Contacts with primary and secondary healthcare before suicide by those under the care of mental health services: case-control, whole-population-based study using person-level linked routine data in Wales, UK during 2000-2015.

BACKGROUND: People under the care of mental health services are at increased risk of suicide. Existing studies are small in scale and lack comparisons. AIMS: To identify opportunities for suicide prevention and underpinning data enhancement in people with recent contact with mental health services. METHOD: This population-based study includes people who died by suicide in the year following a mental health services contact in Wales, 2001-2015 (cases), paired with similar patients who did not die by suicide (controls). We linked the National Confidential Inquiry into Suicide and Safety in Mental Health and the Suicide Information Database - Cymru with primary and secondary healthcare records. We present results of conditional logistic regression. RESULTS: We matched 1031 cases with 5155 controls. In the year before their death, 98.3% of cases were in contact with healthcare services, and 28.5% presented with self-harm. Cases had more emergency department contacts (odds ratio 2.4, 95% CI 2.1-2.7) and emergency hospital admissions (odds ratio 1.5, 95% CI 1.4-1.7), but fewer primary care contacts (odds ratio 0.7, 95% CI 0.6-0.9) and out-patient appointments (odds ratio 0.2, 95% CI 0.2-0.3) than controls. Odds ratios were larger in females than males for injury and poisoning (odds ratio: 3.3 (95% CI 2.5-4.5) v. 2.6 (95% CI 2.1-3.1)). CONCLUSIONS: We may be missing existing opportunities to intervene, particularly in emergency departments and hospital admissions with self-harm presentations and with unattributed self-harm, especially in females. Prevention efforts should focus on strengthening routine care contacts, responding to emergency contacts and better self-harm care. There are benefits to enhancing clinical audit systems with routinely collected data.

Patterns of engagement in a digital mental health service during COVID-19: a cohort study for children and young people.

Introduction: The COVID-19 pandemic increased public use of digital mental health technologies. However, little is known about changes in user engagement with these platforms during the pandemic. This study aims to assess engagement changes with a digital mental healthcare service during COVID-19. Methods: A cohort study based on routinely collected service usage data from a digital mental health support service in the United Kingdom. Returning users aged 14-25 years that interacted for a maximum of two months were included. The study population was divided into pre-COVID and COVID cohorts. Demographic and usage information between cohorts were compared and usage clusters were identified within each cohort. Differences were tested using Chi-squared, two-sample Kolmogorov-Smirnov tests and logit regressions. Results: Of the 624,103 users who joined the service between May 1, 2019, and October 1, 2021, 18,889 (32.81%) met the inclusion criteria: 5,048 in the pre-COVID cohort and 13,841 in the COVID cohort. The COVID cohort wrote more journals; maintained the same focus on messaging practitioners, posting discussions, commenting on posts, and having booked chats; and engaged less in writing journals, setting personal goals, posting articles, and having ad-hoc chats. Four usage profiles were identified in both cohorts: one relatively disengaged, one focused on contacting practitioners through chats/messages, and two broadly interested in writing discussions and comments within the digital community. Despite their broad similarities, usage patterns also exhibited differences between cohorts. For example, all four clusters had over 70% of users attempting to have ad-hoc chats with practitioners in the pre-COVID cohort, compared to one out of four clusters in the COVID cohort. Overall, engagement change patterns during the COVID-19 pandemic were not equal across clusters. Sensitivity analysis revealed varying strength of these defined clusters. Discussion: Our study identified changes in user activity and engagement behavior within a digital mental healthcare service during the COVID-19 pandemic. These findings suggest that usage patterns within digital mental health services may be susceptible to change in response to external events such as a pandemic. Continuous monitoring of engagement patterns is important for informed design and personalized interventions.

Trends in socioeconomic inequalities in incidence of severe mental illness - A population-based linkage study using primary and secondary care routinely collected data between 2000 and 2017.

