Socioeconomic factors can affect healthcare management.

The aim was to investigate if patient educational attainment is associated with management of bipolar disorder.

We included patients with bipolar disorder type 1 (n = 4289), type 2 (n = 4020) and not otherwise specified (n = 1756), from the Swedish National Quality Register for Bipolar Disorder (BipoläR). The association between patients’ educational level and pharmacological and psychological interventions was analysed by binary logistic regression. We calculated odds ratios after adjusting for demographic and clinical variables.

Higher education was associated with increased likelihood of receiving psychotherapy (adjusted odds ratio 1.34, 95% CI 91.22–1.46) and psychoeducation (adjusted odds ratio 1.18, 95% CI 1.07–1.46), but with lower likelihood of receiving first-generation antipsychotics (adjusted odds ratio 0.76, 95% CI 0.62–0.94) and tricyclic antidepressants (adjusted odds ratio 0.76, 95% CI 0.59–0.97). Higher education was also associated with lower risk for compulsory in-patient care (adjusted odds ratio 0.79, 95% CI 0.67–0.93).

Pharmacological and psychological treatment of bipolar disorder differ depending on patients’ educational attainment. The reasons for these disparities remain to be explained.

Patient and public involvement can improve study outcomes, but little data have been collected on why this might be. We investigated the impact of the Feasibility and Support to Timely Recruitment for Research (FAST-R) service, made up of trained patients and carers who review research documents at the beginning of the research pipeline.

To investigate the impact of the FAST-R service, and to provide researchers with guidelines to improve study documents.

A mixed-methods design assessing changes and suggestions in documents submitted to the FAST-R service from 2011 to 2020. Quantitative measures were readability, word count, jargon words before and after review, the effects over time and if changes were implemented. We also asked eight reviewers to blindly select a pre- or post-review participant information sheet as their preferred version. Reviewers’ comments were analysed qualitatively via thematic analysis.

After review, documents were longer and contained less jargon, but did not improve readability. Jargon and the number of suggested changes increased over time. Participant information sheets had the most suggested changes. Reviewers wanted clarity, better presentation and felt that documents lacked key information such as remuneration, risks involved and data management. Six out of eight reviewers preferred the post-review participant information sheet. FAST-R reviewers provided jargon words and phrases with alternatives for researchers to use.

Longer documents are acceptable if they are clear, with jargon explained or substituted. The highlighted barriers to true informed consent are not decreasing, although this study has suggestions for improving research document accessibility.

There is evidence of the effectiveness of multi-professional home treatment models with regards to improving mental health for elderly patients with mental illness. However, there is a lack of studies examining the efficacy with regard to physical health.

To explore the effectiveness of geropsychiatric home treatment for elderly patients with mental illness with regards to improving physical health by assessing the need of physical treatment.

A 1-year retrospective matched-pair cohort study was conducted in four regions of Austria. We compared 91 patients with a broad spectrum of mental disorders in geropsychiatric home treatment by 1:2 matching to 182 patients in treatment as usual regarding number of contacts with health services, prescriptions, hospital discharges and length of hospital stay.

Patients in geropsychiatric home treatment showed significantly lower numbers of consultations with general practitioners (P < 0.001) and specialists (internal medicine, P = 0.022; psychiatry, P < 0.001), and lower numbers of prescriptions (medical drugs except psychotropic drugs, P < 0.001; psychotropic drugs, P < 0.004) compared with patients in treatment as usual. However, there was no significant difference in the number of hospital discharges and length of hospital stays.

Geropsychiatric home treatment has a positive effect on mental and physical parameters, which is discussed in the context of stress reduction.

The coronavirus disease 2019 (COVID-19) pandemic is likely to lead to a significant increase in mental health disorders among healthcare workers (HCW).

We evaluated the rates of anxiety, depressive and post-traumatic stress disorder (PTSD) symptoms in a population of HCW in the UK.

