Care Electronic Health Record Data Research Articles

The Disproportionate Impact of Primary Care Disruption and Telehealth Utilization During COVID-19.

The COVID-19 pandemic not only exacerbated existing disparities in health care in general but likely worsened disparities in access to primary care. Our objective was to quantify the nationwide decrease in primary care visits and increase in telehealth utilization during the pandemic and explore whether certain groups of patients were disproportionately affected. We used a geographically diverse primary care electronic health record data set to examine the following 3 outcomes: (1) change in total visit volume, (2) change in in-person visit volume, and (3) the telehealth conversion ratio defined as the number of pandemic telehealth visits divided by the total number of prepandemic visits. We assessed whether these outcomes were associated with patient characteristics including age, gender, race, ethnicity, comorbidities, rurality, and area-level social deprivation. Our primary sample included 1,652,871 patients from 408 practices. During the pandemic we observed decreases of 7% and 17% in total and in-person visit volume and a 10% telehealth conversion ratio. The greatest decreases in visit volume were observed among pediatric patients (-24%), Asian patients (-11%), and those with more comorbidities (-9%). Telehealth usage was greatest among Hispanic or Latino patients (17%) and those living in urban areas (12%). Decreases in primary care visit volume were partially offset by increasing telehealth use for all patients during the COVID-19 pandemic, but the magnitude of these changes varied significantly across all patient characteristics. These variations have implications not only for the long-term consequences of the COVID-19 pandemic, but also for planners seeking to ready the primary care delivery system for any future systematic disruptions.

Risk calculator for incident atrial fibrillation across a range of prediction horizons

The increasing burden of atrial fibrillation (AF) emphasizes the need to identify high-risk individuals for enrolment in clinical trials of AF screening and primary prevention. We aimed to develop prediction models to identify individuals at high-risk of AF across prediction horizons from 6-months to 10-years. We used secondary-care linked primary care electronic health record data from individuals aged ≥30 years without known AF in the UK Clinical Practice Research Datalink-GOLD dataset between January 2, 1998 and November 30, 2018; randomly divided into derivation (80%) and validation (20%) datasets. Models were derived using logistic regression from known AF risk factors for incident AF in prediction periods of 6 months, 1-year, 2-years, 5-years, and 10-years. Performance was evaluated using in the validation dataset with bootstrap validation with 200 samples, and compared against the CHA2DS2-VASc and C2HEST scores. Of 2,081,139 individuals in the cohort (1,664,911 in the development dataset, 416,228 in the validation dataset), the mean age was 49.9 (SD 15.4), 50.7% were women, and 86.7% were white. New cases of AF were 7,386 (0.4%) within 6 months, 15,349 (0.7%) in 1 year, 38,487 (1.8%) in 5 years, and 79,997 (3.8%) by 10 years. Valvular heart disease and heart failure were the strongest predictors, and association of hypertension with AF increased at longer prediction horizons. The optimal risk models incorporated age, sex, ethnicity, and 8 comorbidities. The models demonstrated good-to-excellent discrimination and strong calibration across prediction horizons (AUROC, 95%CI, calibration slope: 6-months, 0.803, 0.789-0.821, 0.952; 1-year, 0.807, 0.794-0.819, 0.962; 2-years, 0.815, 0.807-0.823, 0.973; 5-years, 0.807, 0.803-0.812, 1.000; 10-years 0.780, 0.777-0.784, 1.010), and superior to the CHA2DS2-VASc and C2HEST scores. The models are available as a web-based FIND-AF calculator. The FIND-AF models demonstrate high discrimination and calibration across short- and long-term prediction horizons in 2 million individuals. Their utility to inform trial enrolment and clinical decisions for AF screening and primary prevention requires further study.

Patterns of multimorbidity in primary care electronic health records: A systematic review.

Multimorbidity, the coexistence of multiple chronic conditions in an individual, is a complex phenomenon that is highly prevalent in primary care settings, particularly in older individuals. This systematic review summarises the current evidence on multimorbidity patterns identified in primary care electronic health record (EHR) data. Three databases were searched from inception to April 2022 to identify studies that derived original multimorbidity patterns from primary care EHR data. The quality of the included studies was assessed using a modified version of the Newcastle-Ottawa Quality Assessment Scale. Sixteen studies were included in this systematic review, none of which was of low quality. Most studies were conducted in Spain, and only one study was conducted outside of Europe. The prevalence of multimorbidity (i.e. two or more conditions) ranged from 14.0% to 93.9%. The most common stratification variable in disease clustering models was sex, followed by age and calendar year. Despite significant heterogeneity in clustering methods and disease classification tools, consistent patterns of multimorbidity emerged. Mental health and cardiovascular patterns were identified in all studies, often in combination with diseases of other organ systems (e.g. neurological, endocrine). These findings emphasise the frequent coexistence of physical and mental health conditions in primary care, and provide useful information for the development of targeted preventive and management strategies. Future research should explore mechanisms underlying multimorbidity patterns, prioritise methodological harmonisation to facilitate the comparability of findings, and promote the use of EHR data globally to enhance our understanding of multimorbidity in more diverse populations.

