Identifying subtypes of type 2 diabetes mellitus with machine learning: development, internal validation, prognostic validation and medication burden in linked electronic health records in 420 448 individuals.

INTRODUCTION: None of the studies of type 2 diabetes (T2D) subtyping to date have used linked population-level data for incident and prevalent T2D, incorporating a diverse set of variables, explainable methods for cluster characterization, or adhered to an established framework. We aimed to develop and validate machine learning (ML)-informed subtypes for type 2 diabetes mellitus (T2D) using nationally representative data. RESEARCH DESIGN AND METHODS: In population-based electronic health records (2006-2020; Clinical Practice Research Datalink) in individuals ≥18 years with incident T2D (n=420 448), we included factors (n=3787), including demography, history, examination, biomarkers and medications. Using a published framework, we identified subtypes through nine unsupervised ML methods (K-means, K-means++, K-mode, K-prototype, mini-batch, agglomerative hierarchical clustering, Birch, Gaussian mixture models, and consensus clustering). We characterized clusters using intracluster distributions and explainable artificial intelligence (AI) techniques. We evaluated subtypes for (1) internal validity (within dataset; across methods); (2) prognostic validity (prediction for 5-year all-cause mortality, hospitalization and new chronic diseases); and (3) medication burden. RESULTS: Development: We identified four T2D subtypes: metabolic, early onset, late onset and cardiometabolic. Internal validity: Subtypes were predicted with high accuracy (F1 score >0.98). Prognostic validity: 5-year all-cause mortality, hospitalization, new chronic disease incidence and medication burden differed across T2D subtypes. Compared with the metabolic subtype, 5-year risks of mortality and hospitalization in incident T2D were highest in late-onset subtype (HR 1.95, 1.85-2.05 and 1.66, 1.58-1.75) and lowest in early-onset subtype (1.18, 1.11-1.27 and 0.85, 0.80-0.90). Incidence of chronic diseases was highest in late-onset subtype and lowest in early-onset subtype. Medications: Compared with the metabolic subtype, after adjusting for age, sex, and pre-T2D medications, late-onset subtype (1.31, 1.28-1.35) and early-onset subtype (0.83, 0.81-0.85) were most and least likely, respectively, to be prescribed medications within 5 years following T2D onset. CONCLUSIONS: In the largest study using ML to date in incident T2D, we identified four distinct subtypes, with potential future implications for etiology, therapeutics, and risk prediction.

Vaccinations, cardiovascular drugs, hospitalisation and mortality in COVID-19 and Long COVID.

OBJECTIVE: To identify highest-risk subgroups for COVID-19 and Long COVID(LC), particularly in contexts of influenza and cardiovascular disease(CVD). METHODS: Using national, linked electronic health records for England(NHS England Secure Data Environment via CVD-COVID-UK/COVID-IMPACT Consortium), we studied individuals(of all ages) with COVID-19 and LC (2020-2023). We compared all-cause hospitalisation and mortality by prior CVD, high CV risk, vaccination status(COVID-19/influenza), and CVD drugs, investigating impact of vaccination and CVD prevention using population preventable fractions. RESULTS: Hospitalisation and mortality were 15.3% and 2.0% among 17,373,850 individuals with COVID-19(LC rate 1.3%), and 16.8% and 1.4% among 301,115 with LC. Adjusted risk of mortality and hospitalisation were reduced with COVID-19 vaccination≥2 doses(COVID-19:HR 0.36 and 0.69; LC:0.44 and 0.90). With influenza vaccination, mortality was reduced, but not hospitalisation(COVID-19:0.86 and 1.01, and LC:0.72 and 1.05). Mortality and hospitalisation were reduced by CVD prevention in those with CVD, e.g. anticoagulants- COVID:19:0.69 and 0.92; LC:0.59 and 0.88; lipid lowering- COVID-19:0.69 and 0.86; LC:0.68 and 0.90. COVID-19 vaccination averted 245044 of 321383 and 7586 of 8738 preventable deaths after COVID-19 and LC, respectively. INTERPRETATION: Prior CVD and high CV risk are associated with increased hospitalisation and mortality in COVID-19 and LC. Targeted COVID-19 vaccination and CVD prevention are priority interventions. FUNDING: NIHR. HDR UK.

A nationwide, population-based study on specialized care for acute heart failure throughout the COVID-19 pandemic.

