Investigation of innovation diffusion within consultant specialties

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Processing activities

An extension to this Data Sharing Agreement until 2025 is required to allow time for Imperial College London to complete analyses and produce outputs. No additional data is being requested under this iteration of the Agreement. No further data will be disseminated under v5 of the Data Sharing Agreement. All previously stated As the research on cardiac care is still at the initial results stage, a two year extension to this Data Sharing Agreement is required to allow time for Imperial College London to complete analyses are now nearly complete and journal articles are produce outputs. No additional data is being requested under submission for publication in respect this iteration of the cancer treatment work. The cardiac care research project is however at a less advanced stage – hence the need for a further 3-year extension to the Data Sharing Agreement. All previously stated analyses are now nearly complete and journal articles are under submission for publication in respect of the cancer treatment work. [21 paragraphs unchanged]

Expected output

[1 paragraph unchanged] The journals being targeted are health service research journals (e.g. Social Science and Medicine and the BMJ) and economics journals (e.g. Journal of Health Economics). One paper has already been published - https://www.sciencedirect.com/science/article/pii/S0277953621000472?via%3Dihub with two others submitted to two other journals. The target date for publication of these papers is 2023 or 2024. Papers for the cardiac care research are still at the research stage. The journals being targeted are health service research journals (e.g. Social Science and Medicine and the BMJ) and economics journals (e.g. Journal of Health Economics). The target date for publication of these is 2024/2025. [7 paragraphs unchanged]

Benefits reported

[10 paragraphs unchanged] YIELDED BENEFITS UPDATE 2026 Barrenho, E, Gautier, E, Miraldo, M, Propper, C and Rose, C (2025) Innovation Diffusion among Coworkers: Evidence from Senior Doctors. Management Science Jan 2025 https://doi.org/10.1287/mnsc.2023.00496 Miraldo M, Propper C, Walsh B, Barrenho E (2021) The importance of surgeons and their peers in adoption and diffusion of innovation: An Observational Study of Laparoscopic Colectomy Adoption and Diffusion in England, Social Science and Medicine, ISSN: 0277-9536

Unchanged: Objective for processing, Expected measurable benefits.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

