D27 - QResearch-Oxford Data Linkage Project

Under the previous version of this data sharing agreement (DARS-NIC-240279-Y2V2N-v1.6), quarterly data releases were replaced by monthly data releases under DARS-NIC-375354-G8V1H in order to support urgent COVID-19 research until Sept 2020. From Sept 2020 monthly releases need continue for the next 6 months since the COVID-19 pandemic has recently worsened (at the time of writing 22.09.2020 the UK is begun the second wave). After this University of Oxford expect to resume quarterly releases. Data released under DARS-NIC-375354-G8V1H can be reused for the purposes described within this DARS-NIC-240279-Y2V2N-v1 data sharing agreement, as set out below.

---

QResearch is a database of linked medical records that has been used and continues to be used by a variety of research projects undertaken by UK universities, from reviewing the safety of antidepressant medicines to studying factors to predict variations in survival rates for cancer patients.

QResearch has also been used to develop and validate risk prediction algorithms such as QRISK2. QRISK2 can be used by clinicians to calculate an individual’s risk of a heart attack or stroke taking account of their individual risk factors such as age, sex, ethnicity, clinical values and diagnoses. The research describing the derivation and validation of QRISK2 has been published in the BMJ (2008) and the software implementing QRISK2 is available as open and closed source software. QRISK2 is recommended by NICE as the risk score for use in its guidance on lipid modification (2014) and it is also recommended for use in the NHS Health Check.

The HES and mortality data are requested to link to the existing QResearch database so that it can be used for medical research. The QResearch database consists of the coded pseudonymised electronic health records from primary care patients registered with approximately 1,500 general practices spread throughout the UK.

QResearch is a not for profit collaboration originally between the University of Nottingham and Egton Medical Information Systems (EMIS) but the University of Nottingham’s roles and responsibilities were transferred to the University of Oxford. Strategic decisions about the GP data are taken by a Management Board representing the interests of EMIS and the University of Oxford. The University of Oxford is the sole data controller for the datasets which are linked to QResearch (deaths, cancer and hospital data) and the single point of access to the data.

The database is widely used for medical research into the causes of disease, its natural history, treatment and outcomes. QResearch was started in 2003 in order to improve access for research to primary care data and will continue for the foreseeable future.

In addition to coded data from the GP electronic record, the QResearch database also contains the linked cause of death derived from the death certificate data which was originally supplied directly by the Office of National Statistics (ONS) but NHS Digital has since assumed ownership for this data, and cancer registration data supplied directly by Public Health England, following approval by Trent MREC and Secretary of State for Health. From October 2018, new mortality data (referred to as Civil Registration data) has been supplied by NHS Digital. Any mortality data originally supplied by ONS data will be considered data supplied by NHS Digital and only NHS Digital can approve ongoing access to this data and its use for specific purposes.

The data linkages for QResearch were extended in 2011 to include additional health information from secondary care including HES. The additional linked HES data enables researchers to analyse additional information on patient characteristics, treatment and outcomes which will improve the epidemiological analyses of studies since the data will be more complete. Without the data linkage, research studies may under-estimate the risk and benefits associated with interventions such as prescribed medicines. For example, QResearch data linked to HES data is being used to undertake an assessment of different types of direct anticoagulant medication which is prescribed in primary care to reduce risk of stroke and manage thrombosis. Adverse effects from anticoagulants include major haemorrhage which can be life threatening or life ending. Haemorrhage can affect the brain, gastrointestinal tract, urinary track or other parts of the body. The primary care data provides information on the prescriptions issued and the linked HES data provides information on haemorrhage which is serious enough to require hospital admission. Another example is a recent study looking at safety of the oral contraceptive pill. The primary care data provide information on exposure to the medication and the linked HES data provides information on thrombosis. These are just two examples of projects which can only be done using the linked data. The results help identify and quantify the risks and benefits associated with different types of medication, used in different patients, at different doses over time because of the outcomes. The results help doctors and patients make better decisions and increase the evidence base to inform guideline development and policy.

The patient level data linked to QResearch is only accessed by academics and students employed by University of Oxford, as well as authorised individuals employed by the Universities included as data processors in this agreement. University of Nottingham. In all cases, data can only be accessed on site at the University of Oxford. However, the researchers involved in a given project (contributing to the research question, design, interpretation and writing of the paper for publication but not handling the data) may be employed by other UK universities. The HES and mortality data stay on site at the University of Oxford and are only handled by University of Oxford academics and students employed by University of Oxford to undertake research, the data processor contracted to the University of Oxford (Dancing House Consulting) and authorised researchers employed by the University of Nottingham. The University of Oxford may have a collaborator at another university on the project team advising on clinical aspects or interpretation of findings, but they will not receive any data. In addition, the external researcher may initiate a project but the University of Oxford has sole autonomy for determining the purposes for which the HES and/or civil registration data will be processed and analysis will be done by University of Oxford academics and students employed by University of Oxford to undertake research, with the data located at the University of Oxford. Data will not be used for any solely commercial purposes and all applications for the use of HES and/or mortality linked data are subject to a governance process explained in the Processing Activities section.

