PLoS ONE plos plosone PLOS ONE 1932-6203 Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0292240 PONE-D-23-05392 Research Article Medicine and health sciencesEpidemiologyMedical risk factors Medicine and health sciencesMedical conditionsCardiovascular diseases Medicine and health sciencesCardiologyCardiovascular medicineCardiovascular diseases Medicine and health sciencesHealth carePrimary care Biology and life sciencesBiochemistryLipidsCholesterol Medicine and health sciencesMedical conditionsCardiovascular diseasesCardiovascular diseases in women Medicine and health sciencesCardiologyCardiovascular medicineCardiovascular diseasesCardiovascular diseases in women Medicine and health sciencesWomen's healthCardiovascular diseases in women Medicine and health sciencesMedical conditionsCardiovascular diseasesCardiovascular disease risk Medicine and health sciencesCardiologyCardiovascular medicineCardiovascular diseasesCardiovascular disease risk Medicine and health sciencesPharmacologyDrugsStatins People and placesGeographical locationsEuropeEuropean UnionUnited KingdomEngland Prioritising cardiovascular disease risk assessment to high risk individuals based on primary care records Primary care records to prioritise for cardiovascular disease risk assessments https://orcid.org/0000-0001-6545-4669 Chung Ryan Data curation Formal analysis Methodology Software Writing – original draft Writing – review & editing 1 * https://orcid.org/0000-0003-1519-6707 Xu Zhe Data curation Formal analysis Methodology Writing – original draft Writing – review & editing 1 Arnold Matthew Data curation Formal analysis Methodology Software Writing – original draft Writing – review & editing 1 ¤ Stevens David Data curation Methodology Software Writing – review & editing 1 Keogh Ruth Methodology Software Writing – review & editing 2 Barrett Jessica Methodology Writing – review & editing 3 Harrison Hannah Writing – review & editing 4 Pennells Lisa Methodology Writing – review & editing 1 https://orcid.org/0000-0002-4552-3820 Kim Lois G. Writing – review & editing 1 5 DiAngelantonio Emanuele Conceptualization Writing – review & editing 1 5 6 7 8 https://orcid.org/0000-0003-0855-9872 Paige Ellie Conceptualization Formal analysis Methodology Writing – original draft Writing – review & editing 9 https://orcid.org/0000-0002-8501-2531 Usher-Smith Juliet A. Conceptualization Writing – review & editing 10 Wood Angela M. Conceptualization Data curation Formal analysis Methodology Software Writing – original draft Writing – review & editing 1 5 6 7 11 British Heart Foundation Cardiovascular Epidemiology Unit, Department of Public Health and Primary Care, University of Cambridge, Cambridge, United Kingdom London School of Hygiene and Tropical Medicine, Faculty of Epidemiology & Population Health, London, United Kingdom Medical Research Council Biostatistics Unit, University of Cambridge, Cambridge, United Kingdom Centre for Cancer Genetic Epidemiology, Department of Public Health and Primary Care, University of Cambridge, Cambridge, United Kingdom National Institute for Health and Care Research Blood and Transplant Research Unit in Donor Health and Behaviour, University of Cambridge, Cambridge, United Kingdom British Heart Foundation Centre of Research Excellence, University of Cambridge, Cambridge, United Kingdom Health Data Research UK Cambridge, Wellcome Genome Campus and University of Cambridge, Cambridge, United Kingdom Health Data Science Research Centre, Human Technopole, Milan, Italy National Centre for Epidemiology and Population Health, Australian National University, Canberra, Australia Primary Care Unit, Department of Public Health and Primary Care, University of Cambridge, Cambridge, United Kingdom Cambridge Centre of Artificial Intelligence in Medicine, Cambridge, United Kingdom Pontikos Nikolas Editor University College London Institute of Ophthalmology, UNITED KINGDOM

During the drafting of the manuscript, M.A. became an employee of AstraZeneca. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Current address: AstraZeneca PLC, Cambridge, United Kingdom

 * E-mail: rkyc2@medschl.cam.ac.uk 29 9 2023 2023 18 9 e0292240 23 2 2023 14 9 2023 2023 Chung et al This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Objective

To provide quantitative evidence for systematically prioritising individuals for full formal cardiovascular disease (CVD) risk assessment using primary care records with a novel tool (eHEART) with age- and sex- specific risk thresholds.

 Methods and analysis

eHEART was derived using landmark Cox models for incident CVD with repeated measures of conventional CVD risk predictors in 1,642,498 individuals from the Clinical Practice Research Datalink. Using 119,137 individuals from UK Biobank, we modelled the implications of initiating guideline-recommended statin therapy using eHEART with age- and sex-specific prioritisation thresholds corresponding to 5% false negative rates to prioritise adults aged 40–69 years in a population in England for invitation to a formal CVD risk assessment.

 Results

Formal CVD risk assessment on all adults would identify 76% and 49% of future CVD events amongst men and women respectively, and 93 (95% CI: 90, 95) men and 279 (95% CI: 259, 297) women would need to be screened (NNS) to prevent one CVD event. In contrast, if eHEART was first used to prioritise individuals for formal CVD risk assessment, we would identify 73% and 47% of future events amongst men and women respectively, and a NNS of 75 (95% CI: 72, 77) men and 162 (95% CI: 150, 172) women. Replacing the age- and sex-specific prioritisation thresholds with a 10% threshold identify around 10% less events.

 Conclusions

The use of prioritisation tools with age- and sex-specific thresholds could lead to more efficient CVD assessment programmes with only small reductions in effectiveness at preventing new CVD events.

 http://dx.doi.org/10.13039/501100000274 British Heart Foundation RG/13/13/30194 http://dx.doi.org/10.13039/501100000274 British Heart Foundation RG/18/13/33946 BHF Cambridge Centre of Research Excellence RE/13/6/30180 http://dx.doi.org/10.13039/501100018956 NIHR Cambridge Biomedical Research Centre BRC-1215-20014 http://dx.doi.org/10.13039/501100000265 Medical Research Council MR/K014811/1 http://dx.doi.org/10.13039/501100000274 British Heart Foundation FS/18/56/34177 https://orcid.org/0000-0001-6545-4669 Chung Ryan Chinese Scholarship Council https://orcid.org/0000-0003-1519-6707 Xu Zhe http://dx.doi.org/10.13039/501100000274 British Heart Foundation RG/18/13/33946 Arnold Matthew http://dx.doi.org/10.13039/501100018956 NIHR Cambridge Biomedical Research Centre BRC-1215-20014 Stevens David http://dx.doi.org/10.13039/501100000265 Medical Research Council MR/L501566/1 Barrett Jessica http://dx.doi.org/10.13039/501100000265 Medical Research Council MC_UU_00002/5 Barrett Jessica NIHR BTRU in Donor Health and Genomics NIHR BTRU-2014–10024) https://orcid.org/0000-0002-4552-3820 Kim Lois G. NIHR BTRU in Donor Health and Behaviour NIHR203337 https://orcid.org/0000-0002-4552-3820 Kim Lois G. NIHR Advanced Fellowship NIHR300861 https://orcid.org/0000-0002-8501-2531 Usher-Smith Juliet A. Innovative Medicines Initiative-2 Joint Undertaking 116074 Wood Angela M. BHF-Turing Cardiovascular Data Science Award BCDSA\100005 Wood Angela M. This work was supported by core funding from the: British Heart Foundation (RG/13/13/30194; RG/18/13/33946), BHF Cambridge Centre of Research Excellence (RE/13/6/30180) and NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. *The views expressed are those of the author(s) and not necessarily those of the NIHR, NHSBT or the Department of Health and Social Care. This work was funded by the Medical Research council (MR/K014811/1). The study funders played no role in the design, analysis or interpretation of the study. R.C. is funded by a BHF PhD studentship (FS/18/56/34177). Z.X. is funded by the Chinese Scholarship Council. M.A. was funded by a British Heart Foundation Programme Grant (RG/18/13/33946). D.S. was funded by the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. J.B. was funded by a Medical Research Council fellowship (MR/L501566/1) and unit programme (MC_UU_00002/5). L.G.K. was funded by the NIHR BTRU in Donor Health and Genomics (NIHR BTRU-2014–10024) and is funded by the NIHR BTRU in Donor Health and Behaviour (NIHR203337) [*]. J.A.U. is funded by an NIHR Advanced Fellowship (NIHR300861). A.M.W. is part of the BigData@Heart Consortium, funded by the Innovative Medicines Initiative-2 Joint Undertaking under grant agreement No 116074. A.M.W. is supported by the BHF-Turing Cardiovascular Data Science Award (BCDSA\100005). Cambridge Service for Data Driven Discovery (CSD3) was provided by Dell EMC and Intel using Tier-2 funding from the Engineering and Physical Sciences Research Council (capital grant EP/P020259/1), and DiRAC funding from the Science and Technology Facilities Council (www.dirac.ac.uk). There was no additional external funding received for this study The funder provided support in the form of salaries for authors M.A., but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data Availability This research has been conducted using the UK Biobank Resource under Application Number 26865. Data from the Clinical Practice Research Datalink (CPRD) were obtained under licence from the UK Medicines and Healthcare products Regulatory Agency (protocol 162RMn2). Supplementary code for eHEART is available online (www.github.com/ryanchung894/eHEART). Outbreaks COVID-19 Introduction

