In silico trials of surgical interventions - using routinely collected data to model trial feasibility and design efficiency in vivo randomised controlled trials

This study will be undertaken by the Leicester Cardiac Surgery Research Group at Department of Cardiovascular Sciences based in the University of Leicester. This study is to establish a database of patients with cardiovascular diagnosis in England using routinely collected clinical and mortality data obtained from NHS England. This database will be used to model trials of surgical interventions in silico and devise a set of pragmatic trial proposals to address the priority research questions in cardiac surgery.

HES DATA

The gold standard for the evaluation of medical treatment is through randomised controlled trials. However, the successful delivery of clinical trials are often limited by the many assumptions that are required for design and planning. Assumptions are often made with respect to recruitment, eligibility, event rates, effect estimates, safety and attrition in real world populations. Hospital Episode Statistics (HES) data contains a wealth of real-world data including demographics, diagnoses, procedures and other clinical information collected prospectively from all NHS hospitals. This routinely collected clinical data enables the University of Leicester to explore the effect of interventions (e.g. open heart vs minimally invasive heart surgery) in a population and across different patient subgroups (e.g. elderly patients, or patients with other medical conditions), and obtain the necessary parameters required for designing a clinical trial. The University of Leicester may also optimize the trial design through sensitivity analyses of the inclusion/exclusion criteria of patient populations and definition of clinical outcomes. Applications of electronic health records to estimate trial outcomes and assess trial feasibility have been reported (Longo 2003, Doods 2014, Mc Cord 2018).

In this project, the University of Leicester propose to use HES data to obtain the granular data required for designing clinical trials assessing trial feasibility, thus minimizing the assumptions imputed and making the process quicker, simpler and more reliable. The University of Leicester will model pragmatic trials of surgical interventions in silico that will, in turn, be used to inform commissioning and funding applications for randomised clinical trials in NHS hospitals.

James Lind Alliance (JLA) Priority Setting Partnership (PSP) in Adult Heart Surgery

The Heart Surgery Priority Setting Partnership (PSP) is a collaboration between the Department of Cardiovascular Sciences at the University of Leicester and the James Lind Alliance (JLA). The JLA is a National Institute for Health Research (NIHR) initiative that aims to bring patients, carers and healthcare professionals together to identify and prioritise top unanswered health research questions. The University of Leicester propose to use HES data to design clinical trials in silico, assess trial feasibility, and devise a set of trial proposals to address the priority research questions identified in the JLA Heart Surgery PSP. The PSP has identified over 40 research questions covering different aspects of cardiac surgery, and the Top 10 research questions identified are:

1. How does a patient’s quality of life (QOL) change (e.g. disability-free survival) following heart surgery and what factors are associated with this?

2. How can we address frailty and improve the management of frail patients in heart surgery?

3. How can we improve the outcomes of heart surgery patients with chronic conditions (obesity, diabetes, hypertension, renal failure, autoimmune diseases etc.)?

4. Does prehabilitation (a programme of nutritional, exercise and psychological interventions before surgery) benefit heart surgery patients?

5. When should heart valve intervention occur for patients without symptoms?

6. How does minimally invasive heart surgery compare to traditional open surgery?

7. How do we minimise damage to organs from the heart-lung machine/heart surgery (heart, kidney, lung, brain and gut)?

8. Can we use 3D bio printing or stem cell technology to create living tissues (heart valves/heart) and repair failing hearts (myocardial regeneration)?

9. What are the most effective ways of preventing and treating postoperative atrial fibrillation?

10.How do we reduce and manage infections after heart surgery including surgical site/sternal wound infection and pneumonia?

Please note: these questions are official wording sources from JLA Heart Surgery PSP.

<https://le.ac.uk/cardiovascular-sciences/about/heart-surgery/top-10-priorities>

STUDY OBJECTIVES

The primary objective of this project is to use routinely collected HES and the linked mortality data to model trials of surgical interventions in silico and devise a set of pragmatic trial proposals to address the national priority research priorities in cardiac surgery.

The secondary objectives are:

• To evaluate the extent to which in silico trials can be conducted using HES data to design new trials.

• To establish methodologies and a systematic framework to carry out trials in silico with HES data so that trial feasibility can be conducted quickly and cost-effectively for a range of research questions.

• To develop capacity in processing and analyzing real-world clinical data using statistical and machine learning methods.

• To gain insights into how we might use HES data sets to support data collection and undertake future pragmatic trials.

REQUESTED DATA

Adult patients (18 years and above) with a cardiovascular diagnosis (ICD10 I00-I99, whether primary or secondary diagnoses) in HES Admitted Patient Care (APC) will form the reference cohort. Request for all HES and Civil Registrations - Deaths data sets are confined to patients identified in the reference cohort. Cumulative reference cohort will be used for annual update. The University of Leicester are requesting the data sets for the past 10 years plus an annual update for the next 3 years. An addition of two earlier years is also requested for HES APC for the purpose of new case ascertainment and defining patients’ co-morbidities and frailty score.

1) HES Admitted Patient Care (APC) AR 2007/08 to AR 2018/19 plus annual update to 2021/22

2) HES Critical care (CC) AR 2009/10 to AR 2018/19 plus annual update to 2021/22

3) HES Accident and Emergency (AE) AR 2009/10 to 2019/20 M12

4) ECDS AR 2020/21 to AR 2021/22.