Analysis of primary care electronic health record data of people living with hepatitis B virus: infection and hepatocellular carcinoma risk associated with socio-economic deprivation

ObjectivesWe set out to characterise chronic hepatitis B (CHB) in the primary care population in England and investigate risk factors for progression to hepatocellular carcinoma (HCC). Study designRetrospective cohort study. MethodsWe identified 8039 individuals with CHB in individuals aged ≥18 years between 1999 and 2019 in the English primary care database QResearch. HCC risk factors were investigated using Cox proportional hazards modelling. ResultsMost of those with a record of CHB were males (60%) of non-White ethnicity (>70%), and a high proportion were in the most deprived Townsend deprivation quintile (44%). Among 7029 individuals with longitudinal data, 161 HCC cases occurred. Increased HCC hazards were significantly associated with male sex (adjusted hazards ratio [aHR] 3.17, 95% confidence interval [95% CI] 1.92–5.23), in the fifth deprivation quintile as compared to the third quintile (aHR 1.69, 95% CI 1.01–2.84), with older age (for age groups 56–65 and ≥66 years, compared to 26–35 years, aHRs 2.82 [95% CI 1.45–5.46] and 3.76 [95% CI 1.79–7.9], respectively), Caribbean ethnicity (aHR 3.32, 95% CI 1.43–7.71, compared to White ethnicity), ascites (aHR 3.15, 95% CI 1.30–7.67), cirrhosis (aHR 6.55, 95% CI 4.57–9.38) and peptic ulcer disease (aHR 2.26, 95% CI 1.45–3.51). ConclusionsTargeting interventions and HCC surveillance at vulnerable groups is essential to improve CHB outcomes and to support progress towards international goals for the elimination of hepatitis infection as a public health threat.

Estimating the prevalence of diagnosed Alzheimer disease in England across deprivation groups using electronic health records: a clinical practice research datalink study

ObjectiveEstimate the prevalence of diagnosed Alzheimer’s disease (AD) and early Alzheimer’s disease (eAD) overall and stratified by age, sex and deprivation and combinations thereof in England on 1 January 2020.DesignCross-sectional.SettingPrimary...

Incidence of immune-mediated inflammatory diseases following COVID-19: a matched cohort study in UK primary care

BackgroundSome patients infected with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) go on to experience post-COVID-19 condition or long COVID. Preliminary findings have given rise to the theory that long COVID may be due in part to a deranged immune response. In this study, we assess whether there is an association between SARS-CoV-2 infection and the incidence of immune-mediated inflammatory diseases (IMIDs).MethodsMatched cohort study using primary care electronic health record data from the Clinical Practice Research Datalink Aurum database. The exposed cohort included 458,147 adults aged 18 years and older with a confirmed SARS-CoV-2 infection and no prior diagnosis of IMIDs. They were matched on age, sex, and general practice to 1,818,929 adults with no diagnosis of confirmed or suspected SARS-CoV-2 infection. The primary outcome was a composite of any of the following IMIDs: autoimmune thyroiditis, coeliac disease, inflammatory bowel disease (IBD), myasthenia gravis, pernicious anaemia, psoriasis, rheumatoid arthritis (RA), Sjogren’s syndrome, systemic lupus erythematosus (SLE), type 1 diabetes mellitus (T1DM), and vitiligo. The secondary outcomes were each of these conditions separately. Cox proportional hazard models were used to estimate adjusted hazard ratios (aHR) and 95% confidence intervals (CI) for the primary and secondary outcomes, adjusting for age, sex, ethnic group, smoking status, body mass index, relevant infections, and medications.Results Six hundred and nighty six (0.15%) and 2230 (0.12%) patients in the exposed and unexposed cohort developed an IMID during the follow-up period over 0.29 person-years, giving a crude incidence rate of 4.59 and 3.65 per 1000 person-years, respectively. Patients in the exposed cohort had a 22% increased risk of developing an IMID, compared to the unexposed cohort (aHR 1.22, 95% CI 1.12 to 1.33). The incidence of three IMIDs was significantly associated with SARS-CoV-2 infection. These were T1DM (aHR 1.56, 1.09 to 2.23), IBD (aHR 1.36, 1.18 to 1.56), and psoriasis (1.23, 1.05 to 1.42).ConclusionsSARS-CoV-2 was associated with an increased incidence of IMIDs including T1DM, IBD and psoriasis. However, these findings could be potentially due to ascertainment bias. Further research is needed to replicate these findings in other populations and to measure autoantibody profiles in cohorts of individuals with COVID-19.

