Operational failures in general practice: a consensus-building study on the priorities for improvement.

BACKGROUND: System problems, known as operational failures, can greatly affect the work of GPs, with negative consequences for patient and professional experience, efficiency, and effectiveness. Many operational failures are tractable to improvement, but which ones should be prioritised is less clear. AIM: To build consensus among GPs and patients on the operational failures that should be prioritised to improve NHS general practice. DESIGN AND SETTING: Two modified Delphi exercises were conducted online among NHS GPs and patients in several regions across England. METHOD: Between February and October 2021, two modified Delphi exercises were conducted online: one with NHS GPs, and a subsequent exercise with patients. Over two rounds, GPs rated the importance of a list of operational failures (n = 45) that had been compiled using existing evidence. The resulting shortlist was presented to patients for rating over two rounds. Data were analysed using median scores and interquartile ranges. Consensus was defined as 80% of responses falling within one value below and above the median. RESULTS: Sixty-two GPs responded to the first Delphi exercise, and 53.2% (n = 33) were retained through to round two. This exercise yielded consensus on 14 failures as a priority for improvement, which were presented to patients. Thirty-seven patients responded to the first patient Delphi exercise, and 89.2% (n = 33) were retained through to round two. Patients identified 13 failures as priorities. The highest scoring failures included inaccuracies in patients' medical notes, missing test results, and difficulties referring patients to other providers because of problems with referral forms. CONCLUSION: This study identified the highest-priority operational failures in general practice according to GPs and patients, and indicates where improvement efforts relating to operational failures in general practice should be focused.

Wellbeing Impact Study of High-Speed 2 (WISH2): Protocol for a mixed-methods examination of the impact of major transport infrastructure development on mental health and wellbeing.

Although research has demonstrated that transport infrastructure development can have positive and negative health-related impacts, most of this research has not considered mental health and wellbeing separately from physical health. There is also limited understanding of whether and how any effects might be experienced differently across population groups, whether this differs according to the stage of development (e.g. planning, construction), and how changes to planned infrastructure may affect mental health and wellbeing. This paper presents a protocol for the Wellbeing Impact Study of HS2 (WISH2), which seeks to address these questions using a high-speed rail development in the UK as an applied example. WISH2 is a 10-year, integrated, longitudinal, mixed-methods project using general practices (primary medical care providers in the UK) as an avenue for participant recruitment and for providing a geographically defined population for which aggregated data on mental health indicators are available. The research comprises: (i) a combined longitudinal and repeated cross-sectional cohort study involving multiple waves of survey data collection and data from medical records; (ii) longitudinal, semi-structured interviews and focus groups with residents and community stakeholders from exposed areas; (iii) analysis of administrative data aggregated at the general practice population level; and (iv) health economic analysis of mental health and wellbeing impacts. The study findings will support the development of strategies to reduce negative impacts and/or enhance positive mental health and wellbeing impacts of high-speed rail developments and other large-scale infrastructure projects.

Public engagement with genomics.

As detailed in its flagship report, Genome UK, the UK government recognises the vital role that broad public engagement across whole populations plays in the field of genomics. However, there is limited evidence about how to do this at scale. Most public audiences do not feel actively connected to science, are oftenunsure of the relevance to their lives and rarely talk to their family and friends about; we term this dis-connection a 'disengaged public audience'. We use a narrative review to explore: (i) UK attitudes towards genetics and genomics and what may influence reluctance to engage with these topics; (ii) innovative public engagement approaches that have been used to bring diverse public audiences into conversations about the technology. Whilst we have found some novel engagement methods that have used participatory arts, film, social media and deliberative methods, there is no clear agreement on best practice. We did not find a consistently used, evidence-based strategy for delivering public engagement about genomics across diverse and broad populations, nor a specific method that is known to encourage engagement from groups that have historically felt (in terms of perception) and been (in reality) excluded from genomic research. We argue there is a need for well-defined, tailor-made engagement strategies that clearly articulate the audience, the purpose and the proposed impact of the engagement intervention. This needs to be coupled with robust evaluation frameworks to build the evidence-base for population-level engagement strategies.

Outcomes following suicidal crisis among hazardous and harmful alcohol users in the Crisis Resolution Team.