An electronic survey was conducted between the 5 June 2020 and 31 July 2020 of all hospital HCW in the West Midlands, UK using clinically validated questionnaires: the 4-item Patient Health Questionnaire(PHQ-4) and the Impact of Event Scale-Revised (IES-R). Univariate analyses and adjusted logistic regression analyses were performed to estimate the strengths in associations between 24 independent variables and anxiety, depressive or PTSD symptoms.

There were 2638 eligible participants who completed the survey (female: 79.5%, median age: 42 years, interquartile range: 32–51). The rates of clinically significant symptoms of anxiety, depression and PTSD were 34.3%, 31.2% and 24.5%, respectively. In adjusted analysis a history of mental health conditions was associated with clinically significant symptoms of anxiety (odds ratio (OR) = 2.3, 95% CI 1.9–2.7, P < 0.001), depression (OR = 2.5, 95% CI 2.1–3.0, P < 0.001) and PTSD (OR = 2.1, 95% CI 1.7–2.5, P < 0.001). The availability of adequate personal protective equipment (PPE), well-being support and lower exposure to moral dilemmas at work demonstrated significant negative associations with these symptoms (P ≤ 0.001).

We report higher rates of clinically significant mental health symptoms among hospital HCW following the initial COVID-19 pandemic peak in the UK. Those with a history of mental health conditions were most at risk. Adequate PPE availability, access to well-being support and reduced exposure to moral dilemmas may protect hospital HCW from mental health symptoms.

Sheltered housing is associated with quality-of-life improvements for individuals with serious mental illness (SMI). However, there are equivocal findings around safety outcomes related to this type of living condition.

We aimed to investigate raw differences in prevalence and incidence of crime victimisation in sheltered housing compared with living alone or with family; and to identify groups at high risk for victimisation, using demographic and clinical factors. We do so by reporting estimated victimisation incidents for each risk group.

A large, community-based, cross-sectional survey of 956 people with SMI completed the Dutch Crime and Victimisation Survey. Data was collected on victimisation prevalence and number of incidents in the past year.

Victimisation prevalence was highest among residents in sheltered housing (50.8%) compared with persons living alone (43%) or with family (37.8%). We found that sheltered housing was associated with increased raw victimisation incidence (incidence rate ratio: 2.80, 95% CI 2.36–3.34 compared with living with family; 1.87, 95% CI 1.59–2.20 compared with living alone). Incidence was especially high for some high-risk groups, including men, people with comorbid post-traumatic stress disorder and those with high levels of education. However, women reported less victimisation in sheltered housing than living alone or with family, if they also reported drug or alcohol use.

The high prevalence and incidence of victimisation among residents in sheltered housing highlights the need for more awareness and surveillance of victimisation in this population group, to better facilitate a recovery-enabling environment for residents with SMI.

Bipolar disorder is a chronic mental health condition, which can result in functional impairment despite medication. A large evidence base supports use of psychological therapies and structured care in the treatment of mood disorders, but these are rarely implemented. e-Pathways are digital structures that inform and record patient progress through a healthcare system, although these have not yet been used for bipolar disorder.

To assess the perceived benefits and costs associated with implementing a collaborative NICE-informed e-pathway for bipolar disorder.

Healthcare professionals and people with bipolar disorder attended a workshop to share feedback on e-pathways. Data were collected through questionnaires (n = 26) and transcription of a focus group, analysed qualitatively by a framework analysis.

Patients and healthcare professionals welcomed the development of an e-pathway for bipolar disorder. There were five elements to the framework: quality and delivery of care, patient–clinician collaboration, flexibility and adaptability, impact on staff and impact on healthcare services.

Identification of benefits and costs ensures that future development of e-pathways addresses concerns of healthcare professionals and people with bipolar disorder, which would be essential for successful implementation. Recommendations for this development include making e-pathways less complicated for patients, ensuring sufficient training and ensuring clinicians do not feel their skills become invalidated. Limitations of the study, and directions for future research, are discussed.