Anxiety and Depression in People with Eczema or Psoriasis: A Comparison of Associations in UK Biobank and Linked Primary Care Data.

Previous research has shown associations between eczema and psoriasis and anxiety and depression. We investigated whether associations are consistent across different settings of ascertainment for depression and anxiety, including interview and survey responses from UK Biobank (a large longitudinal cohort recruiting individuals aged 40-69 years between 2006-2010), and linked primary care data, with the aim of drawing more reliable conclusions through triangulation. In cross-sectional studies, we estimated associations between eczema or psoriasis and anxiety or depression, defining anxiety or depression as 1) self-reported previous diagnosis at UK Biobank recruitment interview; 2) PHQ-9/GAD-7 score indicating depression or anxiety from a UK Biobank mental health follow-up survey in 2016; and 3) diagnosis in linked primary care electronic health record data. We analysed 230,047 people with linked Biobank and primary care data. We found poor agreement between the data sources for eczema, psoriasis, anxiety, and depression. Eg, 9474 had a previous eczema diagnosis in primary care data, 4069 self-reported previous eczema diagnosis at the UK biobank interview, and 1536 had eczema in both data sources (for depression 40,455; 13,320; and 9588 respectively). Having eczema or psoriasis (recorded in primary care or baseline interview) was associated with higher odds of anxiety and depression. Eg, the adjusted odds ratio for depression comparing those with eczema to those without was greater than 1 when defining the outcome from 1) the recruitment interview (1.36, 95% confidence interval 1.27-1.45); 2) the follow-up survey (1.24, 1.09-1.39), and 3) primary care records (1.56, 1.50-1.62). Our findings support increased prevalence of mental illness in people with psoriasis and eczema across multiple data sources, which should be considered in planning of mental health services. However, we found poor agreement in disease ascertainment between settings, with implications for data interpretation in electronic health records.

Statin prescribing patterns in patient-centered medical home patients with NAFLD.

Cardiovascular disease is the leading cause of mortality in patients with nonalcoholic fatty liver disease (NAFLD), and statins play a pivotal role in the primary prevention of cardiovascular events. This study investigates statin prescribing in primary care patients with NAFLD to identify opportunities to address cardiovascular disease risk in this cohort. Retrospective cohort study of primary care electronic health record data from 2012-2018. This cohort included 652 patients with radiographic evidence of hepatic steatosis and no evidence of competing chronic liver disease. A statin prescription identified at any time during the study period was the primary outcome. Univariate and multivariable analyses were performed to evaluate the association of clinical signals and comorbidities with statin prescribing. Of the 652 patients in the NAFLD cohort, 56% received a statin prescription during the study period. Elevations in aminotransferases were not associated with statin prescribing (adjusted odds ratio [AOR], 1.17; 95% CI, 0.78-1.76), whereas older patients (AOR, 1.06; 95% CI, 1.05-1.08) and those with diabetes (AOR, 2.61; 95% CI, 1.73-3.92), hypertension (AOR, 2.76; 95% CI, 1.70-4.48), and a BMI greater than or equal to 30 kg/m2 (AOR, 1.49; 95% CI, 1.01-2.22) had higher odds of having a statin prescribed. Of the 288 patients without a statin prescription, 49% had an indication for statin therapy by atherosclerotic cardiovascular disease risk. In total, 16% of included patients did not have lipid panel results during the study period. This study showed no association between NAFLD and statin prescribing, and the findings highlight opportunities to improve primary prevention of cardiovascular disease in these at-risk patients.

UK research data resources based on primary care electronic health records: review and summary for potential users.

The range and scope of electronic health record (EHR) data assets in the UK has recently increased, which has been mainly in response to the COVID-19 pandemic. Summarising and comparing the large primary care resources will help researchers to choose the data resources most suited to their needs. To describe the current landscape of UK EHR databases and considerations of access and use of these resources relevant to researchers. Narrative review of EHR databases in the UK. Information was collected from the Health Data Research Innovation Gateway, publicly available websites and other published data, and from key informants. The eligibility criteria were population-based open-access databases sampling EHRs across the whole population of one or more countries in the UK. Published database characteristics were extracted and summarised, and these were corroborated with resource providers. Results were synthesised narratively. Nine large national primary care EHR data resources were identified and summarised. These resources are enhanced by linkage to other administrative data to a varying extent. Resources are mainly intended to support observational research, although some can support experimental studies. There is considerable overlap of populations covered. While all resources are accessible to bona fide researchers, access mechanisms, costs, timescales, and other considerations vary across databases. Researchers are currently able to access primary care EHR data from several sources. Choice of data resource is likely to be driven by project needs and access considerations. The landscape of data resources based on primary care EHRs in the UK continues to evolve.