Diffusion of innovation in cancer treatment plays a key role in improving the survival rates and the quality of life of patients affected by cancer. BACKGROUND AND PURPOSE The English National Health System has experienced a slower diffusion of innovation than health systems in other countries. To speed up the diffusion of innovation the Department of Health instituted formal cancer networks covering all NHS Trusts in 2001 and NICE issued guidelines intended to increase adoption of innovative procedures with proven benefits. The aim is to understand the role of the government introduced formal networks and the less formal professional networks of consultant surgeons in determining adherence to these voluntary guidelines. Diffusion of innovation (the diffusion of innovation theory seeks to explain how, why and at what rate new ideas and technology spread) in cancer treatment plays a key role in improving the survival rates and the quality of life of patients affected by cancer. This project focuses on surgical treatment of two specific types of cancer. In both cases surgical treatment of these cancers is covered by clinical guidance issued by NICE. This guidance is intended to increase the rate of diffusion of innovation. The NHS has experienced a slower diffusion of innovation compared to the health systems of other countries. To speed up the diffusion of innovation the Department of Health instituted formal cancer networks covering all NHS Trusts in 2001 and The National Institute for Health and Care Excellence (NICE) issued guidelines intended to increase the adoption of innovative procedures with proven benefits. The aim of the ‘Investigation of Innovation Diffusion within Consultant Specialties’ research project, led by the Imperial College London Business School, is to understand the role of the government-introduced formal networks and the less formal professional networks of consultant surgeons in determining adherence to these voluntary guidelines. 1. Colorectal Cancer: Patients undergoing surgery can undergo two alternative procedures (a) Open Surgery (b) Laparoscopic Surgery (the innovation). Laparoscopic surgery has been shown to result in significant improvements in outcomes for a range of surgical procedures including colon cancer resection compared to open resection, in terms of mortality, and secondary outcomes such as shorter length of stay, reduced surgical complications, reduced bleeding and pain, and lower hospital costs. This research project initially focused on the surgical treatment of two specific types of cancer. In both cases the surgical treatment of these cancers is covered by clinical guidance issued by NICE. In 2018 however, Imperial College London applied to NHS England to extend the analytical scope of the research project to examine the uptake of innovative technology and practice outside of cancer services. Imperial College London opted to focus on cardiac surgery due to the high volume of said surgeries and the fact that cardiac surgeries are subject to considerable variation in the uptake of differing methodologies. The intention of this extension in scope was to examine whether Imperial College London’s findings were generalisable outside of cancer surgery. 2. Breast Cancer Patients who need to have a lymph node dissected have two possible treatments (a) Axillary lymph node dissection (ALND) (b) Sentinel lymph node dissection (SLN) (the innovation). ALND is the procedure traditionally used to assess whether cancer has spread from the breast to neighbouring lymph nodes. SLN is the surgical removal of one or more small lymph glands from the axilla. ALND is associated with significant more side effects and morbidity relative to SLN, yet the use of SLN has plateaued in the NHS. Focusing on the surgical treatment of the two specific types of cancer: This project will examine the impact on diffusion of these innovations of (a) professional networks of consultants (b) the formal spatially based cancer networks of NHS Trusts. The former are examined because of their importance in shaping individual and firm behaviour. 1. Colorectal Cancer: Patients undergoing surgery can undergo two alternative procedures: (a) Open Surgery or (b) Laparoscopic Surgery – a means of performing surgery on the abdomen and pelvis without making a large incision (the innovation). Laparoscopic surgery has been shown to result in significant improvements in outcomes for a range of surgical procedures including colon cancer resection compared to open resection in relation to mortality and secondary outcomes such as shorter length of stay, reduced surgical complications, reduced bleeding and pain, and lower hospital costs. The long training period of consultants, the importance of the Royal Colleges and the dominance of the NHS as an employer means that professional networks are potentially very important in the NHS in determining clinical practice. The latter are examined as they are important policy instruments intended to increase collaboration between hospitals in cancer care. However, they are in operation at the same time as there were, perhaps conflicting, and policies to allow competition for patients between hospitals (the right of patients, with their GPs, to choose the hospital of treatment from 2006 onwards). Therefore the impact of spatial competition between nearby NHS Trusts will also be studied. 2. Breast Cancer: Patients, who need to have a lymph node dissected, have two possible modes of treatment: (a) Axillary lymph node dissection (ALND) – removal of a large number of lymph nodes from the axilla (underarm) or (b) Sentinel lymph node dissection (SLN) – the removal of only the diseased nodes (the innovation). ALND is the procedure traditionally used to assess whether cancer has spread from the breast to neighbouring lymph nodes despite being associated with significantly more side effects and morbidity relative to SLN. The use of SLN has in contrast plateaued across the NHS. Imperial are modelling cancer innovation uptake as a function of a number of factors that are considered to be important in the uptake of innovation literature as well as ones that are specific to the treatments examined. These include: This research project examines the impact on diffusion of these innovations for: (a) professional networks of consultants and (b) the formal spatially based cancer networks of NHS Trusts. (a) the work history of the consultants (hence the need for workforce data) The former is examined because the long training period of consultants, the importance of the Royal Colleges and the dominance of the NHS as an employer means that professional networks are potentially very important in the NHS in determining clinical practice and shaping individual and firm behaviour. (b) the nature of the hospital they work in, including measures of size; patient mix (severity, SES status, age, gender) across a range of specialties (not just oncology or surgery); the extent to which laproscopic techniques are used in other specialties; resources that may be complementary to the innovation (e.g. Use of imaging); decisions to invest in areas other than cancer or general surgery; the extent of