Despite associations between alcohol use and suicidal acts, little research measures prognoses of alcohol-using patients treated by Crisis Resolution Teams (CRTs), an intensive community-based intervention. We estimated the association of alcohol use amongst patients accepted following suicidal acts or ideation in four London-based Crisis Resolution Teams, with death-by-any-cause or recontact with crisis care. We analysed the electronic health records of 1615 CRT patients accepted following suicidal acts or ideation over 38 months, following STROBE guidelines. Using logistic regression we estimated the association of alcohol use (indicated by risk-assessment, AUDIT, or ICD-10 diagnosis) with death-or-recontact at (i) 30-days and (ii) 1-year after treatment start, adjusted for age, sex, ethnicity, psychiatric diagnosis, and severity of need. Hazardous, harmful, or dependent drinking was identified in 270 cases at baseline (16.7%); 73 (4.5%) were alcohol dependent. By 1-year, 622 patients (38.5%) had recontacted crisis care or died. After adjustment, alcohol use at a hazardous, harmful, or dependent level was not associated with increased odds of death-or-recontact at 30-days (AOR 1.17, 95%CI 0.73, 1.88) or 1-year (AOR 1.17, 95%CI 0.85, 1.60). Patients with hazardous, harmful, and dependent alcohol use are a small proportion of CRT patients, despite being more commonly encountered in emergency settings from which patients may be referred to CRTs, indicating a potential gap in provision. Those who are included in CRTs are not at increased risk of death-or-recontact within 1 year of treatment, suggesting that their inclusion can work, at least in a sample with predominantly hazardous or harmful alcohol use.

Assessing the Impact of Developments in Genetic Testing on Insurers' Risk Exposure.

Predictive genetic testing provides individuals with information about their future risk of developing health conditions. Theoretically, predictive genetic tests could have positive or negative impacts on the insurance industry. If genetic test results stimulate actions to reduce health risks, they may reduce costs to insurers. If disclosed to insurers, such information may allow them to better understand individual- and population-level risks and make insurance more affordable. However, if individuals who know they are at high genetic risk of becoming ill or dying are more likely to apply for insurance than those not at high risk, this may lead to an unanticipated increase in claims. It may be exacerbated if people at low genetic risk are less likely to apply for insurance compared to the general population. If this happened on a large scale it could make the insurance market unsustainable. Determining whether a genetic test could affect the insurance industry is complex and needs to be evaluated on a per-test basis. The Cambridge Centre for Health Services Research, a collaboration between RAND Europe and the University of Cambridge, developed a framework for evaluating the potential impacts on the UK insurance industry arising from predictive genetic tests. It considers the characteristics of genetic tests and behavioural aspects that influence their uptake. It is intended to provide a transparent approach for evaluating whether a specific condition for which a test is available could impact the insurance industry, currently or in the future, and understanding the key factors that influence this.

Guiding organisational decision-making about COVID-19 asymptomatic testing in workplaces: mixed-method study to inform an ethical framework.

BACKGROUND: Workplace programmes to test staff for asymptomatic COVID-19 infection have become common, but raise a number of ethical challenges. In this article, we report the findings of a consultation that informed the development of an ethical framework for organisational decision-making about such programmes. METHODS: We conducted a mixed-method consultation - a survey and semi-structured interviews during November-December 2020 in a UK case study organisation that had introduced asymptomatic testing for all staff working on-site in its buildings. Analysis of closed-ended survey data was conducted descriptively. An analysis approach based on the Framework Method was used for the open-ended survey responses and interview data. The analyses were then integrated to facilitate systematic analysis across themes. Inferences were based on the integrated findings and combined with other inputs (literature review, ethical analysis, legal and public health guidance, expert discussions) to develop an ethical framework. RESULTS: The consultation involved 61 staff members from the case study organisation (50 survey respondents and 11 interview participants). There was strong support for the asymptomatic testing programme: 90% of the survey respondents viewed it as helpful or very helpful. Open-ended survey responses and interviews gave insight into participants' concerns, including those relating to goal drift, risk of false negatives, and potential negative impacts for household members and people whose roles lacked contractual and financial stability. Integration of the consultation findings and the other inputs identified the importance of a whole-system approach with appropriate support for the key control measure of isolation following positive tests. The need to build trust in the testing programme, for example through effective communication from leaders, was also emphasised. CONCLUSIONS: The consultation, together with other inputs, informed an ethical framework intended to support employers. The framework may support organisational decision-making in areas ranging from design and operation of the programme through to choices about participation. The framework is likely to benefit from further consultation and refinement in new settings.