Only University of Oxford academics and students employed by University of Oxford to undertake research, their data processor, Dancing House Consulting, and authorised researchers employed by the University of Nottingham will have access to HES and/or mortality record level data and external researchers will only have access to tabular outputs. Record level data are not shared with researchers outside of the University of Oxford. Small numbers are suppressed in line with the HES Analysis Guide.

Research undertaken using the extended database continues to be processed using the existing arrangements with respect to scientific review and annual reports to Trent MREC. Research has to be peer reviewed, original, hypothesis driven or hypothesis testing, intended for publication in an academic peer reviewed journal.

All research undertaken using the QResearch database and linked data are subject to independent peer review and the results of all research are published.

The outputs are research papers which are published in peer reviewer academic scientific journals and presented at academic conferences. All research is published in academic journals with a link from the QResearch website on an ongoing basis. The publications are accompanied by h press releases from the relevant organisations and highlighted on social media.

Examples of conferences include the annual academic conference for the Society of Academic Primary Care (next due July 2021) and the UKRI (Sept 2020); international conferences such as the North American Primary Care Research Group (NAPCRG- US Nov 2021); the annual conferences of the EMIS National User Group (a national education and research charity representing the GP practices who contribute data to QResearch Sept 2020 etc); annual conferences run by cancer charities such as Macmillan Cancer Support and Pancreatic Cancer UK; local and regional conferences run by the Nottingham Biomedical Research Centre.

Results are also shared with policy makers and NICE guideline committees on a regular basis via their stakeholder consultations in order to support development of relevant guidelines.

Examples of expected outputs include:

• University of Oxford have recently completed a study found there was no increase in risk of severe COVID infection in people taking ACE medication – thereby reassuring patients taking these medicines that there is no need for them to stop taking them (as had been feared)

We have recently completed the largest single study ever published to examine the risk of breast cancer among women taking HRT. The study is due for publication in the BMJ (expected Nov 2020) and provides more accurate quantification of the risks of breast cancer according to the type, duration and method of taking HR which University of Oxford anticipate will help inform decisions between GPs and patients regarding treatment options.

At the request of the four UK CMO’s, University of Oxford have recently completed the first part of a study to develop a novel risk prediction tool to estimate risk of poor outcomes from COVID which will soon be used to inform national policy on shielding, targeting of vaccines (expected Nov 2020).

Our research to identify early symptoms of Motor Neurone Disease, funded by the Motor Neurone Association, has identified a set of symptoms which could be used to increase referrals and improve early recognition and diagnosis. This si expected to lead to the development of a decision support tools for use in primary care (excepted 2022)

Our research funded by INNOVATE UK to develop a risk stratification tool to identify those at high risk of oesophageal cancer is likely to lead to a more efficient way of identifying patients who might be eligible for the Cytosponge device (which is an alternative to endoscopy). This is especially important given the limitations on use of endoscopy arising from the COVID-19 pandemic (results e3xpected 2022/2023).

Another project funded by INNOVATE UK investigates the risks and benefits and health economic consequences the pilot lung cancer screening program is expected to inform the development of the national screening program (results expected 2023).

The results tables within the papers will only contain statistical information with cell counts of > 5, being suppressed in line with the ICO code on anonymisation. Outputs will contain aggregate level data with small numbers suppressed in line with the HES analysis guide.

No indicators are produced which show performance of an organisation – indeed the identity of the GP practices contributing to QResearch are not shared with any third party.

The results of research undertaken continues to result in new knowledge and understanding regarding disease epidemiology, health inequalities, drug safety, methods of identifying patients at high risk of serious illnesses. Every year new research is published in high impact international research journals such as the British Medical Journal and the British Journal of General Practice. The research is ongoing with target dates for individual projects rather than one overall target date. A complete list of research papers using QResearch database is published at http://www.qresearch.org/SitePages/publications.aspx

There are many benefits arising from this research - some examples are listed here.

The results of two studies published in the BMJ which described enhanced methods to estimate risk of bleeding (QBleed) and stroke (QStroke) have informed NICE's decision to update their guidelines on atrial fibrillation [CG180]. For more details please see the information published on NICE's website in Sept 2017 on assessment of bleeding and stroke risk https://www.nice.org.uk/guidance/cg180/resources/surveillance-report-2017--atrial-fibrillation-management-2014-nice-guideline-cg180-4597399263/chapter/Surveillance-decision

University of Oxford’s analysis of the uptake and impacts of the NHS health checks program continues to inform the development and delivery of national policy on the NHS health checks program.