5) HES Outpatient (OP) AR 2009/10 to AR 2018/19 plus annual update to 2021/22

6) Civil Registrations - Deaths data 2007/08 to 2018/19 plus annual update to 2021/22

7) Diagnostic imaging data sets (DID) 2007/08 to 2018/19 plus annual update to 2021/22

COHORT

The estimated size of the reference cohort is roughly 1.5 million admissions a year. The University of Leicester project team recognises that this is a large cohort size but will use the requested data to model trial in-silico and address the priority research questions in cardiac surgery identified by the James Lind Alliance process.

It is important to recognise that the James Lind Alliance process provides a set of priorities for research agenda which could be translated into future clinical trials. These research questions represent the areas that are important to those affected by cardiac surgery but are not precisely-worded that can be shared immediately with research funders. Further work is required to translate the research priorities into specific potential researchable questions for research funders to work with. For example, priority 3 ‘How can we improve the outcomes of heart surgery patients with chronic conditions?’, this question can potentially lead to specific research questions like (a) does pre-surgery optimisation of chronic conditions reduce post-operative lung and kidney injury or infection? (b) does minimally invasive approach improve outcomes in patients with chronic diseases? (c) are there specific pre-surgery interventions that can be targeted to patients with specific chronic diseases (e.g. weight loss programme for obese patients, glucose control for diabetic patients, iron supplement for anaemic patients)? Also, some specific question can address more than one research priority, for example, the question “does minimally invasive approach improve outcomes in patients with chronic conditions?” addresses two research priorities in relation to improving outcomes in patients with chronic conditions (priority 3) and minimally invasive cardiac approach (priority 6).

As well as strategic partnership with research groups to facilitate question formulation, the University of Leicester, in collaboration with Cochrane Heart, will commission a series of systematic reviews of the priority research questions to identify the knowledge gaps that can be addressed by clinical trials. In addition, the University of Leicester is organising a one-day Clinical Research Priorities Workshop to pump prime potential research teams who can come together to develop high quality research proposals for research funders. The workshop will bring together patients, carers, and a critical mass of expertise including clinicians, methodologists and scientists from across the UK to form interdisciplinary working groups and identify important trial questions from the research priorities. This Workshop was initially scheduled for July 2020 but has been postponed to early next year due to the Coronavirus pandemic.

In this project, the population of interests are adult patients with cardiovascular diseases who require or potentially require cardiac surgery. Cardiac surgery is performed to fix problems in the heart. It is used to treat a wide variety of cardiovascular diseases including aortic disease, arrhythmia, heart failure, coronary heart disease, cardiomyopathy, valvular heart disease, etc. Sometimes these problems can be addressed with medications or non-surgical procedures. For example, coronary angioplasty is a minimally invasive procedure in which a stent is inserted into a narrowed or blocked coronary artery. There are many types of heart surgery, some of the most common ones include coronary valvular surgery, aortic surgery, arrhythmia surgery, coronary artery bypass graft (CABG) surgery. The identified research questions cover all types of cardiac surgery and encompass all aspects of surgery from pre-operative assessment and risk stratification, to intraoperative management and post-operative outcomes.

The University of Leicester requests hospital admissions (HES APC) from adult patients with cardiovascular diagnoses. The requested data will be used to model a variety of cardiac surgery trials, and the analyses involved are board ranging. The analyses will not be limited to cardiac surgery patients, as it is also imperative to examine the effectiveness of surgical treatment by comparing the outcomes of cardiovascular patients with surgical and non-surgical interventions. For example, the research team is interested in developing trials examining the risks and benefits of CABG bypass surgery vs non-surgical angioplasty in heart failure patients. By limiting the patient cohort to patients who had cardiac surgery would limit the usefulness of the data. It is also not feasible to produce an exhaustive list of diagnosis codes that the cardiac surgery is used for. Therefore, although the focus of this study is cardiac surgery, the study requires a boarder cardiovascular cohort to define the patient population. It is important to note that although the HES APC data is termed the reference cohort, it will be used not only to define the patient populations but also to identify post-operative outcomes which include a wide range of cardiovascular conditions such as stroke, myocardial infarction, atrial fibrillation, in the trial modelling. In addition, the exact research questions for cardiac surgery are still being developed through the Cochrane review and the Clinical Research Priorities Workshop, it is necessary to ensure the requested hospital data covers all cardiovascular conditions so that the design of trials would not be limited by a predefined set of diagnoses.

The University of Leicester also requests hospital admissions within two years prior to the cardiovascular admissions, this data is required to check for patients’ frailty score and co-morbid conditions. These prior hospital episodes need to include ICD codes beyond the cardiovascular codes because the derivation of comorbidity and frailty scores such as Hospital Frailty Risk Score (Gilbert 2018) and Charlson Comorbidity Index (Li 2008) require a board range of diagnoses including both cardiovascular and non-cardiovascular codes. In addition, two of the research priorities identified are related to frailty (priority 2) and patients with chronic conditions (priority 3). Frail patients are often elderly patients with multiple medical conditions. Frailty is currently poorly defined for cardiac surgery. It would be desirable to examine if specific set of ICD codes could be identified to define frailty for cardiac surgery. By limiting our study to only pre-specified comorbid conditions, it would limit the usefulness and ability of the study to inform future cardiac surgery trials.

This project requires adult data defined by patients aged 18 and above, in line with the scope of the James Lind Alliance Priority Setting Partnership for Heart Surgery. Children's data are not needed.