The aim of the present study is to determine whether vulnerable non-psychotic clients presenting in court proceedings do not share the same mortality profile as psychotic patients in similar environments. It is hypothesised that the two display quite separate mortality profiles.

The increased mortality of psychiatric patients and prisoners has been documented but less is known of the outcomes among other vulnerable populations .

The population for study is a consecutive series of assessments in court proceedings of carers of children at risk and violent offenders.

Assistants not involved in the initial assessments transferred data from case notes and this material was transferred to computer files. Statistical analysis SPSS19

Formal psychiatric diagnoses were those agreed in court proceedings. National mortality records were searched and copies of death certificates obtained. A small number of cases known to have returned overseas were excluded. 772 cases were studied. One in five were assessed in prison, twice as many gave a history of violent criminal behaviour. Over a half suffered abuse or neglect or admitted to being unhappy in childhood. Three subgroups have been identified: Vulnerable with no psychotic illness(60%), psychosis with no evidence of personality disorder or of mixed psychosis(18%), linked psychosis(22%). It was found that demographic variables , deprivation factors, adverse childhood experiences and outcomes and clinical variables are in excess among linked psychotics compared with other groups. Linear regression of unnatural death among psychotic patients identifies five risk factors. The distribution of high-risk factors among linked psychosis is more than twice that found in other groups.

Natural mortality is most evident among clients suffering from psychosis without personality disorder or mixed disorder.Unnatural mortality is more than 10 times greater among patients with linked psychosis, compared with those with no psychosis and four times greater than other psychoses. Risk factors for unnatural mortality are: physical illness, stressful relationship, violence to self or others, detained and history of behaviour disorder.

The findings of the present study demonstrate that vulnerable clients without psychosis are less likely to die by unnatural causes than clients who suffer psychosis coexisting with personality disorder or mixed psychosis. The null hypothesis is upheld. The findings suggest that risk assessment of vulnerable populations should take account of risk factors of unnatural death which have been identified in this study.

Positive affect and anhedonia are important but challenging targets for mental health treatments. Previous research indicates the potential of a computerised cognitive training paradigm involving generation of positive mental imagery, termed positive mental imagery training (PMIT), to increase positive affect and reduce anhedonia.

Our main aim was to investigate the feasibility of PMIT as a positive affect-focused, transdiagnostic adjunct to treatment as usual for patients in in-patient mental health settings.

We ran an open feasibility, randomised controlled trial with three parallel arms: treatment as usual; treatment as usual plus PMIT; and treatment as usual plus an active comparator, cognitive control training. Fifty-seven patients from two different in-patient mental health treatment clinics in Germany were randomised in a 1:1:1 ratio. PMIT and cognitive control training comprised an introductory session followed by eight 15-min training sessions over 2 weeks. Clinical outcomes such as positive affect (primary outcome measure) and anhedonia were assessed at pre- and post-training, and at a further 2-week follow-up.

Adherence was good and attrition was low. The patterns of results for the outcome data were not consistent with a specific effect of PMIT on positive affect, but were more consistent with a specific effect on anhedonia.

The results indicate feasibility and potential promise of a larger efficacy trial investigating PMIT as a treatment adjunct in in-patient mental health settings. Limitations include lack of researcher blinding, small sample size and lack of pre-specified feasibility outcomes. Anhedonia may be a more suitable primary outcome for a future larger trial.

Studies have shown that over half of individuals with bipolar disorder experience early-life trauma, which may influence clinical outcomes, including suicidality and presence of psychotic features. However, studies report inconsistent findings regarding the effect of trauma on cognitive outcomes in bipolar disorder.

Our study explores the effect of lifetime trauma on the level of vulnerability to psychosis and cognitive performance in participants with bipolar disorder.