Our study on safety of HRT and thrombosis risk was published in the BMJ in 2019 and to demonstrate important differences between different types of HRT which can help improve treatment decisions.

QRISK, which the applicant continues to develop and enhance, has become the preferred cardiovascular risk assessment tool used across the NHS. It is implemented in every GP practice, used extensively in NHS Health Checks, and a variant is used for the NHS Choices 'Heart Age' tool. It was recognised as an outstanding impact case study in the Research Assessment Exercise in 2014. Use of QRISK2 has led to targeted interventions to reduce CVD risk as shown in NHS health check research published in BMJ Open in 2016. The research to update QRISK3 was shared at meetings with the British Heart Foundation and Public Health England as well with the National Directors of Cardiovascular Disease (Huon Grey and Matt Kearney) and at meetings of Expert Scientific Advisory Group which oversees the NHS Health Checks program and which is chaired by John Newton (PHE). QRISK3 was implemented in the NHS from 2018 and will continue to be updated on regular basis to ensure that the tool gives the most accurate estimation of risk possible.

In response to nationally identified NHS needs, the applicant has developed & updated a suite of risk prediction tools to identify patients at high risk of an adverse condition for intervention to improve outcomes. Many of these tools are recommended in policy, NICE guidelines & widely implemented in clinical practice. Examples include three risk assessment tools known as QFracture, QDiabetes and QAdmissions.

1. The QFracture tool which assesses risk of fracture has been recommended in NICE guidelines on fracture prevention (August 2012). It is the preferred tool in the NICE quality standard (2016) & SIGN in Scotland. It was included as a quality indicator within the GP Quality and Outcomes Framework in 2013/4. The tool is integrated in over 4,300 GP practices. It helps identify patients who have a high risk of fracture who can then be offered interventions to reduce their risk of a fracture. This is especially important for elderly patients for whom a osteoporotic fracture can be life changing or life limiting.

2. QDiabetes is recommended by 2012 NICE guidance on diabetes prevention & used to identify patients for the national Diabetes Prevention Programme for interventions to reduce risk of type 2 diabetes. Interventions to reduce diabetes risk will have benefits for the individual patient who may otherwise have developed diabetes. Reducing the incidence of diabetes is also likely to have a wider benefit for the health system given the high work load and costs associated with caring for people with diabetes. Research on risk assessment for diabetes has fed into recently published NICE guideline on prevention of diabetes (PH38, September 2017) and the update to the “NHS Health Checks best Practice Guidance” published by Public Health England in February 2017. This is now being used by GP practices to identify patients suitable for the diabetes prevention program which includes interventions to reduce risk of diabetes

3. Research on identifying patients at high risk of emergency hospital admissions led to the development of the QAdmissions risk assessment tool. This tool is recommended by 2016 NICE guidance on co-morbidity & is used to delivery NHS England's unplanned admissions Designated Enhanced Service by identifying those at high risk of emergency admissions for targeted interventions to reduce their risk of having an emergency admission. This is important since emergency admissions are distressing for the patient and their families but also put strain on the NHS. It will also help GP practices to identify frail older patients as required by the new changes to the 2017 GP contract.

Research into the early diagnosis of cancer was awarded the 2012 Royal College of General Practitioners (RCGP) paper of the year category. It also led to implementation of new risk assessment tool in over 4300 GP practices to improve early diagnosis of cancer, in partnership with Macmillan Cancer Support. It also led to professor Hippisley-Cox becoming an expert witness to the All Party Parliamentary Group on Pancreatic Cancer (2012, 2017) which produced high profile recommendations on how to improve early diagnosis of pancreatic cancer. Patients who are diagnosed with cancer at an earlier stage have a better chance of accessing treatment which is capable of improving survival and quality of life.

Research on the safety of antidepressants has been published in the British Medical Journal and research on how it could be used to improve prescribing decisions for antidepressants in primary care is supported as part of the new NIHR funded Biomedical Research Centre in Nottingham (2017-2022).

The results of recent research on cardiovascular risk in people with severe mental health illness were shared in August 2017 with the guideline development group for guideline CG178 entitled “The Psychosis and schizophrenia in adults: prevention and management”.

• The results of work on unplanned admissions was shared with the NICE guideline on multi-morbidity [NG56] to inform its review and update process. This guideline was updated in September 2016 and can be found here. https://www.nice.org.uk/guidance/ng56

• Planned research on the risks and benefits of HRT were targeted at the recent NICE guideline on HRT at the review in 2019. Patient representatives have been involved with the development of the research question and the grant application and will advise on the research as it progresses.

Results have been shared with the Parliamentary Enquiry into Pancreatic Cancer (JHC attended as an expert witness in 2017).

