Search

Background and Hypothesis Previous studies show that people with severe mental illness (SMI) are at higher risk of COVID-19 mortality, however limited evidence exists regarding risk postvaccination. We investigated COVID-19 mortality among people with schizophrenia and other SMIs before, during and after the UK vaccine roll-out. Study Design Using the Greater Manchester (GM) Care Record to access routinely collected health data linked with death records, we plotted COVID-19 mortality rates over time in GM residents with schizophrenia/psychosis, bipolar disorder (BD), and/or recurrent major depressive disorder (MDD) from February 2020 to September 2021. Multivariable logistic regression was used to compare mortality risk (risk ratios; RRs) between people with SMI (N = 193 435) and age–sex matched controls (N = 773 734), adjusted for sociodemographic factors, preexisting comorbidities, and vaccination status. Study Results Mortality risks were significantly higher among people with SMI compared with matched controls, particularly among people with schizophrenia/psychosis (RR 3.18, CI 2.94–3.44) and/or BD (RR 2.69, CI 2.16–3.34). In adjusted models, the relative risk of COVID-19 mortality decreased, though remained significantly higher than matched controls for people with schizophrenia (RR 1.61, CI 1.45–1.79) and BD (RR 1.92, CI 1.47–2.50), but not recurrent MDD (RR 1.08, CI 0.99–1.17). People with SMI continued to show higher mortality rate ratios relative to controls throughout 2021, during vaccination roll-out. Conclusions People with SMI, notably schizophrenia and BD, were at greater risk of COVID-19 mortality compared to matched controls. Despite population vaccination efforts that have prioritized people with SMI, disparities still remain in COVID-19 mortality for people with SMI.

IntroductionThis study considers the prediction of the time until two survival outcomes have both occurred. We compared a variety of analytical methods motivated by a typical clinical problem of multimorbidity prognosis.MethodsWe considered five methods: product (multiply marginal risks), dual-outcome (directly model the time until both events occur), multistate models (msm), and a range of copula and frailty models. We assessed calibration and discrimination under a variety of simulated data scenarios, varying outcome prevalence, and the amount of residual correlation. The simulation focused on model misspecification and statistical power. Using data from the Clinical Practice Research Datalink, we compared model performance when predicting the risk of cardiovascular disease and type 2 diabetes both occurring.ResultsDiscrimination was similar for all methods. The product method was poorly calibrated in the presence of residual correlation. The msm and dual-outcome models were the most robust to model misspecification but suffered a drop in performance at small sample sizes due to overfitting, which the copula and frailty model were less susceptible to. The copula and frailty model's performance were highly dependent on the underlying data structure. In the clinical example, the product method was poorly calibrated when adjusting for 8 major cardiovascular risk factors.DiscussionWe recommend the dual-outcome method for predicting the risk of two survival outcomes both occurring. It was the most robust to model misspecification, although was also the most prone to overfitting. The clinical example motivates the use of the methods considered in this study.

Background/Aims Hydroxychloroquine (HCQ) is an antimalarial drug commonly used in autoimmune conditions such as systemic lupus erythematous (SLE) and rheumatoid arthritis (RA). In SLE, it has been shown to significantly improve survival and morbidity, but is known to be rarely associated with sight-threatening retinopathy. Data from the US suggested HCQ retinopathy was significantly more common than previously thought (up to 7%), causing the Royal College of Ophthalmology to recommend increased screening in 2018. Natural language processing (NLP) is a machine learning technique which purports to be able to identify patients with a specific diagnosis or taking specific mediations from unstructured data such as clinic letters. This study aimed to understand whether NLP could be used to (i) identify patients taking HCQ and (ii) determine rates of HCQ retinopathy from clinic letters in a UK tertiary Rheumatology centre. Methods The study was set across the Kellgren Centre for Rheumatology (regional tertiary lupus centre) and the Manchester Eye Hospital. Analysis was performed using CogStack/MedCat, an open source self-supervised NLP framework which extracts SNOMED coded items from unstructured electronic health data such as clinic letters. All clinical documents 2012-2022 were identified for patients attending clinics coded as HCQ retinopathy screening. These were processed through Cogstack/MedCat in two stages. Firstly, it identified all patients exposed to HCQ according to their clinic letters. Subsequently, MedCat searched all clinical documents for those patients for references to HCQ retinopathy, HCQ retinopathy screening or retinopathy. Diabetic retinopathy was used to sense-check for other forms of retinopathy. Finally, manual validation of the HCQ retinopathy positive cases was performed. Results From an original 202,704 documents, HCQ exposure was identified in 3904 unique patients. These patients had 74,377 associated clinical documents, from which the NLP process indicated 2815 referenced HCQ retinopathy or retinopathy alone (including screening). From clinic codes 1299 individual patients (33.3% of the HCQ exposed population) were confirmed to have undergone screening; 1094 (84.2%) were female, mean age 54.8 years at the time of screening. Of those screened, NLP highlighted 568 references to HCQ retinopathy from 507 patients, within which 10 (0.77%) were confirmed to have HCQ retinopathy by manual validation. Conclusion The study has demonstrated that NLP is a promising technique to quickly identify subsets of patients (e.g. those taking HCQ) from large numbers of outpatient clinic letters. Further training of the algorithm may improve its ability to identify rare but important clinical outcomes. In a tertiary SLE centre, roughly one third of patients on HCQ were screened for HCQ retinopathy and 0.77% confirmed HCQ retinopathy. This is much lower than data from the previous studies in the US, but is in keeping with more recent publications in UK data, reflected in recent amendments to screening guidelines. Disclosure Y. Mushtaq: None. G. Tilston: None. K.L. Hyrich: None. A. Barton: None. N. Peek: None. B. Parker: None. D. Griffiths: None. J.H. Humphreys: None.

Background/Aims Rheumatic and musculoskeletal diseases (RMDs) are some of the most common indications for prescribed opioids. It is unclear how opioid prescribing has changed in the UK for RMDs, especially during the COVID-19 pandemic with limited healthcare access and cancelled elective-surgical interventions, which could impact prescribing in either direction. We aimed to investigate trends in opioid prescribing in RMDs and assess the impact of the pandemic in the UK. Methods Adult patients with rheumatoid arthritis (RA), psoriatic arthritis (PsA), axial spondyloarthritis (AxSpA), systemic lupus erythematosus (SLE), osteoarthritis (OA) and fibromyalgia with opioid prescriptions between 01/Jan/2006–31/Aug/2021 without prior cancer in the UK Clinical Practice Research Datalink (CPRD) were included. We calculated age- and gender-standardised yearly rates of people with opioid prescriptions between 2006–2021, and identified change points in trends by checking whether the rate of change of standardised rates crossed zero. For people with opioid prescriptions, monthly measures of mean morphine milligram equivalents (MME)/day were calculated between 2006–2021. To assess the impact of the pandemic, we fitted regression models to the monthly number of people with opioid prescriptions between Jan/2015–Aug/2021. The time coefficient reflects the trend pre-pandemic and the interaction term coefficient represents the change in the trend during the pandemic. Results We included 1,313,519 patients: 36,932 with RA, 12,649 with PsA, 6,811 with AxSpA, 6,423 with SLE, 1,255,999 with OA, and 66,944 with fibromyalgia. People with opioid prescriptions increased from 2006 to 2018 for OA, to 2019 for RA, AxSpA and SLE, to 2020 for PsA, and to 2021 for fibromyalgia, and all plateaued/decreased afterwards. OA patients on opioids increased from 466.8/10,000 persons in 2006 to a peak of 703.0 in 2018, followed by a decline to 575.3 in 2021. From 2006 to 2021, there was a 4.5-fold increase in fibromyalgia opioid users (17.7 vs.78.5/10,000 persons). In this period, MME/day increased for all RMDs, with the highest for fibromyalgia (≥35). During COVID-19 lockdowns, RA, PsA and fibromyalgia showed significant changes in the trend of people with opioid prescriptions. With a decreasing trend for RA (-0.001,95%CI=-0.002,-0.001) and a decreasing-to-flat curve for PsA (0.0010,95%CI=0.0006,0.0015) pre-pandemic until Feb/2020, the trends changed by -0.005 (95%CI=-0.008,-0.002) for RA and -0.003 (95%CI=-0.006,-0.0003) for PsA, leading to steeper decreasing trends during the pandemic (Mar/2020–Aug/2021). Fibromyalgia, conversely, had an increasing trend (0.009,95%CI=0.008,0.009) pre-pandemic, and this trend started decreasing by -0.009 (95%CI=-0.011,-0.006) during the pandemic. Conclusion The plateauing/decreasing trend of people with opioid prescriptions in RMDs after 2018 may reflect the efforts to tackle the rising opioid prescribing in UK primary care. Of all RMDs, fibromyalgia patients had the highest MME/day throughout the study period. COVID-19 lockdowns contribute to fewer people on opioids for most RMDs, reassuring there was no sudden increase in opioid prescribing during the pandemic. Disclosure Y. Huang: None. D.A. Jenkins: None. B.B. Yimer: None. J. Benitez-Aurioles: None. N. Peek: None. M. Lunt: None. W.G. Dixon: Consultancies; WGD has received consultancy fees from Google unrelated to this work. M. Jani: None.