AIMS: The COVID-19 pandemic disrupted the delivery of care for patients with heart failure (HF), leading to fewer HF hospitalizations and increased mortality. However, nationwide data on quality of care and long-term outcomes across the pandemic are scarce. METHODS AND RESULTS: We used data from the National Heart Failure Audit (NHFA) linked to national records for hospitalization and deaths. We compared pre-COVID (2018-2019), COVID (2020), and late/post-COVID (2021-2022) periods. Data for 227 250 patients admitted to hospital with HF were analysed and grouped according to the admission year and the presence of HF with (HFrEF) or without reduced ejection fraction (non-HFrEF). The median age at admission was 81 years (interquartile range 72-88), 55% were men (n = 125 975), 87% were of white ethnicity (n = 102 805), and 51% had HFrEF (n = 116 990). In-hospital management and specialized cardiology care were maintained throughout the pandemic with an increasing percentage of patients discharged on disease-modifying medications over time (p < 0.001). Long-term outcomes improved over time (hazard ratio [HR] 0.92, 95% confidence interval [CI] 0.90-0.95, p < 0.001), mainly driven by a reduction in cardiovascular death. Receiving specialized cardiology care was associated with better long-term outcomes both for those who had HFrEF (HR 0.79, 95% CI 0.77-0.82, p < 0.001) and for those who had non-HFrEF (HR 0.87, 95% CI 0.85-0.90, p < 0.001). CONCLUSIONS: Despite the disruption of healthcare systems, the clinical characteristics of patients admitted with HF were similar and the overall standard of care was maintained throughout the pandemic. Long-term survival of patients hospitalized with HF continued to improve after COVID-19, especially for HFrEF.

Risk factors, outcomes and healthcare utilisation in individuals with multimorbidity including heart failure, chronic kidney disease and type 2 diabetes mellitus: a national electronic health record study.

BACKGROUND: Heart failure (HF), type 2 diabetes (T2D) and chronic kidney disease (CKD) commonly coexist. We studied characteristics, prognosis and healthcare utilisation of individuals with two of these conditions. METHODS: We performed a retrospective, population-based linked electronic health records study from 1998 to 2020 in England to identify individuals diagnosed with two of: HF, T2D or CKD. We described cohort characteristics at time of second diagnosis and estimated risk of developing the third condition and mortality using Kaplan-Meier and Cox regression models. We also estimated rates of healthcare utilisation in primary care and hospital settings in follow-up. FINDINGS: We identified cohorts of 64 226 with CKD and HF, 82 431 with CKD and T2D, and 13 872 with HF and T2D. Compared with CKD and T2D, those with CKD and HF and HF and T2D had more severe risk factor profile. At 5 years, incidence of the third condition and all-cause mortality occurred in 37% (95% CI: 35.9%, 38.1%%) and 31.3% (30.4%, 32.3%) in HF+T2D, 8.7% (8.4%, 9.0%) and 51.6% (51.1%, 52.1%) in HF+CKD, and 6.8% (6.6%, 7.0%) and 17.9% (17.6%, 18.2%) in CKD+T2D, respectively. In each of the three multimorbid groups, the order of the first two diagnoses was also associated with prognosis. In multivariable analyses, we identified risk factors for developing the third condition and mortality, such as age, sex, medical history and the order of disease diagnosis. Inpatient and outpatient healthcare utilisation rates were highest in CKD and HF, and lowest in CKD and T2D. INTERPRETATION: HF, CKD and T2D carry significant mortality and healthcare burden in combination. Compared with other disease pairs, individuals with CKD and HF had the most severe risk factor profile, prognosis and healthcare utilisation. Service planning, policy and prevention must take into account and monitor data across conditions.

Identifying subtypes of chronic kidney disease with machine learning: development, internal validation and prognostic validation using linked electronic health records in 350,067 individuals.

