Modelling tool to support decision-making in the NHS Health Check programme: workshops, systematic review and co-production with users

Stakeholder mapping and recruitment

We developed a stakeholder recruitment grid based on our extensive public health networks at the local, regional and national level. The workHORSE project team identified relevant organisations and individuals from these organisations were added to the recruitment grid. The final recruitment grid contained a diverse group of stakeholders from different organisations, including PHE (national and regional levels), the British Heart Foundation (London, UK), Diabetes UK (London, UK), Alzheimer’s Research UK (Cambridge, UK), the National Institute for Health and Care Excellence (NICE) (London, UK), the British Medical Association (London, UK), Alcohol Change UK (London, UK), the North West Coast Strategic Clinical Networks (Warrington, UK), directors of public health, the Local Government Association (London, UK), Clinical Commissioning Groups (CCGs), LAs, GPs, pharmacies and academics. Inviting a cross-section of stakeholders representing local, regional and national perspectives provided a broad skill set and diverse perspectives for the process of co-producing the tool.

Stakeholders were sent an e-mail invitation to attend the workshops. If specific stakeholders were unable to attend, we used snowballing techniques to identify other individuals at their organisation to invite. Depending on the objectives of the specific workshop session, we would either group stakeholders from similar organisations or mix stakeholders from a broad range of perspectives (e.g. NHS HCP, local, regional and national decision-makers).

Workshop design

We delivered four workshops across the duration of the 2-year project. Workshops were delivered in months 4, 8, 16 and 24. We developed a systematic and pragmatic approach to the planning, development and delivery of all four workshops.

The organisation for each workshop commenced at least 3 months before delivery. Development and planning were undertaken in face-to-face research team meetings. The initial planning meeting would include reporting of stakeholder recruitment (i.e. number and organisation) and discussion about the duration, objectives, outcomes, format and activities for the workshop. Based on the initial discussion, the workshop co-ordinators (FLW and LH) would develop a draft workshop programme for further development and refinement at subsequent meetings. The final workshop plan included the purpose, aims, time required and allocated to each activity, materials required, roles and tasks for each team member during the workshop according to their skills and expertise, and outputs of the workshop. A week before the actual workshop, a full rehearsal took place to ensure that the workshop would be delivered efficiently and effectively to maximise the co-production process.

The design of the workshop programme was theory based, using the Cairney–Oliver key co-production principles. 35,36 These included co-identifying the requirements of the decision-support tool based on stakeholders’ current views and experience and future requirements, working iteratively over the lifespan of the project to co-steer the decision-support tool content and outputs, and co-developing interpretations of the decision-support tool and implications for dissemination and end use.

The workHORSE project workshops had the overall aim of co-producing the web-based decision-support tool with stakeholders and included a series of small-group exercises with specific objectives and outputs. Exercises were designed in the form of scripts. 37 We adapted previously validated scripts to our specific needs and context (Scriptapedia38), based on the work of Hovmand et al. ,39 as part of a general framework. This allowed the modelling team to engage with stakeholders in the co-design of qualitative and quantitative models. Each script contained a succession of elements, including descriptions of the exercise, purpose, time, materials needed, inputs, outputs, team roles required, steps and evaluation criteria. The scripts included a series of small-group exercises with specific objectives, questions, activities and outputs.

The workshops were iterative in their approach and involved an independent facilitator in their delivery. Immediate feedback was obtained using Post-it^® Notes (3M, Cynthiana, KY, USA), flip charts and small-group and plenary discussions.

Group model building

The use of scripts enabled a better design of the workshops and more useful sessions, leading to a more comprehensive and user-friendly workHORSE modelling tool and ‘buy-in’ from stakeholders. Furthermore, the activities enabled the team to engage with stakeholders in the co-design of the decision-support tool, facilitating open discussion and opportunities for stakeholders to provide additional feedback afterwards. An example script for workshop 2 is provided as additional material [see workHORSE workshop programme and script examples via the NIHR Journals Library project web page URL: www.journalslibrary.nihr.ac.uk/programmes/hta/1616501/#/ (accessed 15 March 2021)].

A summary of the aims and activities for each workshop are shown in Table 1.

Ethics approval

Ethics approval for the workshops was granted by the Health and Life Sciences Committee on Research Ethics (Psychology, Health and Society), University of Liverpool, Liverpool, UK, on 14 September 2017 (reference number 2242). Written consent was obtained from stakeholders before the workshop. All data were anonymised and stored in locked filing cabinets and on password-protected computers.

Evaluation

Both stakeholders and the modelling team completed questionnaires with open-ended and closed questions to evaluate the co-production process. At the end of each workshop, stakeholders completed stakeholder engagement questionnaires to explore their views and experiences throughout the process. Questions included their reasons for attending the workshops, their expectations and what they had gained from attending, and the perceived added value of their involvement.

The modelling team also completed questionnaires to explore their expectations before and their experiences after the workshops. Questions included the added value of having a series of workshops, the process of co-production and how the decision-support tool benefited from stakeholder involvement.

At the end of the project, the modelling team and stakeholders were e-mailed a final questionnaire that was tailored appropriately to identify their overall experience of the co-production process.

Thematic analysis

The qualitative information obtained from the questionnaires and notes from meetings with the lay advisers were analysed using the principles of thematic analysis, as described by Braun and Clarke. 40 Familiarisation of the data was carried out: reading through all of the data and generating initial codes based on the responses to the open questions. These data were then grouped into meaningful categories and further searched and reviewed for themes. The responses were then categorised into a sufficiently small set of broad categories, which were then coded and subsequently indexed.

Workshop 1

Workshop 1 took place in Liverpool in February 2018. Fifteen stakeholders participated, representing the local, regional and national perspective, with attendees from LAs, CCGs, general practices, academia, PHE and third-sector organisations.

We delivered two key activities. Activity 1 focused on developing a shared understanding of the NHS HCP and asked stakeholders to identify aspects of the programme that were working well or not that well, and their future hopes for the programme. During activity 2, stakeholders were asked to identify the key features that the workHORSE modelling tool should include that would make the tool useful for the decision-making process.

Each activity was completed individually, followed by both table and whole-group discussions. Each table had a mix of local, regional and national stakeholders to stimulate discussion. Stakeholders provided written feedback on Post-it Notes and the table and group discussions were audio-recorded and summarised on flip chart paper.

Workshop 2

Workshop 2 took place in Liverpool in June 2018. Seventeen stakeholders participated, representing local, regional and national perspectives, with attendees from LAs, CCGs, general practices, academia, PHE, NICE and third-sector organisations.

The workshop activities commenced with a presentation reviewing the workshop 1 findings of ‘what will make workHORSE a useful tool’, based on the MoSCoW approach (i.e. what we must/should/could/would do). Stakeholders were reminded of their earlier comments and given a summary of the findings. This enabled stakeholders to understand how their proposals, so far, linked to the seven parameters of the model and what the model could provide, based on feedback, using the MoSCoW prioritisation approach.

We then delivered two activities. Activity 1 focused on enabling stakeholders to consider alternative NHS Health Check implementations and then practicing modelling these implementations on a laptop computer provided for each mixed-specialty group. Individual groups were invited to feedback on how they had modelled their NHS HCP scenario, followed by a plenary discussion.

In activity 2, stakeholders were asked to rank the importance of different model outputs and visualisations that would make workHORSE a useful tool. Again, feedback was provided by individual groups, followed by a plenary discussion.

Workshop 3

Workshop 3 took place in Liverpool in February 2019. Ten stakeholders participated, representing the local, regional and national perspective, with attendees from LAs, CCGs, academia, PHE and third-sector organisations.

We delivered two key activities. Activity 1 focused on helping stakeholders understand how to create and interpret realistic scenarios. In activity 2, we wanted to obtain feedback on the alpha version of the model, focusing on output specifications. Stakeholders were asked to explore the importance of model outputs and visualisations that would make workHORSE a useful tool. Each activity was completed in small groups, followed by both table and plenary discussions. Each table had a mix of local, regional and national stakeholders to stimulate discussion.

Workshop 4

Workshop 4 took place in in Liverpool in October 2019. Eleven stakeholders participated, representing local, regional and national perspectives, with attendees from LAs, CCGs, academia, PHE and third-sector organisations.

Workshop 4 was the concluding workshop and an opportunity to showcase the final model and provide stakeholders with information about the next steps. The aim of workshop 4 was to influence the adoption of the model, engage with influencers, exploit the model as an academic product and demonstrate its added value. Again, the programme was interactive and included (1) lay advisers talking about their experience and involvement in workHORSE and providing tips for lay involvement in future projects, (2) showcasing the model and talking through the changes made, including usage, understanding its application and maximising the use of tutorials, and (3) stakeholders (one from a local perspective and one from a national perspective) demonstrating the model’s capabilities, in terms of model usage and interpretation of results. Stakeholders were also informed about the implementation plan for the model and the next steps required to confirm support for the dissemination of the workHORSE model.

Perceived benefits of participating in the workHORSE project

Respondents commented on what they had gained from attending the stakeholder workshops. The overarching theme was being able to meet, communicate and co-produce work with other stakeholders and the project team:

The interactive group exercises provided a platform for different ideas to be discussed among the stakeholders, and this is also more efficient in identifying key questions and in formulating the most helpful suggestions or recommendations.

SH2

Some of the comments on language and ease of usability – it was really useful to hear from others who are addressing these issues in the real world daily. Consultation is critical if the tool is to be fit for purpose and used, rather than another technical programme which does not get traction in the real world and its use is not maximised.

SH3

Ideas from other participants for offering and delivering Health Checks.

SH4

Opportunity to shape and develop a new resource and tool. Gave me a greater understanding of barriers to SROI [social return on investment] and modelling tools. Opportunity to network with others.