The University of Leicester requests national data as this project targets at designing multi-centre pragmatic trials to evaluate clinical effectiveness in a real-world setting. Access to and outcomes of cardiac surgery vary across geographical regions. Such variations may reflect difference in patient case-mix, centralisation of care into specialist hospitals, variation in practices, and other factors. By limiting the analysis to specific geographical regions would affect the generalisable of the findings in multiple settings. It is also important to recognise that certain heart operations such as the Ross procedure, Transcatheter aortic valve implantation (TAVI) are only practised by limited heart centres in the UK. In addition, an important part of the analysis is to enable a detailed understanding of the characteristics of patient populations and the estimation of treatment effects across patient groups stratified by age, comorbid conditions or frailty thus enabling the identification of targeted populations for specific interventions. Large volume national level data is therefore needed to ensure sufficient patient size to carry out sub-group analysis.

The University of Leicester requests various HES datasets and Civil Registrations – Deaths data to evaluate short (in hospital, within 1 month), medium (3-6 months) and long term (1 to 5 years) outcomes of cardiac surgery patients. Cardiac surgery is a complex operation with all heart surgery patients requiring intensive care support immediately after the surgery. It is necessary to include organ support data in critical care (HES CC) and post-op complications and in-hospital mortality in the index episodes (HES APC) to evaluate the short-term outcomes of patients after surgery. Also, HES APC datasets will be longitudinally linked to track short to medium term outcomes including readmission due to cardiovascular causes and repeat of heart operation. Together with the HES A&E data, unplanned readmissions could be identified which could serve as an indicator of adverse outcomes post surgery. The request of Civil registration mortality data will enable the project team to undertake survival analysis and evaluate patients’ survival in short, medium and long term basis. The project team is planning to track the outcomes within 5 years after surgery as long term outcomes. Most of the existing clinical trials have focused on reporting short terms outcomes. Data on long term outcomes are lacking, although this is important to determine the comparative effectiveness of different surgical and non-surgical interventions. The University of Leicester requests 10 years of patient data in the initial cohort, and this will provide 5 years of data with long term outcome measures.

In addition to clinical outcomes, this study will include analysis to evaluate healthcare resource use following an intervention. As well as the extraction of resources use data during the index admission, post-discharge healthcare use data will be obtained on hospital readmissions, visits to Accident and Emergency, outpatient attendances, and imaging tests. Resources including length of hospital stay and various levels of care after surgery will be obtained with HES APC. As all heart surgery patients need to be followed up in outpatient clinics post discharge and cardiac imaging such as echocardiography and cardiovascular magnetic resonance are used to assess cardiovascular function after cardiac surgery, it is necessary to include HES outpatient (OP) and imaging (DID) data sets as part of the outcome analysis. Also, HES A&E data is needed to identify unplanned medical visits as indication of healthcare resource use resulting from post surgery complications.

In summary, patient population, identified by the index episodes receiving the relevant surgical /procedural interventions, along with their baseline patient characteristics (demographics, co-morbid conditions, frailty scores, etc) will first be defined using HES APC. Short term outcomes including post-operative complications and in-hospital mortality would be tracked using HES CC and APC. Medium term healthcare resource use following the surgical intervention will be tracked with HES AE, OP and DID data sets. Long term outcomes including 1-year and 5-year survival will be tracked using Civil Registrations - Deaths data. The University of Leicester has considered data minimisation to ensure the data requested is justified and limited to the study objectives.

The requested HES data sets will be individual records and pseudonymised with unique identifiers generated by NHS England. No identifiable data (name, address, NHS number, etc) will be included in the data sets. Civil Registrations - Deaths, and DIDs data sets are linked to the HES data sets through bridging files. HES data set will be longitudinally linked through the pseudo identifiers. All the analyses will be carried out within the data sets requested in this application. There will be no linking of these data sets to any external data sets.

The University of Leicester is the sole data controller and will process the data for this project. No data processing will be carried out by other organisations. The project team will share and discuss the results in form of summary statistics with the collaborators. No data processing would be undertaken by them and all decisions about the data analysis would remain with the university. James Lind Alliance is not involved in this project. However, the University will feedback to James Lind Alliance for any successfully funded trials resulting from the work of this project.

PATIENT AND PUBLIC INVOLVEMENT (PPI)

The University of Leicester will engage with patients and the public for dissemination and communication of the main findings. This will be facilitated through the established Patient and Public Involvement (PPI) networks with the Leicester Cardiac Surgery Research Group and the Heart Surgery Priority Setting Partnership Steering Committee. Through the involvement and recommendations regarding the dissemination of findings by the PPI groups, this will ensure the outputs are interpret-able to the wider patient and public community.

LEGAL BASIS

This project is managed by the University of Leicester and will be conducted in accordance with all applicable regulatory guidelines. The University of Leicester will lawfully be processing personal data on the basis of GDPR Article 6.1(e) - the processing is necessary for the performance of a task in the public interest. Research is a task that the University of Leicester performs in the public interest, as part of the core functions as a university.

This project will involve processing data related to patients’ ethnicity (there are marked ethnic differences in risk of cardiovascular diseases). The University will lawfully be processing special categories of personal data on the basis of GDPR Article 9.2(j) - the processing is necessary for research purposes or statistical purposes. The University will be processing pseudonymised data and the data sets will be stored and processed in accordance with the University Information Security Policy, College of Life Sciences Information Governance Policy, UK General Data Protection Regulation (GDPR) and the UK Data Protection Act (2018).

REFERENCES

1/ Longo et al. Can randomised trials rely on existing electronic data? A feasibility study to explore the value of routine data in health technology assessment. Health Technol Assess. 2003;7(26):iii, v-x, 1-117.