BACKGROUND: Although chronic kidney disease (CKD) is associated with high multimorbidity, polypharmacy, morbidity and mortality, existing classification systems (mild to severe, usually based on estimated glomerular filtration rate, proteinuria or urine albumin-creatinine ratio) and risk prediction models largely ignore the complexity of CKD, its risk factors and its outcomes. Improved subtype definition could improve prediction of outcomes and inform effective interventions. METHODS: We analysed individuals ≥18 years with incident and prevalent CKD (n = 350,067 and 195,422 respectively) from a population-based electronic health record resource (2006-2020; Clinical Practice Research Datalink, CPRD). We included factors (n = 264 with 2670 derived variables), e.g. demography, history, examination, blood laboratory values and medications. Using a published framework, we identified subtypes through seven unsupervised machine learning (ML) methods (K-means, Diana, HC, Fanny, PAM, Clara, Model-based) with 66 (of 2670) variables in each dataset. We evaluated subtypes for: (i) internal validity (within dataset, across methods); (ii) prognostic validity (predictive accuracy for 5-year all-cause mortality and admissions); and (iii) medications (new and existing by British National Formulary chapter). FINDINGS: After identifying five clusters across seven approaches, we labelled CKD subtypes: 1. Early-onset, 2. Late-onset, 3. Cancer, 4. Metabolic, and 5. Cardiometabolic. Internal validity: We trained a high performing model (using XGBoost) that could predict disease subtypes with 95% accuracy for incident and prevalent CKD (Sensitivity: 0.81-0.98, F1 score:0.84-0.97). Prognostic validity: 5-year all-cause mortality, hospital admissions, and incidence of new chronic diseases differed across CKD subtypes. The 5-year risk of mortality and admissions in the overall incident CKD population were highest in cardiometabolic subtype: 43.3% (42.3-42.8%) and 29.5% (29.1-30.0%), respectively, and lowest in the early-onset subtype: 5.7% (5.5-5.9%) and 18.7% (18.4-19.1%). MEDICATIONS: Across CKD subtypes, the distribution of prescription medication classes at baseline varied, with highest medication burden in cardiometabolic and metabolic subtypes, and higher burden in prevalent than incident CKD. INTERPRETATION: In the largest CKD study using ML, to-date, we identified five distinct subtypes in individuals with incident and prevalent CKD. These subtypes have relevance to study of aetiology, therapeutics and risk prediction. FUNDING: AstraZeneca UK Ltd, Health Data Research UK.

The impact of the COVID-19 pandemic on cardiovascular disease prevention and management.

How the Coronavirus Disease 2019 (COVID-19) pandemic has affected prevention and management of cardiovascular disease (CVD) is not fully understood. In this study, we used medication data as a proxy for CVD management using routinely collected, de-identified, individual-level data comprising 1.32 billion records of community-dispensed CVD medications from England, Scotland and Wales between April 2018 and July 2021. Here we describe monthly counts of prevalent and incident medications dispensed, as well as percentage changes compared to the previous year, for several CVD-related indications, focusing on hypertension, hypercholesterolemia and diabetes. We observed a decline in the dispensing of antihypertensive medications between March 2020 and July 2021, with 491,306 fewer individuals initiating treatment than expected. This decline was predicted to result in 13,662 additional CVD events, including 2,281 cases of myocardial infarction and 3,474 cases of stroke, should individuals remain untreated over their lifecourse. Incident use of lipid-lowering medications decreased by 16,744 patients per month during the first half of 2021 as compared to 2019. By contrast, incident use of medications to treat type 2 diabetes mellitus, other than insulin, increased by approximately 623 patients per month for the same time period. In light of these results, methods to identify and treat individuals who have missed treatment for CVD risk factors and remain undiagnosed are urgently required to avoid large numbers of excess future CVD events, an indirect impact of the COVID-19 pandemic.

Harmonising electronic health records for reproducible research: challenges, solutions and recommendations from a UK-wide COVID-19 research collaboration.

BACKGROUND: The CVD-COVID-UK consortium was formed to understand the relationship between COVID-19 and cardiovascular diseases through analyses of harmonised electronic health records (EHRs) across the four UK nations. Beyond COVID-19, data harmonisation and common approaches enable analysis within and across independent Trusted Research Environments. Here we describe the reproducible harmonisation method developed using large-scale EHRs in Wales to accommodate the fast and efficient implementation of cross-nation analysis in England and Wales as part of the CVD-COVID-UK programme. We characterise current challenges and share lessons learnt. METHODS: Serving the scope and scalability of multiple study protocols, we used linked, anonymised individual-level EHR, demographic and administrative data held within the SAIL Databank for the population of Wales. The harmonisation method was implemented as a four-layer reproducible process, starting from raw data in the first layer. Then each of the layers two to four is framed by, but not limited to, the characterised challenges and lessons learnt. We achieved curated data as part of our second layer, followed by extracting phenotyped data in the third layer. We captured any project-specific requirements in the fourth layer. RESULTS: Using the implemented four-layer harmonisation method, we retrieved approximately 100 health-related variables for the 3.2 million individuals in Wales, which are harmonised with corresponding variables for > 56 million individuals in England. We processed 13 data sources into the first layer of our harmonisation method: five of these are updated daily or weekly, and the rest at various frequencies providing sufficient data flow updates for frequent capturing of up-to-date demographic, administrative and clinical information. CONCLUSIONS: We implemented an efficient, transparent, scalable, and reproducible harmonisation method that enables multi-nation collaborative research. With a current focus on COVID-19 and its relationship with cardiovascular outcomes, the harmonised data has supported a wide range of research activities across the UK.