We evaluated lifetime trauma history in 236 participants with a diagnosis of bipolar disorder type 1 or 2, using the Structured Clinical Interview for DSM-IV and the Childhood Trauma Questionnaire. We classified trauma types based on the Substance Abuse and Mental Health Services Administration's concept of trauma, which characterises the type of experienced trauma (e.g. interpersonal and intentional, accidental or naturally occurring). Our primary outcome measures of interest were vulnerability to psychosis (Schizotypal Personality Questionnaire), cognitive performance (MATRICS Consensus Cognitive Battery) and social functioning (Social Adjustment Scale Self-Report).

Multivariate analysis of covariance showed a significant effect of trauma type on the Schizotypal Personality Questionnaire cognitive–perceptual domain (F(3) = 6.7, P < 0.001). The no-trauma group had lower cognitive–perceptual schizotypal features compared with the accidental and intentional trauma (P < 0.001) and interpersonal and intentional trauma (P = 0.01) groups.

Our results highlight the need for careful trauma inquiry in patients with bipolar disorder, and consideration of how trauma-focused or -informed treatments may be an integral part of treatment planning to improve outcomes in bipolar disorder.

No co-productive narrative synthesis of system-level facilitators and barriers to personal recovery in mental illness has been undertaken.

To clarify system-level facilitators and barriers to personal recovery of people with mental illness.

Qualitative study guided by thematic analysis. Data were collected through one focus group, which involved seven service users and three professionals. This group had 11 meetings, each lasting 2 h at a local research institute, between July 2016 to January 2018.

The analysis yielded three themes: barriers inhibiting positive interaction within personal relationship networks, roots of barriers from mental health systems and the social cultural context, and possible solutions to address the roots. Barriers were acknowledged as those related to sense of safety, locus of control within oneself and reunion with self. The roots of barriers were recognised within mental health services, including system without trauma sensitivity, lack of advocacy support and limited access to psychosocial approaches. Roots from social cultural context were also found. There were no narratives relating to facilitators. A possible solution was to address the roots from systems. Social cultural change was called for that makes personalised goals most valued, with an inclusive design that overcomes stigma, to achieve an open and accepting community.

The analysis yielded system-level barriers specific to each recovery process. Roots of barriers that need transformation to facilitate personal recovery were identified within mental health services. Social interventions should be further explored to translate the suggested social cultural changes into action.

A review of Australian mental health services identified a gap in routine outcome measures addressing social, emotional and behavioural domains for pre-schoolers and infants. A Child and Adolescent Mental Health Information Development Expert Advisory Panel working group developed the Health of the Nation Outcome Scales for Infants (HoNOSI), a clinician-reported routine outcome measure for infants 0–47 months. Prior face validity testing showed that the HoNOSI was considered useful in measuring mental health outcomes.

To examine the concurrent validity of the HoNOSI.

Mental health clinicians providing assessment and treatment to infants in routine clinical practice participated in the study. The mental health status of 108 infants were rated by a minimum of 26 clinicians with the HoNOSI, the Parent-Infant Relationship Global Assessment Scale (PIR-GAS) and measures of symptom severity and distress.

The HoNOSI was statistically significantly correlated with the PIR–;GAS, rs = −0.73; Clinical Worry, rs = 0.77; and Severity Judgement ratings, rs = 0.85; P < 0.001. A good level of internal consistency was found. Using the COsensus-based Standards for the selection of health Measurement INstruments (COSMIN) criteria for judging instrument acceptability, the HoNOSI meets the standard for both concurrent validity and internal consistency.

There has been a clear need for a routine outcome measure for use with infants. This study provides positive evidence of aspects of validity. These findings, along with those from the prior face validity study, support a controlled release of the HoNOSI accompanied by further research and development.

Homelessness in England and Wales is on the rise together with the mortality rate among homeless people. Many homeless people have a mental illness, which is a risk factor for suicide.

This study used data from the National Confidential Inquiry into Suicide and Safety in Mental Health to examine demographic and clinical characteristics of homeless people who died by suicide and were in recent contact with mental health services.