Results are also regularly shared with patient participants on the QResearch Advisory Board and PPI representatives on individual research projects.

As of this V1 data sharing agreement, quarterly data releases under this agreement have been replaced by monthly data releases under DARS-NIC-375354-G8V1H in order to support urgent COVID-19 research. Data released under DARS-NIC-375354-G8V1H can be reused for the purposes described within this DARS-NIC-240279-Y2V2N-v1 data sharing agreement, as set out below.

---

QResearch is a database of linked medical records that has been used and continues to be used by a variety of research projects undertaken by UK universities, from reviewing the safety of antidepressant medicines to studying factors to predict variations in survival rates for cancer patients.

QResearch has also been used to develop and validate risk prediction algorithms such as QRISK2. QRISK2 can be used by clinicians to calculate an individual’s risk of a heart attack or stroke taking account of their individual risk factors such as age, sex, ethnicity, clinical values and diagnoses. The research describing the derivation and validation of QRISK2 has been published in the BMJ (2008) and the software implementing QRISK2 is available as open and closed source software. QRISK2 is recommended by NICE as the risk score for use in its guidance on lipid modification (2014) and it is also recommended for use in the NHS Health Check.

The HES and mortality data are requested to link to the existing QResearch database so that it can be used for medical research. The QResearch database consists of the coded pseudonymised electronic health records from primary care patients registered with approximately 1,500 general practices spread throughout the UK.

QResearch is a not for profit collaboration originally between the University of Nottingham and Egton Medical Information Systems (EMIS) but the University of Nottingham’s roles and responsibilities are being transferred to the University of Oxford. Strategic decisions about the GP data are taken by a Management Board representing the interests of EMIS and the University of Oxford. The University of Oxford is the sole data controller for the datasets which are linked to QResearch (deaths, cancer and hospital data) and the single point of access to the data.

The database is widely used for medical research into the causes of disease, its natural history, treatment and outcomes. QResearch was started in 2003 in order to improve access for research to primary care data and will continue for the foreseeable future.

In addition to coded data from the GP electronic record, the QResearch database also contains the linked cause of death derived from the death certificate data which was originally supplied directly by the Office of National Statistics (ONS) but NHS Digital has since assumed ownership for this data, and cancer registration data supplied directly by Public Health England, following approval by Trent MREC and Secretary of State for Health. From October 2018, new mortality data (referred to as Civil Registration data) will be supplied by NHS Digital. Any mortality originally supplied by ONS data will be considered data supplied by NHS Digital and only NHS Digital can approve ongoing access to this data and its use for specific purposes.

The data linkages for QResearch were extended in 2011 to include additional health information from secondary care including HES. The additional linked HES data enables researchers to analyse additional information on patient characteristics, treatment and outcomes which will improve the epidemiological analyses of studies since the data will be more complete. Without the data linkage research studies may under-estimate the risk and benefits associated with interventions such as prescribed medicines. For example, QResearch data linked to HES data is being used to undertake an assessment of different types of direct anticoagulant medication which is prescribed in primary care to reduce risk of stroke and manage thrombosis. Adverse effects from anticoagulants include major haemorrhage which can be life threatening or life ending. Haemorrhage can affect the brain, gastrointestinal tract, urinary track or other parts of the body. The primary care data provides information on the prescriptions issued and the linked HES data provides information on haemorrhage which is serious enough to require hospital admission. Another example is a recent study looking at safety of the oral contraceptive pill. The primary care data provide information on exposure to the medication and the linked HES data provides information on thrombosis. These are just two examples of projects which can only be done using the linked data. The results help identify and quantify the risks and benefits associated with different types of medication, used in different patients, at different doses over time because of the outcomes. The results help doctors and patients make better decisions and increase the evidence base to inform guideline development and policy.

The patient level data linked to QResearch is only accessed by academics employed by University of Oxford, as well as authorised individuals employed by the University of Nottingham. In all cases, data can only be accessed on site at the University of Oxford. However, the researchers involved in a given project (contributing to the research question, design, interpretation and writing of the paper for publication but not handling the data) may be employed by other UK universities. The HES and mortality data stay on site at the University of Oxford and are only handled by University of Oxford staff, the data processor contracted to the University of Oxford (Dancing House Consulting) and authorised researchers employed by the University of Nottingham. The University of Oxford may have a collaborator at another university on the project team advising on clinical aspects or interpretation of findings, but they will not receive any data. In addition, the external researcher may initiate a project but the University of Oxford has sole autonomy for determining the purposes for which the HES and/or civil registration data will be processed and analysis will be done by University of Oxford staff with the data located at the University of Oxford. Data will not be used for any solely commercial purposes and all applications for the use of HES and/or mortality linked data are subject to a governance process explained in the Processing Activities section.