Using national electronic health records for pandemic preparedness: validation of a parsimonious model for predicting excess deaths among those with COVID-19-a data-driven retrospective cohort study.

OBJECTIVES: To use national, pre- and post-pandemic electronic health records (EHR) to develop and validate a scenario-based model incorporating baseline mortality risk, infection rate (IR) and relative risk (RR) of death for prediction of excess deaths. DESIGN: An EHR-based, retrospective cohort study. SETTING: Linked EHR in Clinical Practice Research Datalink (CPRD); and linked EHR and COVID-19 data in England provided in NHS Digital Trusted Research Environment (TRE). PARTICIPANTS: In the development (CPRD) and validation (TRE) cohorts, we included 3.8 million and 35.1 million individuals aged ≥30 years, respectively. MAIN OUTCOME MEASURES: One-year all-cause excess deaths related to COVID-19 from March 2020 to March 2021. RESULTS: From 1 March 2020 to 1 March 2021, there were 127,020 observed excess deaths. Observed RR was 4.34% (95% CI, 4.31-4.38) and IR was 6.27% (95% CI, 6.26-6.28). In the validation cohort, predicted one-year excess deaths were 100,338 compared with the observed 127,020 deaths with a ratio of predicted to observed excess deaths of 0.79. CONCLUSIONS: We show that a simple, parsimonious model incorporating baseline mortality risk, one-year IR and RR of the pandemic can be used for scenario-based prediction of excess deaths in the early stages of a pandemic. Our analyses show that EHR could inform pandemic planning and surveillance, despite limited use in emergency preparedness to date. Although infection dynamics are important in the prediction of mortality, future models should take greater account of underlying conditions.

COVID-19 trajectories among 57 million adults in England: a cohort study using electronic health records.

BACKGROUND: Updatable estimates of COVID-19 onset, progression, and trajectories underpin pandemic mitigation efforts. To identify and characterise disease trajectories, we aimed to define and validate ten COVID-19 phenotypes from nationwide linked electronic health records (EHR) using an extensible framework. METHODS: In this cohort study, we used eight linked National Health Service (NHS) datasets for people in England alive on Jan 23, 2020. Data on COVID-19 testing, vaccination, primary and secondary care records, and death registrations were collected until Nov 30, 2021. We defined ten COVID-19 phenotypes reflecting clinically relevant stages of disease severity and encompassing five categories: positive SARS-CoV-2 test, primary care diagnosis, hospital admission, ventilation modality (four phenotypes), and death (three phenotypes). We constructed patient trajectories illustrating transition frequency and duration between phenotypes. Analyses were stratified by pandemic waves and vaccination status. FINDINGS: Among 57 032 174 individuals included in the cohort, 13 990 423 COVID-19 events were identified in 7 244 925 individuals, equating to an infection rate of 12·7% during the study period. Of 7 244 925 individuals, 460 737 (6·4%) were admitted to hospital and 158 020 (2·2%) died. Of 460 737 individuals who were admitted to hospital, 48 847 (10·6%) were admitted to the intensive care unit (ICU), 69 090 (15·0%) received non-invasive ventilation, and 25 928 (5·6%) received invasive ventilation. Among 384 135 patients who were admitted to hospital but did not require ventilation, mortality was higher in wave 1 (23 485 [30·4%] of 77 202 patients) than wave 2 (44 220 [23·1%] of 191 528 patients), but remained unchanged for patients admitted to the ICU. Mortality was highest among patients who received ventilatory support outside of the ICU in wave 1 (2569 [50·7%] of 5063 patients). 15 486 (9·8%) of 158 020 COVID-19-related deaths occurred within 28 days of the first COVID-19 event without a COVID-19 diagnoses on the death certificate. 10 884 (6·9%) of 158 020 deaths were identified exclusively from mortality data with no previous COVID-19 phenotype recorded. We observed longer patient trajectories in wave 2 than wave 1. INTERPRETATION: Our analyses illustrate the wide spectrum of disease trajectories as shown by differences in incidence, survival, and clinical pathways. We have provided a modular analytical framework that can be used to monitor the impact of the pandemic and generate evidence of clinical and policy relevance using multiple EHR sources. FUNDING: British Heart Foundation Data Science Centre, led by Health Data Research UK.