SH6

. . . being informed throughout the process and improving my understanding of the limitations of such a project and being able to ask questions in real time.

SH8

Benefits of being consulted were gaining insight into how and why the tool was being developed and having the opportunity to contribute towards shaping it, and having this knowledge and experience means I will be more likely to use it.

SH9

workHORSE e-lab online platform

Four of the respondents had used the workHORSE e-lab online platform, with mixed views about its value for the project. All agreed that it had enhanced their understanding of the workHORSE tool and three felt that it had enabled them to contribute to tool development. However, it was not perceived as the best platform for networking and discussing with other stakeholders or the project team about the workHORSE tool or the NHS HCP.

Seven respondents had not used the workHORSE e-lab. The reasons given were lack of time (n = 6) and not being relevant to their organisation (n = 2).

Stakeholder engagement for future research projects

Although there were mixed views about the e-lab, seven respondents commented that having a project website, such as e-lab, earlier in the project would have been beneficial. Other suggestions for improving stakeholder engagement included (1) having more workshops during the project (n = 3); (2) having workshops of shorter duration (n = 3); (3) having more opportunities to communicate with other stakeholders between workshops (n = 2); and (4) being consulted before and during the design of the research project (n = 2).

Planned usage of the workHORSE tool

Eight respondents commented on how the workHORSE tool will be used within their organisation for the NHS HCP. The tool was an asset for the commissioning process, especially relating to projecting the effectiveness and cost-effectiveness of different scenarios and future impact. SH5 commented on the potential power of the tool to provide evidence at the local level and provide an NHS Health Check service tailored to local population requirements:

To gain an understanding of the current effectiveness and cost-effectiveness projection of the NHS Health Check and how these could be affected by various scenarios (e.g. reduced funding).

SH2

To support awareness of its value in commissioning and delivering evidence-based practice through our national influencing work and also through the support that our regional teams give to local commissioners.

SH3

Projections.

SH4

In the south, there has never been an appetite or strategic leadership to embed the programme as a key enabler to CVD action at scale. I have long promoted the use of HEAs [health equity assessments] to understand the equity of access and outcomes and use of this data to work with local stakeholders to design approaches to implementation that suit local population need and local system priorities (e.g. detection of those with HT [hypertension]). This tool, if made available, could help LAs demonstrate and engage locally to do this. It could also be used to demonstrate how impact of the programme is massively restrained if primary care doesn’t provide appropriate clinical follow up in line with NICE guidance (e.g. offer of statins to all over 10% risk). Tool should make it possible to demonstrate clearly how increasing such take up impacts, which could be powerful when working with CCGs, STPs [sustainability and transformation partnerships] and ICSs [integrated care systems] on their NHS Long Term Plan CVD ambitions and 5-year plan commitments.

SH5

Have kept as a regular agenda item on NW [north west] network meetings and hope to demonstrate at a NW network meeting with NW Health Check commissioners.

SH6

To establish the cost-effectiveness of one version v[ersus] another version of the HC [Health Check] programme . . . this will inform future decisions based on funding available.

SH8

To model various scenarios to show impacts.

SH9

Inform decision-making, allow for easier projections to be made and possibly use the tool as an enabler for better coverage across our patch.

SH11

Stakeholder requirements for implementing the workHORSE tool

Eight respondents requested additional support for the implementation of the workHORSE decision-support tool. The requests included having a user manual, telephone/e-mail support and training:

User manual, FAQs [frequently asked questions], short video clips for the demonstration of key elements and possibly some sample scenarios.

SH2

Some clear promotional material to describe the value of the tool and some additional support/training materials (or workshops) to assist those using the tool.

SH3

Telephone support, online/e-mail support.

SH4

I need a better understanding of where the platform will be hosted, who and how it can be accessed and what guidance there will be around using it with LAs and other local stakeholders.

SH5

It would be wonderful to have someone to present the tool at my NW [north west] network meeting in February.

SH6

I would like a future training session on the various ways the tool can be used as we were only given a basic demonstration on how to use.

SH8

Adding some demo[nstration] scenarios such as those demonstrated at the last workshop might make the tool more user-friendly. The user could run these scenarios and then tweak as required and save as a custom report.

SH9

Accreditation of the programme tool may be vital to ensure confidence of the outputs are acknowledged and useful.

SH11

Stakeholder engagement: areas for improvement

Five respondents commented on how their experiences in shaping the workHORSE tool could have been improved. Two respondents commented on the location of the workshops (i.e. all four held in Liverpool) and the time commitment of travelling and attendance. The remaining feedback related to workshop delivery and the decision-support tool.

Aim of patient and public involvement in workHORSE

Lay advisers were recruited and involved in the workHORSE project to:

• contribute to the design of the research

• contribute to the management of the research through the Study Steering Committee

• contribute to the content and delivery of the stakeholder workshops

• contribute to the reporting of the research

• summarise messages for lay audiences

• ensure that the perspective of the public was represented.

Methods

We recruited four lay advisers through the National Institute for Health Research (NIHR) Patient and Public Involvement (PPI) Network and local Healthwatch. All of the lay advisers had personal experience of using the NHS HCP and were interested in how the development of the decision-support tool would benefit Health Check provision.

We continuously evaluated our PPI with our lay advisers throughout the project, at meetings and by e-mail, to assess its impact and to identify areas for improvement in lay adviser involvement.

Preparation of the workHORSE application for funding

We organised a meeting with all four lay advisers and research team members to (1) provide the opportunity for lay advisers to meet each other and the research team; (2) gather lay advisers’ views on the readability of the Plain English summary and Abstract; (3) provide the opportunity for lay advisers to share their early thoughts about the proposed research; and (4) obtain lay advisers’ observations about how the proposed research might be improved.

The lay advisers provided constructive feedback on the content of the Plain English summary and offered suggestions on additional stakeholders we should invite. The meeting enabled discussion and agreement of their role in the project (i.e. their own experiences and skills in helping interpret the results and acting as a voice to disseminate the research to a broad audience).

workHORSE project delivery

The lay advisers were active members of our Study Steering Committee, providing feedback on their perspectives (primarily on workshop development and delivery). We also ensured ongoing communication with the lay advisers between workshops by e-mail and at face-to-face meetings. The lay advisers provided advice on the content of the four workshops and attended workshops 1, 2 and 4, where, as members of the research team, they ensured the timely delivery of activities and recorded and fed back their observations. In workshop 4, the lay advisers presented on their experience of the workHORSE project and what a project needs for effective PPI.

Towards the end of the project, we held an event with the lay advisers and research team to discuss their perspectives on their experience and involvement in the workHORSE project, what went well and what could be carried out better in the future.

workHORSE project dissemination

The lay advisers have been involved in the writing of academic publications for peer-reviewed journals. They also helped us to write clear, understandable literature for dissemination within the research community and engagement with the wider public.

Study results

The lay advisers felt valued in their involvement in the project. They perceived their role as acting as public consultants and translators of information for a wider audience, and as advocates adding value from the public perspective and observing the ‘return on investment’ for the research funding provided from the public purse.

Lay advisers saw their involvement in writing the research proposal as a positive approach to co-producing with the public:

. . . beginning was excellent in terms of involvement. I loved commenting on the bid. The first meeting was taking our views and making a key contribution to how the bid would look . . . although it felt a bit over my head, we thought it was looking at NHS Health Checks, not specifically modelling. It took time to figure out the idea of modelling . . . but I did enjoy it.

LA1

However, as illustrated by the comment above, the lay advisers felt that the nature of the project inhibited their involvement. They commented that:

The project was quantitative, and we look more into the qualitative . . . PPI is important because it influences the care standard of what the patient is receiving.

LA2

Conversely, they did identify their valuable role in dissemination to ensure that the research findings reached the public realm, including co-producing summaries of the research for publication in, for example, local government newsletters.

Discussion

From the perspective of the workHORSE research team, the involvement of the lay advisers in the project provided valuable feedback in terms of having a public perspective on the NHS HCP and how the decision-support tool could improve the patient experience.

The lay advisers did comment that the workHORSE project was not typical of projects that they had previously been advisers for, as it was not research involving patients and the public. However, as the project progressed, they very much perceived their role as ‘translators’ of the research as being at the interface of a research project and the ultimate beneficiaries of the project outcomes.

As active members of the Study Steering Committee, the lay advisers were able to have ongoing involvement in and input to the content and delivery of the workshops. They also provided valuable input in the writing of project materials, especially those requiring plain English. The lay advisers’ participation in the workshops enabled us to gain perspectives of model usage from the public who would ultimately benefit from the model.

The workHORSE project was not a conventional project in terms of having patients and the public as the research subjects. Therefore, the role the lay advisers had envisaged (and had previous experience of) was not utilised to its full extent, leading to an initial mismatch of the project requirements and lay advisers’ expectations.

Researchers embarking on building decision-support tools need to look at the project aims, outcomes and delivery, and embed PPI activities that are best suited to the project to maximise their role and project impact.

Sampling local authorities

We selected the best-performing LAs based on data from the NHS Health Checks Fingertips website [URL: https://fingertips.phe.org.uk/profile/nhs-health-check-detailed (accessed 4 November 2019)]. For the original sample, we looked at the performance of LAs during the complete 5-year cycle, from 2013 to 2017.

When considering the 5-year cycle, including the new quarters in 2018, the top-performing LAs remained the same. In addition, Gateshead joined the top-performing LAs, having improved most in the previous year, and was also included. The best-performing LAs were judged based on the percentage of eligible people who received an NHS Health Check. The percentage of eligible people who received an NHS Health Check varied from 17% to 95% across all LAs. PHE had an initial target for LAs to work towards 66% uptake of NHS Health Checks to improve coverage.