OBJECTIVE: In 2008, the UK entered a period of economic recession followed by sustained austerity measures. We investigate changes in inequalities by area deprivation and urbanicity in incidence of severe mental illness (SMI, including schizophrenia-related disorders and bipolar disorder) between 2000 and 2017. METHODS: We analysed 4.4 million individuals from primary and secondary care routinely collected datasets (2000-2017) in Wales and estimated the incidence of SMI by deprivation and urbanicity measured by the Welsh Index of Multiple Deprivation (WIMD) and urban/rural indicator respectively. Using linear modelling and joinpoint regression approaches, we examined time trends of the incidence and incidence rate ratios (IRR) of SMI by the WIMD and urban/rural indicator adjusted for available confounders. RESULTS: We observed a turning point of time trends of incidence of SMI at 2008/2009 where slope changes of time trends were significantly increasing. IRRs by deprivation/urbanicity remained stable or significantly decreased over the study period except for those with bipolar disorder sourced from secondary care settings, with increasing trend of IRRs (increase in IRR by deprivation after 2010: 1.6 % per year, 95 % CI: 1.0 %-2.2 %; increase in IRR by urbanicity 1.0 % per year, 95 % CI: 0.6 %-1.3 %). CONCLUSIONS: There was an association between recession/austerity and an increase in the incidence of SMI over time. There were variations in the effects of deprivation/urbanicity on incidence of SMI associated with short- and long-term socioeconomic change. These findings may support targeted interventions and social protection systems to reduce incidence of SMI.

An external validation of coding for childhood maltreatment in routinely collected primary and secondary care data.

Validated methods of identifying childhood maltreatment (CM) in primary and secondary care data are needed. We aimed to create the first externally validated algorithm for identifying maltreatment using routinely collected healthcare data. Comprehensive code lists were created for use within GP and hospital admissions datasets in the SAIL Databank at Swansea University working with safeguarding clinicians and academics. These code lists build on and refine those previously published to include an exhaustive set of codes. Sensitivity, specificity and positive predictive value of previously published lists and the new algorithm were estimated against a clinically assessed cohort of CM cases from a child protection service secondary care-based setting-'the gold standard'. We conducted sensitivity analyses to examine the utility of wider codes indicating Possible CM. Trends over time from 2004 to 2020 were calculated using Poisson regression modelling. Our algorithm outperformed previously published lists identifying 43-72% of cases in primary care with a specificity ≥ 85%. Sensitivity of algorithms for identifying maltreatment in hospital admissions data was lower identifying between 9 and 28% of cases with high specificity (> 96%). Manual searching of records for those cases identified by the external dataset but not recorded in primary care suggest that this code list is exhaustive. Exploration of missed cases shows that hospital admissions data is often focused on the injury being treated rather than recording the presence of maltreatment. The absence of child protection or social care codes in hospital admissions data poses a limitation for identifying maltreatment in admissions data. Linking across GP and hospital admissions maximises the number of cases of maltreatment that can be accurately identified. Incidence of maltreatment in primary care using these code lists has increased over time. The updated algorithm has improved our ability to detect CM in routinely collected healthcare data. It is important to recognize the limitations of identifying maltreatment in individual healthcare datasets. The inclusion of child protection codes in primary care data makes this an important setting for identifying CM, whereas hospital admissions data is often focused on injuries with CM codes often absent. Implications and utility of algorithms for future research are discussed.

Neurological and psychiatric disorders among autistic adults: a population healthcare record study.

BACKGROUND: Co-occurring psychiatric disorders are common in autism, with previous studies suggesting 54-94% of autistic individuals develop a mental health condition in their lifetime. Most studies have looked at clinically-recruited cohorts, or paediatric cohorts followed into adulthood, with less known about the autistic community at a population level. We therefore studied the prevalence of co-occurring psychiatric and neurological conditions in autistic individuals in a national sample. METHODS: This retrospective case-control study utilised the SAIL Databank to examine anonymised whole population electronic health record data from 2001 to 2016 in Wales, UK (N = 3.6 million). We investigated the prevalence of co-occurring psychiatric and selected neurological diagnoses in autistic adults' records during the study period using International Classification of Diseases-10 and Read v2 clinical codes compared to general population controls matched for age, sex and deprivation. RESULTS: All psychiatric conditions examined were more common amongst adults with autism after adjusting for age, sex and deprivation. Prevalence of attention-deficit hyperactivity disorder (7.00%), bipolar disorder (2.50%), obsessive-compulsive disorder (3.02%), psychosis (18.30%) and schizophrenia (5.20%) were markedly elevated in those with autism, with corresponding odds ratios 8.24-10.74 times the general population. Depression (25.90%) and anxiety (22.40%) were also more prevalent, with epilepsy 9.21 times more common in autism. CONCLUSIONS: We found that a range of psychiatric conditions were more frequently recorded in autistic individuals. We add to understanding of under-reporting and diagnostic overshadowing in autism. With increasing awareness of autism, services should be cognisant of the psychiatric conditions that frequently co-occur in this population.

