Neighbourhoods and Dementia

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

Objective for processing

[7 paragraphs unchanged] Furthermore, the ways in which the processing of data will should be of benefit to the public – thereby demonstrating that the processing [72 words unchanged] categories of personal data. The ways in which the processing of data will should be of benefit to the public - primarily NHS staff, dementia patients [5 words unchanged] described in section 5dii: Expected Measurable Benefits to Health and/or Social Care. [7 paragraphs unchanged] There is minimal risk of potential public harm from dissemination of the project’s findings. There is a potential that the findings will should highlight a continued pattern of poorer hospital care for people with dementia, [49 words unchanged] to suppress dissemination if doing so helps to perpetuate inequalities in care. [1 paragraph unchanged] The study will should use pseudonymised HES data for England only. It covers all patients with [177 words unchanged] (e.g., a random selection) was not a viable option. Key hospital outcomes will are also to be compared to the University of Manchester and Lancaster's own national training [9 words unchanged] at an individual hospital level. The survey was administered in January 2017. [9 paragraphs unchanged] Only the three analysis years (i.e., 2010/11, 2012/13 and 2016/17) were required for the HES OP and HES A&E data where only health economics analyses will should be undertaken. In addition, civil registration mortality was used purely at NHS [5 words unchanged] whether a patient died in hospital or within 30 days after discharge. [3 paragraphs unchanged]

Processing activities

[2 paragraphs unchanged] HES admissions data has been used to derive the following patient-level outcomes: [62 words unchanged] discharge. The use of hospital services data (LoS, A&E and OP visits) will is to be costed using the currency and service codes and the NHS Reference costs for the cost analysis models. [2 paragraphs unchanged] Descriptive statistics (mean, SD, median and interquartile range) will should be produced to summarize each outcome and costs for each financial year 2010/11 and 2012/13 and 2016/17 for those with dementia versus control patients. Summary results will are to be split by key variables including level of dementia care training, elective versus emergency admission, geographical region and discharge destination (e.g., care home). An additional descriptive analysis will should examine the trend in identification of patients with dementia at admission across [8 words unchanged] analysis of outcomes has now been completed. Work on costs is on-going. Multi-level models will should be used for the main statistical analyses, which will should investigate the impact of training variables on outcomes and costs for dementia patients compared to those without dementia, controlling for confounding covariates. The LoS outcome will can be analysed using a multi-level model for continuous outcome data or a survival approach, as appropriate. The mortality and re-admissions outcomes will can be analysed by multi-level models for binary outcomes. The cost models will should include the costs of the admission and associated use of A&E and OP visits. The costs will should be analysed using the best fit distribution for skewed data (e.g., gamma [8 words unchanged] the regression parameters (coefficients or odds ratios), confidence intervals and associated P-values will should be reported. Only data aggregated at the hospital level will is to be reported; no patient-record level data will is to be produced as an output at any stage. The statistical analyses for [7 words unchanged] using the preferred multi-level models. Work on the cost models is ongoing. For the analysis of the smaller survey of staff at 20-30 hospitals, the HES-based measures will should be aggregated to the hospital-level and reported descriptively. Individual hospitals will should not be identified in any publication. Other than those already specified in the agreement The University of Manchester will can not link this data to any other patient level dataset or attempt [5 words unchanged] to calculate dates of death using the data supplied by NHS Digital.