Ethnic differences in the indirect effects of the COVID-19 pandemic on clinical monitoring and hospitalisations for non-COVID conditions in England: a population-based, observational cohort study using the OpenSAFELY platform.

The COVID-19 pandemic disrupted healthcare and may have impacted ethnic inequalities in healthcare. We aimed to describe the impact of pandemic-related disruption on ethnic differences in clinical monitoring and hospital admissions for non-COVID conditions in England. In this population-based, observational cohort study we used primary care electronic health record data with linkage to hospital episode statistics data and mortality data within OpenSAFELY, a data analytics platform created, with approval of NHS England, to address urgent COVID-19 research questions. We included adults aged 18 years and over registered with a TPP practice between March 1, 2018, and April 30, 2022. We excluded those with missing age, sex, geographic region, or Index of Multiple Deprivation. We grouped ethnicity (exposure), into five categories: White, Asian, Black, Other, and Mixed. We used interrupted time-series regression to estimate ethnic differences in clinical monitoring frequency (blood pressure and Hba1c measurements, chronic obstructive pulmonary disease and asthma annual reviews) before and after March 23, 2020. We used multivariable Cox regression to quantify ethnic differences in hospitalisations related to diabetes, cardiovascular disease, respiratory disease, and mental health before and after March 23,2020. Of 33,510,937 registered with a GP as of 1st January 2020, 19,064,019 were adults, alive and registered for at least 3 months, 3,010,751 met the exclusion criteria and 1,122,912 were missing ethnicity. This resulted in 14,930,356 adults with known ethnicity (92% of sample): 86.6% were White, 7.3% Asian, 2.6% Black, 1.4% Mixed ethnicity, and 2.2% Other ethnicities. Clinical monitoring did not return to pre-pandemic levels for any ethnic group. Ethnic differences were apparent pre-pandemic, except for diabetes monitoring, and remained unchanged, except for blood pressure monitoring in those with mental health conditions where differences narrowed during the pandemic. For those of Black ethnicity, there were seven additional admissions for diabetic ketoacidosis per month during the pandemic, and relative ethnic differences narrowed during the pandemic compared to the White ethnic group (Pre-pandemic hazard ratio (HR): 0.50, 95% confidence interval (CI) 0.41, 0.60, Pandemic HR: 0.75, 95% CI: 0.65, 0.87). There was increased admissions for heart failure during the pandemic for all ethnic groups, though highest in those of White ethnicity (heart failure risk difference: 5.4). Relatively, ethnic differences narrowed for heart failure admission in those of Asian (Pre-pandemic HR 1.56, 95% CI 1.49, 1.64, Pandemic HR 1.24, 95% CI 1.19, 1.29) and Black ethnicity (Pre-pandemic HR 1.41, 95% CI: 1.30, 1.53, Pandemic HR: 1.16, 95% CI 1.09, 1.25) compared with White ethnicity. For other outcomes the pandemic had minimal impact on ethnic differences. Our study suggests that ethnic differences in clinical monitoring and hospitalisations remained largely unchanged during the pandemic for most conditions. Key exceptions were hospitalisations for diabetic ketoacidosis and heart failure, which warrant further investigation to understand the causes. LSHTM COVID-19 Response Grant (DONAT15912).

Development, validation and outcomes associated with a risk prediction algorithm for incident atrial fibrillation: retrospective national cohort study of 2 081 139 individuals