We used the 66% cut-off point for the best-performing LAs. In 2018, 12 LAs were reaching this target (Table 2), mapping well to the workHORSE model parameters and enabling scenario analysis, as described in Chapter 5.

Additionally, we wanted to ensure that we had a representative sample of best-performing LAs. We identified five additional LAs that improved most over the previous year/four quarters (2017/18) compared with the previous year (2016/17) using the percentage of eligible people who received an NHS Health Check (Table 3).

Model parameters and inclusion criteria

An objective was to produce real-world input parameters of NHS Health Checks for the workHORSE model. The model had six input parameters related to NHS Health Checks, including interventions, implementations or deliveries affecting:

1. eligibility criteria

2. coverage

3. uptake

4. average risk profile

5. diagnosis and treatment

6. referrals for brief interventions and lifestyle services.

To inform our workHORSE model, we selected the best-performing LAs based on their performance concerning coverage and uptake. These parameters were selected because what happens during and after an NHS Health Check is not routinely published. Only patchy data relating to the other parameters exist, and these are discussed in each LA summary in Results. Further information on the methodology can be found in Appendix 1.

Coverage

Conclusions

The approaches adopted for improving coverage and uptake of NHS HCP varied across all of the top-performing LAs we contacted. The LAs varied in terms of population profile and numbers, and levels of social deprivation. These factors influenced how the LAs designed and implemented strategies to increase coverage and uptake. It was therefore not possible to establish a typical pattern to identify a set of effective approaches that can be recommended to LAs that are not performing as well. However, it was apparent that all of the LAs had taken a strategic and sometimes innovative approach based on their population profile to achieve the targets set.

In terms of the workHORSE project, the information obtained provided valuable case examples for possible scenarios. The information also suggested that uptake-based scenarios for analysis will provide support to model different ‘implementations’ or optimisation of the programme to analyse with the workHORSE tool and to inform the content of the stakeholder engagement workshops.

Introduction

Among the strategies developed to tackle the preventable burden of NCDs, screening programmes are proposed and in place for many conditions: primarily cancers and infectious diseases. Screening programmes are used to detect diseases in an earlier stage in asymptomatic people who may have an increased risk of disease and can lead to better chances of successful treatment. In turn, earlier diagnosis reduces premature morbidity and mortality. 44 However, several biases, relating to screening programmes, have been identified, including overdiagnosis, length time bias, lead time bias and reaching the worried well. 44

The introduction of screening programmes depends on the effectiveness, acceptability and cost of the intervention. Globally, there are many types of screening programmes that target different populations. Screening programmes for adults are mainly cancer related, including breast, cervical and colorectal cancer, as feasible and efficient evidence-based strategies exist for these diseases, which are cost-effective. 45,46 However, variability exists in approaches to screening in different countries, as shown in a report on cancer screening in the European Union. 47 Differences exist concerning the level at which the screening takes place (regional or national), the screening test used [i.e. an immunochemical faecal occult blood test (FOBT) vs. a guaiac-based FOBT for colorectal screening], the interval at which screening takes place and the population targeted (age range). 47

For screening programmes to be successful, there needs to be a system and not just a test. Programmes need to have an infrastructure to provide support throughout the entire process, from inviting people to attend the screening through to treatment and follow-up. 44 However, to maximise the impact of screening programmes, high uptake, compliance and diagnosis are also essential. Uptake represents the most important factor in determining the success of screening programmes. Unfortunately, for breast and cervical cancer screening, a trend is emerging internationally that a smaller proportion of eligible women are being screened. This trend is also observed in the UK, with none of the cancer screening programmes meeting its agreed standard targets in 2017/18. 48

Reasons for low uptake included low awareness of screening benefits, low acceptability of certain screening tests [e.g. Pap (Papanicolaou) test], difficulty in accessing services, language or cultural barriers, perceived costs and structural barriers. 44,48 Some ethnic minority groups have significantly lower uptake and people with disabilities or mental health problems also tend to have lower uptake than the general population. Screening has the potential to reduce health inequalities. However, this is not certain with the current design and therefore it is crucial to promote equitable access for underserved groups. 11,48,49

High participation rates in screening programmes targeting NCDs are instrumental in achieving full screening benefits; however, uptake enhancement in screening programmes remains underused, especially among vulnerable populations.

We conducted an umbrella review to assess the type of approaches screening programmes use to maximise uptake, the effectiveness of the approaches and the impact on equity.

Methods

Results

In total, 5286 records were identified through the database searches. After removing 2207 duplicates, 3133 records were left for the title and abstract screening. During that process, 2955 records were excluded, leaving 178 records for full-text review. We included a total of 61 reviews in this umbrella literature review. Of these 61 reviews, 38 included more than two interventions or screening programmes (detailing 180 outcomes).

The main interventions identified included patient education, patient invitations and reminders, provider interventions, reducing out-of-pocket client costs, reducing structural barriers and multiple interventions. Definitions are provided in Table 5.

Most screening programmes identified focused on breast, cervical or colorectal cancer, with a few examples of other cancers (e.g. testicular) or infectious conditions [e.g. human immunodeficiency virus (HIV)].

Most of the systematic reviews and systematic reviews with meta-analyses included were reviewing RCTs, quasi-experimental and observational designs. In general, the individual reviews were at high risk of bias. The outcomes reported included screening uptake, participation, adherence and utilisation, test utilisation and some looked at guidelines’ adherence and compliance. A summary of the main findings is provided in Table 6.

Discussion

High-level description

The epidemiological engine of workHORSE is a discrete-time dynamic stochastic microsimulation. The epidemiological engine of workHORSE consists of three modules: (1) the sociodemographic module, (2) the exposure module and (3) the disease module.

Within the workHORSE epidemiological engine, each unit is a synthetic individual (simulant) represented by a record containing a unique identifier and a set of associated attributes. The microsimulation then projects the life course of each synthetic individual.

The attributes for each synthetic individual include the sociodemographic characteristics, exposures to risk factors, acquired diseases and cause of death, if relevant.

Specific attributes included the following:

• Age, sex, ethnicity, the LA of residence, education, quintile groups of equivalised income, Townsend Deprivation Index and Index of Multiple Deprivation (IMD) as sociodemographic exposures.

• Alcohol intake, smoking status (i.e. current smoker/ex-smoker/never smoked), smoking duration, smoking intensity, environmental tobacco exposure, fruit consumption, vegetable consumption and physical activity as behavioural risk exposure variables.

• Body mass index (BMI), systolic blood pressure (SBP), total serum cholesterol and high-density lipoprotein as biological risk exposures.

• Type 2 diabetes mellitus (T2DM), atrial fibrillation (AF), CHD, stroke, post-stroke dementia, chronic obstructive pulmonary disease (COPD), lung cancer, colon cancer and breast cancer as diseases. It is worth noting that some diseases have dual roles, additionally acting as risk exposures for other diseases (e.g. T2DM and AF are risks for stroke).

• Mortality from CHD, stroke, COPD, lung cancer, colon cancer, breast cancer or any other cause is recorded if it occurs.

• Statin utilisation, antihypertension medication utilisation and corticosteroid medication utilisation are each quantified.

• Family history of CVD, family history of diabetes, family history of cancer, personal history of any cancer, chronic kidney disease, rheumatoid arthritis and the additional number of comorbidities are also captured as auxiliary attributes.

All of these attributes are updated in discrete annual steps according to a set of stochastic rules. We structured these rules based on well-established epidemiological principles. Specifically, behavioural risk exposures are conditional on sociodemographic exposures, biological risk exposures are conditional on behavioural and sociodemographic exposures, and diseases are conditional on biological, behavioural and sociodemographic exposures. Finally, mortality is conditional on sociodemographic and disease exposures.

The life course of synthetic individuals is simulated as many times as the number of scenarios specified in the GUI, using the same random numbers for all policy scenarios to reduce stochastic noise. One of the scenarios is always the ‘baseline’ scenario with which all remaining policy scenarios are compared. The comparison of the disease outcomes under the life courses under the baseline scenario compared with the policy scenarios generates the health impact of the policy scenarios. The output of the epidemiological engine is a data set that contains the adult life course of the simulated synthetic individuals, with all of the attributes mentioned above recorded on an annual basis for every scenario.

The epidemiological engine of workHORSE consists of three modules: (1) the sociodemographic module, (2) the exposure module and (3) the disease module. In the below paragraphs, we will describe these three modules. Table 7 summarises the key assumptions and limitations of the workHORSE microsimulation model.

Sociodemographic module

The first year of every simulation in workHORSE is 2013, reflecting that the LA commissioning of the NHS HCP has been a statutory requirement since 2013 as part of the Health and Social Care Act. 122 When the user selects the geographic area for simulation in the GUI, the algorithm in the module proceeds with the following:

• The algorithm identifies the lower-layer super output areas (LSOAs) that constitute the user area selection.

• The algorithm draws 200,000 synthetic individuals, aged 30–89 years, from the joint age and sex distribution of the identified LSOAs. This is a default value that can be modified by the user. The joint age and sex distribution for each LSOA for 2013 is informed by the Office for National Statistics (ONS) population estimates. 123

• The algorithm assigns each synthetic individual an IMD score and a Townsend Deprivation Index score based on their LSOA.

• The algorithm probabilistically assigns each synthetic individual an ethnicity based on their age group, sex and LSOA. The ethnicity mixture of each LSOA is informed by the 2011 Census. 124 We include nine ethnicities in the model: (1) white, (2) Indian, (3) Pakistani, (4) Bangladeshi, (5) other Asian, (6) black Caribbean, (7) black African, (8) Chinese and (9) other.