Expected output

Outputs will include a final report to the The research programme funder ESRC, - the ESRC - does not require a project final report. Instead, an agreement has been made for the Open University Press to publish the findings from all work-packages of the Neighbourhoods and Dementia programme in book form, of which the research based on HES will form one chapter. Other outputs include papers in peer-reviewed medical journals, and presentations at appropriate health research conferences. The initial expected date of submission plan as specified in the DARS application in July 2016 was for the final report was towards the end of the funded programme, in Spring 2019. There was to be three main peer-reviewed journal publications: the first concerning the analysis of the [25 words unchanged] 2018; and the third on the findings of the hospital staff survey. This timeline However, for publications was specified in the original DARS application in July 2016, and assumed the project would receive the HES data in the latter half of 2016. However, due reasons given below UoM were unable to issues at Manchester in identifying and making ready a server for hosting the data that met all the security requirements, the project was not in a position keep to receive the 2010/11 and 2012/13 HES data until July 2017. This delay made it not feasible to undertake the data manipulation, analysis and first journal paper development within the original timeline. The project has also experienced additional delays due to frequent periods of downtime due to instabilities in the server (the TRE) and several weeks of sick leave for the project RA, who is the sole analyst. With the delay, it no longer made sense to publish the 2010/11 plus 2012/13 analysis, and the 2016/17 analysis in separate papers (since that strategy assumed the prior analysis would be submitted whist still working on the latter). Therefore, the project will combine all three financial years (FYs) together in publications. However, due to the very large volume of results, the project plans to publish these in three papers: the first on trends in hospital outcomes for people with and without dementia; the second on relationships between these trends and hospital staff training and other dementia initiatives; and a third paper providing more descriptive detail on differences in hospital admissions between the group. A fourth paper will cover the findings of the hospital staff survey. The project has also recently heard that ESRC (the funder) will not be requiring a separate final report. this publication plan. The project's previous expected deadline above timeline assumed the project would receive the HES data in the latter half of 2016 but due to issues at Manchester in identifying and making ready a server for completion hosting the data that met all the security requirements, the project was unable to receive the 2010/11 and 2012/13 HES data until July 2017. The 2016/17 data was received in the first quarter of 2018. Additional delays were then experienced due to instabilities in the great majority server (the TRE) on which the analysis depended and several weeks of sick leave for the analyses project RA, who was by 31/03/2019 - the sole analyst. When the original DSA came to an end date of in March 2019 UoM obtained a two-year extension, taking the previous agreement - and that the finalising of manuscripts for submission DSA to journals would go beyond that, with paper submission in Summer and Autumn 2019. Unfortunately, this proved not possible, as 31/3/2021. However, across 2019, 2019 there were continuing instabilities and long periods of downtime of the UoM TRE system on which the data was held occurred. system. The TRE was eventually taken out of service as the hardware had come to end of life and the research funding for it had dried up. It was therefore decided to move research projects that required a highly restricted In December 2019 the project data platform and files were moved to the UoM centrally provided DSH Data Save Haven service. The migration took place in December 2019, the DSA was amended accordingly, and the data on the TRE was destroyed (with a destruction certificate provided to NHS-Digital). Data analysis proceeded in January 2020. However, this 2020 but was slow as teething problems with the new DSH service were experienced and there was no facility to run jobs “in the background” or when not logged into the system (such that overnight runs were not possible). again disrupted In March 2020, the work was further disrupted 2020 by the Covid 19 pandemic, which resulted in all project staff moving to “working from home”. After the university closed on 17th March 2020, UoM did not receive home” and a pause until NHS-Digital gave permission from NHS-Digital for remote access to the data until on 5th May. At the same time, suitable equipment for secure remote access [13 words unchanged] staff needed time to familiarise themselves with the new ways of working. In all, The pandemic also impacted on senior project staff working across multiple projects disrupted by covid-19, resulting in additional administrative and research demands upon their time. Furthermore, it emerged that the new DSH service could not run analyses continuously for more than 24 hours due to data security requirements, but the sophisticated multi-level regression models UoM desired to apply usually took days and sometimes weeks. UoM therefore conducted provisional analyses using simpler but less than ideal “single-level” models while exploring ways around 3 months of processing time was lost. this limitation. Progress since has been slower than hoped for two main reasons. First, senior project staff are working across multiple projects that have been disrupted by covid-19 and resulted in additional administrative and research demands upon their time; this has particularly affected the health economics component of the study. Second, the new DSH service had an unexpected limitation such that analysis jobs could not run continuously for more than 24 hours, due to data security requirements. UoM's NHS-Digital datasets include many millions of rows of data and the sophisticated multi-level regression models that were wanting to be applied often take much longer than 24 hours. UoM therefore conducted provisional analyses using simpler “single-level” models, but these are not ideal and considerable time was spent exploring alternatives around this limitation. The slow progress in 2020 resulted in UoM requesting a further one-year DSA extension in March 2021, which was approved. Since then, the University Research IT unit has resolved their analysis-time limitations of the DSH and UoM have redone their analyses using the preferred multi-level regression models. The time delays led UoM to revise their planned peer-review publications. It no longer made sense to publish the 2010/11 plus 2012/13 analysis, and the 2016/17 analysis in separate papers. Instead, all three financial years (FYs) are combined in publications. Due to the large volume of results, UoM should publish three papers: the first on trends in hospital outcomes for people with and without dementia and the relationships between these and hospital staff training and other dementia initiatives; the second on differences in rates and types of elective admissions between the patient groups; the third on the cost analysis of inpatient admissions. The first publication has been completed and a paper submitted to BMC Medicine. The analysis for the second paper, on elective admissions, should be completed by March. The current situation is that the statistical However, analysis required for the planned publications is almost complete, though restricted to single-level models. However, an exception has just been approved which will allow UoM continuous 5-day processing to re-analyse using preferred multi-level models. Drafts of papers are in progress. The associated health economics cost analysis is still on-going. Progress on this paper has been particularly hit by covid-19, but also since the HE research fellow a combination of factors. The RA for this component left the University in February 2019 and has yet was not replaced. Prof Linda Davies took over the analysis herself, but was ill for periods of 2020 and reduced her working hours to 0.6 FTE from April 2021. UoM are still aiming to complete the cost analysis by the end of March 2022, though achieving that may be replaced. The plan remains that these papers will be published in open-access journals where they will be publicly and freely available. Target journals include PlosOne and BMC Public Health and target conferences will include the Health Services Research UK Meeting, the British Society of Gerontology conference, the joint Royal College of Nursing/British Geriatrics Society conference, the UK Dementia Congress conference and Kings Fund conferences. tight. The aim now is to then submit these the outstanding papers for peer-reviewed publication during summer/autumn 2021. in summer 2022. UoM are therefore requesting a 12-month extension until 31st March 2022, as it is essential to allow time for any additional analyses/corrections that reviewers may request. The University will set up a ‘Neighbourhoods and Dementia’ project website hosted at the University of Manchester and provide regular postings and invites for project engagement and interaction. This will include a blog and will allow for comments to be posted and the principal investigator will take responsibility for coordinating inputs onto the site. The research programme will take advantage of other social media outlets, such as twitter. The University will monitor access and record comments for evidence of impact. All papers should be published in open-access journals where they will be publicly and freely available. Target journals include BMC Medicine and PlosOne. and target conferences should include the Health Services Research UK Meeting, the British Society of Gerontology conference, the joint Royal College of Nursing/British Geriatrics Society conference, the UK Dementia Congress and Alzheimer Europe conferences. Draft results of the now-complete analysis of patient outcomes were presented by the programme lead in a plenary session at the (virtual) Alzheimer Europe conference in October 2020. The In addition to peer-reviewed publications and conferences, results are being disseminated through various other routes. These include the ‘Neighbourhoods and Dementia’ project website hosted at the University will disseminate the project through a variety of publication resources from high impact peer reviewed journals through to Manchester and social media outlets such as twitter; practice and professional outputs, communications, such as briefing updates distributed through INVOLVE, DeNDRoN, Age UK, the Alzheimer’s Society and Alzheimer Scotland. The University will present Initial results were presented at an international conference on ‘Dementia-Friendly Neighbourhoods’ at the work at international, national end of the research programme that bought together researchers, practitioners and local conferences, sharing the conference stage with people with dementia and carers at every opportunity. The results of the now-complete analysis of patient outcomes were presented by the programme lead in a plenary session at the (virtual) Alzheimer Europe conference in October 2020. carers. The University will engage with television and media outlets and will look to develop a series of features on ‘neighbourhoods’ work on dementia in national newspapers, e.g., the Guardian’s Society page. In Sweden, similar media outlets and impacts will be sought. The University will host an international conference on ‘Dementia-Friendly Neighbourhoods’ at the end of the research programme to bring each work programme together; people with dementia and carers will be planners, coordinators and speakers at this event. UoM note that all outputs should report only results aggregated across large numbers of patients, for example in the form of means, variances and regression coefficients. Graphs such as scatterplots may display derived values for individual hospitals, but without any hospital identifying information. Small numbers should be suppressed in line with HES analysis guidelines. Also, only the statisticians and health economists from the University of Manchester have access to the HES data. Their University of Lancaster collaborative colleagues have access to the aggregated outputs for the purpose of journal and conference abstract submissions. These outputs will not be given to a third party. In addition to the above, the research programme also includes a Dementia Use Involvement stream, which has provided a co-researcher education programme to a number of people living with dementia to enable them to participate as co-researchers in the study and to facilitate their further participation as co-developers of user-engagement outputs. An Impact on Policy conference is also planned for the end of the research programme with the Rt Hon Hazel Blears and Prof Alistair Burns. Outputs will report only results aggregated across all patients in any particular analysis, for example in the form of means, variances and regression coefficients. Graphs such as scatterplots may display derived values for individual hospitals, but without any hospital identifying information. Small numbers will be suppressed in line with HES analysis guidelines. Only the statisticians and health economists from the University of Manchester will have access to the HES data. Their University of Lancaster collaborative colleagues will have access to the aggregated outputs for the purpose of journal and conference abstract submissions. These outputs will not be given to a third party.