We have compared 514 patients (2% of the total sample) who died by suicide and who were reported as being homeless or having no fixed abode by their clinicians with patients in stable accommodation between 2000 and 2016 to identify differences in sociodemographic characteristics and clinical care.

Our analysis suggests that homeless patients who died by suicide had more acute (alcohol: 47% v. 25%, P < 0.01, drug: 39% v. 15%, P < 0.01) and chronic (alcohol: 72% v. 44%, P > 0.01, drug: 64% v. 31%) substance misuse issues than patients in stable accommodation. Homeless patients were also more likely to die as in-patients (21% v. 10%, P < 0.01) or within 3 months of discharge (32% v. 19%, P < 0.01).

Homeless patients who died by suicide more often had known risk factors for suicide than patients in stable accommodation. As a result of the higher percentages of post-discharge and in-patient suicides in homeless patients as well as the high prevalence of substance misuse, this study recommends closer integration of services as well as awareness of risks during in-patient admission and in the weeks immediately after discharge.

To audit the investigation, identification and treatment of vitamin D deficiency within Women's Secure Services.

It has been suggested that vitamin D and vitamin D deficiency may play a role in the pathogenesis of psychiatric illness. There is evidence that vitamin D inadequacy is prevalent among patients in long-term hospital settings. Patients within secure hospitals are considered to be at high risk due to their often lengthy admissions, having been transferred from other hospital or prison settings. Ardenleigh in Birmingham is a blended female secure unit. Here we present the findings of an audit, completed in 2019, of vitamin D monitoring and treatment in this service.

A retrospective review of electronic patient records, for all inpatients admitted within women's secure services at Ardenleigh as of 1st September 2019 (n = 27). Standards were based on the Trust accepted guidelines for management of vitamin D deficiency.

Key findings included:-

The majority of inpatients were Caucasian (44%) and African-Caribbean (41%). Median age was 31 years (range 20–56).

Approximately two-thirds (60%) had been in hospital for over a year.

89% of patients had their vitamin D level checked at some point during admission.

Of those checked, 25% were tested within 1 week of admission. Seven patients were tested after being in hospital for over one year (30%).

Only 25% of patients tested were found to have adequate vitamin D levels. Nine patients were found to have insufficient levels of vitamin D (37.5%) or deficiency (37.5%).

89% of those identified as requiring treatment were prescribed supplementation, of which the majority was prescribed at the correct dose for the appropriate duration (94%). One patient refused treatment. Of those with sufficient levels, 67% were prescribed ongoing maintenance treatment due to previously detected deficiency.

Of those found to have sufficient vitamin D in the last 12 months (n = 14), 71% were continued on maintenance treatment.

We identified a high prevalence of vitamin D insufficiency in women admitted to secure services. Testing was delayed for a number of patients from the point of admission. However, once identified, the vast majority of those in need of treatment were managed appropriately by the medical team. We advise that vitamin D be considered an essential routine blood test at the point of admission to minimise delays in identifying those with deficiency and establishing necessary supplementation.

The First Episode Rapid Early Intervention for Eating Disorders (FREED) service model is associated with significant reductions in wait times and improved clinical outcomes for emerging adults with recent-onset eating disorders. An understanding of how FREED is implemented is a necessary precondition to enable an attribution of these findings to key components of the model, namely the wait-time targets and care package.

This study evaluated fidelity to the FREED service model during the multicentre FREED-Up study.

Participants were 259 emerging adults (aged 16–25 years) with an eating disorder of <3 years duration, offered treatment through the FREED care pathway. Patient journey records documented patient care from screening to end of treatment. Adherence to wait-time targets (engagement call within 48 h, assessment within 2 weeks, treatment within 4 weeks) and care package, and differences in adherence across diagnosis and treatment group were examined.