Abstract Funding Acknowledgements Type of funding sources: Foundation. Main funding source(s): British Heart Foundation Background An algorithm that identifies individuals with a digital electronic health record (EHR) signature homologous to patients with atrial fibrillation (AF) could delineate a subpopulation that may benefit from early interventions to reduce future adverse events. Purpose We aimed to train and test a scalable algorithm to identify individuals at higher risk of incident AF in the short-term, and quantify associations with AF and a range of other conditions. Methods We used UK primary care EHR data from individuals aged ≥30 years without known AF in the CPRD-GOLD dataset (Jan 2, 1998, Nov 30, 2018), randomly divided into training (80%) and testing (20%) datasets. We trained a random forest classifier using age, sex, ethnicity and comorbidities (FIND-AF). Performance was evaluated in the testing dataset with internal bootstrap validation with 200 samples, and compared against the CHA2DS2-VASc and C2HEST scores. We calculated the cumulative incidence rate for AF, heart failure, valvular heart disease (and specifically aortic stenosis), MI, stroke or TIA, peripheral vascular disease, CKD, diabetes and COPD. Incident diagnoses were the first record of that condition in primary or secondary care records from any diagnostic position. We excluded individuals for the analysis of each condition who had a preceding diagnosis of that condition. Fine and Gray’s models with competing risk of death were fit for each condition between higher and lower predicted AF risk. Results FIND-AF could be applied to 100% of records for 2 081 139 individuals in the cohort. In the testing dataset (n = 416 228), individuals at higher predicted AF risk had similar baseline characteristics to individuals who developed incident AF (Table 1). Prediction performance for AF was strongest for FIND-AF (AUROC 0·824, 95% CI 0·813-0·829; Brier score 0.069) compared with CHA2DS2-VASc (0·784, 0·773-0·794; 0.093) and C2HEST (0·757, 0·744-0·770; 0.102). FIND-AF demonstrated favourable reclassification and superior net benefit on decision curve analysis, with robust performance in both sexes and across ethnic groups. The higher predicted risk cohort, compared to lower predicted risk, had a 20-fold higher 6-month incidence rate for AF and higher long-term risk of AF (HR 8·75, 95% CI 8·44-9·06), but also incident heart failure (HR 12.54, 95% CI 12.08-13.01) aortic stenosis (9.98, 9.16-10.87), stroke/TIA (8.07, 7.80-8.34), CKD (6.85, 6.70-7.00), peripheral vascular disease (6.62, 6.28-6.98), valvular heart disease (6.49, 6.14-6.85), MI (5.02, 4.82-5.22), diabetes (2.05, 2.00-2.10) and COPD (2.02, 2.00-2.05) (Figure 1). This cohort were also at higher risk of death (10.45, 10.23-10.68), accounting for 71% of cardiovascular deaths. Conclusions FIND-AF is applicable to national electronic health records data, identifies people at higher risk of incident AF within the next 6 months with good performance, and predicts risk of a range of other conditions and death.

Characteristics of patients with asthma overprescribed short-acting beta-agonist (SABA) reliever inhalers stratified by blood eosinophil count in North East London: a cross-sectional observational study.

Overprescription of short-acting beta-agonist (SABA) inhalers and blood eosinophil count have strong associations with exacerbation risk in asthma. However, in the authors' recent publication only a minority of patients overprescribed SABA (≥6 inhalers in 12 months) were eosinophilic (≥0.3 x 109 cells/l). To compare the characteristics of eosinophilic and non-eosinophilic patients with asthma overprescribed SABA inhalers, and identify latent classes using clinical variables available in primary care. Cross-sectional analysis of patients with asthma in North East London, England, using primary care electronic health record data. Unadjusted and adjusted multi-variate regression models and latent class analysis. Eosinophilia was significantly less likely in female patients (P = 0.004), those with multiple mental health comorbidities (P<0.001), and those with SABA on repeat prescription (P<0.001). Latent class analysis identified the following three classes of patients overprescribed SABA: class 1, which represents classical uncontrolled asthma (oral steroids required for exacerbations, step 2-3 asthma medications, high probability of being eosinophilic); class 2, which represents mild asthma (low exacerbation frequency, low asthma medication step, low probability of being eosinophilic); and class 3, which represents difficult asthma (high exacerbation frequency despite high-strength preventer inhalers, low probability of being eosinophilic). The mild asthma class was the largest. Many patients being overprescribed SABA were non-eosinophilic with a low exacerbation frequency, suggesting disproportionately high SABA prescription compared with other asthma control markers. Potential reasons for high SABA prescription in these patients included repeat prescription (being dispensed but not taken) and use of SABA for non-asthma breathlessness (for example, breathing pattern disorders with anxiety). Further research is needed into management of SABA overuse in patients without other markers of uncontrolled asthma.