Expected measurable benefits

According to a NAD dementia survey carried out by the Royal College [67 words unchanged] foundation level dementia training, and a number of other local and national training initiatives have been launched. However, there is no available evidence, apart from anecdotal, on the impact such training is and other initiatives are having on hospital outcomes for patients with dementia, on the cost-effectiveness of [5 words unchanged] on what aspects of dementia care training hospital staff find most valuable. This study aims to produce that evidence, and in addition, to identify those components of hospital staff training in dementia care that are particularly influential in affecting outcomes. However, it may be that it is found the impact of the training initiatives being delivered to be negligible relative to all other internal and external factors impinging on patient outcomes. In either case, it is envisaged that the published findings will should inform policy at both the national and Hospital Trust level. Research findings will should help Trusts to ensure that the dementia training care they provide is cost-effective and best meets the needs of both staff and patients. The results will should also identify Trusts where care outcomes for patients with dementia are particularly [23 words unchanged] other dementia initiatives, and other factors, in influencing quality of care. Consideration will should be taken on how to sensitively feed the findings back to specific [19 words unchanged] dementia. At the national policy level, it is envisaged that the research will shoud feed into the development of guidance on dementia awareness training and the care of dementia patients in hospital, and into the development of recommendations around future research.

Benefits reported

Preliminary findings suggested that the increased length of stay for people with dementia appeared to be mainly a result of different demographics, e.g., gender and age, pre-existing health, and the reasons for hospital admission. Thus, concerns that longer stays may be due to differences in care whilst in hospital may be unfounded. However, a tangential discovery was that people with dementia appeared less likely have elective admissions for a range of interventions they might benefit from, such as hip replacements and cardiac surgery. The completed analysis of patient outcomes of emergency admissions has revealed that for all key measures - length of stay, emergency re-admissions, and deaths – the apparently poorer outcomes for people with dementia can mostly be accounted for by patient characteristics rather than differences in care. These headline results have been presented by the programme lead in a plenary session at the (virtual) Alzheimer Europe conference in October 2020. Further analyses have been conducted that examine the impact of staff dementia care training and other hospital initiatives on the remaining differences in patient outcomes. In essence, no evidence for any beneficial impact has been found, though UoM were unable to examine quality of daily interpersonal care as an outcome. The findings are the subject of the paper recently submitted to BMC Medicine. The now complete analysis of patient outcomes reveals that for all key measures - length of stay, emergency re-admissions, and deaths – the apparently poorer outcomes for people with dementia can mostly be accounted for by patient characteristics rather than differences in care. These headline results have been presented by the programme lead in a plenary session at the (virtual) Alzheimer Europe conference in October 2020. Further analyses have now been conducted that examine the impact of staff dementia care training or other hospital initiatives on the remaining differences in patient outcomes. In essence, no evidence for any beneficial impact has been found, though UoM were unable to examine quality of daily interpersonal care as an outcome. Together with the findings regarding elective admissions, these results are important in that they challenge existing assumptions and priorities, while revealing previously unrecognised major concerns around timely access to hospital care for people with dementia. Implications for policy and practice: The project is still in the process of conducting finalised analyses of the data due to numerous delays explained in section 5c. More yielded benefits are therefore anticipated in the future, once all the outputs have been produced. Both policy and practice should give less emphasis to attempts to reduce LoS and ERAs for PwD, and give more focus to achieving appropriate, rather than short, lengths of stay. Higher hospital mortality rates for PwD may be more effectively addressed by keeping PwD out of hospital, particularly by identifying individuals at terminal stages before admission, with a shift of emphasis to advance care planning,48 admission avoidance,49 and referral to palliative care services. This may require some relocation of resources into the community. Hospital staff could facilitate earlier discharge back to home through specific training in recognising the terminal phase of the condition and on end of life care in dementia. Implications for research: Better understanding is needed of the higher hospital mortality rates for PwD, especially the roles played by pre- and post-admission factors such as unrecognised terminal illness and post-discharge community support. Research is also needed into how best to use hospital avoidance approaches and advance care planning to minimise unnecessary hospital admissions. Comparison of hospital outcomes between PwD and PwoD require comprehensive case-mix adjustment to produce accurate results and avoid mis-interpretation. Better alternatives to LoS, ERAs and mortality as metrics for outcomes of care are needed. Further analysis needs to be repeated using more recent data. A further discovery was that people with dementia appear less likely have elective admissions for a range of interventions they might benefit from, such as hip replacements and cardiac surgery. Together with other findings, the results are important in that they challenge existing assumptions and priorities, while revealing previously unrecognised major concerns around timely access to hospital care for people with dementia. A comparison of the costs of hospital care for people with and without dementia is still ongoing but UoM anticipate that this analysis should add further value to the evidence base. In particular, this analysis should also analyse data on the costs of A&E and out-patient visits and how these compare to inpatient costs and between patient groups.