There were significant increases (16–40%) in adherence to the wait-time targets following the introduction of FREED, irrespective of diagnosis. Receiving FREED under optimal conditions also increased adherence to the targets. Care package use differed by component and diagnosis. The most used care package activities were psychoeducation and dietary change. Attention to transitions was less well used.

This study provides an indication of adherence levels to key components of the FREED model. These adherence rates can tentatively be considered as clinically meaningful thresholds. Results highlight aspects of the model and its implementation that warrant future examination.

The COVID-19 pandemic and associated restrictions are expected to affect the mental health of the population, especially people with intellectual disability and/or autism spectrum disorder, because of a variety of biological and psychosocial reasons.

This study aimed to estimate if COVID-19 restrictions are associated with a change in number of total consultations carried out by psychiatrists and prescription of psychotropic medication in people with intellectual disability and/or autism spectrum disorder, within a community intellectual disability service.

A quantitative observational study was conducted, involving retrospective and prospective data collection before and during lockdown. Data was collected on a spreadsheet and emailed to all psychiatrists working within the Coventry and Warwickshire Partnership NHS Trust-wide community intellectual disability service. Variables included total consultations, medication interventions, types of medications used, multidisciplinary team input and clinical reasons for medication interventions. Data was analysed separately for child and adolescent mental health services (CAMHS) and adult intellectual disability teams, and for the whole service.

During the lockdown period, total consultations in the community intellectual disability service increased by 19 per week and medication interventions increased by two per week. Multidisciplinary team input increased in CAMHS from 0.17 to 0.71 per week and in adult intellectual disability from 5.7 to 6.5 per week. Hypnotics and benzodiazepines were the most commonly prescribed psychotropic medications during the lockdown period.

COVID-19-related lockdown resulted in an increase in medication interventions, total consultations and involvement of multidisciplinary teams to manage mental health and behavioural issues in people with intellectual disability and/or autism spectrum disorder.

Little is known about the effects of depression before birth on the quality of the mother–infant interaction.

To understand whether depression, either in pregnancy or in lifetime before pregnancy, disrupts postnatal mother–infant interactions.

We recruited 131 pregnant women (51 healthy, 52 with major depressive disorder (MDD) in pregnancy, 28 with a history of MDD but healthy pregnancy), at 25 weeks’ gestation. MDD was confirmed with the Structured Clinical Interview for DSM-IV Disorders. Neonatal behaviour was assessed at 6 days with the Neonatal Behavioural Assessment Scale, and mother–infant interaction was assessed at 8 weeks and 12 months with the Crittenden CARE-Index.

At 8 weeks and 12 months, dyads in the depression and history-only groups displayed a reduced quality of interaction compared with healthy dyads. Specifically, at 8 weeks, 62% in the depression group and 56% in the history-only group scored in the lowest category of dyadic synchrony (suggesting therapeutic interventions are needed), compared with 37% in the healthy group (P = 0.041); 48% and 32%, respectively, scored the same at 12 months, compared with 14% in the healthy group (P = 0.003). At 6 days, neonates in the depression and history-only groups exhibited decreased social-interactive behaviour, which, together with maternal socioeconomic difficulties, was also predictive of interaction quality, whereas postnatal depression was not.

Both antenatal depression and a lifetime history of depression are associated with a decreased quality of mother–infant interaction, irrespective of postnatal depression. Clinicians should be aware of this, as pregnancy provides an opportunity for identification and intervention to support the developing relationship.

Efficient handovers are integral to patient care. Challenges to handover for wards include high patient turnover and varied handover approaches between wards, as well as admissions out of hours. Patients on Old Age Wards often have multiple comorbidities and can deteriorate rapidly without coordinated care. Our focus was on improving handover of patients transferred between the Old Age Admissions Ward and Rehabilitation Ward. We aimed to create a ward handover protocol to improve compliance with documenting a pretransfer plan and ensure there was an 80% compliance with completing this plan within 3 months.