A retrospective cohort study predicting and validating impact of the COVID-19 pandemic in individuals with chronic kidney disease.

Chronic kidney disease (CKD) is associated with increased risk of baseline mortality and severe COVID-19, but analyses across CKD stages, and comorbidities are lacking. In prevalent and incident CKD, we investigated comorbidities, baseline risk, COVID-19 incidence, and predicted versus observed one-year excess death. In a national dataset (NHS Digital Trusted Research Environment [NHSD TRE]) for England encompassing 56 million individuals), we conducted a retrospective cohort study (March 2020 to March 2021) for prevalence of comorbidities by incident and prevalent CKD, SARS-CoV-2 infection and mortality. Baseline mortality risk, incidence and outcome of infection by comorbidities, controlling for age, sex and vaccination were assessed. Observed versus predicted one-year mortality at varying population infection rates and pandemic-related relative risks using our published model in pre-pandemic CKD cohorts (NHSD TRE and Clinical Practice Research Datalink [CPRD]) were compared. Among individuals with CKD (prevalent:1,934,585, incident:144,969), comorbidities were common (73.5% and 71.2% with one or more condition[s] in respective data sets, and 13.2% and 11.2% with three or more conditions, in prevalent and incident CKD), and associated with SARS-CoV-2 infection, particularly dialysis/transplantation (odds ratio 2.08, 95% confidence interval 2.04-2.13) and heart failure (1.73, 1.71-1.76), but not cancer (1.01, 1.01-1.04). One-year all-cause mortality varied by age, sex, multi-morbidity and CKD stage. Compared with 34,265 observed excess deaths, in the NHSD-TRE and CPRD databases respectively, we predicted 28,746 and 24,546 deaths (infection rates 10% and relative risks 3.0), and 23,754 and 20,283 deaths (observed infection rates 6.7% and relative risks 3.7). Thus, in this largest, national-level study, individuals with CKD have a high burden of comorbidities and multi-morbidity, and high risk of pre-pandemic and pandemic mortality. Hence, treatment of comorbidities, non-pharmaceutical measures, and vaccination are priorities for people with CKD and management of long-term conditions is important during and beyond the pandemic.

Evaluation of antithrombotic use and COVID-19 outcomes in a nationwide atrial fibrillation cohort.

OBJECTIVE: To evaluate antithrombotic (AT) use in individuals with atrial fibrillation (AF) and at high risk of stroke (CHA2DS2-VASc score ≥2) and investigate whether pre-existing AT use may improve COVID-19 outcomes. METHODS: Individuals with AF and CHA2DS2-VASc score ≥2 on 1 January 2020 were identified using electronic health records for 56 million people in England and were followed up until 1 May 2021. Factors associated with pre-existing AT use were analysed using logistic regression. Differences in COVID-19-related hospitalisation and death were analysed using logistic and Cox regression in individuals with pre-existing AT use versus no AT use, anticoagulants (AC) versus antiplatelets (AP), and direct oral anticoagulants (DOACs) versus warfarin. RESULTS: From 972 971 individuals with AF (age 79 (±9.3), female 46.2%) and CHA2DS2-VASc score ≥2, 88.0% (n=856 336) had pre-existing AT use, 3.8% (n=37 418) had a COVID-19 hospitalisation and 2.2% (n=21 116) died, followed up to 1 May 2021. Factors associated with no AT use included comorbidities that may contraindicate AT use (liver disease and history of falls) and demographics (socioeconomic status and ethnicity). Pre-existing AT use was associated with lower odds of death (OR=0.92, 95% CI 0.87 to 0.96), but higher odds of hospitalisation (OR=1.20, 95% CI 1.15 to 1.26). AC versus AP was associated with lower odds of death (OR=0.93, 95% CI 0.87 to 0.98) and higher hospitalisation (OR=1.17, 95% CI 1.11 to 1.24). For DOACs versus warfarin, lower odds were observed for hospitalisation (OR=0.86, 95% CI 0.82 to 0.89) but not for death (OR=1.00, 95% CI 0.95 to 1.05). CONCLUSIONS: Pre-existing AT use may be associated with lower odds of COVID-19 death and, while not evidence of causality, provides further incentive to improve AT coverage for eligible individuals with AF.