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

Objective for processing

[3 paragraphs unchanged] Furthermore, 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 (Article 9(2)(j)). Evidence has accumulated in recent years that a high percentage of dementia is not recognised upon hospital admission and most hospital staff lack the knowledge of how best to care for patients with dementia, potentially leading to poorer outcomes of care. In 2013 Health Education England (HEE) was mandated to ensure training is made available so that all NHS staff looking after patients with dementia have foundation level dementia training, and a number of other local and national training schemes have been launched. This study aims to assess the impact of these initiatives in addressing inequalities in care and outcomes, identify good practice, and via the published findings help inform policy at both the national and Hospital Trust level. Public Authority: The University of Manchester and Lancaster University are both public authorities. 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 Act 2000”. The FOI Act 2000 Part 1, section 3 (1)(a)(i) specifies that a public authority means any body which is listed in Schedule 1. Schedule 1 of the FOI Act 2000 lists “Maintained schools and further and higher education institutions” as public authorities. There is minimal risk of potential public harm from dissemination of the project’s findings. There is a potential that the findings will highlight a continued pattern of poorer hospital care for people with dementia, despite the training initiatives, thus reducing the confidence of people with dementia and their carers in hospital services. Nonetheless, such findings – should they result – can lead to a re-evaluation of the approaches being taken which can have longer-term benefits. Arguably it would be morally or ethically wrong to suppress dissemination if doing so helps to perpetuate inequalities in care. Public Task: There is a mix of training packages in dementia care being delivered in NHS general hospital settings as shown from the findings from the 2010 and 2012 NAD national surveys. Given the rapid development in this area much more understanding is needed about the impact of training on hospital outcomes for patients with dementia compared to those without dementia, the cost-benefits of training, and what types of training hospital staff find most useful. To the best of the University of Manchester and Lancaster's knowledge this is the first investigation of the effect of hospital training on hospital outcomes for patients with dementia using a large national sample of hospital admissions data. Section 8 of the Data Protection Act 2018 clarifies that “In Article 6(1) of the GDPR (lawfulness of processing), the reference in point (e) to processing of personal data that is necessary for the performance of a task carried out in the public interest or in the exercise of the controller’s official authority includes processing of personal data that is necessary for… (d) the exercise of a function of the Crown, a Minister of the Crown or a government department”. The University of Manchester has a royal charter (See: http://documents.manchester.ac.uk/display.aspx?DocID=16239) which states "The University shall further the prosecution of original research and shall be a teaching, assessment and awarding body. Its objects shall be to advance education, knowledge and wisdom by research, scholarship, learning and teaching, for the benefit of individuals and society at large." Additionally, the University of Lancaster has a royal charter (See: https://www.lancaster.ac.uk/media/lancaster-university/content-assets/documents/strategic-planning--governance/governance/council-key-documents/Charter-Statutes-Ordinances.pdf) which states " The objects of the University shall be to advance knowledge, wisdom and understanding by teaching and research and by the example and influence of its corporate life." The study will use pseudonymised HES data to construct measures of key hospital outcomes (see below) for patients with dementia, and also for matched patients without dementia, and compare these to hospital-level information on staff training from the National Audit on Dementia (NAD) administered in 2010, 2012 and 2016, plus the University of Manchester and Lancaster's own national training survey which captures information about availability of training at an individual hospital level. The survey was administered in January 2017. The University of Manchester and Lancaster will also undertake an in-depth survey of random samples of individual staff at 20-30 hospitals and will be comparing the hospital-level HES-based outcomes data to that data for these specific hospitals, to examine relationships to staff knowledge and confidence in caring for patients with dementia. Necessity: Throughout the application process, the necessity of the processing for the performance of the task has been assessed. This included but was not limited to ensuring appropriate minimisation of the data to ensure that only the minimum amount of data required are processed. During the application process it has been considered whether the information that the processing aims to determine is already available from other sources or whether the task could be performed using publicly available data or data from alternative sources than NHS Digital. Consideration has been given to whether the volume of data being requested is proportionate to the expected benefit and, through examination of the expected benefits consideration has been given to whether the task is itself necessary. The focus of the statistical and Health Economics analyses of admissions data will be on the financial years 2010/11 and 2012/13 in the first instance. The analysis is repeated including the final 2016/17 data when it became available. Furthermore, 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 - is covered by Article 9(2)(j) of the GDPR (processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes). 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 - primarily NHS staff, dementia patients and their families - are described in section 5dii: Expected Measurable Benefits to Health and/or Social Care. The Health Economics analyses will in addition use A&E and OP data for these financial years. In accordance with GDPR Article 89(1) processing is subject to appropriate safeguards. These include: HES APC from 2005-6 (and the following 4 years) is required as part of the process for identifying patients with dementia in 2010. Dementia is not routinely coded when a person enters hospital, therefore the usual process for determining dementia is to look for an ICD10 code for dementia in one of the diagnosis fields for any spell in the previous 5 years: this method has been previously used by the Care Quality Commission and the Centre for Health Economics at York. For this reason, the University of Manchester requested HES APC data starting from 2005 to correspond to the first analysis year 2010/11. This includes spells in mental health hospitals as well as acute hospitals. However, the only data the University of Manchester are asking for over this period is the diagnosis code data and fields to identify spell dates. i. The data will be pseudonymised prior to dissemination by NHS Digital to the data recipient The statistical and health economics analyses of HES APC data will only be on the financial years 2010/11, 2012/13 and 2016/17 to correspond with the hospital-level information on staff training from the National Audit on Dementia surveys and also the University's own national hospital-level survey of dementia administered in January 2017. The statistical analysis will be based on comparing outcomes for patients with and without dementia in acute hospitals during these three years. For the statistical methods a matched cohort study of HES APC patients in acute hospitals with dementia will be matched to controls using several variables such as age and sex. The other variables requested will be used to construct outcomes such as length of stay or used as covariates in the statistical models where it is important to control for confounding. ii. The data recipient’s technical and organisational measures to safeguard the data have been assessed and meet NHS Digital’s acceptance criteria (see sections 2 and 5b of this application for further details); Descriptive analysis will examine the trend in identification of patients with dementia at admission (under the HES APC dataset) across the period 2010/11 to 2016/17. iii. The requested data has been assessed as proportionate to the aim pursued (see section 5a of this application for further details); Only the three analysis years (i.e., 2010/11, 2012/13 and 2016/17) are required for the HES OP and HES A&E data where only health economics analyses will be undertaken. In addition, civil registration mortality will be used purely at NHS Digital to derive indicators on whether a patient died in hospital or within 30 days after discharge. iv. Controls, data retention and