The WHO estimate that the majority of premature cardiovascular disease (CVD) events are preventable through lifestyle choices and pharmacotherapy that target modifiable risk factors [1]. Many countries implement population-wide CVD risk assessment programmes in primary care that recommend use of established risk prediction tools to identify individuals at high risk of CVD and guide patient and clinical decision-making [2–8]. More recently, to reduce programme running costs and address rising health inequalities in access to preventative care, guidelines and policy makers have advocated using primary care records for systematic prioritisation of individuals before formal CVD risk assessment [7]. This has become even more important given the current COVID-19 related burdens on national health care provision. As an example, in addition to recommendations to formally assess all individuals in England, individuals are recommended to be prioritised for a formal CVD risk assessment if their estimated risk, using existing data from primary care health records, is greater than 10% risk over ten years [7]. However, quantitative evidence of the health impact of prioritisation when using a fixed 10% threshold in clinical practice is limited, and no specific risk tool is recommended for systematic prioritisation of risk.

Therefore, to help in the formulation of evidence-based CVD risk assessment programmes, we aim to evaluate two key aspects. First, we compare the population-wide formal CVD risk assessment approach against approaches that systematically prioritise individuals before full formal CVD risk assessment. Here, we evaluate two distinct prioritisation tools based on primary care records: a novel tool (eHEART), developed in this manuscript to leverage the sparse and sporadically observed longitudinal data on conventional CVD risk factors recorded primary care records; and the other, being an existing CVD risk tool (QRISK2) [9], that has been derived on 2.3 million individuals from the QResearch database and which implements a substitution approach for any missing conventional risk factors based on age, sex and ethnicity. Second, we compare the use of a fixed 10% prioritisation threshold versus age- and sex-specific prioritisation thresholds.

 Methods CPRD data source

To derive eHEART we used anonymised person-level primary care records from general practices contributing data to Clinical Practice Research Datalink GOLD (CPRD) with linked information on hospital episodes from Hospital Episode Statistics (HES), and death registrations from the Office of National Statistics (ONS) [10]. Primary care records were extracted for individuals from the latest of: date of their registration at general practice plus 6 months, their 30th birthday, date when the general practice provided “up-to-standard” data [11], or 1st April 2004 (the date of introduction of the Quality and Outcomes Framework) [12], until the earliest of: date of their first (i.e. “incident”) newly recorded CVD event, date of de-registration at the general practice, their 85th birthday, date of death, last contact date for the practice with CPRD, or 31 May 2019 (the end of data availability). All analyses focus on individuals without known pre-existing CVD and without prescribed statins. The data used in this study were obtained under licence from the UK Medicines and Healthcare Products Regulatory Agency (protocol 162RMn2).

 UK Biobank data source

Data from UK Biobank, a prospective cohort study with detailed baseline information and linked primary care record data (from the The Phoenix Partnership (TPP), Egton Medical Information Systems (EMIS) and Vision GP system suppliers) for 177,361 individuals [13, 14], was used to model the implications of prioritising individuals for formal CVD assessment and subsequent initiation of guideline-recommended statin therapy in a primary care setting. UK Biobank was chosen due to the complete data available at baseline to replicate being able to perform a full risk assessment. Population health modelling focussed on individuals without known pre-existing CVD and without prescribed statins at baseline. This research has been conducted using the UK Biobank Resource under Application Number 26865.

 Definition of CVD outcome

We ascertained individuals with a first ever incident CVD, defined as nonfatal or fatal events of coronary heart disease (CHD) (including myocardial infarction and angina), stroke, and transient ischemic attack, as those with a relevant Read-code or ICD-10 code (listed in Table 1 in S2 File) appearing in the primary care data, hospital episodes (main or secondary diagnostic code position in the admitted patient care component of the Hospital Episode Statistics data), or death registry (underlying or contributing cause of death) during follow-up.

 CVD risk predictors

Conventional and commonly recorded CVD risk predictors [5, 15] in primary care were pre-selected for inclusion into the estimated eHEART prioritisation tool and included: age (in years), sex (male or female), diabetes status (yes or no), hypertension medication (yes or no), systolic blood pressure (SBP) (mmHg), total cholesterol (mmol/litre) and high-density lipoprotein (HDL) cholesterol (mmol/litre) and smoking status (current smoker or not) (details of measurements have been previously described) [16].

The following measurements were considered biologically implausible and were changed to missing (∼0.4% of measurements): SBP <60 or >250 mmHg; total cholesterol <1.75 or >20 mmol/litre; and HDL cholesterol <0.3 or >3.1 mmol/litre [17].

 Statistical analysis The eHEART prioritisation tool

To optimise the nature of longitudinal primary care records, we used a landmark-age approach to derive 90 age- (ie, 40, 41, 42, …to 84 years) and sex-specific Cox models as previously described [18]. Each model utilised primary care records from persons alive without pre-existing CVD and without prescribed statins at the landmark-age by using available risk predictors recorded before the landmark-age to predict 10-year risk of incident CVD outcomes (Fig 1 in S1 File).

Each model was developed in two stages. In the first stage, a multivariate mixed-effects model [19] was fitted to repeat measures of smoking status, SBP, total cholesterol and HDL cholesterol recorded before the landmark-age and used to estimate risk predictor values at the landmark-age amongst individuals who had at least one past record of any one predictor. The model included fixed- and correlated random-intercepts and linear age fixed-effects (quadratic terms were not significant) for each predictor, as well as an interaction between SBP and an age-varying binary covariate for hypertension treatment, and an interaction between total cholesterol and an age-varying binary covariate for statin treatment (but excluded those with pre-statins for Cox model). The model intrinsically accounts for missing data in the predictors with the assumption that predictor values from individuals with incomplete data are from the same multivariate normal distribution for predictor values as individuals with observed data. Further details are given in Text 1 in S2 File.

In the second stage, we derived Cox models with time since landmark age as the underlying time scale and the following predictors: diabetes status, treatment for hypertension, and the estimated values of SBP, total cholesterol, HDL cholesterol, and smoking status from the multivariate mixed-effects model. The four estimated predictors were entered as linear terms. The outcome was incident CVD and censoring occurred for individuals at end of their follow-up, either date of CVD event (cases), the study end date or date of death (from non-CVD causes). No violation of the proportional hazards assumption was identified. The Cox model was used to estimate 10-year CVD risks, constituting the “eHEART risk”.

 Model internal validation and assessment

The eHEART prioritisation model was derived using primary care records from 2/3 of general practices and internally validated on the remaining 1/3 of general practices. In the validation dataset, we calculated measures of predictive accuracy using the Brier score [20] and risk discrimination using the C-index [21]. Calibration was assessed visually by plotting mean predicted risk against mean observed risk by deciles of predicted risk [22, 23] and by assessing the calibration slope, estimated from the linear regression of log transformed observed risk and predicted risk in predicted risk decile groups.