• The algorithm assigns each synthetic individual a Strategic Health Authority (SHA) based on their LSOA. A SHA is the smallest geographical area that is accessible in the Health Survey for England (HSE) series. We use this in the exposure module, as described in the next section (see Exposure module).

• The algorithm probabilistically assigns each synthetic individual their highest educational qualification conditional on their age, sex, quintile of the Index of Multiple Deprivation (QIMD), SHA and ethnicity. We used six levels for this variable [(1) National Vocational Qualification (NVQ) 4/NVQ 5/degree or equivalent, (2) higher education below degree level, (3) NVQ 3/General Certificate of Education (GCE) Advanced Level equivalent, (4) NVQ 2/GCE Ordinary Level equivalent, (5) NVQ 1/Certificate of Secondary Education other grade equivalent, (5) foreign/other and (6) no qualification], which was informed by the HSE ‘topqual3’ variable. 121 Specifically, we first fitted an ordinal regression model to the HSE data using topqual3 as the dependent variable and year, age, sex, QIMD, SHA and ethnicity as the independent variables. Then we sampled from the distribution of this model to simulate the highest educational qualification variable in the synthetic population. The approach binds the synthetic individuals to their highest educational qualification that remains constant over the simulated years. This is unlikely to introduce any substantial bias, as most adults aged > 30 years have already achieved their highest educational qualification.

So far, the algorithm has created a synthetic population that is a snapshot of the population in 2013. The following steps of the algorithm create backward and forward projections of the synthetic population that are essential to model exposure time trends and time lags between exposures and diseases.

The backward projection of the synthetic population goes back to 2003 and therefore the maximum time lag we allow in the model is 10 years. As everyone alive and aged > 30 years in 2013 was alive in 2003, the algorithm simply creates the back projections by appropriately reducing the age of the synthetic individuals, while keeping all other variables constant.

Similarly, for the forward projections, we project until the year 2041 and the algorithm increases the age of the synthetic individuals while keeping all other variables constant. For forward projections, mortality needs to be considered. We describe mortality within the disease module as disease-specific mortality, which is closely related to disease prevalence. workHORSE follows an open cohort approach. For every simulated year from 2013 onwards, a new cohort of 30-year-old synthetic individuals enters the model. The same sources inform the size of the cohort and the joint age, sex and ethnicity distribution that we described above. For example, in 2014, the new 30-year-old cohort will be informed by the population size and the joint age, sex and ethnicity distribution of those who were 29 years old in 2013. The approach may be crude; however, the final model outputs are directly standardised to ONS population projections estimates by LA or nationally. 125,126

Exposure module

This module simulates the adult life course exposures of the synthetic individuals based on the HSE series between 2003 and 2014. 127–138 For all simulated exposures, we follow the same general principles. First, we fit an appropriate statistical model to the HSE data, with the exposure of interest as the dependent variable and some functions of year, age, sex, QIMD, ethnicity and SHA as independent variables. Then, we use the statistical model to predict the exposure level of every synthetic individual in the simulation based on their sociodemographic characteristics that were estimated from the sociodemographic module.

The inclusion of year as an independent variable in our exposure model allows us to extract the trends from the HSE series and project them into the future. Furthermore, it allows us to make backward projections of exposures that we use when we simulate time lags. For example, for a synthetic female individual aged 30 years in 2013, we can estimate her BMI in 2003 when she was aged 20 years and in 2033 when she will be aged 50 years. To avoid excessively fast changes in exposure trends, and to reflect our belief that decays and growths in natural phenomena are rarely linear, we included the natural logarithm of years in the statistical models, assuming logarithmic trends.

The inclusion of SHA as an independent variable in our exposure models allow us to perform small-area estimation from the SHA level that is available in HSE to the LA level using individual-level modelling. 139 Essentially, we apply the exposure level observed in HSE at SHA level to the LA level weighted for the sociodemographic characteristics of the population at each LA. Therefore, we assume that differences in mean exposures between LAs belonging in the same SHA are because of differences in their sociodemographic composition.

For exposures that were recorded as ordinal categorical variables in HSE, we used logit ordinal regression to model them. For all other exposures, we used generalized additive model for location, scale and shape (GAMLSS). 140,141 These are flexible statistical models that can produce all parameters of an assumed distribution for the dependent variable, conditional to some function of the independent variables. For example, GAMLSS can model both the mean and the standard deviation of a dependent normally distributed variable, whereas a linear regression models only the mean. Appendix 4, Table 36 summarises our modelling approach for all of the exposures in the model.

The approach described above provides us with equations to estimate the distribution of an exposure for a given time and sociodemographic characteristics of a synthetic individual. When the synthetic individual enters the simulation, a set of random numbers between 0 and 1 and of size equal to the number of the modelled exposures is allocated to them. Each of them represents the percentile of the relevant exposure distribution. The principle is that synthetic individuals retain their percentiles throughout their life course (this is known as the rank stability assumption). 142 For example, in 2013, a 40-year-old male synthetic individual living in a QIMD 3 area with SBP of 120 mmHg has a SBP percentile of 0.52. Twenty years later, the same synthetic individual has retained his percentile score for SBP. However, his SBP is now estimated to be 137.6 mmHg because the SBP distribution has changed to reflect the SBP of 60-year-old men living in a QIMD 3 area in 2033 (Figure 2). In workHORSE, we allow the percentiles of the synthetic individuals to fluctuate every year using random walks to relax the rank stability assumption.

FIGURE 2.

Plot of the percentile against the SBP (cumulative distribution) of a male synthetic individual living in QIMD 3 area for aged 20 and 70 years.

Disease module

The previous two modules for demographics and exposure generate a dynamic close-to-reality synthetic population that is composed of the adult life course exposures of each of the synthetic individuals. The disease module then translates these exposures to disease incidence, using a population attributable risk approach [i.e. population attributable fraction (PAF)]. 144 We will first describe how the disease incidence is simulated in the model and then how the model simulates mortality.

Economics methods

The main economic analysis objective was to enable users to estimate the cost-effectiveness of each scenario within the model. The potential scenarios are a range of real-world and hypothetical scenarios around performance on NHS Health Checks in England as a whole or in individual LA areas. The eligible population is typically adults aged 40–74 years without pre-existing conditions. The broader context is to allow decision-makers to test scenarios around NHS Health Checks so that they can optimise the programme in their area in terms of understanding cost-effectiveness and equity impact; and informing budget allocation, payment and incentive strategies, and performance management.

Cost-effectiveness can be estimated within the tool with a range of ICER willingness-to-pay thresholds or QALY valuations (e.g. £30,000 based on NICE recommendations or £60,000 based on the UK Treasury Green Book29) and a range of perspectives (i.e. health care, health and social care, and societal).

Overall, the method was cost-effectiveness analysis. However, we enabled decision-makers to use a range of outcome measures and economic perspectives. First, the health-care perspective included intervention costs, health-care cost–consequences (e.g. disease costs) and net QALYs. Second, the health and social care perspective included the same as the health-care perspective with the addition of social care costs. Finally, the societal perspective included the same as the health and social care perspective with the addition of net informal care costs and production costs (i.e. household production and earnings).

The QALYs were calculated based on population norms from Janssen and Szende,158 adjusted using the equations from the UK EuroQol-5 Dimensions (EQ-5D) Medical Expenditure Panel Survey (MEPS) catalogue. 159 The workHORSE model adjusts for comorbidities, which is not possible with typical cohort or life table economic models.

The health-care disease costs were assembled based on searching for excess costs of diseases (as opposed to total running costs of individuals that do not separate out the costs of disease from other costs and comorbidities) and data obtained from recent high-quality UK studies.

Social care costs were based on Office for Budget Responsibility (OBR) estimates by age, adjusted for prevalence of dementia and stroke. 160,161 We did not find strong evidence for excess social care costs for other specific diseases. However, there are additional social care costs in the model that were associated with ageing.

Informal care costs were based on a study that predicted informal care costs using Health Outcomes Data Repository data for people discharged from hospital in Wales based on their age and EQ-5D scores. 162 Production costs (i.e. household production and earnings) were based on Appendix B of a paper by Claxton et al. 163 that presents a wealth of information about wider social benefits and has informed the approach taken by the UK Department of Health and Social Care in its impact assessments. The earnings and production estimates from this paper were updated with more recent data from the ONS. Unlike other costs that are deficits, production is a benefit measure in the model. Therefore, production decreases as rates of disease increase, whereas health, social and informal care costs all increase as rates of disease increase.

Costs were measured in Great British pounds in 2019 prices. The discount year was therefore 2019. Costs and QALYs that were gained or lost in years before 2019 were inversely discounted in the model. Costs and QALYs are given a range of potential discount rates in the model so that decision-makers can use 3.5% for both costs and QALYs (as typically used by NICE), 1.5% for QALYs and 3.5% for costs (as typically used by the UK Treasury) or 0% for both if they want to estimate the actual undiscounted costs and QALYs in constant prices.

More detail of costs and QALYs is given in Appendix 6.

Outcomes

The workHORSE simulation tool produces a range of outcomes, including disease case-years prevented or postponed (CYPP); economic outcomes, such as ICERs; and equity-related outcomes, such as a change in slope index of inequality and relative index of inequality. This approach enables distributive equity analysis, as we have demonstrated in our published papers using this model. 11,164

Overall results are presented as benefit–cost ratios, ICERs and net monetary benefit (NMB). This is to assist decision-makers in understanding both the ratio of costs and benefits, and the absolute magnitude of the benefits, as, for example, decision-makers may favour a programme that delivers slightly larger absolute benefits across the population over a programme that has a better cost–benefit ratio but smaller absolute returns, particularly if they have a fixed budget to allocate to NHS Health Checks.