innovation in other areas; the position of the hospital as a training facility and the position of the hospital in various networks for specialist care (e.g. Stroke networks). What is important is often not just whether a hospital does something but how much it does in absolute and relative terms. The latter is examined as cancer networks of NHS Trusts are important policy instruments intended to increase collaboration between hospitals in cancer care. However, they are in operation at the same time as policies (perhaps conflicting policies) facilitating competition for patients between hospitals (the right of patients, with their GPs, to choose the hospital of treatment from 2006 onwards). Therefore, the impact of spatial competition between nearby NHS Trusts has also been studied. (c) The rest of HES is useful in that it allows issues with missing data that come from coding and recording practices. For example, it may appear that some consultants do very low volume in one year preceded and followed by very high volumes. The low volume may in fact not be correct and may reflect HES coding practices in a hospital. Imperial will be able to examine this by looking at the patterns of consultant recording in other specialties. It means sophisticated Monte Carlo methods to deal with missing data under various assumptions can be utilised. This is very important when using administrative data. In relation to the work on cardiac care, by evaluating cardiac specialists, the research project is able to explore the distribution of patient outcomes resulting from cardiac care. The research project may then be able to evaluate to what degree these outcomes can be explained by the same factors assessed in relation to cancer care: clinician networks, competition, and doctor experience. This allows Imperial College London to test whether the factors they find are specific to cancer treatment or can be generalised to other settings. This will greatly increase the external validity of the research project. (d) Finally, the approach has the added benefit that rapid progress in cancer treatment can be analysed, Imperial can look at spillovers from these into other innovations in related areas such as surgery. This will increase the value to the patients and user community of the research. DATA SUMMARY Most of these variables in (b), (c) and (d) will be measured by use of the HES data outside oncology and surgery. For that reason the whole HES data set is required (excluding maternity and mental health). In order to meet the objectives of the research project, data from three different sources, was required. These sources do not always match up at the consultant-year level. The GMC data is for all consultants registered with the GMC. The workforce data convers those individuals who were consultants in 1992 or became a consultant before 2012. The inpatient HES data the team use to study innovation adoption and patient outcomes only contains a consultant identifier if the consultant did inpatient activity in that year. The literature on innovation shows that context matters and therefore how much the rest of the HES data set is needed to examine this context as well as address a range of statistical issues. The whole purpose of the study would be seriously invalidated if Imperial were not able to have the data to establish this and to allow for miscoding and recording. A number of the data sources came from administrative records whereas others were collected as part of the research process. All data has been matched at the consultant and/or site and/or trust level – the level of data was determined by the type of data collected. In 2018 the study extended their analysis to look at the same factors that might determine the uptake of innovative technology and practice, but outside of cancer services, using the same dataset and requiring no further data. Imperial College London hold the following data under this Data Sharing Agreement (all data disseminated is record-level, pseudonymised data): By evaluating cardiac specialists, the study will be able to explore the distribution of patient outcomes resulting from cardiac care. The study will then be able to evaluate to what degree these outcomes can be explained by the same factors they were looking at in cancer care: clinician networks, competition, and doctor experience. This allows the study to test whether the factors they find are specific to cancer treatment or can be generalised to other settings. This will greatly increase the external validity of the study. - Workforce data - current and past employment of consultants in the relevant specialties in NHS trusts. The variables used are the trust and specialty in which each relevant consultant is employed in each year, covering the period 1992-1993 to 2014-2015. Imperial College London downloaded the data around the 2nd of August 2016. These are as noted on page 20 of the original DSA agreement. The datasets and the variable in each are detailed below. The primary outcome of interest, mortality, will remain the same, along with secondary outcomes: length of stay, surgical complications, reduced bleeding / pain and lower hospital costs. The statistical analyses the study will undertake will be the same as those for the cancer treatment study. The factors that the study will examine will be the same factors as in the original study focus. These are: GMC_NUMBER (a) Work history YEAR (b) The nature of the trust that the consultant is employed by Most recent appearance age-band SPECIALTY_DESCRIPTION Reporting Org code HCHS_GROUP_EQUIVALENT PAYSCALE_DESCRIPTION Group Reporting Org Name - Hospital Episode Statistics (HES) Admitted Patient Care data - the procedures undertaken by the target consultant population, covering the period from 2000-2001 to 2014-2015. The HES and Workforce data are required to allow Imperial College London to study several factors considered important in the uptake of innovation literature as well as factors that are specific to the treatments examined. These include: (a) the work history of the consultants (hence the need for workforce data). This is because, those in the Workforce data in any given year, may not appear in the inpatient HES data. They will only be in HES if they treated inpatients at some point during that year. A consultant may spend time away from inpatient treatment because they are only treating outpatients, or working in Accident and Emergency, have a management role, or may be on leave (say spending a year abroad). During that time they will be in the Workforce data.  