2/ Doods et al. A European inventory of common electronic health record data elements for clinical trial feasibility. Trials. 2014 Jan 10; 15:18.

3/ Mc Cord et al. Routinely collected data for randomized trials: promises, barriers, and implications. Trials. 2018 Jan 11;19(1):29.

4/ Gilbert et al. Development and validation of a Hospital Frailty Risk Score focusing on older people in acute care settings using electronic hospital records: an observational study. Lancet. 2018 May 5;391(10132):1775-1782.

5/ Li et al. Risk adjustment performance of Charlson and Elixhauser comorbidities in ICD-9 and ICD-10 administrative databases. BMC Health Serv Res. 2008 Jan 14;8:12.

The University of Leicester will use the data for the research purposes specified in the application. This project will set out a methodological framework for conducting in silico trials using routinely collected HES and the linked death data. The work will provide granular data required for designing surgical trials and lead to the production of a portfolio of pragmatic trial proposals addressing the top priorities research questions in heart surgery.

The project team will first work on two candidate trials, which have been selected so that the strengths and the weaknesses of the in-silico trials approach can be identified. The two trials are:

1. Benefits of re-vascularisation (bypass surgery vs minimally invasive angioplasty) in heart failure patients – this work will model the comparative effectiveness of bypass surgery vs angioplasty in people with heart failure (a chronic condition) and coronary artery disease. The trial will address research priorities including improving outcomes in patients with chronic conditions (priority 3) and comparative effectiveness of minimally invasive vs open surgery (priority 6).

2. Outcomes of single vs multiple arterial grafts in women undergoing bypass surgery - this trial will address improving outcomes of heart surgery patients in relation to long-term quality of life outcome (priority 1) and frailty (priority 2).

As well as the outputs for these two trials, this work will facilitate the development of the in-silico methodological framework and contribute to the production of a master protocol that will describe the methods, strengths and limitations of conducting in-silico trials using HES data.

The project team will work on modelling other trials in-silico when the exact research questions are formulated after the Clinical Research Priorities webinars.

Trial proposals, funding applications and all research reports and presentations resulting from this project will contain only summary aggregated data with small numbers suppressed in line with the NHS England HES Analysis Guide. Research presentations may consist of oral presentations, poster and published abstract.

Scientific findings will be disseminated by usual academic channels, i.e. presentation at academic conferences and publication in peer-reviewed journals. No identifiable information will be presented.

The James Lind Alliance Priority Setting Partnership in Heart Surgery identified over 40 research questions covering different aspects of cardiac surgery. The priority will be given to address the top 10 questions, but it is important to recognise that some of the remaining questions are also important research questions. These include geographical variation in outcomes of heart surgery. The project team will use an observational study design to examine the short and long term outcomes of heart surgery patients by geographical regions and the factors associating with the variations. Peer reviewed scientific publications will be produced from analysing the data.

Proposed digital tool for designing trials of cardiovascular diseases

The University of Leicester Cardiac Surgery Research Group is planning to develop a web-based digital tool to support the planning of clinical trials in cardiac surgery. The tool is to be a web-based interactive tool to display the data summary and analyses that the University of Leicester Cardiac Surgery Research Group do for the research questions. The tool would keep multidimensional databases (in the same vein as an online analytical processing (OLAP) cube. An OLAP cube is a computer-based technique of analyzing data to look for insights. The term cube here refers to a multi-dimensional data set) for important trial parameters like patient numbers, outcome events stratified by dimensions including surgery/intervention, patient age and sex, frailty score, geographical region, etc that the users are able to be specified. Users will be asked to select/enter the targeted patient population (based on age, diagnosis, medical conditions etc), geographical region, the intervention, the comparator, trial design, primary outcome. The digital tool will then output the trial parameters including the occurrence of the primary outcome, expected treatment effect, sample size required, availability of patient populations by hospital etc. The tool will also help users to explore the trial parameters for different patient groups.

The tool is intended to be put in a public domain, to be used by UK healthcare professionals and researchers inside and outside the University of Leicester. The tool is a web application accessible via a web address. The data will sit on a secure server owned by the University of Leicester. With respect to access control, users need to register and login to gain access to the tool.

The tool will make no attempt to link NHS England data with any other data sets. All outputs are aggregated statistics with small number suppressed as per the HES guide.

Target dates for outputs:

Short term (1 - 2 year) - a methodological paper that will describe the methods, strengths and limitations of an in-silico trial approach: the research team has published the in-silico methodology, together with the results of the first candidate trial modelling bypass surgery vs stenting to treat heart failure patients with ischemic heart diseases, in European Heart Journal. doi: 10.1093/eurheartj/ehac670 ;

Medium term (2 - 5 years) - a series of pragmatic trial proposals, including the two candidate trials, that answer the priority research questions identified by JLA Priority Setting Partnership in Heart Surgery.

The research team has so far applied the in-silico method in HES data and supported the design and the funding application of these clinical trials.

(a) BCIS-4: CABG vs PCI in heart failure patients requiring revascularisation

(b) PROPHESY: transfusion with prothrombin complex concentrates (PCC) vs fresh frozen plasma (FFP) in cardiac surgery patients with intra-operative bleeding

(c) EVOCC: endovascular vs open surgery in severe occlusive aorto-iliac disease

(d) PROTECT: anticoagulation treatments after mitral valve repair surgery

The outcomes of this project lead to the production of a portfolio of cardiac surgery trial proposals that address important research questions in cardiac surgery. We have supported the design and funding applications of four cardiac surgery trials so far. Whilst it will take time for these clinical trials to be implemented and evaluated, the results of these trials could lead to improved clinical care and outcomes for heart surgery patients.