Assessing the Implementation of Digital Innovations in Response to the COVID-19 Pandemic to Address Key Public Health Functions: Scoping Review of Academic and Nonacademic Literature.

BACKGROUND: Digital technologies have been central to efforts to respond to the COVID-19 pandemic. In this context, a range of literature has reported on developments regarding the implementation of new digital technologies for COVID-19-related surveillance, prevention, and control. OBJECTIVE: In this study, scoping reviews of academic and nonacademic literature were undertaken to obtain an overview of the evidence regarding digital innovations implemented to address key public health functions in the context of the COVID-19 pandemic. This study aimed to expand on the work of existing reviews by drawing on additional data sources (including nonacademic sources) by considering literature published over a longer time frame and analyzing data in terms of the number of unique digital innovations. METHODS: We conducted a scoping review of the academic literature published between January 1, 2020, and September 15, 2020, supplemented by a further scoping review of selected nonacademic literature published between January 1, 2020, and October 13, 2020. Both reviews followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) approach. RESULTS: A total of 226 academic articles and 406 nonacademic articles were included. The included articles provided evidence of 561 (academic literature) and 497 (nonacademic literature) unique digital innovations. The most common implementation settings for digital innovations were the United States, China, India, and the United Kingdom. Technologies most commonly used by digital innovations were those belonging to the high-level technology group of integrated and ubiquitous fixed and mobile networks. The key public health functions most commonly addressed by digital innovations were communication and collaboration and surveillance and monitoring. CONCLUSIONS: Digital innovations implemented in response to the COVID-19 pandemic have been wide ranging in terms of their implementation settings, the digital technologies used, and the public health functions addressed. However, evidence gathered through this study also points to a range of barriers that have affected the successful implementation of digital technologies for public health functions. It is also evident that many digital innovations implemented in response to the COVID-19 pandemic are yet to be formally evaluated or assessed.

Return of genomic results does not motivate intent to participate in research for all: Perspectives across 22 countries.

PURPOSE: The aim of this study was to determine how attitudes toward the return of genomic research results vary internationally. METHODS: We analyzed the "Your DNA, Your Say" online survey of public perspectives on genomic data sharing including responses from 36,268 individuals across 22 low-, middle-, and high-income countries, and these were gathered in 15 languages. We analyzed how participants responded when asked whether return of results (RoR) would motivate their decision to donate DNA or health data. We examined variation across the study countries and compared the responses of participants from other countries with those from the United States, which has been the subject of the majority of research on return of genomic results to date. RESULTS: There was substantial variation in the extent to which respondents reported being influenced by RoR. However, only respondents from Russia were more influenced than those from the United States, and respondents from 20 countries had lower odds of being partially or wholly influenced than those from the United States. CONCLUSION: There is substantial international variation in the extent to which the RoR may motivate people's intent to donate DNA or health data. The United States may not be a clear indicator of global attitudes. Participants' preferences for return of genomic results globally should be considered.

Substance use in psychiatric crisis: relationship to violence.

BACKGROUND: Substance use and psychiatric illness, particularly psychotic disorders, contribute to violence in emergency healthcare settings. However, there is limited research regarding the relationship between specific substances, psychotic symptoms and violent behaviour in such settings. We investigated the interaction between recent cannabinoid and stimulant use, and acute psychotic symptoms, in relation to violent behaviour in a British emergency healthcare setting. METHODS: We used electronic medical records from detentions of 1089 individuals under Section 136 of the UK Mental Health Act (1983 amended 2007), an emergency police power used to detain people for 24-36 h for psychiatric assessment. The relationship between recent cannabinoids and/or stimulant use, psychotic symptoms, and violent behaviour, was estimated using logistic regression. FINDINGS: There was evidence of recent alcohol or drug use in 64.5% of detentions. Violent incidents occurred in 12.6% of detentions. Psychotic symptoms increased the odds of violence by 4.0 [95% confidence intervals (CI) 2.2-7.4; p < 0.0001]. Cannabinoid use combined with psychotic symptoms increased the odds of violence further [odds ratios (OR) 7.1, 95% CI 3.7-13.6; p < 0.0001]. Recent use of cannabinoids with stimulants but without psychotic symptoms was also associated with increased odds of violence (OR 3.3, 95% CI 1.4-7.9; p < 0.0001). INTERPRETATION: In the emergency setting, patients who have recently used cannabinoids and exhibit psychotic symptoms are at higher risk of violent behaviour. Those who have used both stimulants and cannabinoids without psychotic symptoms may also be at increased risk. De-escalation protocols in emergency healthcare settings should account explicitly for substance use.