To ensure that the team do not incorrectly classify such consultants as having left the NHS, the team need data to follow them over time. (b) the nature of the hospitals the consultants work in, including measures of size; patient mix (severity, socio-economic status (SES), age, gender) across a range of specialties (not just oncology or surgery); the extent to which laparoscopic techniques are used in other specialties; resources that may be complementary to the innovation (e.g. use of imaging); decisions to invest in areas other than cancer or general surgery; the extent of innovation in other areas; the position of the hospital as a training facility and the position of the hospital in various networks for specialist care (e.g. stroke networks) - what is important is often not just whether a hospital does something but how much it does in absolute and relative terms. (c) The HES data is useful in that it allows Imperial College London to identify issues with missing data that may have come from a hospital’s coding and recording practices. For example, it may appear that some consultants perform a very low volume of clinical work in one year but a very high volume the year after. The low volume may in fact not be correct and reflect incorrect HES coding practices in a hospital. Imperial College London are able examine this by looking at the patterns of consultant recording in other specialties. It means sophisticated methods to deal with missing data under various assumptions can be utilised. This is very important when using administrative data. (d) Finally, the approach has the added benefit of allowing for analysis of rapid progress in cancer treatment. Imperial College London can look at spillovers from these into other innovations in related areas such as cardiac surgery. This may increase the value to the patients and user community of the research project. In relation to the research project on cardiac surgeries, the primary outcome of interest (mortality) remains the same, along with secondary outcomes: length of stay, surgical complications, reduced bleeding / pain and lower hospital costs. The statistical analyses the research project undertakes are the same as those for the cancer treatment research project. The factors that the research project examines are the same factors as that of the original research project. These are: (a) Work history, (b) The nature of the trust that the consultant is employed by, [1 paragraph unchanged] The majority of the variables outlined in points (b), (c) and (d) above are measured by use of HES data outside oncology and surgery. For that reason, the whole HES data set is required (excluding maternity and mental health data). The literature on the diffusion of innovation also shows that context matters and therefore the majority of the HES Admitted Patient Care dataset is needed to examine the wider context as well as address a range of statistical issues. The whole purpose of this research project would be seriously invalidated if Imperial College London was not able to receive data to: (1) examine the wider context and (2) allow for miscoding and recording. Imperial College London also holds General Medical Council (GMC) data - qualifications and date of qualifications of consultants. The variables used are the year of qualification, age band, gender, place, covering the period from 2005-2006 onwards. Under the Medical Act 1983 the GMC have a valid legal basis to disclose registered doctors’ Level 1 information (eg. Gender, name, year and place of primary medical qualification, date of registration etc.) to any enquirer if the information is already published on the List of Registered Medical Practitioners (LRMP). Registered doctors provide their consent for their Level 1 information to be used in this way. For the purposes of this Agreement, Level 1 information is shared by the GMC with Imperial College London. Imperial College London have a signed Agreement in place with the GMC permitting them to share the entire LRMP dataset with for the purposes of linkage by NHS England with HES Admitted Patient Care and Workforce data. For the purposes of this Agreement, no new information will be gathered on consultants - this is part of the workforce data previously held by the applicant. The final sources are publicly available statistics which provide information on, for example, Medium Super Lower Output characteristics, NHS Trust financial performance, details of cancer networks and guidelines on use the of laparoscopic surgery in treatment of colorectal cancer and SLN in breast cancer. DATA CONTROLLERSHIP / FUNDING Imperial College London is the sole Data Controller that will also process the NHS England data disseminated. The research project is partially funded and, to that end, some internal resources are being used for the ongoing processing of the data and the cost of extending this Data Sharing Agreement with NHS England. The research project was originally funded by the Health Foundation under their Efficiency research programme for the period April 2015 – September 2019. It is now partially funded (for Professor Propper’s time only) by her European Research Council Advanced Investigator Grant entitled Healthcare Labour (#788529). The head of research support for Imperial College London, has confirmed that the European Research Council will have no input in how the data is used. Imperial College London are supporting the study with internal funding. LEGAL BASIS FOR PROCESSING Imperial College London rely upon the following legal bases under the General Data Protection Regulation (GDPR) to process personal data and special categories of personal data: Article 6.1.e - processing is necessary for the performance of a task in the public interest or in the exercise of official authority vested in the controller. This research is in the public interest because innovation in the NHS effects both patients, in terms of the care they receive, and the wider public as more innovation makes the NHS more efficient. However, the take up of innovation in the NHS can be slow. It is therefore important to understand what affects adoption of new innovations. The research project focuses on an important innovation treatment Imperial College London is a public authority. The Data Protection Act 2018 s7(1)(a) defines ‘public bodies’, for the purpose of the GDPR, as “a public authority as defined by the Freedom of Information (FOI) Act 2000”. The FOI Act 2000 Part 1, section 3 (1)(a)(i) specifies that a public authority comprises any body listed in Schedule 1. Schedule 1 of the FOI Act 2000 lists “maintained schools and further and higher education institutions” as public authorities. Imperial College London has a Royal Charter. It includes several references to research including: (1) to provide the highest specialised instruction and the most advanced training, education, research and scholarship in science, technology and medicine; (2) freedom to carry out and publish research work without any interference, subject to accepted professional principles including professional responsibility and subject to national and international recognised professional principles of intellectual rigour, scientific inquiry and research ethics. Article 9.2.j - processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes in accordance with Article 89(1). The data are required for research purposes in the public interest – meeting the conditions in the DPA 2018 Schedule 1 Part 1 (4) - which GDPR Recital 52(2) determines is an appropriate derogation from the prohibition on processing special categories of personal data. The ways in which the processing of data will be of benefit to the public – thereby demonstrating that the processing is in the public interest – are described in section 5d.