 The QRISK2 formal risk assessment

We calculated formal 10-year CVD risks using QRISK2 (recommended by UK NICE guidelines until May 2023) using the values at the baseline assessment of UK Biobank’s and published coefficients. Chronic kidney disease, family history and historical prescriptions of statins from primary care records were supplemented in addition to the baseline records, and missing values were imputed as described above.

 Population health modelling

We compared the potential public health impact of using eHEART to prioritise individuals for invitation to a formal CVD risk assessment with QRISK2 (Fig 1) versus the whole-population strategy to invite all adults for formal CVD risk assessment. Amongst 174,715 individuals in UK Biobank with linked primary care records prior to baseline, we calculated eHEART and QRISK2 (using complete data available at baseline survey) risk at the individual’s age at baseline.

10.1371/journal.pone.0292240.g001 Flow chart of implementation of eHEART and QRISK2 as a prioritisation tool for formal cardiovascular disease assessments.

Abbreviations: BMI, body mass index; CVD, cardiovascular disease; HDL, high density lipoprotein.

Due to UK Biobank being a cohort of healthier individuals than the UK primary care population, we recalibrated the eHEART and QRISK2 models using age-group- and sex-specific risk factor levels and CVD incidence rates from UK Biobank, and then rescaled the predicted eHEART and QRISK2 risks to achieve 10-year CVD risk distributions that would be expected in a UK primary care setting, using methods previously described [23]. Details are provided in Text 2 in S2 File.

We modelled a population of 50,000 men and 50,000 women aged 40–69 years with age profiles matching that of the contemporary UK population (2017 mid-year population) [24], and CVD incidence rates as observed in CPRD individuals without pre-existing CVD. We assumed a policy of statin initiation for individuals at ≥10% predicted QRISK2 10-year risk as recommended by National Institute for Health and Care Excellence (NICE) guidelines [7] and assumed statin allocation would reduce CVD risk by 20% [25]. We assumed 50% statin compliance and a 50% invitation uptake when formally inviting all individuals for a formal assessment. We then modelled the impact of prioritising individuals for QRISK2 formal assessment, based on a fixed prioritisation threshold of ≥10% estimated eHEART 10-year risk as currently recommended by NICE guidelines [7] and also on age- and sex-specific prioritisation thresholds selected to correspond to 5% false negative rates in UK Biobank; these were chosen by ranking the estimated eHEART 10-year risks in UK Biobank individuals who go on to have a CVD event in the next 10 years. The prioritisation thresholds were chosen as the minimum 10-year risk such that 5% of events would be missed. Population health impact was assessed using the number needed to screen (i.e., undergo formal assessment) to prevent one CVD event, the number of events identified and the number need to invite for formal assessment to prevent one CVD event assuming an increased invitation uptake of 55% after prioritisation [26].

In sensitivity analyses, first we repeated population health-analyses by using QRISK2 (using primary care records) (Text 3 in S2 File) to prioritise individuals for invitation to a formal CVD risk assessment with QRISK2 (using complete data available at baseline-survey). Second, we repeated population-health analyses including all individuals, including those without a primary care record for any one of SBP, HDL, total cholesterol and smoking status. For the eHEART tool, we assumed all these individuals would be directly invited for a formal risk assessment, and for the estimated QRISK2 prioritisation tool, additional missing values were imputed as described above. Third, to assess the use of prioritisation tools at different formal risk assessments thresholds, we repeated the analyses assuming a 5% formal risk assessment threshold in combination with a fixed 5% prioritisation threshold and age- and sex-specific prioritisation thresholds selected to correspond to 2.5% false negative rates in UK Biobank.

Analyses were performed with R x64 3.6.1 [27] and Stata version 15. This study follows the RECORD (Table 2 in S2 File) and TRIPOD (Table 3 in S2 File) reporting guidelines [28, 29]. R code to predict eHEART for new individuals is provided online.

 Patient and public involvement

Patients and the public were not directly involved in designing or implementing this research project. Written informed consent was obtained from all participants

 Results Study population and baseline characteristics using CPRD

We identified 2,154,089 individuals from 398 practices contributing to the CPRD database with linked hospital episode statistics and ONS death registrations in England and who met our inclusion criteria. For our primary analysis we excluded 511,591 (24%) without at least 1 record of SBP, total cholesterol, HDL cholesterol or smoking status, leaving 1,642,498 individuals. A total of 263 practices were randomly assigned to the derivation dataset and the remainder (n = 135) to a validation cohort. Overall, 1,120,053 and 522,445 individuals were included in the derivation and validation cohorts respectively (Fig 2 in S1 File). Compared to those included in our analysis, individuals without any measurement on SBP, cholesterol, or smoking were slightly older, more likely to be male, and had higher CVD incidence rate (Table 1 in S1 File).

The baseline characteristics of individuals are summarised in Table 1 (split by the derivation and validation cohorts in Table 2 in S1 File). Overall, 45% were men and 55% were women; 3% of men and 3% of women had a reported diabetes diagnosis and 16% of men and 23% of women were prescribed anti-hypertension medication before the study entry (Table 1). The number of individuals with CVD risk factors recorded in primary care varied by age and sex (for example, from 21,674 men at age 85 to 188,959 women at age 46) (Figs 3 and 4 in S1 File). There were more repeated SBP than cholesterol measurements recorded (Table 1 and Figs 3 and 4 in S1 File). Characteristics were similar for individuals in the derivation and validation cohorts (Table 2 in S1 File).

10.1371/journal.pone.0292240.t001 Key characteristics of individuals in CPRD cohort.
Men, N = 746,386 Women, N = 896,112
Characteristics Mean (SD) or n (%) No. (%) of persons with value Median (IQR) of measures per person Mean (SD) or n (%) No. (%) of persons with value Median (IQR) of measures per person
Earliest age during follow-up, years 50.4 (12.7) 746,836 (100) - 51.3 (13.6) 896,112 (100) -
History of diabetes§ 21,936 (3.0) 746,836 (100) - 20,662 (3.0) 896,112 (100) -
Blood pressure-lowering medication prescriptions^ 119,230 (16.0) 746,836 (100) - 203,394 (22.7) 896,112 (100) -
Current smoker, n (%)# 192,509 (25.8) 438,638 (59) 3 (2–6) 176,719 (19.7) 424,268 (47) 4 (2–7)
Systolic blood pressure mm Hg, mean (SD)# 136.7 (18.5) 699,229 (93) 4 (2–11) 131.6 (20.3) 870,075 (97) 6 (3–14)
Total cholesterol mmol/litre, mean (SD)# 5.4 (1.0) 532,620 (71) 2 (1–5) 5.6 (1.0) 624,189 (70) 2 (1–5)
HDL cholesterol mmol/litre, mean (SD)# 1.3 (0.4) 490,003 (66) 2 (1–4) 1.6 (0.4) 573,842 (64) 2 (1–4)

* Included 1,642,498 individuals from Clinical Practice Research Datalink, Hospital Episode Statistics, and the Office for National Statistics, England, United Kingdom, 2004–2019, aged 40–85 years, without prevalent CVD and statin treatment before study entry, and had least 1 measurement on any of systolic blood pressure, total cholesterol, HDL cholesterol, or smoking status between their study entry and study exit dates.

§Defined as ever having recorded diagnosis of diabetes before the study entry

^Defined as ever having recorded use of blood-pressure lowering medications before the study entry

#Proportion or mean (standard deviation) of the first ever measurement

In the derivation cohort, 104,830 (6%) individuals had an incident CVD event recorded in either primary care, HES or death registry (Fig 5 in S1 File) over a median 9 years of follow-up (IQR: 5–11). The crude CVD incidence rates per 1000 person years increased with age (from 2.13 in 40-year-olds to 54.88 in those aged 85) and were higher among men than women (Fig 6 in S1 File). In the validation cohort, 32,862 (6%) individuals had an incident CVD event over a median 9 years of follow-up (IQR: 5–11).