processing activities have been assessed to ensure respect to the essence of the right to data protection (see sections 5a, 5b and 8a of this application for further details); The University of Manchester and University of Lancaster are joint data controllers because they jointly determine the purposes and means of the processing the data. This includes decisions on what and how variables are constructed from the HES, such as a diagnosis of dementia and other medical conditions. In addition the project brings together hospital survey data on staff training collected by UoL with HES patient outcome measures generated by UoM, where joint decisions are made on how these are analysed in combination. v. Measures to protect the rights and freedoms of data subjects have been assessed including transparency (fair processing) publishing subject’s rights to withdraw consent and/or have their data erased or rectified, etc. (further information on fair processing is provided within this Abstract below and in section 4 of this application). Evidence has accumulated in recent years that a high percentage of dementia is not recognised upon hospital admission and most hospital staff lack the knowledge of how best to care for patients with dementia, potentially leading to poorer outcomes of care. In 2013 Health Education England (HEE) was mandated to ensure training is made available so that all NHS staff looking after patients with dementia have foundation level dementia training, and a number of other local and national training schemes have been launched. This study aims to assess the impact of these initiatives in addressing inequalities in care and outcomes, identify good practice, and via the published findings, help inform policy at both the national and Hospital Trust level. There is minimal risk of potential public harm from dissemination of the project’s findings. There is a potential that the findings will highlight a continued pattern of poorer hospital care for people with dementia, despite the training initiatives, thus reducing the confidence of people with dementia and their carers in hospital services. Nonetheless, such findings – should they result – can lead to a re-evaluation of the approaches being taken which can have longer-term benefits. Arguably, it would be morally or ethically wrong to suppress dissemination if doing so helps to perpetuate inequalities in care. There is a mix of training packages in dementia care being delivered in NHS general hospital settings as shown from the findings from the 2010 and 2012 NAD national surveys. Given the rapid development in this area, much more understanding is needed about the impact of training on hospital outcomes for patients with dementia compared to those without dementia, the cost-benefits of training, and what types of training hospital staff find most useful. To the best of the University of Manchester and Lancaster's knowledge this is the first investigation of the effect of hospital training on hospital outcomes for patients with dementia using a large national sample of hospital admissions data. The study will use pseudonymised HES data for England only. It covers all patients with a diagnosis of dementia within the financial years under study (2010/11 n= 142,129; 2012/13 n=170,988; and 2016/17 n=227,131). The numbers of patients without dementia included in the analysis are 985,731 for 2010/11; 1,030152 for 2012/13; and 1,538,205 for 2016/17. A primary objective of the analysis is to relate outcomes of hospital stays to staff training in care for patients with dementia. The latter is measured using data from the National Audits of Dementia (NADS), conducted in 2010, 2012 and 2016. These audits were national surveys including all general hospitals in England. For the analysis, HES data for the same set of hospitals in the same years is therefore required. All three years are needed in order to examine the longitudinal trends over the period 2010 to 2017, during which dementia training for staff was introduced. The NADS training scores are only at the hospital level, hence the statistical power of the analysis is dictated by the number of hospitals, which is around 180. This size of sample provides only limited power and hence analysing a smaller sample (e.g., a random selection) was not a viable option. Key hospital outcomes will also be compared to the University of Manchester and Lancaster's own national training survey survey which captures information about availability of training at an individual hospital level. The survey was administered in January 2017. The University of Manchester and Lancaster has also undertaken an in-depth survey of random samples of individual staff at 20-30 hospitals and compared the hospital-level HES-based outcomes data to that data for these specific hospitals, to examine relationships to staff knowledge and confidence in caring for patients with dementia. To minimise the data request, alongside restricting the years requested, only data for patients age 60 or over that attended NHS trust hospitals was requested for analysis. For 2010/11, 2012/13 and 2016/17 all fields were provided (FieldList), but for all other years, only a small subset of fields was provided to further minimise the data request. The analysis requires data at the individual patient level, to examine patient factors (eg gender, age, comorbidities) related to hospital outcomes such as length of stay and emergency re-admission. Factors need to be controlled for one-another in the analysis, hence tabulated data would not suffice for the objectives of the analysis. The focus of the statistical and Health Economics analyses of admissions data is on the financial years 2010/11, 2012/13 and 2016/17 data. The Health Economics analyses in addition uses A&E and OP data for these financial years. HES APC from 2005-6 (and the following 4 years) was required as part of the process for identifying patients with dementia in 2010. Dementia is not routinely coded when a person enters hospital, therefore the usual process for determining dementia is to look for an ICD10 code for dementia in one of the diagnosis fields for any spell in the previous 5 years: this method has been previously used by the Care Quality Commission and the Centre for Health Economics at York. For this reason, the University of Manchester requested HES APC data starting from 2005 to correspond to the first analysis year 2010/11. This includes spells in mental health hospitals as well as acute hospitals. However, the only data the University of Manchester asked for over this period is the diagnosis code data and fields to identify spell dates. The statistical and health economics analyses of HES APC data covers the financial years 2010/11, 2012/13 and 2016/17 to correspond with the hospital-level information on staff training from the National Audit on Dementia surveys and also the University's own national hospital-level survey of dementia administered in January 2017. The statistical analysis is based on comparing outcomes for patients with and without dementia in acute hospitals during these three years. For the statistical methods, a matched cohort study of HES APC patients in acute hospitals with dementia is matched to controls using several variables such as age and sex. The other variables requested are used to construct outcomes such as length of stay or used as covariates in the statistical models where it is important to control for confounding. The statistical analyses for this are largely complete. However, due to IT issues, the university's preferred multi-level regression models have been unable to run, up until now. There is now the facility (i.e. longer continuous processing time) to apply those models, and have made a start on fitting these models. The health economics analysis has been more badly affected by staff changes/recruitment issues, staff illness and covid. A protocol for the HE analysis has been produced and work has been progressing, albeit slowly. Descriptive analysis have been conducted to examine the trend in identification of patients with dementia at admission (under the HES APC dataset) across the period 2010/11 to 2016/17. This analysis has been completed. Only the three analysis years (i.e., 2010/11, 2012/13 and 2016/17) were required for the HES OP and HES A&E data where only health economics analyses will be undertaken. In addition, civil registration mortality was used purely at NHS Digital to derive indicators on whether a patient died in hospital or within 30 days after discharge. The University of Manchester and University of Lancaster are joint data controllers because they jointly determine the purposes and means of the processing the data. This includes decisions on what and how variables are constructed from the HES, such as a diagnosis of dementia and other medical conditions. In addition, the project brings together hospital survey data on staff training collected by UoL with HES patient outcome measures generated by UoM, where joint decisions are made on how these are analysed in combination. [2 paragraphs unchanged]