Background/Aims There has been a sharp rise in the use of opioids for non-cancer pain globally. Despite increased awareness of adverse effects, they remain commonly prescribed in the UK. Clinical prediction models offer the possibility of assessing individual risk for a given outcome allowing better allocation of resources towards those at risk. Machine learning (ML) approaches can address nonlinear relationships and complex interactions between variables and are increasingly used to develop these models. Our objective is to develop, validate, and compare the performance of three clinical prediction models based on regression and ML, which leverage primary care data to estimate the risk of opioid-related death in patients prescribed opioids for non-cancer pain. Methods Patients ≥18 years old without prior cancer who were prescribed any opioid between 01/01/2006 and 31/12/2017 were identified in the Clinical Practice Research Datalink (CPRD). Only new opioid users were included. Index date was date of first prescription, with censoring at withdrawal from the CPRD or after not having an opioid prescription for two years. Baseline data were extracted from each patient’s records, including demographic information, comorbidities, concomitant medications, and the opioid type being prescribed, collecting 49 candidate predictors. These were used to train three competing risk models: a Fine&Gray regression model with LASSO regularisation, a survival random forest (RF), and a neural network (DeepHit). The outcome was opioid-related mortality and other cause mortality the competing event, defined using a curated ICD-10 codelist. Predictive performance of the models, such as the area under the receiver characteristic operator curve (AUC-ROC), were calculated using 5-fold cross validation. Results We included a total of 1,029,681 patients, of which 1,240 experienced an opioid-related death, and 52,833 experienced a competing death. The Fine&Gray, RF and DeepHit models achieved average AUC-ROC values of 0.83(95% CI: 0.81-0.85), 0.78(0.77-0.79) and 0.81(0.80-0.82) respectively. At the optimum risk cut point, as per Youden’s index, the models achieved sensitivities of 0.82(0.78-0.85), 0.75(0.67-0.82) and 0.80(0.78-0.83), and specificities of 0.78(0.73-0.82), 0.75(0.68-0.83) and 0.78(0.75-0.80) when predicting 12-month risk, respectively. In the Fine&Gray model, factors associated with an increased risk were history of substance use disorder (hazards ratio [HR]: 3.40, 95% CI:3.12-3.69) and alcohol abuse (HR:3.07, 95% CI:2.93-3.22). COPD (HR:1.53, 95% CI:1.48-1.58) and moderate liver disease (HR:1.31, 95% CI:0.99-1.63) were the comorbidities associated with highest risk. Morphine (HR:2.39, 95% CI:2.08-2.69) and oxycodone (HR:1.10, 95% CI:1.00-1.20) at initiation and concomitant gabapentinoids (HR:1.99, 95% CI:1.80-2.18) and benzodiazepines (HR:1.30, 95% CI:1.24-1.36) were associated with an increased risk. HR for rheumatologic diseases was 1.08 (95% CI:1.01-1.14). Conclusion The Fine&Gray and DeepHit models exhibited comparable discriminative performance. Substance abuse, lung and liver comorbidities, morphine or oxycodone at initiation and co-prescription of gabapentinoids and benzodiazepines, were some of the factors associated with a higher risk of opioid-related mortality. Disclosure J. Benitez-Aurioles: None. D. Jenkins: None. Y. Huang: None. C. Ramirez Medina: None. N. Peek: None. M. Jani: None.

UNSTRUCTURED International deployment of remote monitoring and virtual care (RMVC) technologies would efficiently harness their positive impact on outcomes. Since Canada and the UK have similar populations, healthcare systems, and digital health landscapes, transferring digital health innovations between them should be relatively straightforward. Yet, examples of successful attempts are scarce. In a workshop, we identified six differences that may complicate RMVC transfer between Canada and the UK, and provide recommendations for addressing. These key differences include: 1) minority groups, (2) physical geography, (3) clinical pathways, (4) value propositions, (5) governmental priorities and support for digital innovation, (6) regulatory pathways. We detail four broad recommendations to plan for sustainability including: the need to formally consider how highlighted country-specific recommendations may impact RMVC and contingency planning to overcome challenges; the need to map which pathways are available as an innovator to support cross-country transfer; the need to report on and apply learnings from regulatory barriers and facilitators so that everyone may benefit; and explore existing guidance to successfully transfer digital health solutions while developing further guidance (e.g., extending the NASSS framework for cross-country transfer). Finally, we present an ecosystem readiness checklist. Considering these recommendations will contribute to successful international deployment and increased positive impact of RMVC technologies.

Objectives: No clear guidance exists on handling missing data at each stage of developing, validating and implementing a clinical prediction model (CPM). We aimed to review the approaches to handling missing data that underly the CPMs currently recommended for use in UK healthcare.Study Design and Setting: A descriptive cross-sectional meta-epidemiological study aiming to identify CPMs recommended by the National Institute for Health and Care Excellence (NICE), which summarized how missing data is handled across their pipelines.Results: A total of 23 CPMs were included through “sampling strategy.” Six missing data strategies were identified: complete case analysis (CCA), multiple imputation, imputation of mean values, k-nearest neighbours imputation, using an additional category for missingness, considering missing values as risk-factor-absent. 52% of the development articles and 48% of the validation articles did not report how missing data were handled. CCA was the most common approach used for development (40%) and validation (44%). At implementation, 57% of the CPMs required complete data entry, whilst 43% allowed missing values. Three CPMs had consistent paths in their pipelines.Conclusion: A broad variety of methods for handling missing data underly the CPMs currently recommended for use in UK healthcare. Missing data handling strategies were generally inconsistent. Better quality assurance of CPMs needs greater clarity and consistency in handling of missing data.

ObjectivesHaving multiple long term health conditions (MLTCs), also known as multimorbidity, is becoming increasingly common as populations age. Understanding how clusters of diseases are likely to lead to other diseases and the effect of multimorbidity on healthcare resource use (HRU) will be of great importance as this trend continues. ApproachGraph-based approaches, also called network analysis in the literature, have been used previously to study multimorbidity. The use of hypergraphs, which are generalisations of graphs where edges can connect to any number of nodes, and their application to the problem of understanding multimorbidity will be discussed. Analysis using hypergraphs was carried out using a population-scale cohort of people in the Secure Anonymised Information Linkage (SAIL) Databank to find the diseases and disease sets which are most important based on a measure of prevalence and measures of healthcare resource utilisation in secondary care. ResultsThe most important sets of diseases based on the centrality of a hypergraph weighted by a measure of prevalence featured hypertension, and the most important was hypertension and diabetes. The most important sets of diseases based on the centrality of a hypergraph weighted by a measure of unplanned inpatient HRU were arrhythmia, heart failure and hypertension while for a measure of outpatient HRU the most important set of diseases was diabetes and hypertension. ConclusionHypergraphs are very flexible and general mathematical objects and there is still a great deal of development that can be done to make them more useful in epidemiological settings and beyond.