 Derivation of the eHEART model

We estimated person-level risk predictor values for smoking status, SBP, HDL and total cholesterol at each age of follow-up; all estimated model coefficients are provided (Table 3 in S1 File). Mean levels of estimated SBP and HDL cholesterol were higher at older ages, as were the proportions of individuals on hypertensive treatment and with diagnosed diabetes. Estimated levels of total cholesterol remained stable across age, whilst the proportion of smokers declined with age (Fig 7 in S1 File).

The eHEART age- and sex-specific hazard ratios for SBP, HDL, total cholesterol, smoking and history of diabetes steadily attenuated towards 1 at older ages, whilst the hazard ratios for hypertensive treatment declined to a lesser extent, especially amongst women (Fig 2). Predicted eHEART risks increased at older ages and changed in distribution from a positive skew to bimodal (due to the contribution of the risk factor for hypertensive treatment) (Fig 8 in S1 File).

10.1371/journal.pone.0292240.g002 Age and sex-specific hazard ratios of eHEART risk predictors with cardiovascular disease in derivation cohort.

Hazard ratios and 95% confidence intervals (shown as vertical lines) for association of cardiovascular disease with: systolic blood pressure, total cholesterol, HDL cholesterol, smoking status, anti-hypertension medication and history of diabetes by age, in men and women in the derivation dataset, Clinical Practice Research Datalink, Hospital Episode Statistics, and the Office for National Statistics, England, United Kingdom, 2004–2019. Hazard ratios are given per standard-deviation increase for SBP, total cholesterol, and HDL cholesterol. Hazard ratios and 95% confidence intervals are shown on the natural log scale.

 Internal validation of eHEART

In the validation cohort, the overall C-index for eHEART was 0.771 (95% CI: 0.769, 0.772) and was higher in women (C-index = 0.786, 95% CI:0.784, 0.788) than in men (C-index = 0.741, 95% CI: 0.739, 0.742) and higher in younger individuals (Fig 3). The Brier score was also lower (better) in women (0.2354, 95% CI: 0.2340, 0.2368) than in men (0.3085, 95% CI: 0.3068, 0.3102) (Table 4 in S1 File). Calibration plots by decile of eHEART risk showed good agreement with observed risk (Figs 9 and 10 in S1 File). The calibration slope was more deviatory from 1 for older ages (Fig 11 in S1 File).

10.1371/journal.pone.0292240.g003 Age and sex-specific C indices of eHEART in validation cohort.

C-indices and 95% confidence intervals (shown as vertical lines) from eHEART for the prediction of 10-year cardiovascular disease by age and for all ages, in men and women in the validation dataset, Clinical Practice Research Datalink, Hospital Episode Statistics, and the Office for National Statistics, England, United Kingdom, 2004–2019. The overall C-indices account for the discriminatory information in age, hence are higher than the age-specific C-indices.

 Population health modelling

Overall, 119,137 individuals from UK Biobank, with primary care records available, were used to model a representative population of 50,000 men and 50,000 women aged 40–69 in the UK (Fig 12 and Table 5 in S1 File). We estimated that there would be 3,566 and 1,819 CVD events over the next 10 years in men and women respectively. If QRISK2 was used as a formal assessment on the whole population, then 2,704 (75.8%) men and 895 (49.2%) women with CVD events over the next 10 years would be classified at high risk, ie QRISK2 ≥10% (Table 2). These percentages of “events captured” varied considerably by age group; for example, only 19% of events amongst 40–49-year-old men but 98% of events amongst 60–69-year-old men would be captured. Assuming statin therapy would be initiated for persons with predicted QRISK2 ≥10%, the number needed to screen to prevent one CVD event in men and in women would be 93 and 279 respectively. Assuming 50% invitation uptake, the number needed to invite to prevent one CVD event in men and in women would be 185 and 559 respectively (Table 6 in S1 File).

10.1371/journal.pone.0292240.t002 Number needed to screen to prevent one event, and number of events captured when prioritising with eHEART in a hypothetical population of 100,000 individuals in England assuming all individuals invited for formal assessment attend.
Hypothetical population of 100,000 Full formal assessment only (QRISK2 threshold = 10%) Prioritisation tool (eHEART threshold = 10%) followed by full formal assessment (QRISK2 threshold = 10%) Prioritisation tool (eHEART threshold corresponding to age- and sex-specific 5% false negative rates) followed by full formal assessment (QRISK2 threshold = 10%)
Sex Age group Expected N Expected CVD events in 10 years Events captured N (%) Number needed to screen to prevent 1 CVD event Individuals prioritised N Events captured N (%) Number needed to screen to prevent 1 CVD event Individuals prioritised N Prioritisation threshold for eHEART Events captured N (%) Number needed to screen to prevent 1 CVD event
Men 40–49 17673 516 96 (18.6%) 1832.8 (1335.2, 2168.8) 598 32 (6.3%) 183.2 (76.5, 236.5) 11054 3.0% 91 (17.7%) 1204.7 (849.1, 1435.1)
50–59 18061 1294 844 (65.2%) 214.1 (204.3, 223.2) 6348 518 (40.0%) 122.7 (112.6, 131.0) 14576 6.2% 810 (62.6%) 180.0 (170.7, 188.1)
60–69 14266 1756 1714 (97.6%) 83.2 (82.6, 83.8) 13391 1659 (94.5%) 80.8 (79.8, 81.6) 13391 10% 1659 (94.5%) 80.8 (79.8, 81.6)
Total 50000 3566 2654 (74.4%) 188.4 (185.2, 191.5) 20338 2209 (61.9%) 92.1 (90.4, 93.9) 39021 NA 2560 (71.8%) 152.4 (149.6, 155.1)
Women 40–49 17488 277 17 (6.1%) 10605.6 (643.6, 14255.1) 78 7 (2.5%) 123.7 (0.0, 183.1) 5495 2.0% 16 (5.7%) 3610.2 (0.0, 4933.9)
50–59 17986 607 116 (19.1%) 1552.2 (1254.2, 1787.3) 427 28 (4.5%) 156.4 (75.4, 198.2) 10519 3.3% 113 (18.7%) 930.5 (741.5, 1075.0)
60–69 14526 936 699 (74.7%) 207.9 (201.1, 214.4) 5174 437 (46.7%) 118.4 (110.5, 124.6) 11850 5.9% 677 (72.4%) 175.0 (168.7, 180.5)
Total 50000 1819 831 (45.7%) 601.6 (578.8, 622.7) 5680 471 (25.9%) 120.6 (113.1, 127.2) 27864 NA 805 (44.3%) 346.0 (332.1, 358.4)

Age structure of hypothetical population extrapolated from Office for National Statistics, England, United Kingdom 2017. Expected events at 10 years based on extrapolation of incidence rates from CPRD, 2014–2019. Age group and sex specific prioritisation thresholds were defined as the minimum of 10% and the level such that the expected false negative rate is controlled to be 5%. All individuals have at least one CVD risk factor (systolic blood pressure, smoking, total and/or HDL cholesterol) recorded for eHEART.

Number needed to screen based on assuming that all individuals are formally assessed. Under each scenario, some individuals are not formally assessed and may go on to have events. Statin compliance assumed to be equal to 50%.

If the eHEART tool with a prioritisation threshold of 10% was used to prioritise QRISK2 assessment in the population (and assuming all individuals had at least one primary care record of either smoking, SBP, HDL or total cholesterol) then 2,258 (63.3%) men and 573 (31.5%) women with CVD events over the next 10 years would be classified at high risk (Table 2). The number needed to screen to prevent one CVD event in men and in women would be 45 and 50 respectively. The number needed to invite to prevent one CVD event in men and in women would be 82 and 90 respectively (Table 6 in S1 File). In contrast, using age- and sex-specific prioritisation thresholds corresponding to 5% false negative rates would classify 2,616 (73.4%) men and 861 (47.3%) women with CVD events over the next 10 years as high risk. The number needed to screen to prevent one CVD event in men and in women would be 75 and 162 respectively. Assuming 55% invitation uptake after prioritisation, the number need to invite to prevent one CVD event in men and in women would be 136 and 294 respectively (Table 6 in S1 File).