Processing activities

[2 paragraphs unchanged] HES admissions data will be has been used to derive the following patient-level outcomes: length of stay (LoS); (LoS: in total days and stays>=15 days); emergency re-admission within 30 days after discharge and measures of care whilst [7 words unchanged] potentially avoidable conditions (eg UTIs and bed sores). In addition, NHS Digital provide provided an outcome variable (completed for the last episode) indicating whether a patient [27 words unchanged] service codes and the NHS Reference costs for the cost analysis models. A number of essential covariates will also be have been defined, including age, gender, additional diagnoses, and place of residence (e.g., care home). These variables will then be linked have been at the hospital level to the NAD training data and the University of Manchester and Lancaster's own national training survey in order that the analyses can could be undertaken. A matched cohort study will be has been used to compare outcomes and costs for patients with a known diagnosis of dementia to matched patients without a known diagnosis, with a focus on how hospital training [45 words unchanged] Hospital in England during the previous 5 years prior to their admission will be is considered as having to have dementia. Control The control patients will be are those with no recorded dementia diagnosis in the last five years. A [5 words unchanged] on age, sex and other covariates to the group of dementia patients will be selected for the main analysis. has also been analysed. Diagnosis classification will be was carried forward to subsequent A&E, admissions and OP visits. Data for the same patient across time-points and products will be was linked by using a common encrypted HESID. The statistical analysis of the relationships between staff training scores and hospital outcomes has been completed (using single-level models). The HE analysis is ongoing. Descriptive statistics (mean, SD, median and interquartile range) will be produced to summarize each outcome and cost costs for each financial year 2010/11 and 2012/13 and 2016/17 for those with dementia versus matched control patients. Summary results will also be split by key variables including level of dementia care training, elective [21 words unchanged] of patients with dementia at admission across the period 2010/11 to 2016/17. The descriptive statistical analysis of outcomes has now been completed. Work on costs is on-going. Multi-level models will be used for the main statistical analyses, which will [75 words unchanged] costs will be analysed using the best fit distribution for skewed data (eg (e.g., gamma or poisson) For all statistical and cost models the regression parameters [23 words unchanged] patient-record level data will be produced as an output at any stage. The statistical analyses for this have been completed, but not yet using the preferred multi-level models. Work on the cost models is ongoing. For the analysis of the smaller survey of staff at 20-30 hospitals, the HES-based measures will be aggregated to the hospital-level, hospital-level and reported descriptively. Individual hospitals will not be identified in any publication. [1 paragraph unchanged]