Widening Excess Mortality During the COVID-19 Pandemic in Individuals Who Self-Harmed.

Background: Studies on COVID-19 pandemic-associated changes in mortality following self-harm remain scarce and inconclusive. Aims: To compare mortality risks in individuals who had self-harmed to those for individuals who had not, before and during the COVID-19 pandemic (Waves 1 and 2) in Wales, the United Kingdom, using population-based routinely collected data. Method: We linked whole population health data to all-cause mortality following an episode of self-harm between April 2016 and March 2021. Propensity score matching, Cox regression, and difference-in-differences were applied to compute changes in excess mortality (as ratios of hazard ratios, RHRs) before and during the pandemic for individuals who self-harmed. Results: The difference in mortality for individuals who self-harmed compared to those who did not widened during Wave 1 (RHR = 2.03, 95% CI: 1.04-4.03) and Wave 2 (RHR = 2.19, 95% CI: 1.12-4.29) from before the pandemic. Stratification by sex and age group produced no significant subgroup differences although risk for younger than 65 years group were higher. Limitations: Limitations include small sample size and incomplete data on cause-specific deaths during the pandemic. Conclusion: Our results underscore continuous monitoring of mortality of individuals who self-harm and effective interventions to address any increases in mortality.

Evidence of increasing recorded diagnosis of autism spectrum disorders in Wales, UK: An e-cohort study.

LAY ABSTRACT: Autism spectrum disorders (autism) are thought to be relatively common, with analyses estimating 1% in the population could meet diagnostic criteria. New services for adult diagnosis have been set up in Wales, UK; however, no studies have examined for the proportion of adults with autism in Wales. In this study, we take anonymised healthcare record data from more than 3.6 million people to produce a national estimate of recorded autism diagnoses. We found the overall prevalence rate of autism in healthcare records was 0.51%. The number of new-recorded cases of autism increased from 0.188 per 1000 person-years in 2001 to 0.644 per 1000 person-years in 2016. The estimate of 0.51% prevalence in the population is lower than suggested by population survey and cohort studies, but comparable to other administrative records. From 2001 to 2016, the number of autism services for adults has increased, and autism is more widely known in society, while concurrently in healthcare records, there was a >150% increase autism diagnoses in the years 2008-2016. An increasing number of diagnoses were among women and those aged over 35 years. This study suggests that while the number of people being diagnosed with autism is increasing, many are still unrecognised by healthcare services.

Association of school absence and exclusion with recorded neurodevelopmental disorders, mental disorders, or self-harm: a nationwide, retrospective, electronic cohort study of children and young people in Wales, UK.

BACKGROUND: Poor attendance at school, whether due to absenteeism or exclusion, leads to multiple social, educational, and lifelong socioeconomic disadvantages. We aimed to measure the association between a broad range of diagnosed neurodevelopmental and mental disorders and recorded self-harm by the age of 24 years and school attendance and exclusion. METHODS: In this nationwide, retrospective, electronic cohort study, we drew a cohort from the Welsh Demographic Service Dataset, which included individuals aged 7-16 years (16 years being the school leaving age in the UK) enrolled in state-funded schools in Wales in the academic years 2012/13-2015/16 (between Sept 1, 2012, and Aug 31, 2016). Using the Adolescent Mental Health Data Platform, we linked attendance and exclusion data to national demographic and primary and secondary health-care datasets. We identified all pupils with a recorded diagnosis of neurodevelopmental disorders (ADHD and autism spectrum disorder [ASD]), learning difficulties, conduct disorder, depression, anxiety, eating disorders, alcohol or drugs misuse, bipolar disorder, schizophrenia, other psychotic disorders, or recorded self-harm (our explanatory variables) before the age of 24 years. Outcomes were school absence and exclusion. Generalised estimating equations with exchangeable correlation structures using binomial distribution with the logit link function were used to calculate odds ratios (OR) for absenteeism and exclusion, adjusting for sex, age, and deprivation. FINDINGS: School attendance, school exclusion, and health-care data were available for 414 637 pupils (201 789 [48·7%] girls and 212 848 [51·3%] boys; mean age 10·5 years [SD 3·8] on Sept 1, 2012; ethnicity data were not available). Individuals with a record of a neurodevelopmental disorder, mental disorder, or self-harm were more likely to be absent or excluded in any school year than were those without a record. Unadjusted ORs for absences ranged from 2·1 (95% CI 2·0-2·2) for those with neurodevelopmental disorders to 6·6 (4·9-8·3) for those with bipolar disorder. Adjusted ORs (aORs) for absences ranged from 2·0 (1·9-2·1) for those with neurodevelopmental disorders to 5·5 (4·2-7·2) for those with bipolar disorder. Unadjusted ORs for exclusion ranged from 1·7 (1·3-2·2) for those with eating disorders to 22·7 (20·8-24·7) for those with a record of drugs misuse. aORs for exclusion ranged from 1·8 (1·5-2·0) for those with learning difficulties to 11·0 (10·0-12·1) for those with a record of drugs misuse. INTERPRETATION: Children and young people up to the age of 24 years with a record of a neurodevelopmental or mental disorder or self-harm before the age of 24 years were more likely to miss school than those without a record. Exclusion or persistent absence are potential indicators of current or future poor mental health that are routinely collected and could be used to target assessment and early intervention. Integrated school-based and health-care strategies to support young peoples' engagement with school life are required. FUNDING: The Medical Research Council, MQ Mental Health Research, and the Economic and Social Research Council. TRANSLATION: For the Welsh translation of the abstract see Supplementary Materials section.