Where the cumulative costs and QALYs are as follows, the benefit–cost ratio would be 10.6, representing a return on investment of £9.60 for every £1 spent. Strictly speaking, cost–benefit ratios have no units because they are a ratio, but they are often presented as ‘£10.60 in benefits for every £1 spent’.

The populations in the model are for England and 150 English upper-tier LAs, by age, sex, QIMD and ethnicity.

The decision-support tool allows a range of time horizons from 2013 to 2041, from 1 to 28 years. The reason the tool starts from 2013 is to allow decision-makers to model the cost-effectiveness of historical NHS Health Checks performance retrospectively, which they may want to compare with future scenarios.

Crucially, the model is a dynamic open-cohort microsimulation model (i.e. the model is trying to estimate the actual cost-effectiveness of the scenarios within a dynamic population where people are born, people age, people’s risk factors change and people die). The detailed modelling of the population dynamics in our model is therefore different from many economic models, which are often closed cohort, meaning that they follow the same population over time and often have a lifetime horizon.

The tool allows the user to filter outcomes for specific subgroups if they wish. This can be by age, gender, QIMD or ethnicity so that this can input into a subgroup analysis or be used for an equity audit.

Health equity methods

To estimate the impact of NHS Health Checks on existing absolute and relative socioeconomic inequalities in QALYs experienced across the population, we used two regression-based metrics inspired by the slope index of inequality: (1) the absolute equity slope index and (2) the relative equity slope index. 27,165 The absolute equity slope index measures the impact of an intervention on absolute inequality and the relative equity slope index considers the pre-existing socioeconomic gradient of disease burden and measures the impact of an intervention on relative inequality. For both metrics, positive values mean the intervention reduces inequality and negative values mean the intervention increases inequality. The impact on relative socioeconomic inequalities is therefore meaningful when the intervention tackles absolute inequalities (i.e. the absolute equity slope index is positive) only.

We used fifths of the national IMD as a marker of socioeconomic stratification. Some LAs have populations that are skewed towards certain deprivation quintiles. Each IMD fifth (quintile group) is characterised by a ridit value that corresponds to the average cumulative frequency of the IMD fifth to account for population size differences. This means that the regression line gradient is adjusted for the proportion of the population in each fifth and it represents an estimate of the population social gradient, rather than the most and least deprived group. In addition, within the user interface, we give users the option to use local IMD group quintiles in place of national quintiles (where each local quintile contains 20% of the LA population).

These calculations allow decision-makers to weigh up the equity benefits of a scenario against the health benefits generated, which are shown on the user interface on the health equity plane, as in Figure 3. In Figure 3, scenario 2 gains slightly greater monetary benefits and larger reductions in absolute inequalities. 166

FIGURE 3.

The health equity plane.

Dealing with uncertainty in health economics inputs

There are three main uncertainties around health state utility values.

The first uncertainty is around the model and the sample that was used to produce the indices. The MEPS data were modelled with censored least absolute deviation regression based on 79,000 individuals.

The second uncertainty is around the mean health state utility index value for a given condition (or age, gender, deprivation category), which would be driven by the sample size. Therefore, a small sample size may pick up people whose CVD is more or less severe than the general population in that category. As the sample gets bigger, the standard error gets smaller and tends towards zero, and the average should tend towards the population average. The standard errors in our sources are reassuringly small, which indicates a high degree of certainty around the mean utility decrements each disease.

The third uncertainty is the individual-level variation (unobserved heterogeneity), as described by the standard deviation. As the sample size gets bigger, this may be reduced, but with larger sample size the standard deviation will tend towards the true population standard deviation. In workHORSE, we have included the uncertainty around the mean decrement as measured by the standard error. In addition, because the initial sample was from the USA – albeit matched to UK preference scores – this may introduce another level of uncertainty that we cannot measure.

We estimate the probability over time for the scenario to be cost-effective, cost saving or for reducing health inequalities. We added a visual aid in the plane graphs (see Figure 10) to show that the high upfront cost of the checks is weighted against future health gains because of the time lags between exposure change and disease risk change.

Simulation parameters tab: basic settings

To start using the workHORSE tool, the user needs to provide some necessary information in the simulation parameter sheet, including the geographical area of interest, the time horizon to simulate and how many scenarios to test (including the baseline or business-as-usual scenario). Finally, the user needs to decide on the perspective of the health economics analysis (i.e. societal, health and social care, or health care) (Figure 4).

FIGURE 4.

The simulation parameters tab.

In addition, the user can decide whether to use the national IMD quintiles or area-specific IMD quintiles. Local commissioners sometimes prefer area-specific quintiles because it means that there is an equal proportion (roughly 20%) of the population in each quintile, whereas areas will often have an uneven distribution of the population across national IMD quintiles.

Setting the scenario parameters tab

Once the user has defined the number of scenarios to evaluate, the user needs to define the scenarios in the scenario parameter sheet. To set any scenario, including the baseline scenario, the user needs to provide information about the model parameters. Each parameter has a tab in the scenario parameters sheet (Figure 5).

FIGURE 5.

The scenario parameters tab.

All scenario parameters have default values to give the user an indication of the magnitude of the value it would be expected under ‘normal’ use. For example, in the ‘Health Checks received’ parameter it is usually > 50% in most areas. However, the user is still allowed to input any plausible value for all parameters. We recommend that users have as much real-world data as possible to inform scenarios, particularly the baseline scenario.

Using the general parameters tab, the user can define (1) the name of the scenario (e.g. scenario A, baseline scenario), (2) whether or not this scenario is the baseline against which all other policy scenarios will be compared and (3) the starting year of implementation for this scenario.

Once the scenario has been defined, it can be saved so that it can be used later as a template for a new scenario. To do this, the user has to go to the general parameters tab and click the ‘save scenario’ button. If the user needs to use this scenario as a template later, it can be loaded by selecting ‘load scenario’ in the general parameters tab. All scenario parameters are saved in YAML (YAML Ain’t Markup Language) files that can be edited in any text editor.

Using the eligibility criteria tab (see Figure 5), the user can alter (1) the age of eligibility to be invited, (2) how often (in years) NHS Health Checks are offered and (3) whether known diabetics or hypertensive individuals are eligible. This tab already has default values to reflect current practice at the time that this report was written (2020).

In the appointments offered yearly tab, the user can specify changes related to coverage of the NHS HCP by changing the parameter ‘invitations’ (i.e. percentage of the eligible population) and the ‘cost per invitation’. If the user wishes to vary the above parameters by IMD, this can be carried out by clicking ‘detailed input’.

In the NHS Health Checks received tab, the user can specify changes related to uptake of the NHS HCP by changing the parameter ‘proportion of invitees attending an NHS Health Check’ and changes related to the payment providers received for each participant by changing the ‘cost per completed NHS Health Check’. This way, the user can accommodate different payment mechanisms for NHS Health Checks. For instance, they can select ‘block contract’ or ‘payment by results’ or a combination of both, provided that the user can derive these costs outside the workHORSE GUI. In the detailed input boxes, the user can define uptake by age group, sex, QIMD and risk.

The prescription rate tab (see Figure 5) relates to what happens after the NHS Health Check has taken place. Here, the user can specify changes to prescription rates of statins and antihypertensive medication for those participants with a QRISK 2 score > 10 and raised total cholesterol, or participants with hypertension, respectively. As can be done with the other parameters described above, the user can use the ‘detailed input’ button to vary these parameters by IMD and risk, if data are available.

In the impact on lifestyle tab, the user can evaluate the potential outcomes of referrals to lifestyle services, such as smoking cessation, weight management, alcohol consumption and physical activity programmes. The user would need to specify the percentage of people successfully achieving smoke cessation, losing weight, reducing alcohol consumption or increasing their physical activity, and the associated overhead and per-participant costs. Additionally, for the weight management, alcohol consumption and physical activity programmes, the user would need to specify the mean weight loss (kg), mean percentage reduction of alcohol consumption or the number of days physical activity increased, respectively.

Inspecting the model outputs

After setting up the scenarios, the user runs the simulation. The outputs tab will show on the dashboard. Here, the user will find, summarised in the banner at the top of the screen, the headline results for the scenarios analysed, a cost-effectiveness plane and an equity plane. During our stakeholder engagement workshops, the participants chose the type of graphs and other outputs to be presented and the way of presenting the information in the dashboard (Figure 6).

FIGURE 6.

The outputs tab ‘dashboard’.

The model outputs dashboard provides more in-depth results for health economics, effectiveness and equity analysis. For example, the user can choose to inspect a breakdown of the disease cases prevented or postponed (CPP) in graphical form (Figure 7).

FIGURE 7.

Example of the advanced effectiveness tab.

For advanced users, the results can be further tailored to specific age groups or results using ‘filters’.

The graph can also be exported for use in reports or presentations as high-quality Portable Network Graphics files.

Technical implementation

We developed the GUI in R Shiny (URL: https://shiny.rstudio.com/). For the plots, we used the R Plotly library (https://plot.ly/) and for the tables we used the R DT library, a wrapper for the JavaScript DataTables library. All of the dependencies of the workHORSE app can be found at the source code of the app [URL: https://github.com/ChristK/workHORSE/blob/master/dependencies.yaml (accessed 25 February 2021)].

Conclusions

This prototype GUI is an example of what can be achieved by building on top of the simulation engine. The user interface then represents a baseline scenario for a health-care preventative intervention: in this case, the NHS HCP.

The use of open-source technologies enables further development of the interface and adaptation to specific needs for a user.

Despite the simplicity, the approach used in the scenario parameter tab allows us to implement scenario analysis for all key aspects of programmes such as the NHS HCP. The key design decision was to use the basic parameters as the key entry point for user interaction. The user can then specify parameters that represent process and key performance indicators in the programme design [i.e. who will participate (eligibility), what mechanism will be used to optimise attendance (e.g. appointments offers or invitations), plus additional parameters describing participation (uptake) and provider delivery (prescriptions)]. Users are then empowered to develop scenarios to inform changes in these parameters ‘off-model’ and use the information to refine the scenarios or specify new ones.