People living with rheumatic and musculoskeletal diseases (RMDs) are frequently prescribed opioids to manage pain and represent a large subpopulation of patients at increased risk of long-term opioid use.1 A proportion of patients with RMDs newly prescribed an opioid will transition to long-term opioid use, which can be associated with opioid dependence, abuse and harm.2 Most studies define long-term opioid use as ≥90 days of use, but, the definitions in the literature vary considerably in frequency of use and duration.3 Variations in definitions can lead to challenges in characterising the scale of long-term opioid use, the opioid crisis and targeting the appropriate groups with the greatest benefit of interventions to improve safety.3 4 A more stringent definition would mean some might have missed opportunities to review, titrate or stop opioid therapy. A broader definition, however, would flag patients prescribed opioids intermittently or for acute episodes, and who might be at low risk of harm. Regardless, contemporary estimates of the scale of long-term opioid use remain unquantified in patients with RMDs. Our aim was to assess the proportion of patients transitioning to long-term opioid use in patients newly initiated on an opioid across six RMD conditions using varying definitions from the literature. This study included patients aged ≥18 years registered in the Clinical Practice Research Datalink (CPRD) with a diagnosis of rheumatoid arthritis (RA), psoriatic arthritis (PsA), axial spondyloarthritis (AxSpA), systemic lupus erythematosus (SLE), osteoarthritis (OA) and fibromyalgia and without prior cancer. CPRD is a database of anonymised UK primary care electronic health records representative of the national …

Background: In clinical prediction modelling, missing data can occur at any stage of the model pipeline; development, validation or deployment. Multiple imputation is often recommended yet challenging to apply at deployment; for example, the outcome cannot be in the imputation model, as recommended under multiple imputation. Regression imputation uses a fitted model to impute the predicted value of missing predictors from observed data, and could offer a pragmatic alternative at deployment. Moreover, the use of missing indicators has been proposed to handle informative missingness, but it is currently unknown how well this method performs in the context of clinical prediction models. Methods : We simulated data under various missing data mechanisms to compare the predictive performance of clinical prediction models developed using both imputation methods. We consider deployment scenarios where missing data is permitted or prohibited, imputation models that use or omit the outcome, and clinical prediction models that include or omit missing indicators. We assume that the missingness mechanism remains constant across the model pipeline. We also apply the proposed strategies to critical care data. Results : With complete data available at deployment, our findings were in line with existing recommendations; that the outcome should be used to impute development data when using multiple imputation and omitted under regression imputation. When missingness is allowed at deployment, omitting the outcome from the imputation model at the development was preferred. Missing indicators improved model performance in many cases but can be harmful under outcome-dependent missingness. Conclusion: We provide evidence that commonly taught principles of handling missing data via multiple imputation may not apply to clinical prediction models, particularly when data can be missing at deployment. We observed comparable predictive performance under multiple imputation and regression imputation. The performance of the missing data handling method must be evaluated on a study-by-study basis, and the most appropriate strategy for handling missing data at development should consider whether missing data are allowed at deployment. Some guidance is provided.

Background The effectiveness and cost-effectiveness of biologic therapies for psoriasis are significantly compromised by variable treatment responses. Thus, more precise management of psoriasis is needed. Objectives We aim to identify subgroups of psoriasis patients treated with biologic therapies, based on changes in their disease activity over time, that may better inform patient management. Methods Here we apply latent class mixed modelling, to identify trajectory-based patient subgroups from longitudinal, routine clinical data on disease severity, as measured by the Psoriasis Area and Severity Index (PASI), from 3546 patients in the British Association of Dermatologists Biologics and Immunomodulators Register, as well as in an independent cohort of 2889 subjects pooled across four clinical trials. Results We discovered four, discrete classes of global response trajectories, each characterised in terms of time to response, size of effect and relapse. Each class was associated with differing clinical characteristics, e.g. Body Mass Index, baseline PASI, and prevalence of different manifestations. Our results were verified in a second cohort of clinical trials subjects, where similar trajectories following initiation of biologic therapy were identified. Further, we found differential associations of the genetic marker HLA-C*06:02 between our registry-identified trajectories. Conclusion These subgroups, defined by change in disease over time, may be indicative of distinct endotypes driven by different biological mechanisms and may help inform the management of patients with psoriasis. Future work will aim to further delineate these mechanisms by extensively characterising the subgroups with additional molecular and pharmacological data.

Wearable and mobile technology provides new opportunities to manage health conditions remotely and unobtrusively. For example, healthcare providers can repeatedly sample a person's condition to monitor progression of symptoms and intervene if necessary. There is usually a utility-tolerability trade-off between collecting information at sufficient frequencies and quantities to be useful, and over-burdening the user or the underlying technology, particularly when active input is required from the user. Selecting the next sampling time adaptively using previous responses, so that people are only sampled at high frequency when necessary, can help to manage this trade-off. We present a novel approach to adaptive sampling using clustered continuous-time hidden Markov models. The model predicts, at any given sampling time, the probability of moving to an "alert" state, and the next sample time is scheduled when this probability has exceeded a given threshold. The clusters, each representing a distinct sub-model, allow heterogeneity in states and state transitions. The work is illustrated using longitudinal mentalhealth symptom data in 49 people collected using ClinTouch, a mobile app designed to monitor people with a diagnosis of schizophrenia. Using these data, we show how the adaptive sampling scheme behaves under different model parameters and risk thresholds, and how the average sampling can be substantially reduced whilst maintaining a high sampling frequency during high-risk periods.

BackgroundThe COVID-19 pandemic has placed an unprecedented demand on global healthcare resources. Data on whole population healthcare utilisation (HCU) across both primary and secondary care during the COVID-19 pandemic are lacking.AimTo describe primary and secondary HCU stratified by LTCs and deprivation, during the first 19 months of COVID-19 pandemic across a large urban area in the United Kingdom.MethodsObservational HCU data between 30th-December-2019 and 1st-August-2021 were extracted from the Greater Manchester Care Record. Primary care HCU (incident prescribing and recording of healthcare information) and secondary care HCU (planned and unplanned admissions) were assessed.ResultsThe first national lockdown was associated with reductions in all primary HCU measures, ranging from 24.7% (24.0% to 25.5%) for incident prescribing to 84.9% (84.2% to 85.5%) for cholesterol monitoring. Secondary HCU also dropped significantly for planned (47% (42.9% to 51.5%)) and unplanned admissions (35.0% (28.3% to 41.6%)). Only secondary care had significant reductions in HCU during the second national lockdown. Primary HCU measures had not recovered to pre-pandemic levels by the end of the study. The secondary admission rate ratio between multi-morbid patients and those without LTCs increased during the first lockdown by a factor of 2.4 (2.0 to 2.9;pConclusionMajor changes in primary and secondary HCU have been observed during the COVID-19 pandemic. Secondary HCU reduced more in those without LTCs and the ratio of utilisation between the most and least deprived increased for the majority of HCU measures. Overall primary HCU measures and secondary care HCU for some LTC groups had not returned to pre-pandemic levels by the end of the study.

Health research increasingly requires effective ways to identify existing datasets and assess their suitability for research. We sought to test whether researchers could use an existing metadata catalogue to assess the suitability of datasets for addressing specified research questions. Five datasets were described in the National Institute for Health Research Health Informatics Collaborative metadata catalogue, and for each dataset five associated research questions were formulated, some of which were answerable with the dataset while others were not. Thirteen researchers each assessed whether the ten questions associated with two randomly selected datasets were answerable with the described datasets. After removing instances where participants misunderstood the question or lacked subject matter knowledge to make the assessment, we found that 87 out of 109 assessments (80%) were correct. Participants particularly struggled with one dataset which consisted of EHR data. The most common reason for incorrect assessments was the inability to find the relevant information in the metadata catalogue.

Health research increasingly requires effective ways to identify existing datasets and assess their suitability for research. We sought to test whether researchers could use an existing metadata catalogue to assess the suitability of datasets for addressing specified research questions. Five datasets were described in the National Institute for Health Research Health Informatics Collaborative metadata catalogue, and for each dataset five associated research questions were formulated, some of which were answerable with the dataset while others were not. Thirteen researchers each assessed whether the ten questions associated with two randomly selected datasets were answerable with the described datasets. After removing instances where participants misunderstood the question or lacked subject matter knowledge to make the assessment, we found that 87 out of 109 assessments (80%) were correct. Participants particularly struggled with one dataset which consisted of EHR data. The most common reason for incorrect assessments was the inability to find the relevant information in the metadata catalogue.

The fourth industrial revolution is based on cyber-physical systems and the connectivity of devices. It is currently unclear what the consequences are for patient safety as existing digital health technologies become ubiquitous with increasing pace and interact in unforeseen ways. In this paper, we describe the output from a workshop focused on identifying the patient safety challenges associated with emerging digital health technologies. We discuss six challenges identified in the workshop and present recommendations to address the patient safety concerns posed by them. A key implication of considering the challenges and opportunities for Patient Safety Informatics is the interdisciplinary contribution required to study digital health technologies within their embedded context. The principles underlying our recommendations are those of proactive and systems approaches that relate the social, technical and regulatory facets underpinning patient safety informatics theory and practice.