Unravelling the potential of social prescribing in individual-level type 2 diabetes prevention: a mixed-methods realist evaluation

BackgroundSocial prescribing (SP) usually involves linking patients in primary care with services provided by the voluntary and community sector. Preliminary evidence suggests that SP may offer a means of connecting patients with community-based health promotion activities, potentially contributing to the prevention of long-term conditions, such as type 2 diabetes (T2D).MethodsUsing mixed-methods realist evaluation, we explored the possible contribution of SP to individual-level prevention of T2D in a multi-ethnic, socio-economically deprived population in London, UK. We made comparisons with an existing prevention programme (NHS Diabetes Prevention Programme (NDPP)) where relevant and possible. Anonymised primary care electronic health record data of 447,360 people 18+ with an active GP registration between December 2016 and February 2022 were analysed using quantitative methods. Qualitative data (interviews with 11 primary care clinicians, 11 social prescribers, 13 community organisations and 8 SP users at high risk of T2D; 36 hours of ethnographic observations of SP and NDPP sessions; and relevant documents) were analysed thematically. Data were integrated using visual means and realist methods.ResultsPeople at high risk of T2D were four times more likely to be referred into SP than the eligible general population (RR 4.31 (95% CI 4.17–4.46)), with adjustment for socio-demographic variables resulting in attenuation (RR 1.33 (95% CI 1.27–1.39)). More people at risk of T2D were referred to SP than to NDPP, which could be explained by the broad referral criteria for SP and highly supportive (proactive, welcoming) environments. Holistic and sustained SP allowed acknowledgement of patients’ wider socio-economic constraints and provision of long-term personalised care. The fact that SP was embedded within the local community and primary care infrastructure facilitated the timely exchange of information and cross-referrals across providers, resulting in enhanced service responsiveness.ConclusionsOur study suggests that SP may offer an opportunity for individual-level T2D prevention to shift away from standardised, targeted and short-term strategies to approaches that are increasingly personalised, inclusive and long-term. Primary care-based SP seems most ideally placed to deliver such approaches where practitioners, providers and commissioners work collectively to achieve holistic, accessible, sustained and integrated services.

Real-World Primary Care Data Comparing ALT and FIB-4 in Predicting Future Severe Liver Disease Outcomes.

Alanine aminotransferase (ALT) has long provided a cue for chronic liver disease (CLD) diagnostic evaluation, but the Fibrosis-4 Index (FIB-4), a serologic score used for predicting advanced fibrosis risk in CLD, may provide an alternative signal. Compare the predictive performance of FIB-4 with ALT for severe liver disease (SLD) events while adjusting for potential confounders. Retrospective cohort study of primary care electronic health record data from 2012 to 2021. Adult primary care patients with at least two sets of ALT and other lab values necessary for calculating two unique FIB-4 scores, excluding those patients with an SLD prior to their index FIB-4 value. The occurrence of an SLD event, a composite of cirrhosis, hepatocellular carcinoma, and liver transplantation, was the outcome of interest. Categories of ALT elevation and FIB-4 advanced fibrosis risk were the primary predictor variables. Multivariable logistic regression models were developed to evaluate the association of FIB-4 and ALT with SLD, and the areas under the curve (AUC) for each model were compared. The cohort of 20,828 patients included 14% with an abnormal index ALT (≥40IU/L) and 8% with a high-risk index FIB-4 (≥2.67). During the study period, 667 (3%) patients suffered an SLD event. Adjusted multivariable logistic regression models demonstrated an association between high-risk FIB-4 (OR 19.34; 95%CI 15.50-24.13), persistently high-risk FIB-4 (OR 23.85; 95%CI 18.24-31.17), abnormal ALT (OR 7.07; 95%CI 5.81-8.59), and persistently abnormal ALT (OR 7.58; 95%CI 5.97-9.62) with SLD outcomes. The AUC of the index FIB-4 (0.847, p < 0.001) and combined FIB-4 (0.849, p < 0.001) adjusted models exceeded the index ALT adjusted model (0.815). High-risk FIB-4 scores demonstrated superior performance compared to abnormal ALT in predicting future SLD outcomes.

Prediction of short-term atrial fibrillation risk using primary care electronic health records

ObjectiveAtrial fibrillation (AF) screening by age achieves a low yield and misses younger individuals. We aimed to develop an algorithm in nationwide routinely collected primary care data to predict the...

BLOod Test Trend for cancEr Detection (BLOTTED): protocol for an observational and prediction model development study using English primary care electronic health record data

BackgroundSimple blood tests can play an important role in identifying patients for cancer investigation. The current evidence base is limited almost entirely to tests used in isolation. However, recent evidence suggests combining multiple types of blood tests and investigating trends in blood test results over time could be more useful to select patients for further cancer investigation. Such trends could increase cancer yield and reduce unnecessary referrals. We aim to explore whether trends in blood test results are more useful than symptoms or single blood test results in selecting primary care patients for cancer investigation. We aim to develop clinical prediction models that incorporate trends in blood tests to identify the risk of cancer.MethodsPrimary care electronic health record data from the English Clinical Practice Research Datalink Aurum primary care database will be accessed and linked to cancer registrations and secondary care datasets. Using a cohort study design, we will describe patterns in blood testing (aim 1) and explore associations between covariates and trends in blood tests with cancer using mixed-effects, Cox, and dynamic models (aim 2). To build the predictive models for the risk of cancer, we will use dynamic risk modelling (such as multivariate joint modelling) and machine learning, incorporating simultaneous trends in multiple blood tests, together with other covariates (aim 3). Model performance will be assessed using various performance measures, including c-statistic and calibration plots.DiscussionThese models will form decision rules to help general practitioners find patients who need a referral for further investigation of cancer. This could increase cancer yield, reduce unnecessary referrals, and give more patients the opportunity for treatment and improved outcomes.