Identification of Suicidal Ideation in the Canadian Community Health Survey-Mental Health Component Using Deep Learning.

Introduction: Suicidal ideation (SI) is prevalent in the general population, and is a risk factor for suicide. Predicting which patients are likely to have SI remains challenging. Deep Learning (DL) may be a useful tool in this context, as it can be used to find patterns in complex, heterogeneous, and incomplete datasets. An automated screening system for SI could help prompt clinicians to be more attentive to patients at risk for suicide. Methods: Using the Canadian Community Health Survey-Mental Health Component, we trained a DL model based on 23,859 survey responses to classify patients with and without SI. Models were created to classify both lifetime SI and SI over the last 12 months. From 582 possible parameters we produced 96- and 21-feature versions of the models. Models were trained using an undersampling procedure that balanced the training set between SI and non-SI; validation was done on held-out data. Results: For lifetime SI, the 96 feature model had an Area under the receiver operating curve (AUC) of 0.79 and the 21 feature model had an AUC of 0.77. For SI in the last 12 months the 96 feature model had an AUC of 0.71 and the 21 feature model had an AUC of 0.68. In addition, sensitivity analyses demonstrated feature relationships in line with existing literature. Discussion: Although further study is required to ensure clinical relevance and sample generalizability, this study is an initial proof of concept for the use of DL to improve identification of SI. Sensitivity analyses can help improve the interpretability of DL models. This kind of model would help start conversations with patients which could lead to improved care and a reduction in suicidal behavior.

Suicide trends in the early months of the COVID-19 pandemic: an interrupted time-series analysis of preliminary data from 21 countries.