Processing activities

The object of this analysis is the time to adoption of the innovative procedure by consultants employed in the NHS in the relevant speciality. The study’s focus is the impact on time to adoption of (a) elapsed time plus the factors our analysis is focused upon, these being:- (b) NHS ‘top down’ mandated tumour specific cancer networks (c) consultant initial professional networks (where and with whom the consultant trained) (d) publication of guidelines on adoption of the innovative procedure (e) spatial location of hospital (whether located near other hospitals that provide similar services). This study will estimate models of time to adoption using primarily a generalised hazard approach (time to failure models allowing for heterogeneity), with a focus on the impact of factors (b)-(e). An extension to this Data Sharing Agreement until 2025 is required to allow time for Imperial College London to complete analyses and produce outputs. No additional data is being requested under this iteration of the Agreement. This study will also examine geographical variation in the diffusion of innovation across consultants and hospitals using Arc-GIS heat maps. In each part of the analysis, statistical regression models will be used to disentangle the contribution of different factors, including formal and informal networks of physician, to the diffusion of innovation and potential confounders which may independently affect the speed of adoption and which may be correlated with the factors (b)-(e) will be controlled for. All previously stated analyses are now nearly complete and journal articles are under submission for publication in respect of the cancer treatment work. The cardiac care research project is however at a less advanced stage – hence the need for a further 3-year extension to the Data Sharing Agreement. From this analysis the primary output is estimates of the absolute and relative impact of the factors (b)-(e) on the rate of adoption of innovative surgery for the two cancer types. There was previously an intention under previous iterations of the Agreement for Ipsos MORI to use the linked data disseminated to Imperial College London to identify a control group of consultants who could be followed up via survey questions. This survey was never administered as Ipsos MORI were unable to identify a sufficient number of consultants to interview from their normal panels. The contract was therefore terminated. The researchers at Imperial College London chose not to, and will not as part of this application, make use of a third party survey provider. Data requirements DATA LINKAGE These analyses require bringing together data from several sources, some of which is from administrative records and some of which will be collected as part of the research. These data will need to be matched at the consultant and/or site and/or trust level – the level determined by the type of data to be collected. As noted in section 5a, the GMC have a valid legal basis under the Medical Act 1983 to disclose registered doctors’ Level 1 information to any enquirer (in this case, Imperial College London) if the information is already published on the List of Registered Medical Practitioners (LRMP). Imperial College London have a signed Agreement in place with the GMC permitting them to share Level 1 information from the LRMP dataset with NHS England who then combined GMC data with HES and Workforce data to form a new dataset. The linked product was then returned to Imperial College London for the purposes of this research project in a pseudonymised format. No identifiable data was supplied to Imperial College London by NHS England. Two sources of data are held by NHS Digital (HES and Workforce data). One is published by the GMC. One is a survey of consultants to be carried out by IPSOS for the research project. The final sources are various publicly available data. The data was linked using the clear consultant code from the HES and Workforce data and the GMC number from the GMC dataset. The consultant code/GMC number was then converted into a study ID to permit Imperial College London to use the linked data without the need to access sensitive/identifiable data. The type of data, the sources, the variables the NHS Digital will construct from each source and the time period are detailed below – Fields provided from those datasets included the age band of the consultant (as opposed to actual age) and recent workplace, training and current trust. These were appended to the HES dataset using the same unique study ID. Workforce data - Current and past employment of consultants in the relevant specialties in NHS trusts. The variables used will be the trust and specialty in which each relevant consultant is employed in each year, covering the period 1992-1993 to 2014-2015. INCLUSION/EXCLUSION CRITERIA Published GMC data - Qualifications and date of qualifications of consultants The data supplied contains all consultants. In the first stage of the research the team made an examination of the data to select the exact innovations to be studied. At this stage the team needed to understand how representative the cohort was of the consultant body. In the original application, after selecting the innovations and the relevant consultants the team were going to carry out a qualitative survey of them. To do this the team were to provide NHS England with the consultant’s pseudo anonymised IDs. NHS England would then share with IPSOS (the designated body carrying out the qualitative survey) who these people were. Using the GMC data, IPSOS would run data collection with questionnaires and then they would provide that data back to NHS England (formerly NHS Digital) to merge that for Imperial College London with HES/Workforce/GMC. However, this survey was never undertaken so this