Comparative Effectiveness of Umeclidinium/Vilanterol versus Indacaterol/Glycopyrronium on Moderate-to-Severe Exacerbations in Patients with Chronic Obstructive Pulmonary Disease in Clinical Practice in England.

Chronic obstructive pulmonary disease (COPD) exacerbations are associated with significant morbidity and mortality and increased economic healthcare burden for patients with COPD. Long-acting muscarinic antagonist (LAMA)/long-acting β2-agonist (LABA) dual therapy is recommended for patients receiving mono-bronchodilator therapy who experience exacerbations or ongoing breathlessness. This study compared two single-inhaler LAMA/LABA dual therapies, umeclidinium/vilanterol (UMEC/VI) and indacaterol/glycopyrronium (IND/GLY), on moderate-to-severe exacerbation rates in patients with COPD in England. This retrospective cohort study used linked primary care electronic health record data (Clinical Practice Research Datalink-Aurum) and secondary care data (Hospital Episode Statistics) to assess outcomes for patients with COPD who had a first prescription for single-inhaler UMEC/VI or IND/GLY (index date) between 1 January 2015 and 30 September 2019 (indexing period). Analyses compared UMEC/VI and IND/GLY on moderate-to-severe, moderate, and severe exacerbations, healthcare resource utilization (HCRU), and direct costs at 6, 12, 18, and 24 months, and time-to-first on-treatment exacerbation up to 24 months post-index date. Following inverse probability of treatment weighting (IPTW), non-inferiority and superiority of UMEC/VI versus IND/GLY were assessed. In total, 12,031 patients were included, of whom 8753 (72.8%) were prescribed UMEC/VI and 3278 (27.2%) IND/GLY. After IPTW, for moderate-to-severe exacerbations, weighted rate ratios were <1 at 6, 12, and 18 months and equal to 1 at 24 months for UMEC/VI; around the null value for moderate exacerbations and <1 at all timepoints for severe exacerbations. UMEC/VI showed lower HCRU incidence rates than IND/GLY for all-cause Accident and Emergency visits and COPD-related inpatient stays and associated all-cause costs at 6 months post-indexing. Time-to-triple therapy was similar for both treatments. UMEC/VI demonstrated non-inferiority to IND/GLY in moderate-to-severe exacerbation reduction at 6, 12 and 18 months. These results support previous findings demonstrating similarity between UMEC/VI and IND/GLY on reduction of moderate-to-severe exacerbations.

Machine Learning for Risk Prediction of Oesophago-Gastric Cancer in Primary Care: Comparison with Existing Risk-Assessment Tools.

Simple SummaryOesophago-gastric cancer is one of the commonest cancers worldwide, yet it can be particularly difficult to diagnose given that initial symptoms are often non-specific and routine screening is not available. Cancer risk-assessment tools, which calculate cancer risk based on symptoms and other risk factors present in the primary care record, can aid decisions on referrals for cancer investigations, facilitating earlier diagnosis. Diagnosing common cancers earlier could help improve survival rates. Using UK primary care electronic health record data, we compared five different machine learning techniques for probabilistic classification of cancer patients against a current widely used UK primary care cancer risk-assessment tool. The machine learning algorithms outperformed the current risk-assessment tool, with a higher overall accuracy and an ability to reasonably identify 11–25% more cancer patients. We conclude that machine-learning-based risk-assessment tools could help better identify suitable patients for further investigation and support earlier diagnosis.Oesophago-gastric cancer is difficult to diagnose in the early stages given its typical non-specific initial manifestation. We hypothesise that machine learning can improve upon the diagnostic performance of current primary care risk-assessment tools by using advanced analytical techniques to exploit the wealth of evidence available in the electronic health record. We used a primary care electronic health record dataset derived from the UK General Practice Research Database (7471 cases; 32,877 controls) and developed five probabilistic machine learning classifiers: Support Vector Machine, Random Forest, Logistic Regression, Naïve Bayes, and Extreme Gradient Boosted Decision Trees. Features included basic demographics, symptoms, and lab test results. The Logistic Regression, Support Vector Machine, and Extreme Gradient Boosted Decision Tree models achieved the highest performance in terms of accuracy and AUROC (0.89 accuracy, 0.87 AUROC), outperforming a current UK oesophago-gastric cancer risk-assessment tool (ogRAT). Machine learning also identified more cancer patients than the ogRAT: 11.0% more with little to no effect on false positives, or up to 25.0% more with a slight increase in false positives (for Logistic Regression, results threshold-dependent). Feature contribution estimates and individual prediction explanations indicated clinical relevance. We conclude that machine learning could improve primary care cancer risk-assessment tools, potentially helping clinicians to identify additional cancer cases earlier. This could, in turn, improve survival outcomes.