BACKGROUND: The COVID-19 pandemic is having profound mental health consequences for many people. Concerns have been expressed that, at their most extreme, these consequences could manifest as increased suicide rates. We aimed to assess the early effect of the COVID-19 pandemic on suicide rates around the world. METHODS: We sourced real-time suicide data from countries or areas within countries through a systematic internet search and recourse to our networks and the published literature. Between Sept 1 and Nov 1, 2020, we searched the official websites of these countries' ministries of health, police agencies, and government-run statistics agencies or equivalents, using the translated search terms "suicide" and "cause of death", before broadening the search in an attempt to identify data through other public sources. Data were included from a given country or area if they came from an official government source and were available at a monthly level from at least Jan 1, 2019, to July 31, 2020. Our internet searches were restricted to countries with more than 3 million residents for pragmatic reasons, but we relaxed this rule for countries identified through the literature and our networks. Areas within countries could also be included with populations of less than 3 million. We used an interrupted time-series analysis to model the trend in monthly suicides before COVID-19 (from at least Jan 1, 2019, to March 31, 2020) in each country or area within a country, comparing the expected number of suicides derived from the model with the observed number of suicides in the early months of the pandemic (from April 1 to July 31, 2020, in the primary analysis). FINDINGS: We sourced data from 21 countries (16 high-income and five upper-middle-income countries), including whole-country data in ten countries and data for various areas in 11 countries). Rate ratios (RRs) and 95% CIs based on the observed versus expected numbers of suicides showed no evidence of a significant increase in risk of suicide since the pandemic began in any country or area. There was statistical evidence of a decrease in suicide compared with the expected number in 12 countries or areas: New South Wales, Australia (RR 0·81 [95% CI 0·72-0·91]); Alberta, Canada (0·80 [0·68-0·93]); British Columbia, Canada (0·76 [0·66-0·87]); Chile (0·85 [0·78-0·94]); Leipzig, Germany (0·49 [0·32-0·74]); Japan (0·94 [0·91-0·96]); New Zealand (0·79 [0·68-0·91]); South Korea (0·94 [0·92-0·97]); California, USA (0·90 [0·85-0·95]); Illinois (Cook County), USA (0·79 [0·67-0·93]); Texas (four counties), USA (0·82 [0·68-0·98]); and Ecuador (0·74 [0·67-0·82]). INTERPRETATION: This is the first study to examine suicides occurring in the context of the COVID-19 pandemic in multiple countries. In high-income and upper-middle-income countries, suicide numbers have remained largely unchanged or declined in the early months of the pandemic compared with the expected levels based on the pre-pandemic period. We need to remain vigilant and be poised to respond if the situation changes as the longer-term mental health and economic effects of the pandemic unfold. FUNDING: None.

Clinical management and mortality risk in those with eating disorders and self-harm: e-cohort study using the SAIL databank.

BACKGROUND: Individuals with eating disorders who self-harm are a vulnerable group characterised by greater pathology and poorer outcomes. AIMS: To explore healthcare utilisation and mortality in those with a record of: self-harm only; eating disorders only; and both co-occurring. METHOD: We conducted a retrospective whole population e-cohort study of individuals aged 10-64 years from 2003 to 2016. Individuals were divided into: record of self-harm only; eating disorders only; both self-harm and eating disorders; and no record of self-harm or eating disorders. We used linked routinely collected healthcare data across primary care, emergency departments, hospital admissions and out-patient appointments to examine healthcare contacts and mortality. RESULTS: We identified 82 627 individuals: n = 75 165 with self-harm only; n = 5786 with eating disorders only; n = 1676 with both combined. Across all groups and settings significantly more individuals attended with significantly more contacts than the rest of the population. The combined group had the highest number of contacts per person (general practitioner, incident rate ratio IRR = 3.3, 95% CI 3.1-3.5; emergency department, IRR = 5.2, 95% CI 4.7-5.8; hospital admission, IRR = 5.2, 95% CI 4.5-6.0; out-patients, IRR = 3.9, 95% CI 3.5-4.4). Standardised mortality ratios showed the highest excess mortality overall in the self-harm only group (SMR = 3.2, 95% CI 3.1-3.3), particularly for unnatural causes of death (SMR = 17.1, 95% CI 16.3-17.9). SMRs and years of life lost showed an increased risk of mortality in younger age groups in the combined group. Adjusted hazard ratios showed increased mortality across all groups (self-harm only, HR = 5.3, 95% CI 5.2-5.5; eating disorders only, HR = 4.1, 95% CI 3.4-4.9; combined group, HR = 6.8, 95% CI 5.4-8.6). CONCLUSIONS: Individuals in all groups had higher healthcare service utilisation than the general population. The increased mortality risk in young people with a record of both eating disorders and self-harm highlights the need for early specialist intervention and enhanced support.

Contacts with primary and secondary healthcare prior to suicide: case-control whole-population-based study using person-level linked routine data in Wales, UK, 2000-2017.