Expected output

Outputs will include a final report to the research programme funder ESRC, papers in peer-reviewed medical journals, and presentations at appropriate health research conferences. The initial expected date of submission for the final report will be submitted was towards the end of the funded programme, in Spring 2019. There will was to be three main peer-reviewed journal publications: the first concerning the analysis of [26 words unchanged] 2018; and the third on the findings of the hospital staff survey. This timeline for publications was specified in the original DARS application in July 2016, and assumed the project would receive the HES data in the latter half of 2016. However, due to issues at Manchester in identifying and making ready a server for hosting the data that met all the security requirements, the project was not in a position to receive the 2010/11 and 2012/13 HES data until July 2017. This delay made it not feasible to undertake the data manipulation, analysis and first journal paper development within the original timeline. The project has also experienced additional delays due to frequent periods of downtime due to instabilities in the server (the TRE) and several weeks of sick leave for the project RA, who is the sole analyst. With the delay, it no longer made sense to publish the 2010/11 plus 2012/13 analysis, and the 2016/17 analysis in separate papers (since that strategy assumed the prior analysis would be submitted whist still working on the latter). Therefore, the project will combine all three financial years (FYs) together in publications. However, due to the very large volume of results, the project plans to publish these in three papers: the first on trends in hospital outcomes for people with and without dementia; the second on relationships between these trends and hospital staff training and other dementia initiatives; and a third paper providing more descriptive detail on differences in hospital admissions between the group. A fourth paper will cover the findings of the hospital staff survey. The project has also recently heard that ESRC (the funder) will not be requiring a separate final report. Update December 2018: The above timeline for publications was specified in the original DARS application in July 2016, and assumed the project would receive the HES data in the latter half of 2016. However, due to issues at Manchester in identifying and making ready a server for hosting the data that met all the security requirements, the project was not in a position to receive the 2010/11 and 2012/13 HES data until July 2017. This delay made it not feasible to undertake the data manipulation, analysis and first journal paper development within the original time-line. The project has also experienced additional delays due to frequent periods of downtime due to instabilities in the server (the TRE) and several weeks of sick leave for the project RA, who is the sole analyst. With the delay, it no longer makes sense to publish the 2010/11 plus 2012/13 analysis, and the 2016/17 analysis, in separate papers (since that strategy assumed the prior analysis would be submitted whist still working on the latter). Therefore, the project will combine all three financial years (FYs) together in publications. However, due to the very large volume of results, the project plans to publish these in three papers: the first on trends in hospital outcomes for people with and without dementia; the second on relationships between these trends and hospital staff training and other dementia initiatives; and a third paper providing more descriptive detail on differences in hospital admissions between the group. A fourth paper will cover the findings of the hospital staff survey. The project has also recently heard the ESRC (the funder) will not be requiring a separate final report. The project's previous expected deadline for completion of the great majority of the analyses was by 31/03/2019 - the end date of the previous agreement - and that the finalising of manuscripts for submission to journals would go beyond that, with paper submission in Summer and Autumn 2019. Unfortunately, this proved not possible, as across 2019, continuing instabilities and long periods of downtime of the UoM TRE system on which the data was held occurred. The TRE was eventually taken out of service as the hardware had come to end of life and the research funding for it had dried up. It was therefore decided to move research projects that required a highly restricted data platform to the UoM centrally provided DSH service. The migration took place in December 2019, the DSA was amended accordingly, and the data on the TRE was destroyed (with a destruction certificate provided to NHS-Digital). The project now plans to have completed the great majority of the analyses by the current agreement end date of 31/03/2019, However, it is anticipated that the finalising of manuscripts for submission to journals will go beyond that, with paper submission in Summer and Autumn 2019. These papers will be published in open-access journals where they will be publicly and freely available. Target journals include PlosOne and BMC Public Health and target conferences will include the Health Services Research UK Meeting, the British Society of Gerontology conference, the joint Royal College of Nursing/British Geriatrics Society conference, the UK Dementia Congress conference and Kings Fund conferences. Data analysis proceeded in January 2020. However, this was slow as teething problems with the new DSH service were experienced and there was no facility to run jobs “in the background” or when not logged into the system (such that overnight runs were not possible). In March 2020, the work was further disrupted by the Covid 19 pandemic, which resulted in all project staff moving to “working from home”. After the university closed on 17th March 2020, UoM did not receive permission from NHS-Digital for remote access to the data until 5th May. At the same time, suitable equipment for secure remote access to the DSH from individual residences needed to be sourced and set-up, and staff needed time to familiarise themselves with the new ways of working. In all, around 3 months of processing time was lost. Progress since has been slower than hoped for two main reasons. First, senior project staff are working across multiple projects that have been disrupted by covid-19 and resulted in additional administrative and research demands upon their time; this has particularly affected the health economics component of the study. Second, the new DSH service had an unexpected limitation such that analysis jobs could not run continuously for more than 24 hours, due to data security requirements. UoM's NHS-Digital datasets include many millions of rows of data and the sophisticated multi-level regression models that were wanting to be applied often take much longer than 24 hours. UoM therefore conducted provisional analyses using simpler “single-level” models, but these are not ideal and considerable time was spent exploring alternatives around this limitation. The current situation is that the statistical analysis required for the planned publications is almost complete, though restricted to single-level models. However, an exception has just been approved which will allow UoM continuous 5-day processing to re-analyse using preferred multi-level models. Drafts of papers are in progress. The associated health economics analysis is still on-going. Progress on this has been particularly hit by covid-19, but also since the HE research fellow left the University in February 2019 and has yet to be replaced. The plan remains that these papers will be published in open-access journals where they will be publicly and freely available. Target journals include PlosOne and BMC Public Health and target conferences will include the Health Services Research UK Meeting, the British Society of Gerontology conference, the joint Royal College of Nursing/British Geriatrics Society conference, the UK Dementia Congress conference and Kings Fund conferences. The aim now is to submit these papers for peer-reviewed publication during summer/autumn 2021. UoM are therefore requesting a 12-month extension until 31st March 2022, as it is essential to allow time for any additional analyses/corrections that reviewers may request. [1 paragraph unchanged] The University will disseminate the project through a variety of publication resources [41 words unchanged] the conference stage with people with dementia and carers at every opportunity. The results of the now-complete analysis of patient outcomes were presented by the programme lead in a plenary session at the (virtual) Alzheimer Europe conference in October 2020. The University will engage with television and media outlets and will look to develop a series of features on ‘neighbourhoods’ work on dementia in national newspapers, e.g. e.g., the Guardian’s Society page. In Sweden, similar media outlets and impacts will [27 words unchanged] dementia and carers will be planners, coordinators and speakers at this event. [3 paragraphs unchanged]

Expected measurable benefits

According to a NAD dementia survey carried out by the Royal College of Psychiatrists Psychiatrists, a quarter of acute hospital beds are occupied by people with dementia [17 words unchanged] the knowledge of how best to care for patients with dementia. In 2013 2013, Health Education England (HEE) was mandated to ensure training is made available [60 words unchanged] on what aspects of dementia care training hospital staff find most valuable. This study aims to produce that evidence, and in addition addition, to identify those components of hospital staff training in dementia care that [45 words unchanged] findings will inform policy at both the national and Hospital Trust level. Research findings will help Trusts to ensure that the dementia training they [92 words unchanged] to patients with dementia. At the national policy level, it is envisaged that the research will feed into the development of guidance on dementia awareness [6 words unchanged] patients in hospital, and into the development of recommendations around future research.

Benefits reported

The project is in process of conducting finalised analyses of the data. A preliminary finding is Preliminary findings suggested that the increased length of stay for people with dementia appears appeared to be mainly a result of different demographics, e.g., gender and age, pre-existing health, and the reasons for hospital admission. Thus Thus, concerns that longer stays may be due to differences in care whilst in hospital may be unfounded. However, a tangential discovery is was that people with dementia appear appeared less likely have elective admissions for a range of interventions they might benefit from, such as hip replacements and cardiac surgery. The project is yet to add their measures of staff dementia training into the analysis models. The now complete analysis of patient outcomes reveals that for all key measures - length of stay, emergency re-admissions, and deaths – the apparently poorer outcomes for people with dementia can mostly be accounted for by patient characteristics rather than differences in care. These headline results have been presented by the programme lead in a plenary session at the (virtual) Alzheimer Europe conference in October 2020. Further analyses have now been conducted that examine the impact of staff dementia care training or other hospital initiatives on the remaining differences in patient outcomes. In essence, no evidence for any beneficial impact has been found, though UoM were unable to examine quality of daily interpersonal care as an outcome. Together with the findings regarding elective admissions, these results are important in that they challenge existing assumptions and priorities, while revealing previously unrecognised major concerns around timely access to hospital care for people with dementia. The project is still in the process of conducting finalised analyses of the data due to numerous delays explained in section 5c. More yielded benefits are therefore anticipated in the future, once all the outputs have been produced.