Our approach provides an excellent balance to enable basic users to extract value from scenario analysis while providing flexibility to accommodate changes affecting programme design or provider delivery. Ideally, it might be useful to provide functionality for full-scenario specification. However, that would result in a less flexible modelling environment, particularly for programmes such as the NHS HCP, which are frequently reviewed, with subsequent nuanced changes to the remit or design.

Analysis 1: optimising Liverpool implementation of the NHS Health Check programme to emulate the best-performing local authorities in the region

We first characterised a baseline scenario representing the current implementation of the NHS HCP in Liverpool in 2020, covering a population of some 552,000 people. We modelled the impact and costs if the NHS HCP continued unchanged in Liverpool throughout the modelled period to 2040. The parameters for this scenario, including costs, were informed from a previous research project we co-produced with the Public Health Department at the Liverpool City Council, but updated with new information when necessary and available. 11 We updated the annual coverage to 12.5% and the annual uptake to 57.5% (PHE Fingertips data for 2018–19172), and we assumed that they would remain stable until 2040. Prescription rates of statins would continue at 5% of the participants and antihypertensive treatments would continue at 6% of the participants, with different weights allocated to the prescription rates for low-, medium- and high-risk patients. The prescription rate for statins and antihypertensive medication use the number of participants as the denominator. Our stakeholders informed us that this is an easier indicator to extract from their data.

The workHORSE algorithm internally converts the prescription rate to one that has the number of eligible participants in the denominator. Eligibility is defined based on current NICE guidelines. Costs were updated to £6 per invitation and £15 per successful participation in the NHS HCP. This scenario also assumed that 1% of smokers would cease, 1% of those who are obese and overweight would lose about 1% of their weight, 1% of participants would increase their physical activity by 1 active day per week and that 1% of the participants who are heavy drinkers would reduce their alcohol intake by 1%.

The alternative ‘optimal’ scenario in this analysis represented the optimal implementation based on the best-performing LA in England in terms of invitation and uptake, plus assuming correspondingly better prescription rates and increased referrals to highly effective lifestyle services. The activities that the best-performing LAs described in our review in Chapter 3 provided a pragmatic justification for such a scenario. We therefore modelled the programme to invite 20% of the eligible population with 96.2% uptake (as achieved by the best-performing LA in England). Prescription rates in this scenario were assumed to be 10% and 12% to reflect current prescription guidelines and optimal provider compliance. We also assumed optimally effective lifestyle services, assuming that 10% of smokers would cease, 10% of obese and overweight participants would lose about 5% of their weight, 10% of participants would increase their physical activity by 1 active day per week and that 10% of the participants who are heavy drinkers would reduce their alcohol intake by 10%. The costs were estimated at £6 per invitation and £20 per participant (recognising that the additional activities would be more costly).

Analysis 2: improved uptake method based on a large randomised controlled trial

In this analysis, we used the model to explore the adoption of a better invitation method, using behaviourally informed invitation letters based on a published RCT. 173 In this pragmatic RCT there was one control arm and two intervention arms that used two types of behaviourally informed letters, one looking at arguments to counteract common barriers to attending the programme and one looking at sunk-cost information provision. 173 The counterargument letter proved to be the better of the two interventions, increasing uptake by 5.5%, whereas the sunk-cost letter increased uptake by 4.3%.

We therefore decided to use the more powerful of the two effective intervention letters (i.e. the counterargument letter). We assumed consistent effects by age, sex and deprivation, and assumed that there were no substantial extra costs associated with the improved invitation letter. This scenario was discussed and refined in workshop 4 with our stakeholders.

For the base case, we used a group of LAs in Northamptonshire, including Corby, Kettering, Daventry, Northamptonshire East and South, and Northampton, with a population of 747,622 people. We use invitation and uptake rates as reported on the Fingertips website for 2018–19,172 with an invitation proportion of 12.0% of the NHS HCP eligible population and an uptake of 51.0%.

For both the base case and the behaviourally informed letter scenario, we assumed that all of the scenario parameters were the same, except that uptake increased from 51.0% to 56.5%. We used the same costs for the implementation of the NHS HCP in both scenarios to see if varying the text in the invitation letter would make a negligible difference to costs.

In addition, we reran the scenario, assuming improved lifestyle services provisions, as in scenario 1, simulating a more comprehensive redesign of the implementation. In our experience of interacting with workHORSE, lifestyle changes were the key to substantially increasing programme effectiveness.

Analysis 3: using detailed versus rough input data to inform model equity assessments

The workHORSE model allows the scenarios to be specified using the risk profiles of the participants and those for whom medication is prescribed. However, almost all stakeholders requested that users should be able to define scenarios without this information. Therefore, when the risk profile of the participants is not explicitly specified, the workHORSE model assumes that, on average, their profile is that of the simulated area. When the risk profile for those for whom medication is prescribed is not explicitly specified, the workHORSE model assumes that their risk profile is like that of those who would be eligible for treatment, based on NICE guidelines. These additional assumptions introduce bias to the workHORSE model. The essence of this analysis is to identify the potential direction and quantify the potential magnitude of the bias. It does not necessarily reflect policy-relevant scenarios but may be used to emphasise the importance of detailed data about who is participating in the programme and how they react to it. Data regarding the risk profile of those participating in NHS HCP and those who are prescribed treatment, as a result, are not currently systematically collected. We obtained our data for these scenarios from Liverpool City Council and Liverpool CCG and have published policy-relevant scenarios in our PLOS Medicine paper. 11

The specific scenarios in this analysis were as follows.

Analysis 1: optimising Liverpool implementation of the NHS Health Check programme to emulate the best-performing local authority in England

Compared with current NHS HCP implementation in Liverpool, the workHORSE model estimated that optimal implementation might result in health gains and that there is more than an 80% probability of becoming cost-effective by 2030 and cost saving by 2031, while reducing socioeconomic inequalities.

Effectiveness

The workHORSE model estimated that the optimal implementation of NHS HCP in Liverpool could prevent or postpone approximately 142 (95% UI 110 to 181) deaths by 2040. Table 8 presents CPP and CYPP by disease cumulatively up to 2040. Negative numbers denote additional cases and additional case-years. This is because, as the optimisation of NHS HCP prevents death and prolongs life, synthetic individuals then become susceptible to other diseases (competing risk framework). Furthermore, specifically for T2DM, the additional cases also reflect the increased risk of patients treated with statins to develop T2DM. Figure 8 depicts the dynamics of disease prevention. T2DM is of particular interest, as the initial beneficial effect of NHS HCP through lifestyle improvements is neutralised and reverts by 2040, primarily because of the increased statin utilisation.

TABLE 8 - Analysis 1: effectiveness by 2040

Disease	CPP	95% UI	CYPP	95% UI
CHD	212	161 to 271	1339	861 to 1812
Stroke	60	40 to 88	380	213 to 588
Post-stroke dementia	5	–1 to 13	32	–6 to 74
COPD	–1	13 to 13	–80	–185 to 10
T2DM	–28	–136 to 90	–34	–935 to 838
Lung cancer	1	–7 to 6	1	–27 to 19
Colon cancer	3	–1 to 7	16	–9 to 41
Breast cancer	5	0 to 11	27	–11 to 74

FIGURE 8.

Analysis 1: CYPP by disease over the simulation period.

Cost-effectiveness

The workHORSE model estimated that under the optimal scenario, 610 (95% UI 455 to 845) QALYs could be gained over the simulation period. (For comparison, the total QALYs over the simulated period for the baseline scenario were 4.93 million.) Table 9 presents the cumulative net costs and NMB. Overall, this scenario was highly likely to be cost saving by 2040 (Figures 9 and 10).

TABLE 9 - Analysis 1: cumulative net costs and NMB estimates for the optimal implementation scenario

Note

NMB includes QALYs valued at £20,000.

Output	GBP (£)	95% UI (£)
Net policy costs	5.09M	5.05M to 5.11M
Net health-care costs	–2.44M	–3.72M to –1.27M
Net social care costs	–33,400	–167,000 to 112,000
Net informal care costs	261,000	–251,000 to 761,000
Net productivity costs	20.0M	13.7M to 29.3M
NMB	29.6M	19.2M to 42.8M

FIGURE 9.

Analysis 1: cost-effectiveness plane, optimal implementation scenario by the year 2040. Light orange, cost increasing or cost decreasing intervention with decreasing QALYs; light blue, cost saving intervention; light purple, cost-effective intervention.

FIGURE 10.

Analysis 1: annual probability of optimal implementation scenario to be cost saving.

Equity

In terms of equity, the optimal scenario was more equitable than the current implementation, that is being more effective among the more deprived both in absolute and relative terms (Figure 11). Liverpool is a deprived LA, with most of the population living in areas in the most and second most deprived IMD quintile. Therefore, a highly effective NHS HCP, as this scenario assumes, would almost certainly reduce at least the absolute socioeconomic inequalities in health.

FIGURE 11.

Analysis 1: equity impact of the optimal implementation using the equity plane. SII, slope index of inequality.

Analysis 2: improved uptake method based on a large randomised controlled trial

Improving uptake using a behaviourally informed letter is unlikely to become cost-effective by the end of the simulation horizon (i.e. 2040), compared with the current implementation. However, adding an optimised lifestyle intervention will become cost-effective by 2028 and cost saving by 2029, with a probability higher than 80%. The effect on both scenarios on socioeconomic health inequalities is very uncertain.