Aims Recent studies showed that exercise-based cardiac rehabilitation (ECR) programmes are often not personalized to individual patient characteristics according to latest recommendations. This study investigates whether a computerized decision support (CDS) system based on latest recommendations and guidelines can improve personalization of ECR prescriptions. Pseudo-randomized intervention study. Methods and results Among participating Dutch cardiac rehabilitation centres, ECR programme characteristics of consecutive patients were recorded during 1 year. CDS was used during a randomly assigned 4-month period within this year. Primary outcome was concordance to latest recommendations in three phases (before, during, and after CDS) for 12 ECR programme characteristics. Secondary outcome was variation in training characteristics. We recruited ten Dutch CR centres and enrolled 2258 patients to the study. Overall concordance of ECR prescriptions was 59.9% in Phase 1, 62.1% in Phase 2 (P = 0.82), and 59.9% in Phase 3 (P = 0.56). Concordance varied from 0.0% to 99.9% for the 12 ECR characteristics. There was significant between-centre variation for most training characteristics in Phases 1 and 2. In Phase 3, there was only a significant variation for aerobic and resistance training intensity (P = 0.01), aerobic training volume (P Conclusion CDS did not substantially improve concordance with ECR prescriptions. As aerobic training volume was often lower than recommended and reduced during the study, a lack of institutional resources might be an important barrier in personalizing ECR prescriptions.

Aims: To examine the risk of gastrointestinal (GI) bleeding, major bleeding, stroke and systemic embolism associated with prescribing nonsteroidal anti-inflammatory drugs (NSAIDs) to adults receiving oral anticoagulant (OAC) therapy. Methods: We conducted a population-based cohort study in adults receiving OAC therapy using linked primary care (Clinical Practice Research Datalink GOLD) and hospital (Hospital Episodes Statistics) electronic health records. We used cause-specific Cox regression models with time-dependent NSAID treatment in a propensity score matched population to estimate the increased risk of GI bleeding, stroke, major bleeding and systemic embolism associated with NSAID use. Results: The matched cohort contained 3177 patients with OAC therapy alone and 3177 with at least 1 concomitant NSAID prescription. Compared with OAC therapy alone, concomitant prescription of NSAIDs with OACs was associated with increased risk of GI bleeding (hazard ratio [HR] 3.01, 95% confidence interval [CI] 1.63 to 5.55), stroke (HR 2.71, 95% CI 1.48 to 4.96) and major bleeding (HR 2.77, 95% CI 1.84 to 4.19). The association with systemic embolism did not reach statistical significance (HR 3.02, 95% CI 0.82 to 11.07). Sensitivity analyses indicated that the results were robust to changes in exclusion criteria and the choice of potential confounding variables. Conclusion: When OACs are coprescribed with NSAIDs, the risk of adverse bleeding events increases and, simultaneously, the protective effect of OACs to prevent strokes reduces. There is a need for interventions that reduce hazardous prescribing of NSAIDs in people receiving OAC therapy.

INTRODUCTION: Research is ongoing to increase our understanding of how much a previous diagnosis of type 2 diabetes mellitus (T2DM) affects someone's risk of becoming seriously unwell following a COVID-19 infection. In this study we set out to determine the relative likelihood of death following COVID-19 infection in people with T2DM when compared to those without T2DM. This was conducted as an urban population study and based in the UK.METHODS: Analysis of electronic health record data was performed relating to people living in the Greater Manchester conurbation (population 2.82 million) who had a recorded diagnosis of T2DM and subsequent COVID-19 confirmed infection. Each individual with T2DM (n = 13,807) was matched with three COVID-19-infected non-diabetes controls (n = 39,583). Data were extracted from the Greater Manchester Care Record (GMCR) database for the period 1 January 2020 to 30 June 2021. Social disadvantage was assessed through Townsend scores. Death rates were compared in people with T2DM to their respective non-diabetes controls; potential predictive factors influencing the relative likelihood of admission were ascertained using univariable and multivariable logistic regression.RESULTS: For individuals with T2DM, their mortality rate after a COVID-19 positive test was 7.7% vs 6.0% in matched controls; the relative risk (RR) of death was 1.28. From univariate analysis performed within the group of individuals with T2DM, the likelihood of death following a COVID-19 recorded infection was lower in people taking metformin, a sodium-glucose cotransporter 2 inhibitor (SGLT2i) or a glucagon-like peptide 1 (GLP-1) agonist. Estimated glomerular filtration rate (eGFR) and hypertension were associated with increased mortality and had odds ratios of 0.96 (95% confidence interval 0.96-0.97) and 1.92 (95% confidence interval 1.68-2.20), respectively. Likelihood of death following a COVID-19 infection was also higher in those people with a diagnosis of chronic obstructive pulmonary disease (COPD) or severe enduring mental illness but not with asthma, and in people taking aspirin/clopidogrel/insulin. Smoking in people with T2DM significantly increased mortality rate (odds ratio of 1.46; 95% confidence interval 1.29-1.65). In a combined analysis of patients with T2DM and controls, multiple regression modelling indicated that the factors independently relating to a higher likelihood of death (accounting for 26% of variance) were T2DM, age, male gender and social deprivation (higher Townsend score).CONCLUSION: Following confirmed infection with COVID-19 a number of factors are associated with mortality in individuals with T2DM. Prescription of metformin, SGLT2is or GLP-1 agonists and non-smoking status appeared to be associated with a reduced the risk of death for people with T2DM. Age, male sex and social disadvantage are associated with an increased risk of death.

BACKGROUND Online consultations (OCs) allow patients to contact their care provider online, and are promoted as a way to address increasing workload and decreasing workforce capacity in primary care. Globally, OCs have been rolled out rapidly due to policy initiatives and COVID-19, though there is a lack of evidence regarding how their design and implementation influence care outcomes. OBJECTIVE Informed by existing theories, synthesise research on: 1) outcomes of OCs in primary care; 2) how these are influenced by OC system design and implementation. METHODS We searched databases from 2010 to November 2021. We included quantitative and qualitative studies of real-world OC use in primary care. Quantitative data were transformed into qualitative themes. For objective 1 we used thematic synthesis informed by the Institute of Medicine’s domains of healthcare quality. For objective 2 we used Framework Analysis informed by the NASSS framework and Realistic Evaluation. Critical appraisal was conducted using the Mixed Methods Appraisal Tool and strength of evidence judged using GRADE-CERQual. RESULTS We synthesised 62 studies in nine countries covering 30 OC systems, 13 of which used Artificial Intelligence (AI). Twenty-six were published in 2020 onwards, 11 were post-COVID-19. There was no quantitative evidence for negative impacts of OCs on patient safety, and qualitative studies suggested varied perceptions of safety. Some participants believed OCs improved safety, particularly when patients could describe their queries using unstructured free-text. Staff workload decreased when sufficient resources were allocated to implement OCs, and patients used them for simple problems or could describe their queries using free-text. Staff workload increased when OCs were not integrated with other software or organisational workflows, and patients used them for complex queries. OC systems that required patients to describe their queries using multiple choice questionnaires (MCQs) increased workload for patients and staff. Health costs reduced when patients used OCs for simple queries, and increased when used for complex ones. Patients using OCs were more likely to be female, younger, native speakers, with higher socioeconomic status than those not using OCs. However, OCs increased primary care access for patients with mental health conditions, verbal communication difficulties, and barriers to attending in-person appointments. Access also increased by providing a timely response to patients’ queries. Patient satisfaction increased when using OCs due to better primary care access, though could decrease when using MCQ formats. CONCLUSIONS This is the first theoretically-informed synthesis of research on OCs in primary care, and includes studies conducted during COVID-19. It contributes new knowledge that in addition to producing positive outcomes such as increased access, OCs can also have negative outcomes such as increased workload. Negative outcomes can be mitigated by appropriate OC system design (e.g. free-text format), incorporating advanced technologies (e.g. AI), and integration into technical and organisational workflows (e.g. timely responses).