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

Unchanged: Objective for processing, Processing activities, Expected output, Expected measurable benefits, Benefits reported.

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

Data controllers: + LANCASTER UNIVERSITY

Objective for processing

This DSA is an extension only, no new data will flow and this just allows the continued analysis of the data previously received. The below text is for contextual information only as to the previous flow of data. [1 paragraph unchanged] The neighbourhood and dementia study works on eight work programmes and has involved the University of Manchester, Lancaster University, the University of Stirling, the University of Liverpool, University College London, and Linköping University in Sweden. The data under this agreement is strictly in relation to Work Programme 5 - Developing the evidence base for evaluating dementia training in NHS hospitals (DEMTRAIN). This work programme is only worked on by the University of Manchester and Lancaster University. No other organisation holds any involvement with this work programme and do not determine how data for this work programme is processed. The lawful basis for processing the data comes from Article 6(1)(e) concerning the performance of a task in the public interest or in the exercise of official authority vested in the controller, which includes processing of personal data that is necessary for the exercise of a function of the Crown, a Minister of the Crown or a government department. Furthermore, 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 (Article 9(2)(j)). Evidence has accumulated in recent years that a high percentage of dementia is not recognised upon hospital admission and most hospital staff lack the knowledge of how best to care for patients with dementia, potentially leading to poorer outcomes of care. In 2013 Health Education England (HEE) was mandated to ensure training is made available so that all NHS staff looking after patients with dementia have foundation level dementia training, and a number of other local and national training schemes have been launched. This study aims to assess the impact of these initiatives in addressing inequalities in care and outcomes, identify good practice, and via the published findings help inform policy at both the national and Hospital Trust level. There is minimal risk of potential public harm from dissemination of the project’s findings. There is a potential that the findings will highlight a continued pattern of poorer hospital care for people with dementia, despite the training initiatives, thus reducing the confidence of people with dementia and their carers in hospital services. Nonetheless, such findings – should they result – can lead to a re-evaluation of the approaches being taken which can have longer-term benefits. Arguably it would be morally or ethically wrong to suppress dissemination if doing so helps to perpetuate inequalities in care. [1 paragraph unchanged] The study will use pseudonymised HES data to construct measures of key hospital outcomes (see below) for [101 words unchanged] relationships to staff knowledge and confidence in caring for patients with dementia. The focus of the statistical and Health Economics (HE) analyses of admissions data will be on the financial years 2010/11 and 2012/13 in the first instance. The analysis will be is repeated including the final 2016/17 data when it becomes became available. The HE Health Economics analyses will in addition use A&E and OP data for these financial years. HES APC from 2005-6 (and the following 4 years) is requested required as part of the process for identifying patients with dementia in 2010. [49 words unchanged] Quality Commission and the Centre for Health Economics at York. For this reason reason, the University of Manchester have requested HES APC data starting from 2005 to correspond to the first analysis year [5 words unchanged] mental health hospitals as well as acute hospitals. However, the only data they the University of Manchester are asking for over this period is the diagnosis code data and fields to identify spell dates. The statistical and health economics analyses of HES APC data will only [58 words unchanged] these three years. For the statistical methods a matched cohort study of HES APC patients in acute hospitals with dementia will be matched to controls [27 words unchanged] in the statistical models where it is important to control for confounding. Only the three analysis years (i.e., 2010/11, 2012/13 and 2016/17) are required for the HES OP and HES A&E data where only health economics analyses will be undertaken. Descriptive analysis will examine the trend in identification of patients with dementia at admission (under the HES APC dataset) across the period 2010/11 to 2016/17. All the data handling and analysis will be conducted in the UK at the University of Manchester. None of the HES data provided by NHS Digital will be made available to third parties including Lancaster University. Only the three analysis years (i.e., 2010/11, 2012/13 and 2016/17) are required for the HES OP and HES A&E data where only health economics analyses will be undertaken. In addition, civil registration mortality will be used purely at NHS Digital to derive indicators on whether a patient died in hospital or within 30 days after discharge. The University of Manchester and University of Lancaster are joint data controllers because they jointly determine the purposes and means of the processing the data. This includes decisions on what and how variables are constructed from the HES, such as a diagnosis of dementia and other medical conditions. In addition the project brings together hospital survey data on staff training collected by UoL with HES patient outcome measures generated by UoM, where joint decisions are made on how these are analysed in combination. University of Manchester is the sole organisation processing the data. All the data handling and analysis is conducted in the UK at the University of Manchester and no project staff based at the University of Lancaster (or elsewhere) have any access to the HES or other data stored on the server at the University of Manchester. The only data shared with University of Lancaster is in the form of summaries (e.g. counts, means, percentages) at the hospital or national level. No data is shared with third parties, but summary statistics at the national level only, such as mean lengths of hospital stay for people with and without dementia, are provided to the project expert advisory panel for discussion. The project is funded by the ESRC, who approved the project protocol. However, ESRC have no direct input into the purpose or means of data processing. The project has convened an expert advisory panel consisting of practitioners and researchers in dementia and people living with dementia themselves. This panel provides purely advisory input to the project team on matters such as the clinical care of people with dementia and likely complications, the organisation of staff training, the patient experience, hospital procedures for coding patient data, and interpretation of findings.

Processing activities

The data will only be accessed by substantive employees of the University of Manchester and only for the purposes described in this document. Data Sharing Agreement. There is no flow of data into NHS Digital and thus there is no flow of identifying personal data at any point. Under a previous iteration of this agreement, pseudonymised personal health data was disseminated to the University of Manchester. This was restricted to the datasets and years specified in this agreement. [7 paragraphs unchanged]

Benefits reported

The project is in process of conducting our finalised analyses of the data. A preliminary finding is that the increased [81 words unchanged] to add their measures of staff dementia training into the analysis models.

Unchanged: Expected output, Expected measurable benefits.