Effectiveness

The workHORSE model estimated that the ‘behavioural’ scenario that assumes that a more effective invitation letter could increase uptake by 5.5% would have a negligible impact on mortality, preventing or postponing 1 (95% UI –22 to 36) death by 2040 in Northamptonshire. The addition of highly effective lifestyle interventions could prevent or postpone 28 (95% UI 7 to 62) deaths. Table 10 presents CPP and CYPP by disease cumulatively up to 2040 for both scenarios, and Figure 12 depicts the dynamics of disease prevention. Interestingly, unlike analysis 1, none of the scenarios is expected to increase T2DM cases because the additional prescription of statins resulting from the increased uptake is small in both scenarios.

TABLE 10 - Analysis 2: effectiveness by 2040

Scenario	Disease	CPP	95% UI	CYPP	95% UI
Behavioural	CHD	7	–19 to 34	53	–202 to 254
	Stroke	2	–15 to 20	14	–128 to 153
	Post-stroke dementia	0	–5 to 6	0	–40 to 42
	COPD	0	–10 to 7	–6	–91 to 66
	T2DM	2	–30 to 29	24	–251 to 256
	Lung cancer	0	–6 to 5	–1	–20 to 18
	Colon cancer	0	–4 to 4	0	–26 to 33
	Breast cancer	0	–6 to 5	0	–39 to 36
Behavioural plus lifestyle	CHD	14	–21 to 42	90	–149 to 293
	Stroke	10	–13 to 31	71	–102 to 228
	Post-stroke dementia	0	–6 to 10	0	–37 to 78
	COPD	1	–10 to 10	–14	–94 to 94
	T2DM	34	–9 to 72	324	–92 to 701
	Lung cancer	0	–6 to 7	0	–23 to 26
	Colon cancer	2	–5 to 8	9	–30 to 45
	Breast cancer	2	–5 to 8	9	–38 to 55

FIGURE 12.

Analysis 2: CYPP by disease and by scenario over the simulation period. (a) Behavioural; and (b) behavioural plus lifestyle.

Cost-effectiveness

The workHORSE model estimated that approximately 16 (95% UI –83 to 150) QALYs could be gained under the behavioural scenario and 130 (95% UI 45 to 260) QALYs could be gained under the behavioural plus lifestyle scenario over the simulation period. (For comparison, the total QALYs over the simulated period for the baseline scenario were 7.45 million.) Table 11 presents the cumulative net costs and NMB. Overall, the workHORSE model estimated that the use of the improved invitation letter could only marginally improve the cost-effectiveness of the programme, whereas much larger benefits are expected by adding highly effective lifestyle interventions (Figures 13 and 14).

TABLE 11 - Analysis 2: cumulative net costs and NMB estimates for the behavioural and behavioural plus lifestyle scenario

Note

NMB includes QALYs valued at £20,000.

Scenario	Output	GBP (£)	95% UI (£)
Behavioural	Net policy costs	342,000	320,000 to 365,000
	Net health-care costs	–57,100	–796,000 to 587,000
	Net social care costs	–10,300	–133,000 to 150,000
	Net informal care costs	19,000	–329,000 to 393,000
	Net productivity costs	399,000	–3,190,000 to 4,960,000
	NMB	331,000	–5,000,000 to 7,400,000
Behavioural plus lifestyle	Net policy costs	340,746	312,000 to 365,000
	Net health-care costs	–413,906	–1,280,000 to 260,000
	Net social care costs	10,523	–229,000 to 170,000
	Net informal care costs	10,154	–329,000 to 472,000
	Net productivity costs	4,007,957	1,010,000 to 10,700,000
	NMB	7.20M	2.20M to 15.2M

FIGURE 13.

Analysis 2: cost-effectiveness plane for behavioural and behavioural plus lifestyle scenarios by 2040. Light orange, cost increasing or cost decreasing intervention with decreasing QALYs; light blue, cost saving intervention; light purple, cost-effective intervention.

FIGURE 14.

Analysis 2: annual probability of the scenarios to be cost-effective.

Equity

Both scenarios are very uncertain in terms of reducing inequalities (Figure 15). Note, however, that we did not specify any of the scenario inputs by QIMD and this is reflected in the high estimated uncertainty.

FIGURE 15.

Analysis 2: equity plane for the comparison of the equity impact of the behavioural invitation method and the behavioural invitation method with improved lifestyles services scenarios. SII, slope index of inequality. Light blue, intervention reduce inequalities and results in positive net monetary benefit; light orange, intervention increase in equalities and reduces net monetary benefit.

Analysis 3: using detailed versus rough input data to inform the model

The aim of analysis 3 was to quantify the bias that is introduced by not specifying the detailed inputs for the scenarios. Therefore, the presentation of the results will focus on this aspect. As a reminder, the baseline scenario in this analysis assumes no NHS HCP. Moreover, the use of the detailed inputs that reflect the risk profiles of the participants in Liverpool is, on average, lower than the population in Liverpool.

Effectiveness

Figure 16 depicts the cumulative CYPP in each scenario over the simulated period. The figure gives the impression that scenarios that do not use the detailed inputs in their specification overestimate the effectiveness of NHS HCP. However, a closer look at Figure 17 reveals that scenarios that do not use the detailed inputs underestimate both positive and negative CYPP, and the net effect of this is an overall overestimation of CYPP.

FIGURE 16.

Analysis 3: cumulative CYPP in each scenario in Liverpool over the simulated period.

FIGURE 17.

Analysis 3: cumulative CYPP by disease in each scenario in Liverpool, over the simulated period.

Cost-effectiveness

Interestingly, the omission of the detailed scenario inputs seems to underestimate both the net utility of the intervention (using QALYs) and the incremental costs, and the underestimation increases as the coverage and uptake of the intervention increases (Figure 18). However, the probability of a scenario to be cost-effective appears less sensitive to whether or not detailed inputs are used (Figure 19).

FIGURE 18.

Analysis 3: cost-effectiveness plane for each scenario by 2040. We present only the median values for easier interpretation. Light orange, cost increasing or cost decreasing intervention with decreasing QALYs; light blue, cost saving intervention; light purple, cost-effective intervention.

FIGURE 19.

Analysis 3: annual probability of the scenarios to be cost-effective.

Equity

When equity is considered, it appears that improvements in absolute equity are underestimated when the detailed inputs are not used (Figure 20). Once again, the probability of a scenario being equitable appears less sensitive to whether or not detailed inputs are used (Figure 21; note that we present only median values to improve readability).

FIGURE 20.

Analysis 3: equity plane for the comparison of the equity impact of the scenarios. SII, slope index of inequality. Light blue, intervention reduce inequalities and results in positive net monetary benefit; light orange, intervention increase in equalities and reduces net monetary benefit.

FIGURE 21.

Analysis 3: annual probability of equitable policy (using absolute equity).

Strengths of this modelling approach

As presented in Chapter 4, the workHORSE model is an advanced, validated, flexible microsimulation of the dynamics of NCD in a population, including important NCDs that are amenable to prevention, and provides support to a range of capabilities to conduct state of the art effectiveness, cost-effectiveness and equity analysis.

Furthermore, including other diseases in the workHORSE model was valuable. The complex dynamic generated by reductions in CVD mortality needs to be modelled for a more realistic estimate of future costs from a societal perspective, which includes competing causes of illness and death. For instance, the model factors in that if CVD incidence is delayed, people may live longer and be more likely to develop any of the other modelled diseases. This was evident in our scenarios, where the number of COPD cases usually increased because of people living longer.

Limitations

All such modelling analyses have limitations. Several of the assumptions, particularly around the effectiveness of improved lifestyle services, might overestimate what could usually be achieved. However, we would suggest that these assumptions might usefully serve as ‘ideal or optimal’ targets to help the user better understand the maximal potential output of the exercise. Users have the option to perform one-way sensitivity analysis to identify influential assumptions and conduct targeted studies in their areas to collect more evidence and inform the workHORSE model. Using local effectiveness data for these services might be a way to improve the estimates when using the workHORSE model. We conducted a very simplistic assumption of the increased costs and benefits that might be observed over the full time course of the simulation. However, local commissioners might prefer to use the tool to refine the estimates by inputting their cost data.

For analysis 2, we considered modelling a multicomponent intervention as the most promising in increasing uptake, as identified in our umbrella literature review in Chapter 3. However, we decided that this was not feasible, as the review reported qualitative effect size only. In truth, any competent cost-effectiveness analysis would require an intervention that was very precisely costed to support a solid interpretation of the outputs. We believe that local commissioners have these detailed costings and the workHORSE model will be used as a guide for data collection. However, the optimal scenarios still provide a reasonable ballpark figure for the potential effectiveness of such interventions.

Diabetes prevalence can increase in some scenarios, consistent with the underlying population trends in obesity and diabetes, and the increased risk of incident diabetes in those patients using statins. 178,179 We explicitly modelled diabetes, including the associated increased risk of diabetes attributable to statins. It is not possible from our analyses to suggest changes in diabetes prevalence attributable to the interventions modelled, in part because of the small population numbers involved, resulting in considerable uncertainty. An in-depth analysis of this question might merit future exploration.