BACKGROUND: Longitudinal studies of patterns of healthcare contacts in those who die by suicide to identify those at risk are scarce. AIMS: To examine type and timing of healthcare contacts in those who die by suicide. METHOD: A population-based electronic case-control study of all who died by suicide in Wales, 2001-2017, linking individuals' electronic healthcare records from general practices, emergency departments and hospitals. We used conditional logistic regression to calculate odds ratios, adjusted for deprivation. We performed a retrospective continuous longitudinal analysis comparing cases' and controls' contacts with health services. RESULTS: We matched 5130 cases with 25 650 controls (5 per case). A representative cohort of 1721 cases (8605 controls) were eligible for the fully linked analysis. In the week before their death, 31.4% of cases and 15.6% of controls contacted health services. The last point of contact was most commonly associated with mental health and most often occurred in general practices. In the month before their death, 16.6 and 13.0% of cases had an emergency department contact and a hospital admission respectively, compared with 5.5 and 4.2% of controls. At any week in the year before their death, cases were more likely to contact healthcare services than controls. Self-harm, mental health and substance misuse contacts were strongly linked with suicide risk, more so when they occurred in emergency departments or as emergency admissions. CONCLUSIONS: Help-seeking occurs in those at risk of suicide and escalates in the weeks before their death. There is an opportunity to identify and intervene through these contacts.

Area deprivation, urbanicity, severe mental illness and social drift - A population-based linkage study using routinely collected primary and secondary care data.

We investigated whether associations between area deprivation, urbanicity and elevated risk of severe mental illnesses (SMIs, including schizophrenia and bipolar disorder) is accounted for by social drift or social causation. We extracted primary and secondary care electronic health records from 2004 to 2015 from a population of 3.9 million. We identified prevalent and incident individuals with SMIs and their level of deprivation and urbanicity using the Welsh Index of Multiple Deprivation (WIMD) and urban/rural indicator. The presence of social drift was determined by whether odds ratios (ORs) from logistic regression is greater than the incidence rate ratios (IRRs) from Poisson regression. Additionally, we performed longitudinal analysis to measure the proportion of change in deprivation level and rural/urban residence 10 years after an incident diagnosis of SMI and compared it to the general population using standardised rate ratios (SRRs). Prevalence and incidence of SMIs were significantly associated with deprivation and urbanicity (all ORs and IRRs significantly >1). ORs and IRRs were similar across all conditions and cohorts (ranging from 1.1 to 1.4). Results from the longitudinal analysis showed individuals with SMIs are more likely to move compared to the general population. However, they did not preferentially move to more deprived or urban areas. There was little evidence of downward social drift over a 10-year period. These findings have implications for the allocation of resources, service configuration and access to services in deprived communities, as well as, for broader public health interventions addressing poverty, and social and environmental contexts.

Alcohol-Specific Mortality in People With Epilepsy: Cohort Studies in Two Independent Population-Based Datasets.

Objectives: The risk of dying by alcohol-specific causes in people with epilepsy has seldom been reported from population-based studies. We aimed to estimate the relative risk of alcohol-specific mortality in people with epilepsy, and the extent to which problematic alcohol use was previously identified in the patients' medical records. Method: We delineated cohort studies in two population-based datasets, the Clinical Practice Research Datalink (CPRD GOLD) in England (January 01, 2001-December 31, 2014) and the Secure Anonymised Information Linkage (SAIL) Databank in Wales (January 01, 2001-December 31, 2014), linked to hospitalization and mortality records. People with epilepsy were matched to up to 20 persons without epilepsy on gender, age (±2 years) and registered general practice. We identified alcohol-specific death from Office for National Statistics (ONS) records using specified ICD-10 codes. We further identified prescriptions, interventions and hospitalisations related to alcohol use. Results: In the CPRD GOLD, we identified 9,871 individuals in the incident epilepsy cohort and 185,800 in the comparison cohort and, in the SAIL Databank, these numbers were 5,569 and 110,021, respectively. We identified a five-fold increased risk of alcohol-specific mortality in people with epilepsy vs. those without the condition in our pooled estimate across the two datasets (deprivation-adjusted HR 4.85, 95%CI 3.46-6.79). Conclusions: People with epilepsy are at increased risk of dying by an alcohol-specific cause than those without the disorder. It is plausible that serious alcohol misuse could either contribute to the development of epilepsy or it could commence subsequent to epilepsy being diagnosed. Regardless of the direction of the association, it is important that the risk of dying as a consequence of alcohol misuse is accurately quantified in people affected by epilepsy. Systematically-applied, sensitive assessment of alcohol consumption by healthcare professionals, at opportunistic, clinical contacts, with rapid access to quality treatment services, should be mandatory and play a key role in reduction of health harms and mortality.

Using Neural Networks with Routine Health Records to Identify Suicide Risk: Feasibility Study.