Only University of Oxford staff, their data processor, Dancing House Consulting, and authorised researchers employed by the University of Nottingham will have access to HES and/or mortality record level data and external researchers will only have access to tabular outputs. Record level data are not shared with researchers outside of the University of Oxford. Small numbers are suppressed in line with the HES Analysis Guide.

Research undertaken using the extended database continues to be processed using the existing arrangements with respect to scientific review and annual reports to Trent MREC. Research has to be peer reviewed, original, hypothesis driven or hypothesis testing, intended for publication in an academic peer reviewed journal.

All research undertaken using the QResearch database and linked data are subject to independent peer review and the results of all research are published.

The outputs are research papers which are published in peer reviewer academic scientific journals and presented at academic conferences. All research is published in academic journals with a link from the QResearch website on an ongoing basis. The publications are accompanied by with press releases from the relevant organisations and highlighted on social media.

Examples of conferences include the annual academic conference for the Society of Academic Primary Care ( next due July 2018) and the NIHR School for Primary Care Research (Sept 2017); international conferences such as the North American Primary Care Research Group (NAPCRG- US Nov 2018); the biannual conferences of the EMIS National User Group (a national education and research charity representing the GP practices who contribute data to QResearch Sept 2017, Feb 2018, Sept 2018 etc); annual conferences run by cancer charities such as Macmillan Cancer Support and Pancreatic Cancer UK; local and regional conferences run by the Nottingham Biomedical Research Centre (Dec 2017).

Results are also shared with policy makers and NICE guideline committees on a regular basis via their stakeholder consultations in order to support development of relevant guidelines. For example,

• The results of recent research on cardiovascular risk in people with severe mental health illness were shared in August 2017 with the guideline development group for guideline CG178 entitled “The Psychosis and schizophrenia in adults: prevention and management”.

• The results of work on unplanned admissions was shared with the NICE guideline on multi-morbidity [NG56] to inform its review and update process. This guideline was updated in September 2016 and can be found here. https://www.nice.org.uk/guidance/ng56

• Research on risk assessment for diabetes has fed into recently published NICE guideline on prevention of diabetes (PH38, September 2017) and the update to the “NHS Health Checks best Practice Guidance” published by Public Health England in February 2017.

• The recent research to update QRISK3 was shared at meetings with the British Heart Foundation and Public Health England as well with the National Directors of Cardiovascular Disease (Huon Grey and Matt Kearney) and at meetings of Expert Scientific Advisory Group which oversees the NHS Health Checks program and which is chaired by John Newton (PHE). QRISK3 will be implemented in the NHS from 2018 and will continue to be updated on regular basis to ensure that the tool gives the most accurate estimation of risk possible.

• Planned research on the risks and benefits of HRT are targeted at the recent NICE guideline on HRT at the next review in 2019. Patient representatives have been involved with the development of the research question and the grant application and will advise on the research as it progresses.

The results of two studies published in the BMJ which described enhanced methods to estimate risk of bleeding (QBleed) and stroke (QStroke) have informed NICE's decision to update their guidelines on atrial fibrillation [CG180]. For more details please see the information published on NICE's website in Sept 2017 on assessment of bleeding and stroke risk https://www.nice.org.uk/guidance/cg180/resources/surveillance-report-2017--atrial-fibrillation-management-2014-nice-guideline-cg180-4597399263/chapter/Surveillance-decision

Results have been shared with the Parliamentary Enquiry into Pancreatic Cancer (JHC attended as an expert witness in 2017).

Results are also regularly shared with patient participants on the QResearch Advisory Board and PPI representatives on individual research projects.

The results tables within the papers will only contain statistical information with cell counts of > 5, being suppressed in line with the ICO code on anonymisation. Outputs will contain aggregate level data with small numbers suppressed in line with the HES analysis guide.

No indicators are produced which show performance of an organisation – indeed the identity of the GP practices contributing to QResearch are not shared with any third party.

The results of research undertaken continues to result in new knowledge and understanding regarding disease epidemiology, health inequalities, drug safety, methods of identifying patients at high risk of serious illnesses. Every year new research is published in high impact international research journals such as the British Medical Journal and the British Journal of General Practice. The research is ongoing with target dates for individual projects rather than one overall target date. A complete list of research papers using QResearch database is published at http://www.qresearch.org/SitePages/publications.aspx

There are many benefits arising from this research - some examples are listed here.

QRISK, which the applicant continues to develop and enhance, has become the preferred cardiovascular risk assessment tool used across the NHS. It is implemented in every GP practice, used extensively in NHS Health Checks, and a variant is used for the NHS Choices 'Heart Age' tool. It was recognised as an outstanding impact case study in the Research Assessment Exercise in 2014. Use of QRISK2 has led to targeted interventions to reduce CVD risk as shown in NHS health check research published in BMJ Open in 2016.