Background Online consultations (OCs) allow patients to contact their care providers on the web. Worldwide, OCs have been rolled out in primary care rapidly owing to policy initiatives and COVID-19. There is a lack of evidence regarding how OC design and implementation influence care quality. Objective We aimed to synthesize research on the impacts of OCs on primary care quality, and how these are influenced by system design and implementation. Methods We searched databases from January 2010 to February 2022. We included quantitative and qualitative studies of real-world OC use in primary care. Quantitative data were transformed into qualitative themes. We used thematic synthesis informed by the Institute of Medicine domains of health care quality, and framework analysis informed by the nonadoption, abandonment, scale-up, spread, and sustainability framework. Strength of evidence was judged using the GRADE-CERQual approach. Results We synthesized 63 studies from 9 countries covering 31 OC systems, 14 (22%) of which used artificial intelligence; 41% (26/63) of studies were published from 2020 onward, and 17% (11/63) were published after the COVID-19 pandemic. There was no quantitative evidence for negative impacts of OCs on patient safety, and qualitative studies suggested varied perceptions of their safety. Some participants believed OCs improved safety, particularly when patients could describe their queries using free text. Staff workload decreased when sufficient resources were allocated to implement OCs and patients used them for simple problems or could describe their queries using free text. Staff workload increased when OCs were not integrated with other software or organizational workflows and patients used them for complex queries. OC systems that required patients to describe their queries using multiple-choice questionnaires increased workload for patients and staff. Health costs decreased when patients used OCs for simple queries and increased when patients used them for complex queries. Patients using OCs were more likely to be female, younger, and native speakers, with higher socioeconomic status. OCs increased primary care access for patients with mental health conditions, verbal communication difficulties, and barriers to attending in-person appointments. Access also increased by providing a timely response to patients’ queries. Patient satisfaction increased when using OCs owing to better primary care access, although it decreased when using multiple-choice questionnaire formats. Conclusions This is the first theoretically informed synthesis of research on OCs in primary care and includes studies conducted during the COVID-19 pandemic. It contributes new knowledge that, in addition to having positive impacts on care quality such as increased access, OCs also have negative impacts such as increased workload. Negative impacts can be mitigated through appropriate OC system design (eg, free text format), incorporation of advanced technologies (eg, artificial intelligence), and integration into technical infrastructure (eg, software) and organizational workflows (eg, timely responses). Trial Registration PROSPERO CRD42020191802; https://tinyurl.com/2p84ezjy

BackgroundOnline consultations (OCs) allow patients to contact their care provider online, and have been promoted as a way to address increasing workload and decreasing workforce capacity in primary care. Globally, OCs have been rolled out rapidly due to policy initiatives and the COVID-19 pandemic, though there is a lack of evidence regarding how their design and implementation influence care outcomes.ObjectiveInformed by existing theories, synthesise quantitative and qualitative research on: 1) outcomes of OCs in primary care; 2) how these are influenced by OC system design and implementation.MethodsWe searched Ovid Medline, Embase, Web of Science, Scopus, NTIS, HMIC, and ZETOC from 2010 to November 2021. We included quantitative and qualitative studies of real-world OC use in primary care, written in English, and published 2010 onwards. Quantitative data were transformed into qualitative themes. For objective 1 we used thematic synthesis informed by the Institute of Medicine’s domains of healthcare quality. For objective 2 we used Framework Analysis informed by the NASSS framework and Realistic Evaluation. Critical appraisal was conducted using the Mixed Methods Appraisal Tool and strength of evidence judged using GRADE-CERQual.ResultsWe synthesised 62 studies (quantitative n=32, qualitative n=12, mixed methods n=18) in nine countries covering 30 unique OC systems, 13 of which used Artificial Intelligence (AI). Twenty-six were published in 2020 onwards, and 11 were post-COVID-19. There was no quantitative evidence for negative impacts of OCs on patient safety, and qualitative studies suggested perceptions of OC safety varied. Some participants believed OCs improved safety, particularly when patients could describe their queries using unstructured free-text. Staff workload decreased when sufficient resources were allocated to implement OCs, and patients used them for simple problems or could describe their queries using free-text. Staff workload increased when OCs were not integrated with other software or organisational workflows, and patients used them for complex queries. OC systems that required patients to describe their queries using multiple choice questionnaires (MCQs) increased workload for both them and staff. Health costs were reduced when patients used OCs for simple queries, and increased when used for complex ones. Patients using OCs were more likely to be female, younger, native speakers, with higher socioeconomic status than those not using OCs. However, OCs increased primary care access for patients with mental health conditions, verbal communication difficulties, and barriers to attending in-person appointments. Access also increased by providing a timely response to patients’ queries. Patient satisfaction increased when using OCs due to better primary care access, though could decrease when using MCQ formats.ConclusionsThis is the first theoretically-informed synthesis of research on OCs in primary care, and includes studies conducted during COVID-19. It contributes new knowledge that in addition to producing positive outcomes such as increased access and patient satisfaction, they can also have negative outcomes such as increased workload and costs. These negative outcomes can be mitigated by appropriate OC system design (e.g. free-text format), incorporating advanced technologies (e.g. AI), and integration into technical and organisational workflows (e.g. timely responses).Study protocolPROSPERO (CRD42020191802).

Objectives: This study aimed to demonstrate how to estimate the value of health gain after patients with a multisystem disease achieve a condition-specific composite response endpoint. Methods: Data from patients treated in routine practice with an exemplar multisystem disease (systemic lupus erythematosus) were extracted from a national register (British Isles Lupus Assessment Group Biologics Register). Two bespoke composite response endpoints (Major Clinical Response and Improvement) were developed in advance of this study. Difference-in-differences regression compared health utility values (3-level version of EQ-5D; UK tariff) over 6 months for responders and nonresponders. Bootstrapped regression estimated the incremental quality-adjusted life-years (QALYs), probability of QALY gain after achieving the response criteria, and population monetary benefit of response. Results: Within the sample (n = 171), 18.2% achieved Major Clinical Response and 49.1% achieved Improvement at 6 months. Incremental health utility values were 0.0923 for Major Clinical Response and 0.0454 for Improvement. Expected incremental QALY gain at 6 months was 0.020 for Major Clinical Response and 0.012 for Improvement. Probability of QALY gain after achieving the response criteria was 77.6% for Major Clinical Response and 72.7% for Improvement. Population monetary benefit of response was £1 106 458 for Major Clinical Response and £649 134 for Improvement. Conclusions: Bespoke composite response endpoints are becoming more common to measure treatment response for multisystem diseases in trials and observational studies. Health technology assessment agencies face a growing challenge to establish whether these endpoints correspond with improved health gain. Health utility values can generate this evidence to enhance the usefulness of composite response endpoints for health technology assessment, decision making, and economic evaluation.

Missing data are much studied in epidemiology and statistics. Theoretical development and application of methods for handling missing data have mostly been conducted in the context of prospective research data and with a goal of description or causal explanation. However, it is now common to build predictive models using routinely collected data, where missing patterns may convey important information, and one might take a pragmatic approach to optimizing prediction. Therefore, different methods to handle missing data may be preferred. Furthermore, an underappreciated issue in prediction modeling is that the missing data method used in model development may not match the method used when a model is deployed. This may lead to overoptimistic assessments of model performance. For prediction, particularly with routinely collected data, methods for handling missing data that incorporate information within the missingness pattern should be explored and further developed. Where missing data methods differ between model development and model deployment, the implications of this must be explicitly evaluated. The trade-off between building a prediction model that is causally principled, and building a prediction model that maximizes the use of all available information, should be carefully considered and will depend on the intended use of the model.

INTRODUCTION: Since early 2020 the whole world has been challenged by the SARS-CoV-2 virus and the associated global pandemic (Covid-19). People with diabetes are particularly at high risk of becoming seriously unwell after contracting this virus.METHODS: This population-based study included people living in the Greater Manchester conurbation who had a recorded diagnosis of type 1 diabetes mellitus (T1DM) or type 2 diabetes mellitus (T2DM) and subsequent Covid-19 infection. Each individual with T1DM (n = 862) or T2DM (n = 13,225) was matched with three Covid-19-infected non-diabetes controls.RESULTS: For individuals with T1DM, hospital admission rate in the first 28 days after a positive Covid-19 test was 10% vs 4.7% in age/gender-matched controls [relative risk (RR) 2.1]. For individuals with T2DM, hospital admission rate after a positive Covid-19 test was 16.3% vs 11.6% in age/gender-matched controls (RR 1.4). The average Townsend score was higher in T2DM (1.8) vs matched controls (0.4), with a higher proportion of people with T2DM observed in the top two quintiles of greatest disadvantage (p CONCLUSION: In a UK population we have confirmed a significantly higher likelihood of admission in people with diabetes following Covid-19 infection. A number of factors mediate that increased likelihood of hospital admission. For T2DM, the majority of factors related to increased admission rate are common to the general population but more prevalent in T2DM. There was a protective effect of metformin in people with T2DM.

