was significantly lower (Q-value = 35.6, P < 0.001) than among psychiatric samples (3.9%, 95% CI 2.2-6.6). CONCLUSIONS: Estimates of the extent of the association between suicidal ideation and later suicide are limited by unexplained between-study heterogeneity. The utility of suicidal ideation as a test for later suicide is limited by a modest sensitivity and low PPV.Declaration interestM.M.L. and C.J.R. have provided expert evidence in civil, criminal and coronial matters. I.B.H. has been a Commissioner in Australia's National Mental Health Commission since 2012. 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. I.B.H. 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 is a Board Member of Psychosis Australia Trust and a member of Veterans Mental Health Clinical Reference group. He was a member of the Medical Advisory Panel for Medibank Private until October 2017. He is the Chief Scientific Advisor to, and an equity shareholder in, InnoWell. InnoWell has been formed by the University of Sydney and PricewaterhouseCoopers to administer the $30 M Australian Government Funded Project Synergy. Project Synergy is a 3-year programme for the transformation of mental health services through the use of innovative technologies.

Impulsivity in the self-harm and suicidal behavior of young people: A systematic review and meta-analysis.

BACKGROUND: Impulsivity is considered a possible phenotype underlying the expression of self-harm and suicidal behaviors. Yet impulsivity is a not a unitary construct and there is evidence that different facets of impulsivity follow different neurodevelopmental trajectories and that some facets may be more strongly associated with such behaviors than others. Moreover, it is unclear whether impulsivity is a useful predictor of self-harm or suicidal behavior in young people, a population already considered to display heightened impulsive behavior. METHODS: A systematic review and meta-analysis of studies published in Medline, PubMed, PsychInfo or Embase between 1970 and 2017 that used a neurocognitive measure to assess the independent variable of impulsivity and the dependent variable of self-harm and/or suicidal behavior among young people (mean age < 30 years old). RESULTS: 6183 titles were identified, 141 full texts were reviewed, and 18 studies were included, with 902 young people with a self-harm or suicidal behavior and 1591 controls without a history of these behaviors. Deficits in inhibitory control (13 studies, SMD 0.21, p-value = 0.002, 95% confidence interval (CI) (0.08-0.34), prediction interval (PI) = 0.06-0.35) and impulsive decision-making (14 studies, SMD 0.17, p-value = 0.008, 95% CI (0.045-0.3), PI = 0.03-0.31) were associated with self-harm or suicidal behavior. There were no significant differences between measures of different facets of impulsivity (ie. delay discounting, risky decision-making, cognitive or response inhibition) and self-harm or suicidal behavior. CONCLUSION: Multiple facets of impulsivity are associated with suicidal behavior in young people. Future suicide research should be designed to capture impulsive states and investigate the impact on different subtypes of impulsivity.