Linkage of primary care prescribing records and pharmacy dispensing Records in the Salford Lung Study: application in asthma.

BACKGROUND: Records of medication prescriptions can be used in conjunction with pharmacy dispensing records to investigate the incidence of adherence, which is defined as observing the treatment plans agreed between a patient and their clinician. Using prescribing records alone fails to identify primary non-adherence; medications not being collected from the dispensary. Using dispensing records alone means that cases of conditions that resolve and/or treatments that are discontinued will be unaccounted for. While using a linked prescribing and dispensing dataset to measure medication non-adherence is optimal, this linkage is not routinely conducted. Furthermore, without a unique common event identifier, linkage between these two datasets is not straightforward. METHODS: We undertook a secondary analysis of the Salford Lung Study dataset. A novel probabilistic record linkage methodology was developed matching asthma medication pharmacy dispensing records and primary care prescribing records, using semantic (meaning) and syntactic (structure) harmonization, domain knowledge integration, and natural language feature extraction. Cox survival analysis was conducted to assess factors associated with the time to medication dispensing after the prescription was written. Finally, we used a simplified record linkage algorithm in which only identical records were matched, for a naïve benchmarking to compare against the results of our proposed methodology. RESULTS: We matched 83% of pharmacy dispensing records to primary care prescribing records. Missing data were prevalent in the dispensing records which were not matched - approximately 60% for both medication strength and quantity. A naïve benchmarking approach, requiring perfect matching, identified one-quarter as many matching prescribing records as our methodology. Factors associated with delay (or failure) to collect the prescribed medication from a pharmacy included season, quantity of medication prescribed, previous dispensing history and class of medication. Our findings indicate that over 30% of prescriptions issued were not collected from a dispensary (primary non-adherence). CONCLUSIONS: We have developed a probabilistic record linkage methodology matching a large percentage of pharmacy dispensing records with primary care prescribing records for asthma medications. This will allow researchers to link datasets in order to extract information about asthma medication non-adherence.

A data-driven typology of asthma medication adherence using cluster analysis.

Asthma preventer medication non-adherence is strongly associated with poor asthma control. One-dimensional measures of adherence may ignore clinically important patterns of medication-taking behavior. We sought to construct a data-driven multi-dimensional typology of medication non-adherence in children with asthma. We analyzed data from an intervention study of electronic inhaler monitoring devices, comprising 211 patients yielding 35,161 person-days of data. Five adherence measures were extracted: the percentage of doses taken, the percentage of days on which zero doses were taken, the percentage of days on which both doses were taken, the number of treatment intermissions per 100 study days, and the duration of treatment intermissions per 100 study days. We applied principal component analysis on the measures and subsequently applied k-means to determine cluster membership. Decision trees identified the measure that could predict cluster assignment with the highest accuracy, increasing interpretability and increasing clinical utility. We demonstrate the use of adherence measures towards a three-group categorization of medication non-adherence, which succinctly describes the diversity of patient medication taking patterns in asthma. The percentage of prescribed doses taken during the study contributed to the prediction of cluster assignment most accurately (84% in out-of-sample data).

Investigating the effect of England's smoke-free private vehicle regulation on changes in tobacco smoke exposure and respiratory disease in children: a quasi-experimental study.