BACKGROUND: Each year, approximately 800,000 people die by suicide worldwide, accounting for 1-2 in every 100 deaths. It is always a tragic event with a huge impact on family, friends, the community and health professionals. Unfortunately, suicide prevention and the development of risk assessment tools have been hindered by the complexity of the underlying mechanisms and the dynamic nature of a person's motivation and intent. Many of those who die by suicide had contact with health services in the preceding year but identifying those most at risk remains a challenge. OBJECTIVE: To explore the feasibility of using artificial neural networks with routinely collected electronic health records to support the identification of those at high risk of suicide when in contact with health services. METHODS: Using the Secure Anonymised Information Linkage Databank UK, we extracted the data of those who died by suicide between 2001 and 2015 and paired controls. Looking at primary (general practice) and secondary (hospital admissions) electronic health records, we built a binary feature vector coding the presence of risk factors at different times prior to death. Risk factors included: general practice contact and hospital admission; diagnosis of mental health issues; injury and poisoning; substance misuse; maltreatment; sleep disorders; and the prescription of opiates and psychotropics. Basic artificial neural networks were trained to differentiate between the suicide cases and paired controls. We interpreted the output score as the estimated suicide risk. System performance was assessed with 10x10-fold repeated cross-validation, and its behavior was studied by representing the distribution of estimated risk across the cases and controls, and the distribution of factors across estimated risks. RESULTS: We extracted a total of 2604 suicide cases and 20 paired controls per case. Our best system attained a mean error rate of 26.78% (SD 1.46; 64.57% of sensitivity and 81.86% of specificity). While the distribution of controls was concentrated around estimated risks < 0.5, cases were almost uniformly distributed between 0 and 1. Prescription of psychotropics, depression and anxiety, and self-harm increased the estimated risk by ~0.4. At least 95% of those presenting these factors were identified as suicide cases. CONCLUSIONS: Despite the simplicity of the implemented system, the proposed methodology obtained an accuracy like other published methods based on specialized questionnaire generated data. Most of the errors came from the heterogeneity of patterns shown by suicide cases, some of which were identical to those of the paired controls. Prescription of psychotropics, depression and anxiety, and self-harm were strongly linked with higher estimated risk scores, followed by hospital admission and long-term drug and alcohol misuse. Other risk factors like sleep disorders and maltreatment had more complex effects.

Premature mortality among people with severe mental illness - New evidence from linked primary care data.

Studies assessing premature mortality in people with severe mental illness (SMI) are usually based in one setting, hospital (secondary care inpatients and/or outpatients) or community (primary care). This may lead to ascertainment bias. This study aimed to estimate standardised mortality ratios (SMRs) for all-cause and cause-specific mortality in people with SMI drawn from linked primary and secondary care populations compared to the general population. SMRs were calculated using the indirect method for a United Kingdom population of almost four million between 2004 and 2013. The all-cause SMR was higher in the cohort identified from secondary care hospital admissions (SMR: 2.9; 95% CI: 2.8-3.0) than from primary care (SMR: 2.2; 95% CI: 2.1-2.3) when compared to the general population. The SMR for the combined cohort was 2.6 (95% CI: 2.5-2.6). Cause specific SMRs in the combined cohort were particularly elevated in those with SMI relative to the general population for ill-defined and unknown causes, suicide, substance abuse, Parkinson's disease, accidents, dementia, infections and respiratory disorders (particularly pneumonia), and Alzheimer's disease. Solely hospital admission based studies, which have dominated the literature hitherto, somewhat over-estimate premature mortality in those with SMI. People with SMI are more likely to die by ill-defined and unknown causes, suicide and other less common and often under-reported causes. Comprehensive characterisation of mortality is important to inform policy and practice and to discriminate settings to allow for proportionate interventions to address this health injustice.

Risk of Unnatural Mortality in People With Epilepsy.