In response to nationally identified NHS needs, the applicant has developed & updated a suite of risk prediction tools to identify patients at high risk of an adverse condition for intervention to improve outcomes. Many of these tools are recommended in policy, NICE guidelines & widely implemented in clinical practice. Examples include three risk assessment tools known as QFracture, QDiabetes and QAdmissions.

1. The QFracture tool which assesses risk of fracture has been recommended in NICE guidelines on fracture prevention (August 2012). It is the preferred tool in the NICE quality standard (2016) & SIGN in Scotland. It was included as a quality indicator within the GP Quality and Outcomes Framework in 2013/4. The tool is integrated in over 4,300 GP practices. It helps identify patients who have a high risk of fracture who can then be offered interventions to reduce their risk of a fracture. This is especially important for elderly patients for whom a osteoporotic fracture can be life changing or life limiting.

2. QDiabetes is recommended by 2012 NICE guidance on diabetes prevention & used to identify patients for the national Diabetes Prevention Programme for interventions to reduce risk of type 2 diabetes. Interventions to reduce diabetes risk will have benefits for the individual patient who may otherwise have developed diabetes. Reducing the incidence of diabetes is also likely to have a wider benefit for the health system given the high work load and costs associated with caring for people with diabetes.

3. Research on identifying patients at high risk of emergency hospital admissions led to the development of the QAdmissions risk assessment tool. This tool is recommended by 2016 NICE guidance on co-morbidity & is used to delivery NHS England's unplanned admissions Designated Enhanced Service by identifying those at high risk of emergency admissions for targeted interventions to reduce their risk of having an emergency admission. This is important since emergency admissions are distressing for the patient and their families but also put strain on the NHS. It will also help GP practices to identify frail older patients as required by the new changes to the 2017 GP contract.

Research into the early diagnosis of cancer was awarded the 2012 Royal College of General Practitioners (RCGP) paper of the year category. It also led to implementation of new risk assessment tool in over 4300 GP practices to improve early diagnosis of cancer, in partnership with Macmillan Cancer Support. It also led to professor Hippisley-Cox becoming an expert witness to the All Party Parliamentary Group on Pancreatic Cancer (2012, 2017) which produced high profile recommendations on how to improve early diagnosis of pancreatic cancer. Patients who are diagnosed with cancer at an earlier stage have a better chance of accessing treatment which is capable of improving survival and quality of life.

Research on the safety of antidepressants has been published in the British Medical Journal and research on how it could be used to improve prescribing decisions for antidepressants in primary care is supported as part of the new NIHR funded Biomedical Research Centre in Nottingham (2017-2022).

QResearch is a database of linked medical records that has been used and continues to be used by a variety of research projects undertaken by UK universities, from reviewing the safety of antidepressant medicines to studying factors to predict variations in survival rates for cancer patients.

QResearch has also been used to develop and validate risk prediction algorithms such as QRISK2. QRISK2 can be used by clinicians to calculate an individual’s risk of a heart attack or stroke taking account of their individual risk factors such as age, sex, ethnicity, clinical values and diagnoses. The research describing the derivation and validation of QRISK2 has been published in the BMJ (2008) and the software implementing QRISK2 is available as open and closed source software. QRISK2 is recommended by NICE as the risk score for use in its guidance on lipid modification (2014) and it is also recommended for use in the NHS Health Check.

The HES and mortality data are requested to link to the existing QResearch database so that it can be used for medical research. The QResearch database consists of the coded pseudonymised electronic health records from primary care patients registered with approximately 1,500 general practices spread throughout the UK.

QResearch is a not for profit collaboration originally between the University of Nottingham and Egton Medical Information Systems (EMIS) but the University of Nottingham’s roles and responsibilities are being transferred to the University of Oxford. Strategic decisions about the GP data are taken by a Management Board representing the interests of EMIS and the University of Oxford. The University of Oxford is the sole data controller for the datasets which are linked to QResearch (deaths, cancer and hospital data) and the single point of access to the data.

The database is widely used for medical research into the causes of disease, its natural history, treatment and outcomes. QResearch was started in 2003 in order to improve access for research to primary care data and will continue for the foreseeable future.

In addition to coded data from the GP electronic record, the QResearch database also contains the linked cause of death derived from the death certificate data which was originally supplied directly by the Office of National Statistics (ONS) but NHS Digital has since assumed ownership for this data, and cancer registration data supplied directly by Public Health England, following approval by Trent MREC and Secretary of State for Health. From October 2018, new mortality data (referred to as Civil Registration data) will be supplied by NHS Digital. Any mortality originally supplied by ONS data will be considered data supplied by NHS Digital and only NHS Digital can approve ongoing access to this data and its use for specific purposes.