Further planned in-silico trials include (a) a platform trial to reduce organ failure in cardiac surgery patients

(b) Normal vs Hypothermia in cardiac surgery involving cardiopulmonary bypass

(c) CABG vs PCI in frail or pre-frail people presenting with acute coronary syndrome.

In addition, the research team has published the in-silico methodology together with the results of the first candidate trial modelling bypass surgery vs stenting to treat heart failure patients with ischemic heart diseases in European Heart Journal. doi: 10.1093/eurheartj/ehac670.

Results of other in-silico trials will be published as appropriate.

This study will be undertaken by the Leicester Cardiac Surgery Research Group at Department of Cardiovascular Sciences based in the University of Leicester. This study is to establish a database of patients with cardiovascular diagnosis in England using routinely collected clinical and mortality data obtained from NHS Digital. This database will be used to model trials of surgical interventions in silico and devise a set of pragmatic trial proposals to address the priority research questions in cardiac surgery.

HES DATA

The gold standard for the evaluation of medical treatment is through randomised controlled trials. However, the successful delivery of clinical trials are often limited by the many assumptions that are required for design and planning. Assumptions are often made with respect to recruitment, eligibility, event rates, effect estimates, safety and attrition in real world populations. Hospital Episode Statistics (HES) data contains a wealth of real-world data including demographics, diagnoses, procedures and other clinical information collected prospectively from all NHS hospitals. This routinely collected clinical data enables the University of Leicester to explore the effect of interventions (e.g. open heart vs minimally invasive heart surgery) in a population and across different patient subgroups (e.g. elderly patients, or patients with other medical conditions), and obtain the necessary parameters required for designing a clinical trial. The University of Leicester may also optimize the trial design through sensitivity analyses of the inclusion/exclusion criteria of patient populations and definition of clinical outcomes. Applications of electronic health records to estimate trial outcomes and assess trial feasibility have been reported (Longo 2003, Doods 2014, Mc Cord 2018).

In this project, the University of Leicester propose to use HES data to obtain the granular data required for designing clinical trials assessing trial feasibility, thus minimizing the assumptions imputed and making the process quicker, simpler and more reliable. The University of Leicester will model pragmatic trials of surgical interventions in silico that will, in turn, be used to inform commissioning and funding applications for randomised clinical trials in NHS hospitals.

James Lind Alliance (JLA) Priority Setting Partnership (PSP) in Adult Heart Surgery

The Heart Surgery Priority Setting Partnership (PSP) is a collaboration between the Department of Cardiovascular Sciences at the University of Leicester and the James Lind Alliance (JLA). The JLA is a National Institute for Health Research (NIHR) initiative that aims to bring patients, carers and healthcare professionals together to identify and prioritise top unanswered health research questions. The University of Leicester propose to use HES data to design clinical trials in silico, assess trial feasibility, and devise a set of trial proposals to address the priority research questions identified in the JLA Heart Surgery PSP. The PSP has identified over 40 research questions covering different aspects of cardiac surgery, and the Top 10 research questions identified are:

1. How does a patient’s quality of life (QOL) change (e.g. disability-free survival) following heart surgery and what factors are associated with this?

2. How can we address frailty and improve the management of frail patients in heart surgery?

3. How can we improve the outcomes of heart surgery patients with chronic conditions (obesity, diabetes, hypertension, renal failure, autoimmune diseases etc.)?

4. Does prehabilitation (a programme of nutritional, exercise and psychological interventions before surgery) benefit heart surgery patients?

5. When should heart valve intervention occur for patients without symptoms?

6. How does minimally invasive heart surgery compare to traditional open surgery?

7. How do we minimise damage to organs from the heart-lung machine/heart surgery (heart, kidney, lung, brain and gut)?

8. Can we use 3D bio printing or stem cell technology to create living tissues (heart valves/heart) and repair failing hearts (myocardial regeneration)?

9. What are the most effective ways of preventing and treating postoperative atrial fibrillation?

10.How do we reduce and manage infections after heart surgery including surgical site/sternal wound infection and pneumonia?

Please note: these questions are official wording sources from JLA Heart Surgery PSP.

<https://le.ac.uk/cardiovascular-sciences/about/heart-surgery/top-10>

STUDY OBJECTIVES

The primary objective of this project is to use routinely collected HES and the linked mortality data to model trials of surgical interventions in silico and devise a set of pragmatic trial proposals to address the national priority research priorities in cardiac surgery.

The secondary objectives are:

• To evaluate the extent to which in silico trials can be conducted using HES data to design new trials.

• To establish methodologies and a systematic framework to carry out trials in silico with HES data so that trial feasibility can be conducted quickly and cost-effectively for a range of research questions.

• To develop capacity in processing and analyzing real-world clinical data using statistical and machine learning methods.

• To gain insights into how we might use HES data sets to support data collection and undertake future pragmatic trials.

REQUESTED DATA

Adult patients (18 years and above) with a cardiovascular diagnosis (ICD10 I00-I99, whether primary or secondary diagnoses) in HES Admitted Patient Care (APC) will form the reference cohort. Request for all HES and Civil Registrations - Deaths data sets are confined to patients identified in the reference cohort. Cumulative reference cohort will be used for annual update. The University of Leicester are requesting the data sets for the past 10 years plus an annual update for the next 3 years. An addition of two earlier years is also requested for HES APC for the purpose of new case ascertainment and defining patients’ co-morbidities and frailty score.