IntroductionSudden cardiac death (SCD) is the leading cause of death in patients with myocardial infarction (MI) and can be prevented by the implantable cardioverter defibrillator (ICD). Currently, risk stratification for SCD and decision on ICD implantation are based solely on impaired left ventricular ejection fraction (LVEF). However, this strategy leads to over- and under-treatment of patients because LVEF alone is insufficient for accurate assessment of prognosis. Thus, there is a need for better risk stratification. This is the study protocol for developing and validating a prediction model for risk of SCD in patients with prior MI.Methods and AnalysisThe EU funded PROFID project will analyse 23 datasets from Europe, Israel and the US (∼225,000 observations). The datasets include patients with prior MI or ischemic cardiomyopathy with reduced LVEFEthics and disseminationLocal ethical approval was obtained. The final model will be disseminated through scientific publications and a web-calculator. Statistical code will be published through open-source repositories.

Background/Aims: The assessment of disease severity in people with diabetes enables identification of patients in need for intensive and targeted therapies, and could help optimise the allocation of healthcare resources. We aimed to identify and appraise measures that have used clinical data to establish systems for grading Type 2 diabetes severity. Methods: MEDLINE, Embase and PubMed were searched between inception-June 2018. Data from studies on diabetes-specific severity grading measures derived by medical data in adult patients with Type 2 diabetes were extracted. Studies reporting only on other diabetes forms were excluded. Detailed characteristics of the eligible severity measures, including design, severity domains, and the association between severity and health-related outcomes, were identified following independent screening. Results: Eighteen severity measures in 17 papers including 32,314 participants were included. Measures’ designs were: diabetes severity index (N = 8 measures); severity categories (N = 7 measures); complications’ count (N = 2 measures); or severity checklist (N = 1 measure). The included domains differed greatly across the identified measures. Glucose or HbA1c levels and/or diabetes-related complications were included in 89% of the measures. Two severity measures were validated in a separate study population. Higher levels of diabetes severity were associated with poorer cognitive function, higher healthcare expenses, and significantly increased risks for hospitalisation and death. Conclusion: Increasing diabetes severity is associated with greater risks for adverse outcomes. Despite that our findings demonstrate the suitability of health records to assess diabetes severity, the clinical uptake of existing measures is poor. The need to advance diabetes severity assessment is important to develop actionable measures that would help benchmark efficient clinical services.

Background: Healing of non-traumatic skin ulcers is often suboptimal. Prognostic tools that identify people at high risk of delayed healing within the context of routine ulcer assessments may improve this, but robust evidence on which factors to include is lacking. Therefore, we scoped the literature to identify which potentially prognostic factors may warrant future systematic reviews and meta-analyses.Methods: We conducted electronic searches in MEDLINE and Embase to identify studies in English published between 1997-2017 that tested the association between healing of the three most common non-traumatic skin ulcers encountered by healthcare professionals (venous leg, diabetic foot and pressure ulcers) and patient characteristics, ulcer characteristics and results from clinical investigations. Results: We included 42 studies that investigated factors that may be associated with healing of venous leg ulcers (n=17), diabetic foot ulcers (n=15), and pressure ulcers (n=10). Across ulcer types, ulcer characteristics were most commonly reported as potential prognostic factors for healing (n=37), including the size of the ulcer area (n=29) and ulcer duration at first assessment (n=16). Thirty-five studies investigated the prognostic value of patient characteristics (n=35), including age (n=31), sex (n=30), diabetes (n=22), smoking status (n=15) and history of deep vein thrombosis (n=13). Twenty three studies reported results from clinical investigations as potential prognostic factors, with the majority regarding vessel quality.Conclusion: Age, sex, diabetes, smoking status, history of deep vein thrombosis, ulcer area, and ulcer duration at time of first assessment warrant a systematic review and meta-analysis to quantify their prognostic value for delayed ulcer healing.

ObjectiveColorectal cancer is a common cause of death and morbidity. A significant amount of data are routinely collected during patient treatment, but they are not generally available for research. The National Institute for Health Research Health Informatics Collaborative in the UK is developing infrastructure to enable routinely collected data to be used for collaborative, cross-centre research. This paper presents an overview of the process for collating colorectal cancer data and explores the potential of using this data source.MethodsClinical data were collected from three pilot Trusts, standardised and collated. Not all data were collected in a readily extractable format for research. Natural language processing (NLP) was used to extract relevant information from pseudonymised imaging and histopathology reports. Combining data from many sources allowed reconstruction of longitudinal histories for each patient that could be presented graphically.ResultsThree pilot Trusts submitted data, covering 12 903 patients with a diagnosis of colorectal cancer since 2012, with NLP implemented for 4150 patients. Timelines showing individual patient longitudinal history can be grouped into common treatment patterns, visually presenting clusters and outliers for analysis. Difficulties and gaps in data sources have been identified and addressed.DiscussionAlgorithms for analysing routinely collected data from a wide range of sites and sources have been developed and refined to provide a rich data set that will be used to better understand the natural history, treatment variation and optimal management of colorectal cancer.ConclusionThe data set has great potential to facilitate research into colorectal cancer.

UNSTRUCTURED The fourth industrial revolution is based on cyber-physical systems and the connectivity of devices. It is currently unclear what the consequences are for patient safety as existing digital health technologies become ubiquitous with increasing pace and interact in unforeseen ways. In this paper, we describe the output from a workshop focused on identifying the patient-safety challenges associated with emerging digital health technologies. We present a definition of Patient Safety Informatics that was informed by the workshop and existing literature, discuss the challenges identified in the workshop, and present recommendations to address the patient-safety concerns posed by them. A key implication of considering the challenges and opportunities for Patient Safety Informatics is the interdisciplinary contribution required to study digital health technologies within their embedded context. The principles underlying our recommendations are those of proactive and systems approaches that relate the social, technical and regulatory facets underpinning patient safety informatics theory and practice.

This work was supported by Health Data Research UK (HDR-9006; CFC0110) and the Medical Research Council (MR/S027750/1). Health Data Research UK is funded by: UK Medical Research Council; Engineering and Physical Sciences Research Council; Economic and Social Research Council; National Institute for Health Research (England); Chief Scientist Office of the Scottish Government Health and Social Care Directorates; Health and Social Care Research and Development Division (Welsh Government); Public Health Agency (Northern Ireland); British Heart Foundation and Wellcome Trust. Introduction Multimorbidity is widely recognised as the presence of two or more concurrent long-term conditions, yet remains a poorly understood global issue despite increasing in prevalence. We have created the Wales Multimorbidity e-Cohort (WMC) to provide an accessible research ready data asset to further the understanding of multimorbidity. Our objectives are to create a platform to support research which would help to understand prevalence, trajectories and determinants in multimorbidity, characterise clusters that lead to highest burden on individuals and healthcare services, and evaluate and provide new multimorbidity phenotypes and algorithms to the National Health Service and research communities to support prevention, healthcare planning and the management of individuals with multimorbidity. Methods and analysis The WMC has been created and derived from multisourced demographic, administrative and electronic health record data relating to the Welsh population in the Secure Anonymised Information Linkage (SAIL) Databank. The WMC consists of 2.9 million people alive and living in Wales on the 1 January 2000 with follow-up until 31 December 2019, Welsh residency break or death. Published comorbidity indices and phenotype code lists will be used to measure and conceptualise multimorbidity. Study outcomes will include: (1) a description of multimorbidity using published data phenotype algorithms/ontologies, (2) investigation of the associations between baseline demographic factors and multimorbidity, (3) identification of temporal trajectories of clusters of conditions and multimorbidity and (4) investigation of multimorbidity clusters with poor outcomes such as mortality and high healthcare service utilisation. Ethics and dissemination The SAIL Databank independent Information Governance Review Panel has approved this study (SAIL Project: 0911). Study findings will be presented to policy groups, public meetings, national and international conferences, and published in peer-reviewed journals. Publisher PDF

ObjectiveHow health researchers find secondary data to analyse is unclear. We sought to describe the approaches that UK organisations take to help researchers find data and to assess the findability of health data that are available for research.MethodsWe surveyed established organisations about how they make data findable. We derived measures of findability based on the first element of the FAIR principles (Findable, Accessible, Interoperable, Reproducible). We applied these to 13 UK health datasets and measured their findability via two major internet search engines in 2018 and repeated in 2021.ResultsAmong 12 survey respondents, 11 indicated that they made metadata publicly available. Respondents said internet presence was important for findability, but that this needed improvement. In 2018, 8 out of 13 datasets were listed in the top 100 search results of 10 searches repeated on both search engines, while the remaining 5 were found one click away from those search results. In 2021, this had reduced to seven datasets directly listed and one dataset one click away. In 2021, Google Dataset Search had become available, which listed 3 of the 13 datasets within the top 100 search results.DiscussionMeasuring findability via online search engines is one method for evaluating efforts to improve findability. Findability could perhaps be improved with catalogues that have greater inclusion of datasets, field-level metadata and persistent identifiers.ConclusionUK organisations recognised the importance of the internet for finding data for research. However, health datasets available for research were no more findable in 2021 than in 2018.