merging was never carried out. The variables used will be the year of qualification; age band; gender, place, covering the period from 2005-2006 onwards. DATA STORAGE HES Admitted Patient Care data - The procedures undertaken by the target consultant population (those undertaking breast and colorectal cancer surgery), covering the period from 2000-2001 to 2014-2015 NHS England data is currently stored by the Centre for Health Economics & Policy Innovation (CHEPI) at the Imperial College London Business School, South Kensington, London. The data is stored and processed on a secure machine, located inside a locked office at the Business School. Access to the dataset is controlled by a distribution list maintained by the Principal Investigator. Users must authenticate themselves as holders of Imperial College London computer accounts in order to gain access. This can include visiting researchers, who will hold honorary contracts, and comply with all data security and information protection requirements for access to the data set. Medium Super Lower Output characteristics, publicly available data on Trust financial performance, details of cancer networks and guidelines on use the of laparoscopic surgery in treatment of colorectal cancer and SLN in breast cancer will also be utilised. DATA ANALYSIS The relevant procedures used for both the older treatment and the innovation of the two new tumour treatments. These are (a) Open Surgery (b) Laparoscopic Surgery (the innovation) in colorectal cancer surgery and (a) ALND and (b) SLN (the innovation) in breast cancer surgery. The term “target consultant population” is all those consultants employed as surgeons in breast or colorectal cancer treatment and any other consultants that perform the “relevant procedures”. The object of analysis is the time to adoption of the innovative procedure by consultants employed in the NHS in the relevant speciality. The research project's focus is the impact on time to adoption plus the factors the analysis is focused upon, these being: The data will be linked using the clear consultant code from the HES and workforce data, and the GMC number from the GMC dataset. Once the linkage has taken place, the consultant code/GMC number will be converted into a study ID that permits Imperial to use the linked data without needing sensitive data. i. NHS ‘top down’ mandated tumour specific cancer networks, The linkage will take place at NHS Digital and Imperial will only receive a unique study ID that will be created and present in the HES dataset together with a limited number of fields from both the GMC file and the Workforce file. Fields to be provided from those datasets include Age Band of the Consultant (as opposed to actual age) and recent workplace, training and current trust. These will be appended to the HES dataset using the same unique study ID. ii. consultant initial professional networks (where and with whom the consultant trained), NHS Digital will be performing all of the linkage and will only be providing pseudo files back to Imperial. iii. publication of guidelines on adoption of the innovative procedure, Using the same linked dataset, Imperial College are also evaluating the same factors and outcomes with respect to cardiac care. This change of dependent variable will allow the study to understand whether the findings from the work on cancer care remain true for another speciality and are potentially generalisable to all medical specialities. iv. spatial location of hospital (whether located near other hospitals that provide similar services). The study already hold these data and are not applying for any additional data. The study are not applying for any new data to flow from NHS Digital to Imperial College. There is therefore no need for any additional processing by NHS Digital of the data. The study will not be surveying cardiac consultants for the extension so there will be no use of a third-party survey provider. This research project has sought to estimate models of time to adoption using primarily a generalised hazard approach (time to failure models allowing for heterogeneity), with a focus on the impact of factors (i)-(iv). In terms of analysis, the study will be undertaking the same statistical analyses and using the same modelling techniques as for the original application. The research project is examining geographical variation in the diffusion of innovation across consultants and hospitals using geographical heat maps. In each part of the analysis, statistical regression models (tools to describe the relationship between different variables) are used to disentangle the contribution of different factors, including formal and informal networks of physician, to the diffusion of innovation and potential confounders which may independently affect the speed of adoption and correlate with the factors (i)-(iv) are controlled for. All organisations party to this agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract ie: employees, agents and contractors of the Data Recipient who may have access to that data)” From this analysis the primary output is estimates of the absolute and relative impact of the factors (i)-(iv) on the rate of adoption of innovative surgery for the two cancer types. There will be no data linkage undertaken with the NHS Digital England data provided supplied under this agreement Agreement that is not already noted in the agreement. Agreement. Data will only be accessed and processed by substantive employees of Imperial College London and or individuals who hold an honorary contract with Imperial College London. Data will not be accessed or processed by any other third parties not mentioned unless referenced in this agreement. Agreement. No attempts have been made to re-identify individuals from the pseudonymised NHS England data disseminated.