 Sensitivity analyses

In comparison to eHEART being used as a prioritisation tool, if estimated QRISK2 with a prioritisation threshold of 10% was used to prioritise formal assessment in the population, then 2,409 (67.6%) men and 725 (39.8%) women with CVD events over the next 10 years would be classified at high risk. This equates to capturing approximately 87% of events when applying formal CVD risk assessment on the whole population (Table 7 in S1 File). The number needed to screen to prevent one CVD event in men and in women would be 48 and 67, a reduction of 48% and 76%, respectively, when applying formal CVD risk assessment to the whole population. Notably, the events captured reduced by at least half amongst men under 50 years and women under 60 years. In contrast, using age- and sex-specific prioritisation thresholds corresponding to 5% false negative rates would classify 2,656 (74.5%) men and 871 (47.9%) women with CVD events over the next 10 years as high risk, around 98% events captured when applying formal CVD risk assessment on the whole population. The number needed to screen to prevent one CVD event in men and in women would be 73 (a reduction of 21%) and 148 (a reduction of 47%) respectively, with greatest reductions observed amongst younger individuals. The number need to invite to prevent one CVD event in men and in women would be 133 and 268 respectively (Table 8 in S1 File).

In sensitivity analyses including all individuals (i.e., all those without a primary care record for any one of SBP, HDL, total cholesterol and smoking status) (Table 9 in S1 File), we found comparable results for eHEART and estimated QRISK2, but as expected, we observed greater numbers needed to screen especially among younger individuals (Tables 10, 11 in S1 File).

Comparable results were observed in additional analyses assuming a 5% (rather than 10%) formal risk assessment threshold (Tables 12 and 13 in S1 File). Using a fixed 5% prioritisation threshold resulted in a greater number needed to screen to prevent one CVD, than using a fixed 10% threshold for both prioritisation and formal assessment, however 88.5% of future CVD events in men and 63.3% in women being captured if prioritising with eHEART, and 91.3% of events in men and 71.8% of events in women being captured if prioritising with estimated QRISK2. Using age- and sex-specific prioritisation thresholds to limit the false negative rate to 2.5% resulted in comparable results for eHEART and estimated QRISK2.

 Discussion

We have provided quantitative evidence for systematically prioritising individuals for invitation for full formal CVD risk assessment using existing longitudinal primary care records, by evaluating a novel prioritisation tool (eHEART) with the use of age- and sex-specific prioritisation thresholds. Compared to conducting formal assessments on all individuals, the prioritisation tool has the potential to reduce the number needed to screen to prevent one CVD event by approximately 50% for women and 20% for men whilst identifying 96–98% of high-risk individuals. We found no added value of using the repeat measures in the eHEART prioritisation tool in comparison to estimates using single measures in QRISK2 as a prioritisation tool. The use of a fixed 10% prioritisation threshold substantially reduces the number of high-risk individuals identified by 10–20%. Our study highlights the importance of using prioritisation thresholds well below formal risk assessment thresholds, and lower than the 10% prioritisation threshold currently recommended by the NICE guidelines in the UK [7]. We demonstrated the advantage of using a pragmatic strategy of selecting age- and sex-specific prioritisation thresholds corresponding to 5% false negative rates, to ensure a balance between efficiency and specificity across age and sex.

This is the first study, to our knowledge, to demonstrate the benefits of age- and sex-specific thresholds for prioritising individuals for full formal CVD risk assessment, although they have been shown to improve specificity when applied directly to the formal CVD risk assessment [30, 31]. Note that whilst CVD risk prediction models already account for age and sex, they are not optimised to predict risk at any single threshold or for all subgroups of the population. Hence, there can be major differences in the specificity and sensitivity of CVD risk models used in combination with a threshold in different subgroups. In our study we specified the age- and sex- specific thresholds to minimise the false negative rate to 5% in each subgroup. However, alternative thresholds that optimise for a different criterion, such as a 1% or 10% false negative rate, could be chosen to achieve varying specificity and efficiency gains.

We have illustrated the use of prioritisation tools within the current United Kingdom population-wide preventative programme, in which all individuals aged between 40 and 74 years are invited into their general practitioners for a National Health Service (NHS) health check to assess an individual’s risk of CVD, diabetes, kidney disease and stroke and dementia every five years [32]. A systematic review reported coverage (percentage eligible for health checks who received one) of 45.6% [26] and invitation uptake (percentage invited for health checks who received one) of 48.2% between 2009–2016 and suggested improved coverage and invitation uptake in populations targeted for prioritisation through general practices, including those with low socioeconomic status or ethnic minorities. During the COVID-19 pandemic, invited health checks in England declined by 82% between the end of 2019 and 2020 [33], leaving a concerning backlog [34–36]. As such, a complementary and automated systematic primary care records based prioritisation approach could be supportive to improve both the coverage and uptake of the programme over the next few years, helping to reduce health inequalities [37].

This is also the first study to provide an independent evaluation of the use of the existing QRISK2 as a prioritisation tool compared with our novel eHEART tool. The derivation, validation and implementation of eHEART accounts for and leverages the sparse and sporadically observed longitudinal primary care records, by estimating current predictor values based on all past primary care records to make 10-year CVD predictions. By implementing a landmark model approach in deriving eHEART, the tool removes the need for complete risk predictor measurements in all individuals. Furthermore, eHEART uses only a small number of conventional risk predictors, thus making it transportable across different primary care electronic systems and country settings. In contrast, whilst the published derivation and validation methods for QRISK2 have used multiple imputation to handle non-recorded risk factors, its’ clinical implementation is based on replacing missing non-recorded values with age, sex and ethnicity-specific population average values. However, QRISK2 contains many more risk factors than eHEART and is already integrated into the primary care electronic systems in England and used to prioritise patient invitation to formal CVD risk assessment. We conducted population health modelling utilising both primary care records and baseline measurements in UK Biobank, and used recalibrated risk estimates and observed incidence rates for estimating the population-health impact in England. Our results show only small differences between the tools, and that future focus should be on the selection and use of appropriate prioritisation thresholds.

Our study has potential limitations. First, we considered the QRISK2 rather than QRISK3 model, which includes risk factors such as disease history in 1st degree relative and stage of chronic kidney disease that are difficult to capture in electronic health records. However, we expect similar population health modelling results for QRISK2 and QRISK3, especially because of recalibration approach. Second, thresholds were selected by age-group and sex; alternative thresholds such as ethnic-specific thresholds could also be considered. Third, optimism may exist due to possible overlap of a small proportion of individuals in UK Biobank and in the databases used in deriving eHEART and QRISK2. Fourth, our population health modelling may be affected by healthy volunteer bias in UK Biobank; however, this is largely overcome by the recalibration and rescaling with more representative risk factor levels and incidence rates. Fifth, our population health modelling was restricted to individuals aged between 40–69 years due to data availability; we expect even larger efficiency gains if such a prioritisation tool was used in individuals <40 years but not >69 years. Sixth, we assumed a fixed level of statin compliance and uptake for all individuals; whilst these are likely to vary by age and sex [38–40], the levels of compliance and uptake are likely to change after prioritisation with age- and sex-specific thresholds is introduced. Seventh, our population health modelling focused on prioritising individuals for a formal CVD assessment at a single point in time, rather than every 5 or 10 years; however such prioritisation tools can be used in real-time with updated primary care records. Finally, our population health modelling was restricted to the population of England assuming use of the QRISK2 formal CVD risk assessment tool; future work should investigate the generalisability of the tools across different countries and in combination with other formal risk assessment tools, including QRISK3.