BACKGROUND Drug prescriptions are often recorded in free-text clinical narratives; making this information available in a structured form is important to support many health-related tasks. Although several natural language processing (NLP) methods have been proposed to extract such information, many challenges remain. OBJECTIVE This study evaluates the feasibility of using NLP and deep learning approaches for extracting and linking drug names and associated attributes identified in clinical free-text notes and presents an extensive error analysis of different methods. This study initiated with the participation in the 2018 National NLP Clinical Challenges (n2c2) shared task on adverse drug events and medication extraction. METHODS The proposed system (DrugEx) consists of a named entity recognizer (NER) to identify drugs and associated attributes and a relation extraction (RE) method to identify the relations between them. For NER, we explored deep learning-based approaches (ie, bidirectional long-short term memory with conditional random fields [BiLSTM-CRFs]) with various embeddings (ie, word embedding, character embedding [CE], and semantic-feature embedding) to investigate how different embeddings influence the performance. A rule-based method was implemented for RE and compared with a context-aware long-short term memory (LSTM) model. The methods were trained and evaluated using the 2018 n2c2 shared task data. RESULTS The experiments showed that the best model (BiLSTM-CRFs with pretrained word embeddings [PWE] and CE) achieved lenient micro F-scores of 0.921 for NER, 0.927 for RE, and 0.855 for the end-to-end system. NER, which relies on the pretrained word and semantic embeddings, performed better on most individual entity types, but NER with PWE and CE had the highest classification efficiency among the proposed approaches. Extracting relations using the rule-based method achieved higher accuracy than the context-aware LSTM for most relations. Interestingly, the LSTM model performed notably better in the reason-drug relations, the most challenging relation type. CONCLUSIONS The proposed end-to-end system achieved encouraging results and demonstrated the feasibility of using deep learning methods to extract medication information from free-text data.

Background: Significant evidence suggests that the cholesterol-lowering statins can affect cognitive function and reduce the risk for Alzheimer's disease (AD) and dementia. These potential effects may be constrained by specific combinations of an individual's sex and apolipoprotein E (APOE) genotype.Methods: Here we examine data from 252,327 UK Biobank participants, aged 55 or over, and compare the effects of statin use in males and females. We assessed difference in statin treatments taking a matched cohort approach, and identified key stratifiers using regression models and conditional inference trees. Using statistical modeling, we further evaluated the effect of statins on survival, cognitive decline over time, and on AD prevalence.Results: We identified that in the selected population, males were older, had a higher level of education, better cognitive scores, higher incidence of cardiovascular and metabolic diseases, and a higher rate of statin use. We observed that males and those participants with an APOE ε4-positive genotype had higher probabilities of being treated with statins; while participants with an AD diagnosis had slightly lower probabilities. We found that use of statins was not significantly associated with overall higher rates of survival. However, when considering the interaction of statin use with sex, the results suggest higher survival rates in males treated with statins. Finally, examination of cognitive function indicates a potential beneficial effect of statins that is selective for APOE ε4-positive genotypes.Discussion: Our evaluation of the aging population in a large cohort from the UK Biobank confirms sex and APOE genotype as fundamental risk stratifiers for AD and cognitive function, furthermore it extends them to the specific area of statin use, clarifying their specific interactions with treatments.

BACKGROUND: Reducing the harms associated with acute kidney injury (AKI) requires addressing a wide range of patient safety issues, including polypharmacy and transitions of care, particularly for vulnerable patient groups. Computerised audit and feedback can transform the way healthcare organisations measure, analyse and learn from quality and safety data across different care settings, potentially improving patient safety.OBJECTIVE: To implement and evaluate an audit and feedback dashboard targeting AKI to improve patient safety, focusing on factors affecting a range of user characteristics in primary care.METHODS: We performed a mixed methods study in three stages. Semi-structured interviews were initially performed with both primary (n = 10) and secondary care (n = 5) staff to gather user requirements for six quality indicators extracted from national guidance on post-discharge AKI care. Modified indicators were implemented in the Performance Improvement plaN GeneratoR (PINGR) audit and feedback dashboard for six months, across 45 general practices in Salford. Primary care professionals were then interviewed again (n = 7) and completed usability questionnaires. This was triangulated with an interrupted time series analysis on indicator performance, alongside software usage statistics.RESULTS: Improvements were observed for the indicators for medication review (+9.01 %; 95 % Confidence Interval (CI), +6.95 % to +11.06 %) and blood pressure measurement (+5.20 %; 95 % CI + 3.61 % to +6.78 %). Variable performance and engagement were observed for other indicators including AKI coding (+0.39 %; 95 % CI -1.88 % to +2.65 %), serum creatinine (-3.40 %; 95 % CI -7.66 % to +0.85 %), proteinuria (-1.08 %; 95 % CI -1.47 % to +0.32 %) and providing patient information (+0.16 %; 95 % CI -0.41 % to +0.73 %). A key facilitator to engagement was the development of 'champions of change', achieved through a raised awareness of high-risk patients, guidelines, inconsistencies in coding practice and evidence for quality and safety performance. Barriers related to the specificity and perceived achievability of indicators, and limitations in resources.CONCLUSION: In a six-month, quasi-experimental evaluation of an electronic audit and feedback dashboard targeting AKI, we found improvements for two out of six quality indicators. While information technology can facilitate improvements in patient safety, further allocation of protected staff time and investment into shared learning are needed to realise those improvements in practice.

BACKGROUND: To date, research on the indirect impact of the COVID-19 pandemic on the health of the population and the health-care system is scarce. We aimed to investigate the indirect effect of the COVID-19 pandemic on general practice health-care usage, and the subsequent diagnoses of common physical and mental health conditions in a deprived UK population.METHODS: We did a retrospective cohort study using routinely collected primary care data that was recorded in the Salford Integrated Record between Jan 1, 2010, and May 31, 2020. We extracted the weekly number of clinical codes entered into patient records overall, and for six high-level categories: symptoms and observations, diagnoses, prescriptions, operations and procedures, laboratory tests, and other diagnostic procedures. Negative binomial regression models were applied to monthly counts of first diagnoses of common conditions (common mental health problems, cardiovascular and cerebrovascular disease, type 2 diabetes, and cancer), and corresponding first prescriptions of medications indicative of these conditions. We used these models to predict the expected numbers of first diagnoses and first prescriptions between March 1 and May 31, 2020, which were then compared with the observed numbers for the same time period.FINDINGS: Between March 1 and May 31, 2020, 1073 first diagnoses of common mental health problems were reported compared with 2147 expected cases (95% CI 1821 to 2489) based on preceding years, representing a 50·0% reduction (95% CI 41·1 to 56·9). Compared with expected numbers, 456 fewer diagnoses of circulatory system diseases (43·3% reduction, 95% CI 29·6 to 53·5), and 135 fewer type 2 diabetes diagnoses (49·0% reduction, 23·8 to 63·1) were observed. The number of first prescriptions of associated medications was also lower than expected for the same time period. However, the gap between observed and expected cancer diagnoses (31 fewer; 16·0% reduction, -18·1 to 36·6) during this time period was not statistically significant.INTERPRETATION: In this deprived urban population, diagnoses of common conditions decreased substantially between March and May 2020, suggesting a large number of patients have undiagnosed conditions. A rebound in future workload could be imminent as COVID-19 restrictions ease and patients with undiagnosed conditions or delayed diagnosis present to primary and secondary health-care services. Such services should prioritise the diagnosis and treatment of these patients to mitigate potential indirect harms to protect public health.FUNDING: National Institute of Health Research.

Integrating data from different sources into homogeneous dataset increases the opportunities to study human health. However, disparate data collections are often heterogeneous, which complicates their integration. In this paper, we focus on the issue of content heterogeneity in data integration. Traditional approaches for resolving content heterogeneity map all source datasets to a common data model that includes only shared data items, and thus omit all items that vary between datasets. Based on an example of three datasets in Systemic Lupus Erythematosus, we describe and experimentally evaluate a probabilistic data integration approach which propagates the uncertainty resulting from content heterogeneity into statistical inference, avoiding the need to map to a common data model.