Point of Care Testing for Infectious Disease in Europe: A Scoping Review and Survey Study.

Background: Point of care testing (POCT) for infectious diseases is testing conducted near the patient. It allows clinicians to offer the most appropriate treatment more quickly. As POCT devices have increased in accuracy and become more cost-effective, their use has grown, but a systematic assessment of their use for clinical and public health management of infectious diseases in EU/EEA countries has not been previously undertaken. Methods: A scoping review of the literature on POCT in EU/ EEA countries as at November 2019, and a survey of key stakeholders. Results: 350 relevant articles were identified and 54 survey responses from 26 EU/EEA countries were analysed. POCT is available for a range of infectious diseases and in all countries responding to the survey (for at least one disease). POCT is commonly available for influenza, HIV/AIDS, Legionnaires' disease and malaria, where it is used in at least half of EU/EEA countries. While POCT has the potential to support many improvements to clinical care of infectious diseases (e.g., faster diagnosis, more appropriate use of antimicrobials), the results suggest POCT is infrequently used to support public health functions (e.g., disease surveillance and reporting). Conclusion: Although POCT is in use to some extent in all EU/EEA countries, the full benefits of POCT in wider public health functions have yet to be realised. Further research on barriers and facilitators to implementation is warranted.

Demonstrating trustworthiness when collecting and sharing genomic data: public views across 22 countries.

BACKGROUND: Public trust is central to the collection of genomic and health data and the sustainability of genomic research. To merit trust, those involved in collecting and sharing data need to demonstrate they are trustworthy. However, it is unclear what measures are most likely to demonstrate this. METHODS: We analyse the 'Your DNA, Your Say' online survey of public perspectives on genomic data sharing including responses from 36,268 individuals across 22 low-, middle- and high-income countries, gathered in 15 languages. We examine how participants perceived the relative value of measures to demonstrate the trustworthiness of those using donated DNA and/or medical information. We examine between-country variation and present a consolidated ranking of measures. RESULTS: Providing transparent information about who will benefit from data access was the most important measure to increase trust, endorsed by more than 50% of participants across 20 of 22 countries. It was followed by the option to withdraw data and transparency about who is using data and why. Variation was found for the importance of measures, notably information about sanctions for misuse of data-endorsed by 5% in India but almost 60% in Japan. A clustering analysis suggests alignment between some countries in the assessment of specific measures, such as the UK and Canada, Spain and Mexico and Portugal and Brazil. China and Russia are less closely aligned with other countries in terms of the value of the measures presented. CONCLUSIONS: Our findings highlight the importance of transparency about data use and about the goals and potential benefits associated with data sharing, including to whom such benefits accrue. They show that members of the public value knowing what benefits accrue from the use of data. The study highlights the importance of locally sensitive measures to increase trust as genomic data sharing continues globally.

Alcohol dependence and heavy episodic drinking are associated with different levels of risk of death or repeat emergency service attendance after a suicide attempt.

