Vertical integration of general practices with acute hospitals in England: rapid impact evaluation

Challenges for general practitioner primary care

The sustainability of primary care has for several years been the subject of increasing concern and debate for policy makers in England, as well as for health-care professionals, service managers and commissioners. In essence, demand for primary care is growing inexorably and the supply of GPs is not keeping pace with that demand. 19,21–25 GP FTEs in England grew by 6.6% in the 3 years to September 2022: from 34,729 in September 2019 to 37,026 in September 2022. This was a combination of a small decline in FTEs of fully qualified GPs (down 626 over that period from 28,182 to 27,556) and a large increase in GPs in training (up 44.6% over the same period from 6547 to 9470). 9 Over the same period, the total number of appointments at general practices increased by a similar amount: up 7.1% over the same 3-year period from 26,420,000 in September 2019 to 28,300,000 in September 2022. 26,27 The similarity between the two growth rates is perhaps unsurprising in a supply-constrained system (i.e. where potential demand exceeds available supply) given that the number of GP appointments is necessarily constrained by the number of FTE GPs available for patients to consult with. But it also suggests that employment in primary care practices of health-care professionals other than GPs did not greatly add to capacity for patient consultations over that period. 28

There has also been a rise in the number of patients with multiple long-term conditions, higher costs, developments in the consulting technology and tightening workforce constraints as result of GP recruitment and retention difficulties. 25,27,29,30 The gatekeeping role of GPs means that increasing difficulty with meeting patient demand in a primary care setting can lead to growing pressure on acute hospital emergency services. A desire on the part of managers at acute trusts to keep some control over the flow of patients to their hospitals was evident in our phase 1 evaluation. 2

The policy response has focused on increasing the numbers of GPs in training, as evidenced by the rapid growth in numbers of trainees mentioned earlier; on providing funding for new staff roles in general practice and on merging and integrating primary care practices horizontally with one another in the hope of increased efficiency (more and/or better patient services delivered per primary care staff member). 2,19,22,31 Nevertheless, patient experience and patient satisfaction with general practice services in England is falling, as summarised in the House of Commons Health and Social Care Select Committee report in October 2022 (quote taken from paragraph 25 of that report):25

Over several years the GP Patient Survey has shown declining access standards, albeit with satisfaction rates remaining high: from 2018 to 2020 the proportion of people who reported having a good overall experience of making an appointment fell slightly from 68.6% to 65.5%, but the proportion of people rating their care as good overall remained 81.8% in 2020. In the latest GP Patient Survey, however, the results are significantly worse and show the level of difficulty patients now face when trying to access general practice: the proportion of people who had a good experience of making an appointment has fallen sharply to 56.2% and the proportion of people rating their overall experience as good has also fallen significantly to 72.4%. 23,25 (Parliamentary Copyright House of Commons 2022. This material is reproduced under the terms of the Open Parliament Licence v3.0, which is published at: https://www.parliament.uk/site-information/copyright-parliament/open-parliament-licence)

Integrating health care

NHS policy has, over the past 10 years, repeatedly stressed an objective of more integrated patient care across primary health care, secondary health care and social care settings. 19,21–23 The policy focus on care integration led to the development of several recommendations in the 2014 NHS ‘five-year forward view’. 21 Among them was to seek stronger integration between primary and secondary care, in what were termed primary and acute care systems (PACS). Some piloting of PACS was funded as a result. The 2016 General Practice Forward View announced increased funding for GP services and plans for developing them further. 22 The 2019 NHS Long Term Plan19 set out the intention that all general practices in England should come together to deliver services as part of primary care networks (PCNs). This is a form of horizontal integration designed to cover populations of 30–50,000 patients per network. Since July 2019, all but a tiny number of practices have become horizontally integrated in that way with other practices, while remaining separate legal entities with separate contracts. Findings from a recent rapid evaluation of PCNs show that there have been a number of facilitators and challenges to horizontal integration to help achieve sustainable primary care, address growing workload issues and improve the availability and coordination of local primary care services. 10