The data linkages for QResearch were extended in 2011 to include additional health information from secondary care including HES. The additional linked HES data enables researchers to analyse additional information on patient characteristics, treatment and outcomes which will improve the epidemiological analyses of studies since the data will be more complete. Without the data linkage research studies may under-estimate the risk and benefits associated with interventions such as prescribed medicines. For example, QResearch data linked to HES data is being used to undertake an assessment of different types of direct anticoagulant medication which is prescribed in primary care to reduce risk of stroke and manage thrombosis. Adverse effects from anticoagulants include major haemorrhage which can be life threatening or life ending. Haemorrhage can affect the brain, gastrointestinal tract, urinary track or other parts of the body. The primary care data provides information on the prescriptions issued and the linked HES data provides information on haemorrhage which is serious enough to require hospital admission. Another example is a recent study looking at safety of the oral contraceptive pill. The primary care data provide information on exposure to the medication and the linked HES data provides information on thrombosis. These are just two examples of projects which can only be done using the linked data. The results help identify and quantify the risks and benefits associated with different types of medication, used in different patients, at different doses over time because of the outcomes. The results help doctors and patients make better decisions and increase the evidence base to inform guideline development and policy.

The patient level data linked to QResearch is only accessed by academics employed by University of Oxford, as well as authorised individuals employed by the University of Nottingham. In all cases, data can only be accessed on site at the University of Oxford. However, the researchers involved in a given project (contributing to the research question, design, interpretation and writing of the paper for publication but not handling the data) may be employed by other UK universities. The HES and mortality data stay on site at the University of Oxford and are only handled by University of Oxford staff, the data processor contracted to the University of Oxford (Dancing House Consulting) and authorised researchers employed by the University of Nottingham. The University of Oxford may have a collaborator at another university on the project team advising on clinical aspects or interpretation of findings, but they will not receive any data. In addition, the external researcher may initiate a project but the University of Oxford has sole autonomy for determining the purposes for which the HES and/or civil registration data will be processed and analysis will be done by University of Oxford staff with the data located at the University of Oxford. Data will not be used for any solely commercial purposes and all applications for the use of HES and/or mortality linked data are subject to a governance process explained in the Processing Activities section.

Only University of Oxford staff, their data processor, Dancing House Consulting, and authorised researchers employed by the University of Nottingham will have access to HES and/or mortality record level data and external researchers will only have access to tabular outputs. Record level data are not shared with researchers outside of the University of Oxford. Small numbers are suppressed in line with the HES Analysis Guide.

Research undertaken using the extended database continues to be processed using the existing arrangements with respect to scientific review and annual reports to Trent MREC. Research has to be peer reviewed, original, hypothesis driven or hypothesis testing, intended for publication in an academic peer reviewed journal.

All research undertaken using the QResearch database and linked data are subject to independent peer review and the results of all research are published.

The outputs are research papers which are published in peer reviewer academic scientific journals and presented at academic conferences. All research is published in academic journals with a link from the QResearch website on an ongoing basis. The publications are accompanied by with press releases from the relevant organisations and highlighted on social media.

Examples of conferences include the annual academic conference for the Society of Academic Primary Care ( next due July 2018) and the NIHR School for Primary Care Research (Sept 2017); international conferences such as the North American Primary Care Research Group (NAPCRG- US Nov 2018); the biannual conferences of the EMIS National User Group (a national education and research charity representing the GP practices who contribute data to QResearch Sept 2017, Feb 2018, Sept 2018 etc); annual conferences run by cancer charities such as Macmillan Cancer Support and Pancreatic Cancer UK; local and regional conferences run by the Nottingham Biomedical Research Centre (Dec 2017).

Results are also shared with policy makers and NICE guideline committees on a regular basis via their stakeholder consultations in order to support development of relevant guidelines. For example,

• The results of recent research on cardiovascular risk in people with severe mental health illness were shared in August 2017 with the guideline development group for guideline CG178 entitled “The Psychosis and schizophrenia in adults: prevention and management”.

• The results of work on unplanned admissions was shared with the NICE guideline on multi-morbidity [NG56] to inform its review and update process. This guideline was updated in September 2016 and can be found here. https://www.nice.org.uk/guidance/ng56

• Research on risk assessment for diabetes has fed into recently published NICE guideline on prevention of diabetes (PH38, September 2017) and the update to the “NHS Health Checks best Practice Guidance” published by Public Health England in February 2017.

• The recent research to update QRISK3 was shared at meetings with the British Heart Foundation and Public Health England as well with the National Directors of Cardiovascular Disease (Huon Grey and Matt Kearney) and at meetings of Expert Scientific Advisory Group which oversees the NHS Health Checks program and which is chaired by John Newton (PHE). QRISK3 will be implemented in the NHS from 2018 and will continue to be updated on regular basis to ensure that the tool gives the most accurate estimation of risk possible.