An MDT discussion took place in order to explore change ideas. Questionnaires were filled out post implementation of protocol. A handover proforma was designed to capture important patient data and continuing plans. A PDSA cycle was designed to deliver a structured handover.

Per patient measures were collected including: whether a handover took place, recording of current medical and psychiatric issues, documentation of plan and was the plan put into action or reviewed.

MDT feedback was collected on satisfaction with the protocol and handover process using open questions and Likert scale.

Prior to the establishment of the proforma there was no verbal or written handover between wards. In 28% of cases prior to the intervention, blood results were checked and medication reviews took place within the timeframe written in the patient's notes. A proforma was initiated and used for 93% of patient handovers between wards. Blood results were checked according to the planned timeframe in 86% of cases. Where the handover proforma took place, 100% of patients had a medications review. Qualitative detail revealed that key patient appointments such as MRI Brain scans and important plans such as fluid restriction limits were missed before implementing the protocol. Aftewards complex patient plans were recorded and implemented accordingly.

Questionnaire feedback was positive and MDT found the proforma to be helpful and to improve patient safety.

The team viewed the new handover pathway as a positive patient safety tool. Compliance with completing the protocol in the longer term and maintaining change is an area for ongoing improvement.

One critical barrier to the uptake of mental health programmes is the so-called physical healthcare gap, a concern raised by the unattended physical comorbidity and early mortality of persons with severe mental illness.

To evaluate the extension of physical healthcare gap among persons with severe mental illness under chronic drug therapies.

A population-based cohort study was carried out, using Lombardy healthcare utilisation databases. Prevalent patients treated with blood pressure-, lipid- or glucose-lowering agents were identified in January 2017. Among these, those who were receiving care for depression, schizophrenia, bipolar disorder or personality disorder formed the study cohort. A reference cohort was randomly selected from prevalent patients treated with chronic therapies without signs of severe mental disorders, to be matched with study cohort members for gender, age and number of previous contacts with the National Health System. One-year adherence to healthcare was measured through the proportion of days covered (drug adherence), and exposure to selected recommendations (clinical control adherence).

The 55 162 patients with severe mental illness were less likely to have high adherence to blood pressure-lowering, lipid-lowering or antidiabetic agents than the reference cohort by −24% (95% CI −26 to −22%), −10% (95% CI −14 to −6%) and −25% (95% CI −29 to −21%), respectively. The 9250 patients with diabetes and severe mental illness had −18% (95% CI −22% to −13%) reduced likelihood to meet recommendations for the clinical management of diabetes, compared with the reference cohort.

Adherence to chronic drug therapies was sensibly worse among patients living with mental illness than those without signs of mental disorders.

Suicide prediction models have been formulated in a variety of ways and are heterogeneous in the strength of their predictions. Machine learning has been a proposed as a way of improving suicide predictions by incorporating more suicide risk factors.

To determine whether machine learning and the number of suicide risk factors included in suicide prediction models are associated with the strength of the resulting predictions.

Random-effect meta-analysis of exploratory suicide prediction models constructed by combining two or more suicide risk factors or using clinical judgement (Prospero Registration CRD42017059665). Studies were located by searching for papers indexed in PubMed before 15 August 2020 with the term suicid* in the title.

In total, 86 papers reported 102 suicide prediction models and included 20 210 411 people and 106 902 suicides. The pooled odds ratio was 7.7 (95% CI 6.7–8.8) with high between-study heterogeneity (I2 = 99.5). Machine learning was associated with a non-significantly higher odds ratio of 11.6 (95% CI 6.0–22.3) and clinical judgement with a non-significantly lower odds ratio of 4.7 (95% CI 2.1–10.9). Models including a larger number of suicide risk factors had a higher odds ratio when machine-learning studies were included (P = 0.02). Among non-machine-learning studies, suicide prediction models including fewer risk factors performed just as well as those including more risk factors.

Machine learning might have the potential to improve the performance of suicide prediction models by increasing the number of included suicide risk factors but its superiority over other methods is unproven.