BACKGROUND: Comprehensive tobacco control policies can help to protect children from tobacco smoke exposure and associated adverse respiratory health consequences. We investigated the impact of England's 2015 regulation that prohibits smoking in a private vehicle with children present on changes in environmental tobacco smoke exposure and respiratory health in children. METHODS: In this quasi-experimental study, we used repeated cross-sectional, nationally representative data from the Health Survey for England from Jan 1, 2008, to Dec 31, 2017, of children aged up to 15 years. We did interrupted time series logistic or ordinal regression analyses to assess changes in prevalence of self-reported respiratory conditions, prevalence of self-reported childhood tobacco smoke exposure (children aged 8-15 years only), and salivary cotinine levels (children aged 2 years or older) before and after implementation of the smoke-free private vehicle regulation on Oct 1, 2015. Children who were considered active smokers were excluded from the analyses of salivary cotinine levels. Our primary outcome of interest was self-reported current wheezing or asthma, defined as having medicines prescribed for these conditions. Analyses were adjusted for underlying time trends, quarter of year, sex, age, Index of Multiple Deprivation quintile, and urbanisation level. FINDINGS: 21 096 children aged 0-15 years were included in our dataset. Implementation of the smoke-free private vehicle regulation was not associated with a demonstrable change in self-reported current wheezing or asthma (adjusted odds ratio 0·81, 95% CI 0·62-1·05; p=0·108; assessed in 13 369 children), respiratory conditions (1·02, 0·80-1·29; p=0·892; assessed in 17 006 children), or respiratory conditions probably affecting stamina, breathing, or fatigue (0·75, 0·47-1·19; p=0·220; assessed in 12 386 children). Self-reported tobacco smoke exposure and salivary cotinine levels generally decreased over the study period. There was no additional change in self-reported tobacco smoke exposure in cars among children aged 8-15 years following the legislation (0·77, 0·51-1·17; p=0·222; assessed in 5399 children). We observed a relative increase in the odds of children having detectable salivary cotinine levels post legislation (1·36, 1·09-1·71; p=0·0074; assessed in 7858 children) and levels were also higher (1·30, 1·04-1·62; p=0·020; ordinal variable). Despite introduction of the regulation, one in 20 children still reported being regularly exposed to tobacco smoke in cars and one in three still had detectable salivary cotinine levels. INTERPRETATION: We found no demonstrable association between the implementation of England's smoke-free private vehicle regulation and changes in children's self-reported tobacco smoke exposure or respiratory health. There is an urgent need to develop more effective approaches to protect children from tobacco smoke in various places, including in private vehicles. FUNDING: Netherlands Lung Foundation, Erasmus MC, Farr Institute, Health Data Research UK, Asthma UK Centre for Applied Research, Academy of Medical Sciences, and Newton Fund.

Predicting asthma attacks in primary care: protocol for developing a machine learning-based prediction model.

INTRODUCTION: Asthma is a long-term condition with rapid onset worsening of symptoms ('attacks') which can be unpredictable and may prove fatal. Models predicting asthma attacks require high sensitivity to minimise mortality risk, and high specificity to avoid unnecessary prescribing of preventative medications that carry an associated risk of adverse events. We aim to create a risk score to predict asthma attacks in primary care using a statistical learning approach trained on routinely collected electronic health record data. METHODS AND ANALYSIS: We will employ machine-learning classifiers (naïve Bayes, support vector machines, and random forests) to create an asthma attack risk prediction model, using the Asthma Learning Health System (ALHS) study patient registry comprising 500 000 individuals across 75 Scottish general practices, with linked longitudinal primary care prescribing records, primary care Read codes, accident and emergency records, hospital admissions and deaths. Models will be compared on a partition of the dataset reserved for validation, and the final model will be tested in both an unseen partition of the derivation dataset and an external dataset from the Seasonal Influenza Vaccination Effectiveness II (SIVE II) study. ETHICS AND DISSEMINATION: Permissions for the ALHS project were obtained from the South East Scotland Research Ethics Committee 02 [16/SS/0130] and the Public Benefit and Privacy Panel for Health and Social Care (1516-0489). Permissions for the SIVE II project were obtained from the Privacy Advisory Committee (National Services NHS Scotland) [68/14] and the National Research Ethics Committee West Midlands-Edgbaston [15/WM/0035]. The subsequent research paper will be submitted for publication to a peer-reviewed journal and code scripts used for all components of the data cleaning, compiling, and analysis will be made available in the open source GitHub website (https://github.com/hollytibble).

Policymaking to preserve privacy in disclosure of public health data: a suggested framework.

Health organisations in Turkey gather a vast amount of valuable individual data that can be used for public health purposes. The organisations use rigid methods to remove some useful details from the data while publishing the rest of the data in a highly aggregated form, mostly because of privacy concerns and lack of standardised policies. This action leads to information loss and bias affecting public health research. Hence, organisations need dynamic policies and well-defined procedures rather than a specific algorithm to protect the privacy of individual data. To address this need, we developed a framework for the systematic application of anonymity methods while reducing and objectively reporting the information loss without leaking confidentiality. This framework acts as a roadmap for policymaking by providing high-level pseudo-policies with semitechnical guidelines in addition to some sample scenarios suitable for policymakers, public health programme managers and legislators.