Expected output

One set of outputs will are anticipated to be statistical analyses to be submitted to peer reviewed academic papers. These outputs will not contain any data which enables identification of individual patients, consultants, sites or NHS trusts. The journals being targeted are health service research journals (e.g. Social Science and Medicine and the BMJ) and economics journals (e.g. Journal of Health Economics). Papers will be submitted during in the final two years of the research (i.e. between 2016 and 2017). The stakeholders interested in this research are diverse and extend beyond the academic community. Given the nature of the benefits, the outputs intend to inform policy making. None of these outputs will contain any data which permit identification of individual patients, consultants, sites or NHS Trusts. Several strategies will be adopted to inform. (i) The project will report regularly to a steering group that includes the Chief Analyst at NHS England. (ii) the funder of the research, the Health Foundation will be used to publish research highlights (iii) individual discussions will be held with policy makers and to make presentations to NHS England and other regulatory bodies. More broadly, the dissemination machinery of Imperial College Business School will be used to disseminate research highlights alumni, the general media and through this, to members of the public. The journals being targeted are health service research journals (e.g. Social Science and Medicine and the BMJ) and economics journals (e.g. Journal of Health Economics). One paper has already been published - https://www.sciencedirect.com/science/article/pii/S0277953621000472?via%3Dihub with two others submitted to two other journals. The target date for publication of these papers is 2023 or 2024. Papers for the cardiac care research are still at the research stage. The journals being targeted are health service research journals (e.g. Social Science and Medicine and the BMJ) and economics journals (e.g. Journal of Health Economics). The target date for publication of these is 2024/2025. All outputs will be published in accordance with the HES Analysis guide and will be aggregated outputs with small numbers suppressed. A diverse range of stakeholders are interested in this research project and extend beyond the academic community: Cardiac treatment (i) The project has reported regularly to a steering group that includes the Chief Analyst at NHS England, The way in which data is handled and the statistical analysis for this extension will be the same as those for the original application. The outputs for this research will be submitted to peer reviewed academic journals. The target journals will be the same as for the cancer treatment project. (ii) The former funder of the research, the Health Foundation has published research highlights, The stakeholders for this research are the same broad set of stakeholders as for the original project. The study will use the same channels to disseminate the findings to these stakeholders. (iii) Individual discussions have been held with policy makers, The findings will be disseminated though a mixture of briefing notes, focus groups, personal meetings, presentations at policy focused conferences, and working papers and eventually peer reviewed academic publications. Imperial College London will also seek discussion with policy officials in NHS England who have oversight of cardiac treatment. (iv) In future, Imperial College London propose to make presentations to NHS England and other regulatory bodies. Data retention is required until 2022 to allow time to complete analyses and produce outputs. Publications in economics takes a considerable amount of time from submission to review to acceptance. Imperial College wish to retain the data for this entire period to ensure that results of the research are published in top journals and can effectively respond to any reviewer’s comments. There was a delay in starting this work as the initial data set from NHS Digital was not received for 6 months . All outputs will be/have been published as aggregated data with small numbers suppressed in line with the HES analysis guide. More broadly, the dissemination machinery of Imperial College Business School will be/has been used to disseminate research highlights to alumni, the general media and through this, to members of the public.

Expected measurable benefits

[1 paragraph unchanged] The cost of hospital treatment is estimated to be in the order of approximately £540 million for colorectal cancer and £500 million for breast cancer per annum. The British Heart Foundation estimated in 2020 that heart and circulatory treatment cost the NHS £7.4 billion a year (https://www.bhf.org.uk/what-we-do/news-from-the-bhf/news-archive/2020/march/government-prioritise-heart-health-budget). Hence, the NHS could achieve large benefits for patients and also cost [8 words unchanged] of cost-effective innovation and the implementation of voluntary guidelines into medical practice. Both innovations that will be examined in this study have been shown to be cost-effective. Hence there is a clear social gain from understanding what makes surgeons adhere to voluntary guidelines on treatment in these two fields. This research will quantify the impact of clinical networks on determining variation in the take-up of innovation in treatment of two common forms of cancer. The ultimate beneficiaries are those patients who receive more effective treatment sooner. Understanding the drivers of cost-effective innovation is important, both for individual patients and the healthcare system as a whole. The research will compare and contrast the impact of these formal networks with the social networks of consultants formed during their training and spatial competition. This has not previously been done: to date analysis has simply focused on the time series of uptake of innovation rather than an assessment of the benefit of networks relative to no networks. More broadly, this research project focuses on understanding clinical practice variation which is a major concern for the NHS. The Department of Health and NICE invest large amounts of resources in producing evidence for best practice in hospital care in the form of NICE Pathways. The Department of Health uses Best Practice Tariffs that link reimbursement to adherence to best practice guidelines in several clinical areas. These outputs will help policy makers to decide where to invest resources. Understanding the relative importance of formal, top down driven networks, compared to organic ones formed during or after training will allow more cost-effective decisions to be made about the best place to invest scarce resources to expedite the uptake of proven innovations. Best practice guidelines that inform the adoption of cost-effective innovations are considered essential instruments for reducing variations in clinical practice, improving quality and containing costs, provided they are implemented and adhered to by decision makers in hospitals. The results of this research project are intended as an input to help UK policy makers to target activities and funding to promote a decision- making culture in hospitals that is more in line with the Department of Health's stated objectives. Better understanding of innovation processes are anticipated to enable the development of targeted policies to speed up the diffusion of innovation with benefits to patients including the mitigation of inequalities for access to innovations. Imperial College London have strong links with NHS Improvement and NHS England and the broader UK policy community. The Health Foundation, the funders of this project, have very strong links to the policy community, and this research will form part of their dissemination of their Health Efficiency programme. Imperial College also has strong links to policy makers, for example, though Imperial College’s Institute for Global Health and through the Centre for Economics and Policy Innovation (CHEPI). It will provide guidance to policy makers as to whether it is more important to target doctors during their medical training period or later on, and whether formal networks are more or less effective than informal ones. The ultimate beneficiaries are those patients who receive more effective treatment sooner. More broadly, this research focuses on understanding clinical practice variation which is a major concern for the NHS. The Department of Health and NICE are currently investing large amounts of resources in producing evidence for best practice in hospital care in the form of NICE Pathways. Further, the Quality, Innovation, Productivity and Prevention programme has recently been established to realise an ambitious plan of quality improvements and cost savings though the implementation of evidence based recommendations to best practice. If successful, the programme was expected to deliver up to £20billion efficiency savings to be reinvested in frontline services in 2014-15. The Department of Health has recently introduced Best Practice Tariffs that link reimbursement to adherence to best practice guidelines in several clinical areas. Best practice guidelines that inform the adoption of cost effective innovations are considered essential instruments for reducing variations in clinical practice, improving quality and containing costs, provided they are implemented and adhered to by decision makers in hospitals. The results of this research are intended as an input to help UK policy makers to target activities and funding to promote a decision making culture in hospitals that is more in line with the Department of Health's stated objectives. Data from the wider HES dataset will be used to construct variables at the level of the trust, including hospitals’ patient casemix such as socioeconomic characteristics, the impact of spatial competition between nearby NHS Trusts, and the culture of the hospital in general in the adoption of new innovations. Controlling for these measures will allow for more robust results and provide policymakers with more causal understanding on the factors determining diffusion of best practice innovations. This research project has greatly benefited from the input of clinicians and bodies involved in the diffusion of the procedures of interest. Thus far, the study have been in contact with a number of leading bodies. As part of the project we have had discussions with: 1. LAPCO (the training unit for laparoscopy surgery in England) programme, and a number of other colorectal surgeons. 2. The Institute of Global Health Innovation at Imperial College London 3. Participants of the focus groups that we are our main formal dissemination strategy in the project for non-academic audiences 4. Formal interviews with 12 colorectal surgeons in the UK to validate the findings 5. Cancer Research UK The study also plan to interview other leading people in this area including the Royal College of Surgeons. In the studies discussions they will not only gather crucial information concerning innovation in the NHS but also explore ways of making our research accessible to practitioners. Cardiac treatment The intention of the extension beyond the treatment of colorectal cancer is to increase the applicability of the research to a wider group of patients and to increase its usefulness in policy making. By examining another group of health care professionals and another area of healthcare, the study will be able to test the extent to which results are generalisable to other patient groups and treatment types. This will allow the study to input into policy making to improve care for patients by showing how generalisable the findings are. The target date for this is 2020.