 Conclusions

The use of a prioritisation tool, based on either utilising repeated measures or last observed values of primary care records, in conjunction with age- and sex-specific thresholds has the potential to systematically identify high-risk individuals for invitation for full formal CVD risk assessment. Our results suggest using age- and sex-specific thresholds can reduce the number needed to screen to prevent one CVD event whilst identifying the majority of high-risk individuals. Such a strategy could be particularly important as nations tackle the many COVID-19 related burdens and backlogs in primary care settings.

 Supporting information Supplementary tables and figures.

(PDF)

 Codelists and supplementary methods.

(PDF)

CPRD uses data provided by patients and collected by the NHS as part of their care and support. This work was supported by Health Data Research UK, which is funded by the UK Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, Department of Health and Social Care (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. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission. This work was performed using resources provided by the Cambridge Service for Data Driven Discovery (CSD3) operated by the University of Cambridge Research Computing Service (www.csd3.cam.ac.uk).

 References WHO/Europe | Cardiovascular diseases ‐ Data and statistics [Internet]. [cited 2022 Apr 1]. Available from: https://www.euro.who.int/en/health-topics/noncommunicable-diseases/cardiovascular-diseases/data-and-statistics National Vascular Disease Prevention Alliance. Guidelines for the management of absolute CVD risk 2012 [Internet]. 2012 [cited 2021 Nov 23]. Available from: https://informme.org.au/en/Guidelines/Guidelines-for-the-assessment-and-management-of-absolute-CVD-risk Visseren FLJ, MacH F, Smulders YM, Carballo D, Koskinas KC, Bäck M, et al. 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice. Eur Heart J [Internet]. 2021 Sep 7 [cited 2021 Nov 23];42(34):3227–337. Available from: https://pubmed.ncbi.nlm.nih.gov/34458905/ New Zealand Ministry of Health. Cardiovascular Disease Risk Assessment and Management for Primary Care 2018. Arnett DK, Blumenthal RS, Albert MA, Buroker AB, Goldberger ZD, Hahn EJ, et al. 2019 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019 Sep 10;140(11):e596–646. Pearson GJ, Thanassoulis G, Anderson TJ, Barry AR, Couture P, Dayan N, et al. 2021 Canadian Cardiovascular Society Guidelines for the Management of Dyslipidemia for the Prevention of Cardiovascular Disease in Adults. Can J Cardiol [Internet]. 2021 Aug 1 [cited 2021 Nov 23];37(8):1129–50. Available from: https://pubmed.ncbi.nlm.nih.gov/33781847/ National Institute for Health and Care Excellence (NICE). Lipid modification: Cardiovascular risk assessment and the modification of blood lipids for the primary and secondary prevention of cardiovascular disease (clinical guideline CG181). 2014. Stewart J, Addy K, Campbell S, Wilkinson P. Primary prevention of cardiovascular disease: Updated review of contemporary guidance and literature. JRSM Cardiovasc Dis [Internet]. 2020 Jan [cited 2022 Apr 1];9:204800402094932. Available from: https://pubmed.ncbi.nlm.nih.gov/32994926/ doi: 10.1177/2048004020949326 32994926 Hippisley-Cox J, Coupland C, Vinogradova Y, Robson J, Minhas R, Sheikh A, et al. Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2. BMJ [Internet]. 2008 Jun 28 [cited 2021 Nov 23];336(7659):1475–82. Available from: https://pubmed.ncbi.nlm.nih.gov/18573856/ doi: 10.1136/bmj.39609.449676.25 18573856 Ltd. IPS. The Health Improvement Network (THIN) | Vision [Internet]. [cited 2021 Nov 23]. Available from: https://www.visionhealth.co.uk/portfolio-items/the-health-improvement-network-thin/ Tate AR, Dungey S, Glew S, Beloff N, Williams R, Williams T. Quality of recording of diabetes in the UK: how does the GP’s method of coding clinical data affect incidence estimates? Cross-sectional study using the CPRD database. BMJ Open [Internet]. 2017 Jan 1 [cited 2021 Nov 26];7(1). Available from: /pmc/articles/PMC5278252/ doi: 10.1136/bmjopen-2016-012905 28122831 NHS Digital. Quality and Outcomes Framework, 2020–21 [Internet]. 2021 [cited 2021 Nov 24]. Available from: https://digital.nhs.uk/data-and-information/publications/statistical/quality-and-outcomes-framework-achievement-prevalence-and-exceptions-data/2020-21 Biobank UK. Protocol for a large-scale prospective epidemiological resource. 2006. UK Biobank Primary Care Linked Data. 2019 [cited 2022 May 30]; Available from: https://www.ukbiobank.ac.uk Damen JAAG, Hooft L, Schuit E, Debray TPA, Collins GS, Tzoulaki I, et al. Prediction models for cardiovascular disease risk in the general population: Systematic review. Vol. 353, BMJ (Online). BMJ Publishing Group; 2016. Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, Staa T van, et al. Data Resource Profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol. 2015 Jun 1;44(3):827–36. doi: 10.1093/ije/dyv098 26050254 Welch C, Petersen I, Walters K, Morris RW, Nazareth I, Kalaitzaki E, et al. Two-stage method to remove population- and individual-level outliers from longitudinal data in a primary care database. Pharmacoepidemiol Drug Saf [Internet]. 2012 Jul [cited 2021 Nov 23];21(7):725–32. Available from: https://pubmed.ncbi.nlm.nih.gov/22052713/ Paige E, Barrett J, Stevens D, Keogh RH, Sweeting MJ, Nazareth I, et al. Landmark Models for Optimizing the Use of Repeated Measurements of Risk Factors in Electronic Health Records to Predict Future Disease Risk. Am J Epidemiol [Internet]. 2018 Jul 1 [cited 2021 Nov 23];187(7):1530–8. Available from: https://academic.oup.com/aje/article/187/7/1530/4952104 doi: 10.1093/aje/kwy018 29584812 Laird NM, Ware JH. Random-Effects Models for Longitudinal Data. Biometrics. 1982 Dec;38(4):963. 7168798 Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology [Internet]. 2010 Jan [cited 2021 Nov 23];21(1):128–38. Available from: https://pubmed.ncbi.nlm.nih.gov/20010215/ doi: 10.1097/EDE.0b013e3181c30fb2 20010215 Lloyd-Jones DM. Cardiovascular risk prediction: Basic concepts, current status, and future directions. Circulation [Internet]. 2010 Apr 20 [cited 2021 Nov 23];121(15):1768–77. Available from: doi: 10.1161/CIRCULATIONAHA.109.849166 20404268 Kaptoge S, Di Angelantonio E, Pennells L, Wood AM, White IR, Gao P, et al. C-reactive protein, fibrinogen, and cardiovascular disease prediction. N Engl J Med. 2012;367(14):1310–20. doi: 10.1056/NEJMoa1107477 23034020 Pennells L, Kaptoge S, Wood A, Sweeting M, Zhao X, White I, et al. Equalization of four cardiovascular risk algorithms after systematic recalibration: Individual-participant meta-analysis of 86 prospective studies. Eur Heart J. 2019;40(7):621–31. doi: 10.1093/eurheartj/ehy653 30476079 Office for National Statistics. Population estimates [Internet]. [cited 2021 Nov 26]. Available from: https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationestimates#datasets Collins R, Reith C, Emberson J, Armitage J, Baigent C, Blackwell L, et al. Interpretation of the evidence for the efficacy and safety of statin therapy. Lancet [Internet]. 2016 Nov 19 [cited 2021 Nov 23];388(10059):2532–61. Available from: http://www.thelancet.com/article/S0140673616313575/fulltext doi: 10.1016/S0140-6736(16)31357-5 27616593 Martin A, Saunders CL, Harte E, Griffin SJ, MacLure C, Mant J, et al. Delivery and impact of the NHS Health Check in the first 8 years: A systematic review. Br J Gen Pract [Internet]. 2018 Jul 1 [cited 2020 Oct 27];68(672):e449–59. Available from: /pmc/articles/PMC6014431/?report=abstract doi: 10.3399/bjgp18X697649 29914882 3.6.1. RDCT. A Language and Environment for Statistical Computing [Internet]. Vol. 2, R Foundation for Statistical Computing. 2018. p. https://www.R-project.org. Available from: http://www.r-project.org Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Peteresen I, et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) Statement. PLOS Med [Internet]. 2015 [cited 2021 Nov 23];12(10):e1001885. Available from: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1001885 Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD): The TRIPOD Statement. Circulation [Internet]. 2015 [cited 2021 Nov 24];131(2):211. Available from: /pmc/articles/PMC4297220/ Yebyo HG, Zappacosta S, Aschmann HE, Haile SR, Puhan MA. Global variation of risk thresholds for initiating statins for primary prevention of cardiovascular disease: A benefit-harm balance modelling study. BMC Cardiovasc Disord [Internet]. 2020 Sep 17 [cited 2022 Jun 9];20(1):1–10. Available from: https://bmccardiovascdisord.biomedcentral.com/articles/10.1186/s12872-020-01697-6 Navar-Boggan AM, Peterson ED, D’Agostino RB, Pencina MJ, Sniderman AD. Using age- and sex-specific risk thresholds to guide statin therapy: One size may not fit all. J Am Coll Cardiol. 2015 Apr 28;65(16):1633–9. doi: 10.1016/j.jacc.2015.02.025 25743624 Public Health England. Guidance overview: NHS Health Checks: applying All Our Health ‐ GOV.UK [Internet]. [cited 2021 Nov 23]. Available from: https://www.gov.uk/government/publications/nhs-health-checks-applying-all-our-health Public Health England. NHS Health Check ‐ Data ‐ OHID [Internet]. [cited 2022 Feb 15]. Available from: https://fingertips.phe.org.uk/profile/nhs-health-check-detailed/data#page/4/gid/1938132726/pat/159/par/K02000001/ati/15/are/E92000001/iid/91733/age/219/sex/4/cat/-1/ctp/-1/yrr/1/cid/4/tbm/1 Health and Social Care Committee. Clearing the backlog caused by the pandemic [Internet]. [cited 2022 Feb 15]. Available from: https://publications.parliament.uk/pa/cm5802/cmselect/cmhealth/599/report.html Iacobucci G. Covid-19: Government must fund extra NHS capacity to tackle backlog, BMA urges. BMJ [Internet]. 2020 Sep 15 [cited 2022 Feb 15];370:m3587. Available from: https://www.bmj.com/content/370/bmj.m3587 Levene LS, Seidu S, Greenhalgh T, Khunti K. Pandemic threatens primary care for long term conditions. BMJ [Internet]. 2020 Oct 5 [cited 2022 Feb 15];371. Available from: https://www.bmj.com/content/371/bmj.m3793 doi: 10.1136/bmj.m3793 33020049 NHS Digital. Appointments in General Practice August 2021 [Internet]. [cited 2021 Nov 24]. Available from: https://digital.nhs.uk/data-and-information/publications/statistical/appointments-in-general-practice/august-2021 Colantonio LD, Rosenson RS, Deng L, Monda KL, Dai Y, Farkouh ME, et al. Adherence to statin therapy among US adults between 2007 and 2014. J Am Heart Assoc [Internet]. 2019 Jan 1 [cited 2022 Apr 26];8(1). Available from: doi: https://www.ahajournals.org/doi/abs/10.1161/JAHA.118.010376 30616455 Hope HF, Binkley GM, Fenton S, Kitas GD, Verstappen SMM, Symmons DPM. Systematic review of the predictors of statin adherence for the primary prevention of cardiovascular disease. PLoS One [Internet]. 2019 Jan 1 [cited 2022 Apr 26];14(1):e0201196. Available from: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0201196 Chamnan P, Simmons RK, Khaw KT, Wareham NJ, Griffin SJ. Estimating the population impact of screening strategies for identifying and treating people at high risk of cardiovascular disease: modelling study. BMJ [Internet]. 2010 Apr 25 [cited 2022 Apr 26];340(7754):1016. Available from: https://www.bmj.com/content/340/bmj.c1693 doi: 10.1136/bmj.c1693 20418545 10.1371/journal.pone.0292240.r001 Decision Letter 0 Pontikos Nikolas Academic Editor 2023 Nikolas Pontikos This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 0