• Planned research on the risks and benefits of HRT are targeted at the recent NICE guideline on HRT at the next review in 2019. Patient representatives have been involved with the development of the research question and the grant application and will advise on the research as it progresses.

The results of two studies published in the BMJ which described enhanced methods to estimate risk of bleeding (QBleed) and stroke (QStroke) have informed NICE's decision to update their guidelines on atrial fibrillation [CG180]. For more details please see the information published on NICE's website in Sept 2017 on assessment of bleeding and stroke risk https://www.nice.org.uk/guidance/cg180/resources/surveillance-report-2017--atrial-fibrillation-management-2014-nice-guideline-cg180-4597399263/chapter/Surveillance-decision

Results have been shared with the Parliamentary Enquiry into Pancreatic Cancer (JHC attended as an expert witness in 2017).

Results are also regularly shared with patient participants on the QResearch Advisory Board and PPI representatives on individual research projects.

The results tables within the papers will only contain statistical information with cell counts of > 5, being suppressed in line with the ICO code on anonymisation. Outputs will contain aggregate level data with small numbers suppressed in line with the HES analysis guide.

No indicators are produced which show performance of an organisation – indeed the identity of the GP practices contributing to QResearch are not shared with any third party.

The results of research undertaken continues to result in new knowledge and understanding regarding disease epidemiology, health inequalities, drug safety, methods of identifying patients at high risk of serious illnesses. Every year new research is published in high impact international research journals such as the British Medical Journal and the British Journal of General Practice. The research is ongoing with target dates for individual projects rather than one overall target date. A complete list of research papers using QResearch database is published at http://www.qresearch.org/SitePages/publications.aspx

There are many benefits arising from this research - some examples are listed here.

QRISK, which the applicant continues to develop and enhance, has become the preferred cardiovascular risk assessment tool used across the NHS. It is implemented in every GP practice, used extensively in NHS Health Checks, and a variant is used for the NHS Choices 'Heart Age' tool. It was recognised as an outstanding impact case study in the Research Assessment Exercise in 2014. Use of QRISK2 has led to targeted interventions to reduce CVD risk as shown in NHS health check research published in BMJ Open in 2016.

In response to nationally identified NHS needs, the applicant has developed & updated a suite of risk prediction tools to identify patients at high risk of an adverse condition for intervention to improve outcomes. Many of these tools are recommended in policy, NICE guidelines & widely implemented in clinical practice. Examples include three risk assessment tools known as QFracture, QDiabetes and QAdmissions.

1. The QFracture tool which assesses risk of fracture has been recommended in NICE guidelines on fracture prevention (August 2012). It is the preferred tool in the NICE quality standard (2016) & SIGN in Scotland. It was included as a quality indicator within the GP Quality and Outcomes Framework in 2013/4. The tool is integrated in over 4,300 GP practices. It helps identify patients who have a high risk of fracture who can then be offered interventions to reduce their risk of a fracture. This is especially important for elderly patients for whom a osteoporotic fracture can be life changing or life limiting.

2. QDiabetes is recommended by 2012 NICE guidance on diabetes prevention & used to identify patients for the national Diabetes Prevention Programme for interventions to reduce risk of type 2 diabetes. Interventions to reduce diabetes risk will have benefits for the individual patient who may otherwise have developed diabetes. Reducing the incidence of diabetes is also likely to have a wider benefit for the health system given the high work load and costs associated with caring for people with diabetes.

3. Research on identifying patients at high risk of emergency hospital admissions led to the development of the QAdmissions risk assessment tool. This tool is recommended by 2016 NICE guidance on co-morbidity & is used to delivery NHS England's unplanned admissions Designated Enhanced Service by identifying those at high risk of emergency admissions for targeted interventions to reduce their risk of having an emergency admission. This is important since emergency admissions are distressing for the patient and their families but also put strain on the NHS. It will also help GP practices to identify frail older patients as required by the new changes to the 2017 GP contract.

Research into the early diagnosis of cancer was awarded the 2012 Royal College of General Practitioners (RCGP) paper of the year category. It also led to implementation of new risk assessment tool in over 4300 GP practices to improve early diagnosis of cancer, in partnership with Macmillan Cancer Support. It also led to professor Hippisley-Cox becoming an expert witness to the All Party Parliamentary Group on Pancreatic Cancer (2012, 2017) which produced high profile recommendations on how to improve early diagnosis of pancreatic cancer. Patients who are diagnosed with cancer at an earlier stage have a better chance of accessing treatment which is capable of improving survival and quality of life.

Research on the safety of antidepressants has been published in the British Medical Journal and research on how it could be used to improve prescribing decisions for antidepressants in primary care is supported as part of the new NIHR funded Biomedical Research Centre in Nottingham (2017-2022).