BACKGROUND: Alcohol use is a multidimensional risk factor for suicidal behaviour. However, suicide prevention strategies often take 'one-size-fits-all' approaches to alcohol use, reflecting an evidence base built on unidimensional measures. Latent Class Analysis can use a range of measures to differentiate distinct patterns of alcohol using behaviour and their associated risks. METHODS: We analysed Electronic Health Record data from 650 suicidal adults detained for up to 36 h using police powers (Section 136 of the Mental Health Act 1983, amended 2007) to facilitate psychiatric assessment at a Health-Based Place of Safety, a dedicated emergency psychiatric care centre in London, UK. We conducted a Latent Class Analysis of alcohol using behaviours at first detention, and used multivariable logistic regression to estimate the association of each identified latent class with subsequent death or recontact with emergency psychiatric care over a median follow-up of 490 days, adjusting for sex, age and past-year psychiatric diagnosis. RESULTS: Three classes of alcohol use were identified: low risk drinkers, heavy episodic drinkers and dependent drinkers. The dependent drinking class had twice the odds of death or recontact with emergency psychiatric care as the low risk drinking class (OR 2.32, 95 %CI 1.62-3.32, p < 0.001). Conversely, the heavy episodic drinking class was associated with lower odds of death or recontact than the low risk drinking class (OR 0.66, 95 %CI 0.53-0.81, p < 0.001). CONCLUSIONS: The risk of adverse outcomes after a suicide attempt are not uniform for different alcohol use classes. Clinical assessment and suicide prevention efforts should be tailored accordingly.

The association of acute alcohol use and dynamic suicide risk with variation in onward care after psychiatric crisis.

INTRODUCTION: Despite the association of alcohol use with recurrent suicidal acts, individuals attempting suicide after drinking alcohol face barriers accessing crisis care following emergency assessment, demonstrated by higher odds of inpatient admission for those whose suicide attempt did not feature alcohol. This disparity may be due to suicidality dissipating more rapidly after a suicide attempt involving alcohol. We investigated the effect of acute alcohol use and ongoing suicidality on onward care decisions after emergency assessment. METHODS: We analysed electronic health records of 650 suicidal adults detained under Section 136 of the Mental Health Act (1983, amended 2007) for up to 36 h at a London psychiatric emergency care centre. We used logistic regression to estimate the association of acute alcohol use and ongoing suicidality (including their interaction) with admission to psychiatric hospital. RESULTS: Fifteen percent of previously intoxicated detainees expressed suicidal intent at detention end, compared to 24% of detainees who had not used alcohol prior to detention. Compared to those who were not previously intoxicated and not suicidal at detention end, acute alcohol use was associated with reduced odds of admission amongst those no longer suicidal (AOR 0.4, 95% CI 0.2, 0.6). Where suicidality persisted, odds of admission rose; however, the magnitude of increase when in combination with prior alcohol use (AOR 3.6, 95% CI 1.9, 7.1) was under half that of when alcohol was not involved (AOR 8.2, 95% CI 3.5, 19.1). DISCUSSION AND CONCLUSIONS: Acute alcohol use is associated with transient suicidality, but this only partially accounts for disparities in care following suicide attempts.

Global Public Perceptions of Genomic Data Sharing: What Shapes the Willingness to Donate DNA and Health Data?

Analyzing genomic data across populations is central to understanding the role of genetic factors in health and disease. Successful data sharing relies on public support, which requires attention to whether people around the world are willing to donate their data that are then subsequently shared with others for research. However, studies of such public perceptions are geographically limited and do not enable comparison. This paper presents results from a very large public survey on attitudes toward genomic data sharing. Data from 36,268 individuals across 22 countries (gathered in 15 languages) are presented. In general, publics across the world do not appear to be aware of, nor familiar with, the concepts of DNA, genetics, and genomics. Willingness to donate one's DNA and health data for research is relatively low, and trust in the process of data's being shared with multiple users (e.g., doctors, researchers, governments) is also low. Participants were most willing to donate DNA or health information for research when the recipient was specified as a medical doctor and least willing to donate when the recipient was a for-profit researcher. Those who were familiar with genetics and who were trusting of the users asking for data were more likely to be willing to donate. However, less than half of participants trusted more than one potential user of data, although this varied across countries. Genetic information was not uniformly seen as different from other forms of health information, but there was an association between seeing genetic information as special in some way compared to other health data and increased willingness to donate. The global perspective provided by our "Your DNA, Your Say" study is valuable for informing the development of international policy and practice for sharing genomic data. It highlights that the research community not only needs to be worthy of trust by the public, but also urgent steps need to be taken to authentically communicate why genomic research is necessary and how data donation, and subsequent sharing, is integral to this.

Incidence and treatment costs of severe bacterial infections among people who inject heroin: A cohort study in South London, England.