Benefits reported

This work has been presented at international academic conferences, but final outputs and peer-reviewed publications are not yet complete. This research project has greatly benefited from the input of clinicians and bodies involved in the diffusion of the procedures of interest. As part of the research project, the study team have had discussions with: 1. The LAPCO (the training unit for laparoscopy surgery in England) programme and a number of other colorectal surgeons. 2. The Institute of Global Health Innovation at Imperial College London. 3. Participants of the focus groups that formed part of Imperial College London’s formal dissemination strategy in the research project for non-academic audiences. 4. Formal interviews with 12 colorectal surgeons in the UK to validate the findings. 5. Cancer Research UK. Imperial College London’s findings on innovation have also been disseminated to policy makers via presentations at the Health Foundation and the Efficiency Research Programme Advisory Board that includes representatives from NHS England and NHS Improvement. During the COVID-19 pandemic the focus of the NHS and other policy makers has been on addressing issues generated by the COVID-19 pandemic and therefore, Imperial College London did not engage in further efforts to translate findings into changes in policy and practice. Going forward, Imperial College London will engage with the groups listed above to disseminate findings widely to non-academic audiences to enable the translation of results into policies and practices. Imperial College London have also been approached by Cancer Research UK who are keen to work with Imperial College London to discuss the further dissemination of results and the best ways to generate a meaningful impact. Imperial College London have explored how formal cancer networks compare to formal guidelines issued to promote the uptake of laparascopic surgery. The research findings compare and contrast the impact of these formal networks with the social networks of consultants formed during their training. These outputs may help policy makers to decide where to invest resources. Understanding the relative importance of formal, top-down driven networks compared to organic ones formed during or after training, hopes to allow cost-effective decisions to be made about the best place to invest scarce resources to expedite the uptake of proven innovations. Both innovations that have been examined in the cancer research project have been shown to be cost-effective. Hence there is a clear social gain from understanding what makes surgeons adhere to voluntary guidelines on treatment in these two fields. In this research project, Imperial College London have quantified the impact of clinical networks in determining variation in the take-up of innovation in treatment of colorectal cancer. Understanding the importance of networks formed during or after training will allow better understanding of the importance of networks in driving innovation. Understanding the importance of these organic networks (shown to be important in many contexts outside healthcare as well) in comparison to the costs of establishing formal, top-down driven networks, will allow more cost-effective decisions to be made about the best place to invest scarce resources to expedite the uptake of proven innovations.