Since 2015, a number of local initiatives have led to trusts running general practices but this was not in response to any particular policy stimulus. Apparently explicit policy-maker interest in NHS trusts running general practices emerged in early 2022. 32,33 The expectation was reported as being that functions, activities and organisations that provide different levels of patient care might be brought together under a single management and result in cost savings in secondary care, as had been reported at one vertical integration site by Yu and colleagues. 2,3,34 This policy-maker interest was despite the findings from our phase 1 evaluation that vertical integration in a health-care system worked when driven by local initiative (rather than central direction) and that many GPs evidently did not wish to become vertically integrated with trusts. 2,3

A subsequent ‘stocktake’ commissioned by NHS England of the state of integration in primary care and published in May 2022 made no mention of vertical integration as such. Instead, it focused on ‘building integrated teams’ at the level of PCNs but broadened the concept to:

teams from across primary care networks (PCNs), wider primary care providers, secondary care teams, social care teams, and domiciliary and care staff [working] together to share resources and information and form multidisciplinary teams (MDTs) dedicated to improving the health and wellbeing of a local community and tackling health inequalities. (Source: NHS England 2022, p. 6)31

Caring for people with multiple long-term conditions

The National Institute for Health and Care Excellence (NICE) defines multimorbidity as ‘the presence of two or more long-term health conditions, which can include:

• defined physical and mental health conditions such as diabetes or schizophrenia

• ongoing conditions such as learning disability

• symptom complexes such as frailty or chronic pain

• sensory impairment such as sight or hearing loss

• alcohol and substance misuse.’35

The prevalence of people with two or more such long-term conditions has recently been estimated at 53% of the adult population in England. 36 People with multiple long-term conditions are higher users of NHS services than other members of the population and would particularly benefit from services focused around the individual patient rather than an individual condition. 24 The World Health Organization describes people with multimorbidity, compared with other members of the population, as facing:

more frequent and complex interactions with health-care services leading to greater susceptibility to failures of care delivery and coordination; the need for clear communication and patient-centred care due to complex patient needs; demanding self-management regimens and competing priorities; more vulnerability to safety issues … (p. 4)37

Multimorbidity is now a high-profile issue for consideration in NHS policy and practice. It is a major theme in the NHS Long Term Plan18 and the Academic Health Science Network, National Institute for Health and Care Research (NIHR) and NHS England report on priorities for innovation and research. 18 In July 2019, the Academy of Medical Sciences, NIHR, Department of Health and Social Care, Medical Research Council and Wellcome Trust jointly declared that: ‘Multimorbidity is recognised as an important priority across all our organisations and we all take a special interest in it’. 38

Evidence for vertical integration

It remains the case that relatively little evidence is available to inform policy makers about vertical integration between acute hospitals and general practices. Yu and colleagues34 found in a statistical study at one location, Wolverhampton in the West Midlands of England, using data from the period April 2015 to March 2019, that vertical integration between an acute hospital and 10 general practices was associated with a statistically significant, reduction in the rate of unplanned hospital admissions and readmissions, but not of A&E attendances. They estimated that this would imply an annual cost saving to the hospital of around £1.7 million. 34 We include in Chapter 4 a comparison of Yu and colleagues’ findings with our analysis for the same vertically integrated practices in Wolverhampton.

Our qualitative, phase 1 rapid evaluation of vertical integration at two locations in England and one in Wales revealed several noteworthy findings. It showed that key to achieving vertical integration is better clinical integration (coordination of treatment services for a patient) and functional integration (strengthening key support functions, such as financial management, human resources and strategic planning). Trust managers at case study sites anticipated that vertical integration would enable better management of patient demand for secondary care services, a view consistent with the findings of Yu and colleagues. 34 Vertical integration can lead to alterations in contractual arrangements and accountability, workforce recruitment, premises and care pathways, which in turn have the potential to create better care and outcomes for the patient. The phase 1 study also found potential downsides to vertical integration, including fears among GPs of loss of autonomy and having to interact with more bureaucratic and slower back-office processes in trusts (finance, procurement, human resources, estate management). It is certainly the case that even in locations where some general practices are integrated with an acute trust, many other general practices choose to remain outside any such arrangement. 2

Work package 1: understanding the current scale of vertical integration in England

WP1 details the extent of vertical integration in England through a desk-based analysis of secondary care statutory financial reporting and primary care GP workforce data. There is triangulation of practices where vertical integration has been identified. Statistics that describe the characteristics of the vertically integrated practices are presented: number of acute hospital trusts managing general practices; the number of general practices managed by the acute hospital trust; practice sizes in terms of patient population, patient demographics and workforce descriptors.