The costs of medications for CHD, stroke, diabetes and hypertension were all included in the broad health-care unit cost estimates. Although the costs of statins for people with no other diseases were not explicitly included, these costs are small, at around £20–40 per patient per annum. Furthermore, we did not measure or estimate the increased well-being and reassurance that some people might experience from undergoing NHS Health Checks that inform them that they are reasonably healthy. Finally, we, likewise, did not measure the ‘pill burden’ of being prescribed statins or other potential negative impacts that can be associated with interventions that involve inviting the general population to a health programme, such as discomfort, labelling or increased anxiety, or costs associated with overdiagnosis or misdiagnosis. Neither did we factor in private patient costs (e.g. cost of travel or childcare) or earnings lost for individuals to attend their Health Check. However, these are likely to be small. 180

When the impact of an intervention is modest and the local level is the focus of interest, the model results typically have a large degree of uncertainty in the cost-effectiveness of the decision, which will be revealed by large UIs that may cross zero. However, we consider that the median estimation still provides potentially valuable information as an indication of the potential for gains, rather than a conclusive statement on the equivalence between the compared scenarios. 11 Finally, we purposely used data from a publicly accessible repository reporting key performance indicators for the NHS HCP. However, there was no public or easy access to detailed data on local effectiveness of therapeutic and lifestyle referrals. This is concerning, given the considerable heterogeneity in the programme implementation that we found, even among the best-performing LAs (see Chapter 3).

As presented in Table 7, the workHORSE model considers only the benefits from primary prevention. We consider this to be appropriate, as NHS HCP is a primary prevention programme by design. However, one could reasonably argue that some of the interventions NHS HCP introduces may have an impact on the severity of the clinical disease. For instance, reducing the BMI of some participants may not only prevent or postpone T2DM, but, for those who will eventually develop the condition, clinical management would be better. Unfortunately, the argument that people leading healthier lifestyles might get less severe NCDs may be intuitive. However, this argument has weak evidence to support or refute it in most cases. The argument that people leading healthier lifestyles might experience more prolonged survival from NCDs is backed by more robust evidence in comparison, and that would reduce the cost-effectiveness of the NHS HCP from the health-care perspective.

Finally, we have not included a scenario where the different implementations of the NHS Health Checks were compared with other structural interventions that can reduce the burden of NCDs.

Key strategic implementation factors

The workHORSE model represents a different paradigm when compared with the decision-support tools generally deployed for use in LAs:

• The workHORSE model can support powerful basic and advanced interface capabilities.

• The workHORSE model can support scenario design features, supporting analysis of the current and alternative implementations of the NHS HCP.

• The workHORSE model is provided as an open-source computer code with a permissive and copyleft licence. This feature enables external audit and quality improvement, customisation, adaptation and evolution.

However, using the tool requires skills in ‘scenario development’. Scenario development is an activity that results in the development of ‘scenario narratives’ that reflect the intervention being considered, prepares the necessary quantitative inputs to represent those interventions through changing parameters in the model and, finally, decides the scope of the analysis (i.e. in terms of time horizon and outcomes to evaluate), as illustrated in Chapter 5.

The prototype user interface described in Chapter 4 was built as an example of how scenario development can be supported by the tool. Therefore, developing more sophisticated and user-friendly interfaces will be an essential component of any implementation, as the interface will need to be able to adapt to the needs of local teams so that it can be used to represent the programme and policy to be assessed as the programme evolves.

We propose that there are five major areas to consider when strategically planning to implement the tool in an organisation: (1) the technical aspects of the implementation, (2) keeping the model updated, (3) training users in scenario development, implementation and interpretation, (4) the resources required in terms of people and expertise and (5) exploiting the possibilities of an open-source approach to future-proof the model. A summary is available in Table 12.

Training users in scenario development

Training is usually a key factor to improve the user experience with software.

We think that a critical skill is training users to develop and interpret scenarios. This includes understanding some theoretical principles on how the model works and what it can do, plus the user’s knowledge and expertise on evaluation. Our stakeholder group identified this issue as a critical need if the tool is to be widely adopted, alongside appropriate investment to support its use.

Information technology and software engineering

Users would require information technology and software engineering to provide support in the local deployment of the software. Existing infrastructure might support local versions of the model, but it might require adaptation. In addition, maintenance and support might be required as needed, including third-party cloud deployment, which requires expertise in the technologies used (e.g. Docker). We have tested solutions using open-source software, reducing the need for a commercial software licence if the cloud provider protocols can interoperate with workHORSE code. Most cloud infrastructures are commercial and therefore operate on a pay-per-use basis.

The prototype user interface might need adaptation for specific user requirements, changes in the programme and to enable analyses that are not possible with the current functionality.

Modelling expertise

Additional technical expertise would be needed if users wished to update synthetic populations, incorporate new effectiveness and epidemiology data, or redesign the programme. A software engineering skill set is needed, including advanced knowledge of programming in R and C++ (Standard C++ Foundation; URL: https://isocpp.org) languages and advanced quantitative skills (i.e. to implement synthetic population approaches and other advanced statistical operations; see Chapter 4).

Data science expertise

Data science expertise is needed to conduct bespoke analysis and use or update additional data sets, mostly to keep the model epidemiology and parameters up to date. Models become obsolete very quickly (e.g. the initial cost-effectiveness analysis for the NHS HCP), usually because the intervention or the programme that the model addresses is changing.

We developed the workHORSE model to be flexible and accessible so that it can evolve alongside the issues addressed. The scenario parameters reflect the basic operations of a ‘detection and control’ individual-level intervention, primarily health care based, as it involves the prescription of drugs or intervention by a qualified professional at some point.

Specific programme expertise

Programme knowledge and expertise are essential for scenario design and interpretation. The tool provides a degree of flexibility in designing scenarios that is unique, requiring a thorough knowledge of the programme to fully exploit the tool’s capabilities. As shown in Chapter 2, working closely with stakeholders involved in the NHS HCP allowed us to develop specific interface features that are most relevant for local questions and decision-making.

Implementation approaches

Depending on resource availability and the mode of use of the tool by users and organisations, the implementation can be carried out in a ‘central to periphery’ strategy or a ‘local strategy’.

The ‘central to periphery’ strategy could be a centralised effort to develop an implementation programme that is adequately resourced centrally and in charge of the technical aspects. It could provide a bespoke standardised user interface, further maintenance and development of the tool, and training and support to users. The benefits of this strategy will include economies of scale, particularly around sourcing highly skilled staff.

The ‘local strategy’ (i.e. taking responsibilities for all the roles and areas at local or even regional level) will provide more flexibility. By implementing, developing and training local users to better satisfy local needs, it will respond best to the trends in evidence use at the local level in England. However, this will require the development of specific local or regional teams with modelling and analysis capabilities, or outsourcing of these activities with a dedicated budget. Furthermore, the increasing collaborative links between universities, local public health teams and public health training schemes can provide the necessary skills and research capabilities with a robust and local focus.

We estimate that a ‘central to periphery’ strategy can be a reasonable approach to resource the necessary skills, with a core team at a central level that is well resourced in modelling and software engineering capabilities and resources to support dedicated analysts at the local level. Alternatively, the ‘local’ implementation can serve local needs best with a small team composed of an analyst (with a data science background) and funded collaborations with local universities and software engineering providers.

The workHORSE model and code can be used for any of these levels of implementation.

Presentations

Hyseni L, Lloyd-Williams F, Guzman-Castillo M, Collins B, Kypridemos C, Capewell S, et al. Engaging with Stakeholders to Inform the Development of a Computer Model for the NHS Health Check Programme: A Qualitative Study. Cardiovascular Disease Prevention Conference 2019: Saving Hearts and Minds Together Conference, Manchester, UK, 14 February 2019.

Hyseni L, Maden M, Boland A, Kypridemos C, Collins B, O’Flaherty M. PP Umbrella Review of Strategies to Improve Uptake of Screening Programmes. Pitch presentation: 12th European Public Health Conference, Marseille, November 2019.

Posters

Collins B, Lloyd-Williams F, Hyseni L, Guzman-Castillo M, Kypridemos C, Capewell S, et al. Evaluating Stakeholder Involvement in Building a Decision Support Tool for NHS Health Checks: Co-Producing workHORSE. PHE Annual Conference, Warwick University, UK, 10 and 11 September 2019, PHE ePoster Library. 9 December 2019; 274322; 131.

Lloyd-Williams F, Hyseni L, Guzman-Castillo M, Kypridemos C, Collins B, Capewell S, et al. P34 evaluating stakeholder involvement in building a decision support tool for NHS Health Checks: co-producing the workHORSE study. J Epidemiol Community Health 2019;73:A87.

Coverage

• What are the processes around inviting people for NHS Health Checks?

  • Have these changed over time?

  • If so, please tell us (1) how (2) why and (3) when they have changed.

  • How have these different processes impacted on uptake?

  • Can you provide us with the most recent documents (if available also previous versions), including reports and audits produced regarding the coverage of NHS Health Checks and their outcomes, and any published individual case studies?

Take-up

• What are/have been the approaches to increase take-up of NHS Health Checks?

• Have the approaches changed since the start of the NHS HCP?

  • If so, please tell us (1) how (2) why and (3) when this changed.

  • Please report any changes in the % of uptake during this time.

    • If available, please include data by socioeconomic status.

  • Can you provide us with the most recent documents (if available also previous versions), including reports and audits produced regarding the take-up of Health Checks and their outcomes, and any published individual case studies?

• If not included above, what are your approaches to invite patients?

• Have these approaches to invite eligible people changed since the start of the HCP?

  • If so, please tell us (1) how (2) why and (3) when this changed.

  • Please report any changes in the % of uptake during this time.

    • If available, please include data by socioeconomic status.

    • Can you provide us with the most recent documents (if available also previous versions), including reports and audits produced regarding the take-up of Health Checks and their outcomes, and any published individual case studies?

Community outreach

• Please describe your community outreach activities.

Lifestyle services

• Do you have lifestyle referral services? Yes/No.

  • If no, why not?

  • If yes, which referral services do you have?

Cost per NHS Health Check

• What is the cost per NHS Health Check?

You have been chosen for this study because your performance in terms of NHS Health Checks coverage is better than other areas. Is there anything else that either commissioner or provider has done that you believe has worked in terms of maximising NHS Health Checks coverage or uptake?