1) HES Admitted Patient Care (APC) AR 2007/08 to AR 2018/19 plus annual update to 2021/22

2) HES Critical care (CC) AR 2009/10 to AR 2018/19 plus annual update to 2021/22

3) HES Accident and Emergency (AE) AR 2009/10 to 2019/20 M12

4) ECDS AR 2020/21 to AR 2021/22.

5) HES Outpatient (OP) AR 2009/10 to AR 2018/19 plus annual update to 2021/22

6) Civil Registrations - Deaths data 2007/08 to 2018/19 plus annual update to 2021/22

7) Diagnostic imaging data sets (DID) 2007/08 to 2018/19 plus annual update to 2021/22

COHORT

The estimated size of the reference cohort is roughly 1.5 million admissions a year. The University of Leicester project team recognises that this is a large cohort size but will use the requested data to model trial in-silico and address the priority research questions in cardiac surgery identified by the James Lind Alliance process.

It is important to recognise that the James Lind Alliance process provides a set of priorities for research agenda which could be translated into future clinical trials. These research questions represent the areas that are important to those affected by cardiac surgery but are not precisely-worded that can be shared immediately with research funders. Further work is required to translate the research priorities into specific potential researchable questions for research funders to work with. For example, priority 3 ‘How can we improve the outcomes of heart surgery patients with chronic conditions?’, this question can potentially lead to specific research questions like (a) does pre-surgery optimisation of chronic conditions reduce post-operative lung and kidney injury or infection? (b) does minimally invasive approach improve outcomes in patients with chronic diseases? (c) are there specific pre-surgery interventions that can be targeted to patients with specific chronic diseases (e.g. weight loss programme for obese patients, glucose control for diabetic patients, iron supplement for anaemic patients)? Also, some specific question can address more than one research priority, for example, the question “does minimally invasive approach improve outcomes in patients with chronic conditions?” addresses two research priorities in relation to improving outcomes in patients with chronic conditions (priority 3) and minimally invasive cardiac approach (priority 6).

As well as strategic partnership with research groups to facilitate question formulation, the University of Leicester, in collaboration with Cochrane Heart, will commission a series of systematic reviews of the priority research questions to identify the knowledge gaps that can be addressed by clinical trials. In addition, the University of Leicester is organising a one-day Clinical Research Priorities Workshop to pump prime potential research teams who can come together to develop high quality research proposals for research funders. The workshop will bring together patients, carers, and a critical mass of expertise including clinicians, methodologists and scientists from across the UK to form interdisciplinary working groups and identify important trial questions from the research priorities. This Workshop was initially scheduled for July 2020 but has been postponed to early next year due to the Coronavirus pandemic.

In this project, the population of interests are adult patients with cardiovascular diseases who require or potentially require cardiac surgery. Cardiac surgery is performed to fix problems in the heart. It is used to treat a wide variety of cardiovascular diseases including aortic disease, arrhythmia, heart failure, coronary heart disease, cardiomyopathy, valvular heart disease, etc. Sometimes these problems can be addressed with medications or non-surgical procedures. For example, coronary angioplasty is a minimally invasive procedure in which a stent is inserted into a narrowed or blocked coronary artery. There are many types of heart surgery, some of the most common ones include coronary valvular surgery, aortic surgery, arrhythmia surgery, coronary artery bypass graft (CABG) surgery. The identified research questions cover all types of cardiac surgery and encompass all aspects of surgery from pre-operative assessment and risk stratification, to intraoperative management and post-operative outcomes.

The University of Leicester requests hospital admissions (HES APC) from adult patients with cardiovascular diagnoses. The requested data will be used to model a variety of cardiac surgery trials, and the analyses involved are board ranging. The analyses will not be limited to cardiac surgery patients, as it is also imperative to examine the effectiveness of surgical treatment by comparing the outcomes of cardiovascular patients with surgical and non-surgical interventions. For example, the research team is interested in developing trials examining the risks and benefits of CABG bypass surgery vs non-surgical angioplasty in heart failure patients. By limiting the patient cohort to patients who had cardiac surgery would limit the usefulness of the data. It is also not feasible to produce an exhaustive list of diagnosis codes that the cardiac surgery is used for. Therefore, although the focus of this study is cardiac surgery, the study requires a boarder cardiovascular cohort to define the patient population. It is important to note that although the HES APC data is termed the reference cohort, it will be used not only to define the patient populations but also to identify post-operative outcomes which include a wide range of cardiovascular conditions such as stroke, myocardial infarction, atrial fibrillation, in the trial modelling. In addition, the exact research questions for cardiac surgery are still being developed through the Cochrane review and the Clinical Research Priorities Workshop, it is necessary to ensure the requested hospital data covers all cardiovascular conditions so that the design of trials would not be limited by a predefined set of diagnoses.

The University of Leicester also requests hospital admissions within two years prior to the cardiovascular admissions, this data is required to check for patients’ frailty score and co-morbid conditions. These prior hospital episodes need to include ICD codes beyond the cardiovascular codes because the derivation of comorbidity and frailty scores such as Hospital Frailty Risk Score (Gilbert 2018) and Charlson Comorbidity Index (Li 2008) require a board range of diagnoses including both cardiovascular and non-cardiovascular codes. In addition, two of the research priorities identified are related to frailty (priority 2) and patients with chronic conditions (priority 3). Frail patients are often elderly patients with multiple medical conditions. Frailty is currently poorly defined for cardiac surgery. It would be desirable to examine if specific set of ICD codes could be identified to define frailty for cardiac surgery. By limiting our study to only pre-specified comorbid conditions, it would limit the usefulness and ability of the study to inform future cardiac surgery trials.