21 Jul 2023

PONE-D-23-05392Prioritising cardiovascular disease risk assessment to high risk individuals based on primary care recordsPLOS ONE

Dear Dr. Chung,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Sep 04 2023 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Nikolas Pontikos

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf.

2. Please ensure that you have specified (1) whether consent was informed and (2) what type you obtained (for instance, written or verbal, and if verbal, how it was documented and witnessed). If your study included minors, state whether you obtained consent from parents or guardians. If the need for consent was waived by the ethics committee, please include this information.

3. Thank you for stating in your Funding Statement:

“his work was supported by core funding from the: British Heart Foundation (RG/13/13/30194; RG/18/13/33946), BHF Cambridge Centre of Research Excellence (RE/13/6/30180) and NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. *The views expressed are those of the author(s) and not necessarily those of the NIHR, NHSBT or the Department of Health and Social Care.”

Please provide an amended statement that declares *all* the funding or sources of support (whether external or internal to your organization) received during this study, as detailed online in our guide for authors at http://journals.plos.org/plosone/s/submit-now.  Please also include the statement “There was no additional external funding received for this study.” in your updated Funding Statement.

Please include your amended Funding Statement within your cover letter. We will change the online submission form on your behalf.

4. Thank you for stating the following in the Competing Interests section:

“During the drafting of the manuscript, M.A. became an employee of AstraZeneca.”

We note that one or more of the authors are employed by a commercial company: AstraZeneca

a.             Please provide an amended Funding Statement declaring this commercial affiliation, as well as a statement regarding the Role of Funders in your study. If the funding organization did not play a role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript and only provided financial support in the form of authors' salaries and/or research materials, please review your statements relating to the author contributions, and ensure you have specifically and accurately indicated the role(s) that these authors had in your study. You can update author roles in the Author Contributions section of the online submission form.

Please also include the following statement within your amended Funding Statement.

“The funder provided support in the form of salaries for authors [insert relevant initials], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.”

If your commercial affiliation did play a role in your study, please state and explain this role within your updated Funding Statement.

b. Please also provide an updated Competing Interests Statement declaring this commercial affiliation along with any other relevant declarations relating to employment, consultancy, patents, products in development, or marketed products, etc. 

Within your Competing Interests Statement, please confirm that this commercial affiliation does not alter your adherence to all PLOS ONE policies on sharing data and materials by including the following statement: "This does not alter our adherence to  PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests) . If this adherence statement is not accurate and  there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include both an updated Funding Statement and Competing Interests Statement in your cover letter. We will change the online submission form on your behalf.

4. PLOS requires an ORCID iD for the corresponding author in Editorial Manager on papers submitted after December 6th, 2016. Please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field. This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager. Please see the following video for instructions on linking an ORCID iD to your Editorial Manager account: https://www.youtube.com/watch?v=_xcclfuvtxQ.

5. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“This work was performed using resources provided by the Cambridge Service for Data Driven Discovery (CSD3) operated by the University of Cambridge Research Computing Service (www.csd3.cam.ac.uk), provided by Dell EMC and Intel using Tier-2 funding from the Engineering and Physical Sciences Research Council”

We note that you have provided additional information within the Acknowledgements Section that is not currently declared in your Funding Statement. Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“This work was supported by core funding from the: British Heart Foundation (RG/13/13/30194; RG/18/13/33946), BHF Cambridge Centre of Research Excellence (RE/13/6/30180) and NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. *The views expressed are those of the author(s) and not necessarily those of the NIHR, NHSBT or the Department of Health and Social Care.