Importance: People with epilepsy are at increased risk of mortality, but, to date, the cause-specific risks of all unnatural causes have not been reported. Objective: To estimate cause-specific unnatural mortality risks in people with epilepsy and to identify the medication types involved in poisoning deaths. Design, Setting, and Participants: This population-based cohort study used 2 electronic primary care data sets linked to hospitalization and mortality records, the Clinical Practice Research Datalink (CPRD) in England (from January 1, 1998, to March 31, 2014) and the Secure Anonymised Information Linkage (SAIL) Databank in Wales (from January 1, 2001, to December 31, 2014). Each person with epilepsy was matched on age (within 2 years), sex, and general practice with up to 20 individuals without epilepsy. Unnatural mortality was determined using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes V01 through Y98 in the Office for National Statistics mortality records. Hazard ratios (HRs) were estimated in each data set using a stratified Cox proportional hazards model, and meta-analyses were conducted using DerSimonian and Laird random-effects models. The analysis was performed from January 5, 2016, to November 16, 2017. Exposures: People with epilepsy were identified using primary care epilepsy diagnoses and associated antiepileptic drug prescriptions. Main Outcomes and Measures: Hazard ratios (HRs) for unnatural mortality and the frequency of each involved medication type estimated as a percentage of all medication poisoning deaths. Results: In total, 44 678 individuals in the CPRD and 14 051 individuals in the SAIL Databank were identified in the prevalent epilepsy cohorts, and 891 429 (CPRD) and 279 365 (SAIL) individuals were identified in the comparison cohorts. In both data sets, 51% of the epilepsy and comparison cohorts were male, and the median age at entry was 40 years (interquartile range, 25-60 years) in the CPRD cohorts and 43 years (interquartile range, 24-64 years) in the SAIL cohorts. People with epilepsy were significantly more likely to die of any unnatural cause (HR, 2.77; 95% CI, 2.43-3.16), unintentional injury or poisoning (HR, 2.97; 95% CI, 2.54-3.48) or suicide (HR, 2.15; 95% CI, 1.51-3.07) than people in the comparison cohort. Particularly large risk increases were observed in the epilepsy cohorts for unintentional medication poisoning (HR, 4.99; 95% CI, 3.22-7.74) and intentional self-poisoning with medication (HR, 3.55; 95% CI, 1.01-12.53). Opioids (56.5% [95% CI, 43.3%-69.0%]) and psychotropic medication (32.3% [95% CI, 20.9%-45.3%)] were more commonly involved than antiepileptic drugs (9.7% [95% CI, 3.6%-19.9%]) in poisoning deaths in people with epilepsy. Conclusions and Relevance: Compared with people without epilepsy, people with epilepsy are at increased risk of unnatural death and thus should be adequately advised about unintentional injury prevention and monitored for suicidal ideation, thoughts, and behaviors. The suitability and toxicity of concomitant medication should be considered when prescribing for comorbid conditions.

Evidence of a Task-Independent Neural Signature in the Spectral Shape of the Electroencephalogram.

Genetic and neurophysiological studies of electroencephalogram (EEG) have shown that an individual's brain activity during a given cognitive task is, to some extent, determined by their genes. In fact, the field of biometrics has successfully used this property to build systems capable of identifying users from their neural activity. These studies have always been carried out in isolated conditions, such as relaxing with eyes closed, identifying visual targets or solving mathematical operations. Here we show for the first time that the neural signature extracted from the spectral shape of the EEG is to a large extent independent of the recorded cognitive task and experimental condition. In addition, we propose to use this task-independent neural signature for more precise biometric identity verification. We present two systems: one based on real cepstrums and one based on linear predictive coefficients. We obtained verification accuracies above 89% on 4 of the 6 databases used. We anticipate this finding will create a new set of experimental possibilities within many brain research fields, such as the study of neuroplasticity, neurodegenerative diseases and brain machine interfaces, as well as the mentioned genetic, neurophysiological and biometric studies. Furthermore, the proposed biometric approach represents an important advance towards real world deployments of this new technology.

Localized component filtering for electroencephalogram artifact rejection.

Blind source separation (BSS) based artifact rejection systems have been extensively studied in the electroencephalogram (EEG) literature. Although there have been advances in the development of techniques capable of dissociating neural and artifactual activity, these are still not perfect. As a result, a compromise between reduction of noise and leakage of neural activity has to be found. Here, we propose a new methodology to enhance the performance of existing BSS systems: Localized component filtering (LCF). In essence, LCF identifies the artifactual time segments within each component extracted by BSS and restricts the processing of components to these segments, therefore reducing neural leakage. We show that LCF can substantially reduce the neural leakage, increasing the true acceptance rate by 22 percentage points while worsening the false acceptance rate by less than 2 percentage points in a dataset consisting of simulated EEG data (4% improvement of the correlation between original and cleaned signals). Evaluated on real EEG data, we observed a significant increase of the signal-to-noise ratio of up to 9%.