BACKGROUND: People who inject drugs often get bacterial infections. Few longitudinal studies have reported the incidence and treatment costs of these infections. METHODS: For a cohort of 2335 people who inject heroin entering treatment for drug dependence between 2006 and 2017 in London, England, we reported the rates of hospitalisation or death with primary causes of cutaneous abscess, cellulitis, phlebitis, septicaemia, osteomyelitis, septic arthritis, endocarditis, or necrotising fasciitis. We compared these rates to the general population. We also used NHS reference costs to calculate the cost of admissions. RESULTS: During a median of 8.0 years of follow-up, 24 % of patients (570/2335) had a severe bacterial infection, most commonly presenting with cutaneous abscesses or cellulitis. Bacterial infections accounted for 13 % of all hospital admissions. The rate was 73 per 1000 person-years (95 % CI 69-77); 50 times the general population, and the rate remained high throughout follow-up. The rate of severe bacterial infections for women was 1.50 (95 % CI 1.32-1.69) times the rate for men. The mean cost per admission was £4980, and we estimate that the annual cost of hospital treatment for people who inject heroin in London is £4.5 million. CONCLUSIONS: People who inject heroin have extreme and long-term risk of severe bacterial infections.

Members of the public in the USA, UK, Canada and Australia expressing genetic exceptionalism say they are more willing to donate genomic data.

Public acceptance is critical for sharing of genomic data at scale. This paper examines how acceptance of data sharing pertains to the perceived similarities and differences between DNA and other forms of personal data. It explores the perceptions of representative publics from the USA, Canada, the UK and Australia (n = 8967) towards the donation of DNA and health data. Fifty-two percent of this public held 'exceptionalist' views about genetics (i.e., believed DNA is different or 'special' compared to other types of medical information). This group was more likely to be familiar with or have had personal experience with genomics and to perceive DNA information as having personal as well as clinical and scientific value. Those with personal experience with genetics and genetic exceptionalist views were nearly six times more likely to be willing to donate their anonymous DNA and medical information for research than other respondents. Perceived harms from re-identification did not appear to dissuade publics from being willing to participate in research. The interplay between exceptionalist views about genetics and the personal, scientific and clinical value attributed to data would be a valuable focus for future research.

Frequency of health-care utilization by adults who use illicit drugs: a systematic review and meta-analysis.

AIMS: To summarize evidence on the frequency and predictors of health-care utilization among people who use illicit drugs. DESIGN: Systematic search of MEDLINE, EMBASE and PsychINFO for observational studies reporting health-care utilization published between 1 January 2000 and 3 December 2018. We conducted narrative synthesis and meta-analysis following a registered protocol (identifier: CRD42017076525). SETTING AND PARTICIPANTS: People who use heroin, powder cocaine, crack cocaine, methamphetamine, amphetamine, ecstasy/3,4-methyl​enedioxy​methamphetamine (MDMA), cannabis, hallucinogens or novel psychoactive substances; have a diagnosis of 'substance use disorder'; or use drug treatment services. MEASUREMENTS: Primary outcomes were the cumulative incidence (risk) and rate of care episodes in three settings: primary care, hospital admissions (in-patient) and emergency department (ED). FINDINGS: Ninety-two studies were included, 84% from North America and Australia. Most studies focused on people using heroin, methamphetamine or crack cocaine, or who had a diagnosis of drug dependence. We were able to conduct a meta-analysis of rates across 25 studies reporting ED episodes and 25 reporting hospital admissions, finding pooled rates of 151 [95% confidence interval (CI) = 114-201] and 41 (95% CI = 30-57) per 100 person-years, respectively; on average 4.8 and 7.1 times more often than the general population. Heterogeneity was very high and was not explained by drugs used, country of study, recruitment setting or demographic characteristics. Predictors of health-care utilization were consistent across studies and included unstable housing, drug injection and mental health problems. Opioid substitution therapy was consistently associated with reduced ED presentation and hospital admission. There was minimal research on health-care utilization by people using ecstasy/MDMA, powder cocaine, hallucinogens or novel psychoactive substances. CONCLUSIONS: People who use illicit drugs are admitted to emergency department or hospital several times more often than the general population.