This project requires adult data defined by patients aged 18 and above, in line with the scope of the James Lind Alliance Priority Setting Partnership for Heart Surgery. Children data are not needed.

The University of Leicester requests national data as this project targets at designing multi-centre pragmatic trials to evaluate clinical effectiveness in a real-world setting. Access to and outcomes of cardiac surgery vary across geographical regions. Such variations may reflect difference in patient case-mix, centralisation of care into specialist hospitals, variation in practices, and other factors. By limiting the analysis to specific geographical regions would affect the generalisable of the findings in multiple settings. It is also important to recognise that certain heart operations such as the Ross procedure, Transcatheter aortic valve implantation (TAVI) are only practised by limited heart centres in the UK. In addition, an important part of the analysis is to enable a detailed understanding of the characteristics of patient populations and the estimation of treatment effects across patient groups stratified by age, comorbid conditions or frailty thus enabling the identification of targeted populations for specific interventions. Large volume national level data is therefore needed to ensure sufficient patient size to carry out sub-group analysis.

The University of Leicester requests various HES datasets and Civil Registrations – Deaths data to evaluate short (in hospital, within 1 month), medium (3-6 months) and long term (1 to 5 years) outcomes of cardiac surgery patients. Cardiac surgery is a complex operation with all heart surgery patients requiring intensive care support immediately after the surgery. It is necessary to include organ support data in critical care (HES CC) and post-op complications and in-hospital mortality in the index episodes (HES APC) to evaluate the short-term outcomes of patients after surgery. Also, HES APC datasets will be longitudinally linked to track short to medium term outcomes including readmission due to cardiovascular causes and repeat of heart operation. Together with the HES A&E data, unplanned readmissions could be identified which could serve as an indicator of adverse outcomes post surgery. The request of Civil registration mortality data will enable the project team to undertake survival analysis and evaluate patients’ survival in short, medium and long term basis. The project team is planning to track the outcomes within 5 years after surgery as long term outcomes. Most of the existing clinical trials have focused on reporting short terms outcomes. Data on long term outcomes are lacking, although this is important to determine the comparative effectiveness of different surgical and non-surgical interventions. The University of Leicester requests 10 years of patient data in the initial cohort, and this will provide 5 years of data with long term outcome measures.

In addition to clinical outcomes, this study will include analysis to evaluate healthcare resource use following an intervention. As well as the extraction of resources use data during the index admission, post-discharge healthcare use data will be obtained on hospital readmissions, visits to Accident and Emergency, outpatient attendances, and imaging tests. Resources including length of hospital stay and various levels of care after surgery will be obtained with HES APC. As all heart surgery patients need to be followed up in outpatient clinics post discharge and cardiac imaging such as echocardiography and cardiovascular magnetic resonance are used to assess cardiovascular function after cardiac surgery, it is necessary to include HES outpatient (OP) and imaging (DID) data sets as part of the outcome analysis. Also, HES A&E data is needed to identify unplanned medical visits as indication of healthcare resource use resulting from post surgery complications.

In summary, patient population, identified by the index episodes receiving the relevant surgical /procedural interventions, along with their baseline patient characteristics (demographics, co-morbid conditions, frailty scores, etc) will first be defined using HES APC. Short term outcomes including post-operative complications and in-hospital mortality would be tracked using HES CC and APC. Medium term healthcare resource use following the surgical intervention will be tracked with HES AE, OP and DID data sets. Long term outcomes including 1-year and 5-year survival will be tracked using Civil Registrations - Deaths data. The University of Leicester has considered data minimisation to ensure the data requested is justified and limited to the study objectives.

The requested HES data sets will be individual records and pseudonymised with unique identifiers generated by NHS Digital. No identifiable data (name, address, NHS number, etc) will be included in the data sets. Civil Registrations - Deaths, and DIDs data sets are linked to the HES data sets through bridging files. HES data set will be longitudinally linked through the pseudo identifiers. All the analyses will be carried out within the data sets requested in this application. There will be no linking of these data sets to any external data sets.

The University of Leicester is the sole data controller and will process the data for this project. No data processing will be carried out by other organisations. The project team will share and discuss the results in form of summary statistics with the collaborators. No data processing would be undertaken by them and all decisions about the data analysis would remain with the university. James Lind Alliance is not involved in this project. However, the University will feedback to James Lind Alliance for any successfully funded trials resulting from the work of this project.

PATIENT AND PUBLIC INVOLVEMENT (PPI)

The University of Leicester will engage with patients and the public for dissemination and communication of the main findings. This will be facilitated through the established Patient and Public Involvement (PPI) networks with the Leicester Cardiac Surgery Research Group and the Heart Surgery Priority Setting Partnership Steering Committee. Through the involvement and recommendations regarding the dissemination of findings by the PPI groups, this will ensure the outputs are interpret-able to the wider patient and public community.

LEGAL BASIS

This project is managed by the University of Leicester and will be conducted in accordance with all applicable regulatory guidelines. The University of Leicester will lawfully be processing personal data on the basis of GDPR Article 6.1(e) - the processing is necessary for the performance of a task in the public interest. Research is a task that the University of Leicester performs in the public interest, as part of the core functions as a university.