Objective Developing clinical prediction models (CPMs) on data of sufficient sample size is critical to help minimize overfitting. Using prostate cancer as a clinical exemplar, we aimed to investigate to what extent existing CPMs adhere to recent formal sample size criteria, or historic rules of thumb of events per predictor parameter (EPP) ≥ 10. Study Design and Setting A systematic review to identify CPMs related to prostate cancer, which provided enough information to calculate minimum sample size. We compared the reported sample size of each CPM against the traditional 10 EPP rule of thumb and formal sample size criteria. Results About 211 CPMs were included. Three of the studies justified the sample size used, mostly using EPP rules of thumb. Overall, 69% of the CPMs were derived on sample sizes that surpassed the traditional EPP ≥ 10 rule of thumb, but only 48% surpassed recent formal sample size criteria. For most CPMs, the required sample size based on formal criteria was higher than the sample sizes to surpass 10 EPP. Conclusion Few of the CPMs included in this study justified their sample size, with most justifications being based on EPP. This study shows that, in real-world data sets, adhering to the classic EPP rules of thumb is insufficient to adhere to recent formal sample size criteria.

Background: In view of the growth of published papers, there is an increasing need for studies that summarise scientific research. An increasingly common review is a 'Methodology scoping review', which provides a summary of existing analytical methods, techniques and software, proposed or applied in research articles, which address an analytical problem or further an analytical approach. However, guidelines for their design, implementation and reporting are limited. Methods: Drawing on the experiences of the authors, which were consolidated through a series of face-to-face workshops, we summarise the challenges inherent in conducting a methodology scoping review and offer suggestions of best practice to promote future guideline development. Results: We identified three challenges of conducting a methodology scoping review. First, identification of search terms; one cannot usually define the search terms a priori and the language used for a particular method can vary across the literature. Second, the scope of the review requires careful consideration since new methodology is often not described (in full) within abstracts. Third, many new methods are motivated by a specific clinical question, where the methodology may only be documented in supplementary materials. We formulated several recommendations that build upon existing review guidelines. These recommendations ranged from an iterative approach to defining search terms through to screening and data extraction processes. Conclusion: Although methodology scoping reviews are an important aspect of research, there is currently a lack of guidelines to standardise their design, implementation and reporting. We recommend a wider discussion on this topic., 22 pages, 2 tables

OBJECTIVE: To determine whether using a reweighted disease activity score that better reflects joint synovitis, i.e., the 2-component Disease Activity Score in 28 joints (DAS28) (based on swollen joint count and C-reactive protein level), produces more clinically relevant treatment outcome trajectories compared to the standard 4-component DAS28.METHODS: Latent class mixed modeling of response to biologic treatment was applied to 2,991 rheumatoid arthritis (RA) patients in whom treatment with a biologic disease-modifying antirheumatic drug was being initiated within the Biologics in Rheumatoid Arthritis Genetics and Genomics Study Syndicate cohort, using both 4-component and 2-component DAS28 scores as outcome measures. Patient groups with similar trajectories were compared in terms of pretreatment baseline characteristics (including disability and comorbidities) and follow-up characteristics (including antidrug antibody events, adherence to treatments, and blood drug levels). We compared the trajectories obtained using the 4- and 2-component scores to determine which characteristics were better captured by each.RESULTS: Using the 4-component DAS28, we identified 3 trajectory groups, which is consistent with previous findings. We showed that the 4-component DAS28 captures information relating to depression. Using the 2-component DAS28, 7 trajectory groups were identified; among them, distinct groups of nonresponders had a higher incidence of respiratory comorbidities and a higher proportion of antidrug antibody events. We also identified a group of patients for whom the 2-component DAS28 scores remained relatively low; this group included a high percentage of patients who were nonadherent to treatment. This highlights the utility of both the 4- and 2-component DAS28 for monitoring different components of disease activity.CONCLUSION: Here we show that the 2-component modified DAS28 defines important biologic and clinical phenotypes associated with treatment outcome in RA and characterizes important underlying response mechanisms to biologic drugs.

The original article [1] contains an omitted grant acknowledgement and affiliation as relates to the contribution of co-author, Rafael Perera-Salazar. As such, the following two amendments should apply to the original article

Serious Mental Illnesses (SMIs) including schizophrenia and bipolar disorder are long term conditions which place major burdens on health and social care services. Locomotor activity is altered in many cases of SMI, and so in the long term wearable activity trackers could potentially aid in the early detection of SMI relapse, allowing early and targeted intervention. To move towards this goal, in this paper we use accelerometer activity tracking data collected from the UK Biobank to classify people as being either in a self-reported SMI group or an age and gender matched control group. Using an ensemble dense neural network algorithm we exploited hourly and average derived features from the wearable activity data and the created model obtained an accuracy of 91.3%.

Background: Machine learning (ML) attracts many attentions with the hope that it could make use of large, routinely collected datasets and deliver accurate personalised prognosis. The aim of this systematic review is to identify and critically appraise the reporting and developing of ML models for predicting outcomes after stroke. Methods: We searched PubMed and Web of Science from 1990 to March 2019, using previously published search filters for stroke, ML, and prediction models. We focused on structured clinical data and excluded image and text analysis. Findings: Eighteen studies were eligible for inclusion. Most studies reported less than half of the terms in the reporting quality checklist. The most frequently predicted stroke outcomes were mortality (7 studies) and functional independence (5 studies). The most commonly used ML methods were random forests (9 studies), support vector machines (8 studies), decision trees (6 studies), and neural networks (6 studies). The median sample size was 475 (range 70-3184), with a median of 22 predictors (range 4-152) considered. All studies evaluated discrimination with thirteen studies using area under the ROC curve whilst calibration was assessed in three studies. Two studies performed external validation. None of the studies described the final model sufficiently well to reproduce it. Interpretation: Interest in using ML for predicting stroke outcomes is increasing. However, few met basic reporting standards for clinical prediction tools. None of them made their models available in a way which could be used or evaluated. Major improvements in ML study conduct and reporting are needed before meaningfully considered for practice. Funding Statement: CDW, NP, VC, AGR, and WW acknowledge the financial support from the Health Foundation. CDW, AD, VC, and YW acknowledge support from the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) based at Guy’s and St Thomas’ National Health Service (NHS) Foundation Trust and King’s College London, and the NIHR Collaboration for Leadership in Applied Health Research and Care (ARC) South London at King’s College Hospital NHS Foundation Trust. NP acknowledges support from the NIHR Manchester BRC. IJM is funded by the Medical Research Council (MRC), through its Skills Development Fellowship program, fellowship MR/N015185/1. Declaration of Interests: The authors stated: "No conflict of interest to declare." Ethics Approval Statement: This review was registered with PROSPERO (CRD42019127154).

BackgroundTemporal relations between clinical events play an important role in clinical assessment and decision making. Extracting such relations from free text data is a challenging task because it lies on between medical natural language processing, temporal representation and temporal reasoning.ObjectivesTo survey existing methods for extracting temporal relations (TLINKs) between events from clinical free text in English; to establish the state-of-the-art in this field; and to identify outstanding methodological challenges.MethodsA systematic search in PubMed and the DBLP computer science bibliography was conducted for studies published between January 2006 and December 2018. The relevant studies were identified by examining the titles and abstracts. Then, the full text of selected studies was analyzed in depth and information were collected on TLINK tasks, TLINK types, data sources, features selection, methods used, and reported performance.ResultsA total of 2834 publications were identified for title and abstract screening. Of these publications, 51 studies were selected. Thirty-two studies used machine learning approaches, 15 studies used a hybrid approaches, and only four studies used a rule-based approach. The majority of studies use publicly available corpora: THYME (28 studies) and the i2b2 corpus (17 studies).ConclusionThe performance of TLINK extraction methods ranges widely depending on relation types and events (e.g. from 32% to 87% F-score for identifying relations between clinical events and document creation time). A small set of TLINKs (before, after, overlap and contains) has been widely studied with relatively good performance, whereas other types of TLINK (e.g., started by, finished by, precedes) are rarely studied and remain challenging. Machine learning classifiers (such as Support Vector Machine and Conditional Random Fields) and Deep Neural Networks were among the best performing methods for extracting TLINKs, but nearly all the work has been carried out and tested on two publicly available corpora only. The field would benefit from the availability of more publicly available, high-quality, annotated clinical text corpora.