Efficient Reuse of Natural Language Processing Models for Phenotype-Mention Identification in Free-text Electronic Medical Records: A Phenotype Embedding Approach.

BACKGROUND: Much effort has been put into the use of automated approaches, such as natural language processing (NLP), to mine or extract data from free-text medical records in order to construct comprehensive patient profiles for delivering better health care. Reusing NLP models in new settings, however, remains cumbersome, as it requires validation and retraining on new data iteratively to achieve convergent results. OBJECTIVE: The aim of this work is to minimize the effort involved in reusing NLP models on free-text medical records. METHODS: We formally define and analyze the model adaptation problem in phenotype-mention identification tasks. We identify "duplicate waste" and "imbalance waste," which collectively impede efficient model reuse. We propose a phenotype embedding-based approach to minimize these sources of waste without the need for labelled data from new settings. RESULTS: We conduct experiments on data from a large mental health registry to reuse NLP models in four phenotype-mention identification tasks. The proposed approach can choose the best model for a new task, identifying up to 76% waste (duplicate waste), that is, phenotype mentions without the need for validation and model retraining and with very good performance (93%-97% accuracy). It can also provide guidance for validating and retraining the selected model for novel language patterns in new tasks, saving around 80% waste (imbalance waste), that is, the effort required in "blind" model-adaptation approaches. CONCLUSIONS: Adapting pretrained NLP models for new tasks can be more efficient and effective if the language pattern landscapes of old settings and new settings can be made explicit and comparable. Our experiments show that the phenotype-mention embedding approach is an effective way to model language patterns for phenotype-mention identification tasks and that its use can guide efficient NLP model reuse.

Causes of hospital admission and mortality among 6683 people who use heroin: A cohort study comparing relative and absolute risks.

BACKGROUND: Mortality in high-risk groups such as people who use illicit drugs is often expressed in relative terms such as standardised ratios. These measures are highest for diseases that are rare in the general population, such as hepatitis C, and may understate the importance of common long-term conditions. POPULATION: 6683 people in community-based treatment for heroin dependence between 2006 and 2017 in London, England, linked to national hospital and mortality databases with 55,683 years of follow-up. METHOD: Age- and sex-specific mortality and hospital admission rates in the general population of London were used to calculate the number of expected events. We compared standardised ratios (relative risk) to excess deaths and admissions (absolute risk) across ICD-10 chapters and subcategories. RESULTS: Drug-related diseases had the highest relative risks, with a standardised mortality ratio (SMR) of 48 (95% CI 42-54) and standardised admission ratio (SAR) of 293 (95% CI 282-304). By contrast, other diseases had an SMR of 4.4 (95% CI 4.0-4.9) and an SAR of 3.15 (95% CI 3.11-3.19). However, the majority of the 621 excess deaths (95% CI 569-676) were not drug-related (361; 58%). The largest groups were liver disease (75 excess deaths) and COPD (45). Similarly, 80% (11,790) of the 14,668 excess admissions (95% CI 14,382-14,957) were not drug-related. The largest groups were skin infections (1073 excess admissions), alcohol (1060), COPD (812) and head injury (612). CONCLUSIONS: Although relative risks of drug-related diseases are very high, most excess morbidity and mortality in this cohort was caused by common long-term conditions.

Trust in genomic data sharing among members of the general public in the UK, USA, Canada and Australia.

Trust may be important in shaping public attitudes to genetics and intentions to participate in genomics research and big data initiatives. As such, we examined trust in data sharing among the general public. A cross-sectional online survey collected responses from representative publics in the USA, Canada, UK and Australia (n = 8967). Participants were most likely to trust their medical doctor and less likely to trust other entities named. Company researchers were least likely to be trusted. Low, Variable and High Trust classes were defined using latent class analysis. Members of the High Trust class were more likely to be under 50 years, male, with children, hold religious beliefs, have personal experience of genetics and be from the USA. They were most likely to be willing to donate their genomic and health data for clinical and research uses. The Low Trust class were less reassured than other respondents by laws preventing exploitation of donated information. Variation in trust, its relation to areas of concern about the use of genomic data and potential of legislation are considered. These findings have relevance for efforts to expand genomic medicine and data sharing beyond those with personal experience of genetics or research participants.