This project will involve processing data related to patients’ ethnicity (there are marked ethnic differences in risk of cardiovascular diseases). The University will lawfully be processing special categories of personal data on the basis of GDPR Article 9.2(j) - the processing is necessary for research purposes or statistical purposes. We will be processing pseudonymised data and the data sets will be stored and processed in accordance with the University Information Security Policy, College of Life Sciences Information Governance Policy, General Data Protection Regulation (GDPR) (EU) 2016/679 and the UK Data Protection Act (2018).

REFERENCES

1/ Longo et al. Can randomised trials rely on existing electronic data? A feasibility study to explore the value of routine data in health technology assessment. Health Technol Assess. 2003;7(26):iii, v-x, 1-117.

2/ Doods et al. A European inventory of common electronic health record data elements for clinical trial feasibility. Trials. 2014 Jan 10; 15:18.

3/ Mc Cord et al. Routinely collected data for randomized trials: promises, barriers, and implications. Trials. 2018 Jan 11;19(1):29.

4/ Gilbert et al. Development and validation of a Hospital Frailty Risk Score focusing on older people in acute care settings using electronic hospital records: an observational study. Lancet. 2018 May 5;391(10132):1775-1782.

5/ Li et al. Risk adjustment performance of Charlson and Elixhauser comorbidities in ICD-9 and ICD-10 administrative databases. BMC Health Serv Res. 2008 Jan 14;8:12.

The University of Leicester will use the data for the research purposes specified in the application. This project will set out a methodological framework for conducting in silico trials using routinely collected HES and the linked death data. The work will provide granular data required for designing surgical trials and lead to the production of a portfolio of pragmatic trial proposals addressing the top priorities research questions in heart surgery.

The project team will first work on two candidate trials, which have been selected so that the strengths and the weaknesses of the in-silico trials approach can be identified. The two trials are:

1. Benefits of re-vascularisation (bypass surgery vs minimally invasive angioplasty) in heart failure patients – this work will model the comparative effectiveness of bypass surgery vs angioplasty in people with heart failure (a chronic condition) and coronary artery disease. The trial will address research priorities including improving outcomes in patients with chronic conditions (priority 3) and comparative effectiveness of minimally invasive vs open surgery (priority 6).

2. Benefits of stratification of re-vascularisation decisions based on objective measures of frailty - this work will model a trial to test the hypothesis that treatment decisions stratified by frailty are likely to result in improved long-term benefits. The trial will address improving outcomes of heart surgery patients in relation to long-term quality of life outcome (priority 1), frailty (priority 2) and minimally invasive vs open surgery (priority 6).

As well as the outputs for these two trials, this work will facilitate the development of the in-silico methodological framework and contribute to the production of a master protocol that will describe the methods, strengths and limitations of conducting in-silico trials using HES data.

The project team will work on modelling other trials in-silico when the exact research questions are formulated after the Cochrane review and the Clinical Research Priorities Workshop.

Trial proposals, funding applications and all research reports and presentations resulting from this project will contain only summary aggregated data with small numbers suppressed in line with the NHS Digital HES Analysis Guide. Research presentations may consist of oral presentations, poster and published abstract.

Scientific findings will be disseminated by usual academic channels, i.e. presentation at academic conferences and publication in peer-reviewed journals. No identifiable information will be presented.

The James Lind Alliance Priority Setting Partnership in Heart Surgery identified over 40 research questions covering different aspects of cardiac surgery. The priority will be given to address the top 10 questions, but it is important to recognise that some of the remaining questions are also important research questions. These include geographical variation in outcomes of heart surgery. The project team will use an observational study design to examine the short and long term outcomes of heart surgery patients by geographical regions and the factors associating with the variations. Peer reviewed scientific publications will be produced from analysing the data.

Proposed digital tool for designing trials of cardiovascular diseases

The University of Leicester Cardiac Surgery Research Group is planning to develop a web-based digital tool to support the planning of clinical trials in cardiac surgery. The tool is to be a web-based interactive tool to display the data summary and analyses that the University of Leicester Cardiac Surgery Research Group do for the research questions. The tool would keep multidimensional databases (in the same vein as an online analytical processing (OLAP) cube. An OLAP cube is a computer-based technique of analyzing data to look for insights. The term cube here refers to a multi-dimensional data set) for important trial parameters like patient numbers, outcome events stratified by dimensions including surgery/intervention, patient age and sex, frailty score, geographical region, etc that the users are able to be specified. Users will be asked to select/enter the targeted patient population (based on age, diagnosis, medical conditions etc), geographical region, the intervention, the comparator, trial design, primary outcome. The digital tool will then output the trial parameters including the occurrence of the primary outcome, expected treatment effect, sample size required, availability of patient populations by hospital etc. The tool will also help users to explore the trial parameters for different patient groups.

The tool is intended to be put in a public domain, to be used by UK healthcare professionals and researchers inside and outside the University of Leicester. The tool is a web application accessible via a web address. The data will sit on a secure server owned by the University of Leicester. With respect to access control, users need to register and login to gain access to the tool.

The tool will make no attempt to link NHS Digital data with any other data sets. All outputs are aggregated statistics with small number suppressed as per the HES guide.

Target dates for outputs:

Short term (1 - 2 year) - a methodological paper or a master protocol that will describe the methods, strengths and limitations of an in-silico trial approach;

Medium term (2 - 5 years) - a series of pragmatic trial proposals, including the two candidate trials, that answer the priority research questions identified by JLA Priority Setting Partnership in Heart Surgery.

Yielded Benefits is not a requirement for new applications.