Rheumatoid arthritis, psoriatic arthritis, and axial spondyloarthritis epidemiology in England from 2004 to 2020: An observational study using primary care electronic health record data.

SummaryBackgroundContemporary data on rheumatoid arthritis (RA), psoriatic arthritis (PsA) and axial spondyloarthritits (SpA) epidemiology in England are lacking. This knowledge is crucial to planning healthcare services. We updated algorithms defining patients with diagnoses of RA, PsA, and axial SpA in primary care and applied them to describe their incidence and prevalence in the Clinical Practice Research Datalink Aurum, an electronic health record (EHR) database covering ∼20% of England.MethodsAlgorithms for ascertaining patients with RA, axial SpA, and PsA diagnoses validated in primary care EHR databases using Read codes were updated (to account for the English NHS change to SNOMED CT diagnosis coding) and applied. Updated diagnosis and synthetic disease-modifying anti-rheumatic drug code lists were devised by rheumatologists and general practitioners. Annual incidence/point-prevalence of RA, PsA, and axial SpA diagnoses were calculated from 2004 to 2020 and stratified by age/sex.FindingsPoint-prevalence of RA/PsA diagnoses increased annually, peaking in 2019 (RA 0·779% [95% confidence interval (CI) 0·773, 0·784]; PsA 0·287% [95% CI 0·284, 0·291]) then falling slightly. Point-prevalence of axial SpA diagnoses increased annually (except in 2018/2019), peaking in 2020 (0·113% [95% CI 0·111, 0·115]). RA diagnosis annual incidence was higher between 2013-2019 (after inclusion in the Quality and Outcomes Framework, range 49·1 [95% CI 47·7, 50·5] to 52·1 [95% CI 50·6, 53·6]/100,000 person-years) than 2004-2012 (range 34·5 [95% CI 33·2, 35·7] to 40·0 [95% CI 38·6, 41·4]/100,000 person-years). Increases in the annual incidence of PsA/axial SpA diagnosis occurred following new classification criteria publication. Annual incidence of RA, PsA and axial SpA diagnoses fell by 40·1%, 67·4%, and 38·1%, respectively between 2019 and 2020, likely reflecting the COVID-19 pandemic's impact on their diagnosis.InterpretationRecorded RA, PsA, and axial SpA diagnoses are increasingly prevalent in England, underlining the importance of organising healthcare services to provide timely, treat-to-target care to optimise the health of >1% of adults in England.FundingNational Institute for Health and Care Research (NIHR300826).

Developing Clinical Prediction Models Using Primary Care Electronic Health Record Data: The Impact of Data Preparation Choices on Model Performance.

To quantify prediction model performance in relation to data preparation choices when using electronic health records (EHR). Cox proportional hazards models were developed for predicting the first-ever main adverse cardiovascular events using Dutch primary care EHR data. The reference model was based on a 1-year run-in period, cardiovascular events were defined based on both EHR diagnosis and medication codes, and missing values were multiply imputed. We compared data preparation choices based on (i) length of the run-in period (2- or 3-year run-in); (ii) outcome definition (EHR diagnosis codes or medication codes only); and (iii) methods addressing missing values (mean imputation or complete case analysis) by making variations on the derivation set and testing their impact in a validation set. We included 89,491 patients in whom 6,736 first-ever main adverse cardiovascular events occurred during a median follow-up of 8 years. Outcome definition based only on diagnosis codes led to a systematic underestimation of risk (calibration curve intercept: 0.84; 95% CI: 0.83-0.84), while complete case analysis led to overestimation (calibration curve intercept: -0.52; 95% CI: -0.53 to -0.51). Differences in the length of the run-in period showed no relevant impact on calibration and discrimination. Data preparation choices regarding outcome definition or methods to address missing values can have a substantial impact on the calibration of predictions, hampering reliable clinical decision support. This study further illustrates the urgency of transparent reporting of modeling choices in an EHR data setting.