This work was funded by the Medical Research council (MR/K014811/1). The study funders played no role in the design, analysis or interpretation of the study. R.C. is funded by a BHF PhD studentship (FS/18/56/34177). Z.X. is funded by the Chinese Scholarship Council. M.A. was funded by a British Heart Foundation Programme Grant (RG/18/13/33946). D.S. was funded by the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. J.B. was funded by a Medical Research Council fellowship (MR/L501566/1) and unit programme (MC_UU_00002/5). L.G.K. was funded by the NIHR BTRU in Donor Health and Genomics (NIHR BTRU-2014–10024) and is funded by the NIHR BTRU in Donor Health and Behaviour (NIHR203337) [*]. J.A.U. is funded by an NIHR Advanced Fellowship (NIHR300861). A.M.W. is part of the BigData@Heart Consortium, funded by the Innovative Medicines Initiative-2 Joint Undertaking under grant agreement No 116074. A.M.W. is supported by the BHF-Turing Cardiovascular Data Science Award (BCDSA\\100005).”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

6. Your ethics statement should only appear in the Methods section of your manuscript. If your ethics statement is written in any section besides the Methods, please move it to the Methods section and delete it from any other section. Please ensure that your ethics statement is included in your manuscript, as the ethics statement entered into the online submission form will not be published alongside your manuscript.

7. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Additional Editor Comments (if provided):

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Thank you for allowing me to review the article. In this study, Chung et al. demonstrated using age- and sex-specific thresholds can lead to more efficient CVD assessment with eHEART. The study provides valuable insights into precision medicine as CVD differs among gender, and age is a significant contributor. The methodology was robust and well-written, and I enjoyed learning from it. However, I have some questions that I would like to discuss.

Why do you choose QRISK2, not the newer QRISK3 for comparison?

While all CVD risk prediction models such as Framingham, PCE, and QRISK3 account for age and sex, it is unclear why using age-sex thresholds still makes a difference. For instance, line 284-286 indicates that the percentages of “events captured” varied considerably by age group, with only 19% of events among 40-49-year-old men but 98% of events among 60-69-year-old men being captured. Even though QRISK2 also takes age and sex into account, why such significant differences are still observed?

Could it be possible that this improvement in prioritization resulted from recalibrating the algorithm on newer data that reflect the younger trend in CVD detection rate over recent years, whereas the data used for developing QRISK2 is old?

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

 10.1371/journal.pone.0292240.r002 Author response to Decision Letter 0 Submission Version 1

5 Sep 2023

Amended funding statement

This work was supported by core funding from the: British Heart Foundation (RG/13/13/30194; RG/18/13/33946), BHF Cambridge Centre of Research Excellence (RE/13/6/30180) and NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. *The views expressed are those of the author(s) and not necessarily those of the NIHR, NHSBT or the Department of Health and Social Care.

This work was funded by the Medical Research council (MR/K014811/1). The study funders played no role in the design, analysis or interpretation of the study. R.C. is funded by a BHF PhD studentship (FS/18/56/34177). Z.X. is funded by the Chinese Scholarship Council. M.A. was funded by a British Heart Foundation Programme Grant (RG/18/13/33946). D.S. was funded by the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) [*]. J.B. was funded by a Medical Research Council fellowship (MR/L501566/1) and unit programme (MC_UU_00002/5). L.G.K. was funded by the NIHR BTRU in Donor Health and Genomics (NIHR BTRU-2014–10024) and is funded by the NIHR BTRU in Donor Health and Behaviour (NIHR203337) [*]. J.A.U. is funded by an NIHR Advanced Fellowship (NIHR300861). A.M.W. is part of the BigData@Heart Consortium, funded by the Innovative Medicines Initiative-2 Joint Undertaking under grant agreement No 116074. A.M.W. is supported by the BHF-Turing Cardiovascular Data Science Award (BCDSA\\100005).

Cambridge Service for Data Driven Discovery (CSD3) was provided by Dell EMC and Intel using Tier-2 funding from the Engineering and Physical Sciences Research Council (capital grant EP/P020259/1), and DiRAC funding from the Science and Technology Facilities Council (www.dirac.ac.uk).

There was no additional external funding received for this study

The funder provided support in the form of salaries for authors M.A., but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.

Amended competing interests statement

During the drafting of the manuscript, M.A. became an employee of AstraZeneca. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Reviewer’s comments

Reviewer 1: Thank you for allowing me to review the article. In this study, Chung et al. demonstrated using age- and sex-specific thresholds can lead to more efficient CVD assessment with eHEART. The study provides valuable insights into precision medicine as CVD differs among gender, and age is a significant contributor. The methodology was robust and well-written, and I enjoyed learning from it. However, I have some questions that I would like to discuss.

 

Why do you choose QRISK2, not the newer QRISK3 for comparison?

Up until May 2023 the NICE guidelines recommended the use of QRISK2 in clinical practice. Since May 2023 QRISK3 is recommended. There are challenges in implementing the QRISK3 algorithm in research studies because it requires information on (i) stage of chronic kidney disease and (ii) angina or heart attack in a 1st degree relative, which are not well coded in electronic health records. For this reason, we retain our comparison against QRISK2. However, we expect similar population health modelling results for QRISK2 and QRISK3, especially because of our recalibration approach.

We have justified our use of QRISK2 by adding the following text to the methods (lines 168-170):

“We calculated formal 10-year CVD risks using QRISK2 (recommended by UK NICE guidelines until May 2023) using the values at the baseline assessment of UK Biobank’s and published coefficients.”

and added the following text to the limitations section in the discussion (lines 380-386 & 402-403):

“First, we considered the QRISK2 rather than QRISK3 model, which includes risk factors such as disease history in 1st degree relative and stage of chronic kidney disease that are difficult to capture in electronic health records. However, we expect similar population health modelling results for QRISK2 and QRISk3, especially because of recalibration approach. “

“future work should investigate the generalisability of the tools across different countries and in combination with other formal risk assessment tools, including QRISK3. “

While all CVD risk prediction models such as Framingham, PCE, and QRISK3 account for age and sex, it is unclear why using age-sex thresholds still makes a difference. For instance, line 284-286 indicates that the percentages of “events captured” varied considerably by age group, with only 19% of events among 40-49-year-old men but 98% of events among 60-69-year-old men being captured. Even though QRISK2 also takes age and sex into account, why such significant differences are still observed?

All CVD risk prediction models are statistically optimised to predict 10-year CVD risk across the full distribution of risk and for all individuals in the training dataset. That does not mean they are optimised at any single risk threshold for all individuals, or for subgroups of individuals. It is also likely that some risk models are less optimal for certain subgroups because of lower numbers of events contributing to the derived model.

Hence, we find differences occurring by age and sex groups at a single 10% threshold.

We have added the following text in the introduction to clarify this point:

“Note that whilst CVD risk prediction models already account for age and sex, they are not optimised to predict risk at any single threshold or for all subgroups of the population. Hence, there can be major differences in the specificity and sensitivity of CVD risk models used in combination with a threshold in different subgroups.”

Could it be possible that this improvement in prioritization resulted from recalibrating the algorithm on newer data that reflect the younger trend in CVD detection rate over recent years, whereas the data used for developing QRISK2 is old?

For the reason above, we believe that the key factor that contributes to the improvement in prioritisation is due to the use of the age-and sex specific thresholds. By comparing the impact of using a fixed 10% threshold, whilst keeping all other conditions fixed, we showed that the choice in risk thresholds has the largest impact on public health measures.

Submitted filename: Response to Reviewers.docx

 10.1371/journal.pone.0292240.r003 Decision Letter 1 Pontikos Nikolas Academic Editor 2023 Nikolas Pontikos This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 1

18 Sep 2023

Prioritising cardiovascular disease risk assessment to high risk individuals based on primary care records

PONE-D-23-05392R1

Dear Dr. Chung,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Nikolas Pontikos

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

The authors seemed to have addressed the reviewer's comments.

Reviewers' comments:

 10.1371/journal.pone.0292240.r004 Acceptance letter Pontikos Nikolas Academic Editor 2023 Nikolas Pontikos This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

20 Sep 2023

PONE-D-23-05392R1

Prioritising cardiovascular disease risk assessment to high risk individuals based on primary care records

Dear Dr. Chung:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Nikolas Pontikos

Academic Editor

PLOS ONE