Work package 2: development of a statistical analysis plan

As part of WP2, the study team identified appropriate control sites and an appropriate approach to coding multiple long-term conditions as part of a detailed plan of analysis. In addition, this WP addresses several methodological questions, for example how to deal analytically with general practices that have merged with other practices during the study time frame, and how to allow for the impact of the COVID-19 pandemic in the analyses of how patients use secondary care.

Work package 3: quantitative analysis of the impact of vertical integration on the use of secondary care services

As part of WP3, we assessed the impact of vertical integration on patients’ use of secondary care services both overall and more specifically for people with multiple long-term conditions. We examined the following outcomes: outpatient attendances, A&E attendances, all inpatient admissions, emergency inpatient admissions, inpatient admissions for ACSC, bed days, readmission within 30 days of discharge; for the identified practices and their controls before and after the identified practices were vertically integrated. We then discuss the financial implications in terms of an overall cost difference for secondary care use between vertically integrated and not vertically integrated practices. We additionally analysed the results of the GPPS questions relating to patient experience of primary care, comparing replies for vertically integrated practices with those for control practices.

Work package 4: staff and patient experiences of care delivery and provision across three purposively selected case study sites

We completed focus groups and interviews across three case study sites in England with key service managers and clinicians from the acute hospital, community care and general practices; and primary qualitative research via interviews, capturing the views of patients from integrated general practices, to understand their experiences of accessing services in areas where models of vertical integration are present.

We present more detailed methods and our findings in the next three chapters. Chapter 3 provides a description of the scale of vertical integration of primary care practices with acute and community trusts that has taken place across England. The subsequent chapter (Chapter 4) presents a quantitative analysis of national routine data to explore the impact of vertical integration on patient satisfaction and use of health-care services. Chapter 5 sets out an analysis of interviews and focus groups with key stakeholders and patients across three case study sites to explore qualitatively the impact of vertical integration on patient experience of care, particularly focusing on patients with multiple long-term conditions.

Stage 1a: secondary care statutory financial reporting

The principal source of information on vertical integration where trusts run general practices, was trusts’ annual reports. We hand searched the annual reports of all 230 NHS trusts listed on the nhs.uk website. 40 Each NHS foundation and non-foundation trust must publish annual reports and accounts to allow scrutiny of the year’s operations and outcomes. Trust annual reports are typically published in a PDF format on the trust’s website. 41 We first searched the reports from 2020/21 (the most recent available) and then backed this up by searching the reports from 2019/20, which revealed several additional vertical arrangements not evident from the later year’s reports.

One member of the research team (FO) used the ‘find and retrieve’ search option for the terms ‘general practitioner’, ‘GP’ ‘general practice’ and ‘primary care’ within each annual report and determined from the corresponding text whether the trust had reported a financial/ownership relationship with a provider of GP services, and thus would be considered vertically integrated with that practice or practices. For trusts with no vertically integrated practices, our expectation was that we would find less frequent use of the search terms in the report. We found that the search terms were frequently mentioned in the pension reporting sections but with no reference anywhere else in the report in question that indicated that the trust owned or managed any general practices. These were, consequently, not instances of vertical integration. The researcher (FO) was otherwise independent of the study and had no prior knowledge of vertically integrated practices, which helped to ensure an unbiased assessment of where vertical integration was present. All trust reports with identified vertically integrated general practices were then reviewed for a second time by another member of the research team. A further 20% of trusts identified by the first reviewer as not having vertically integrated practices were also second-reviewed as a control. The second review confirmed that these trusts were not integrated with general practices.

We checked the websites of practices identified by this process to confirm their connections with NHS trusts. We additionally searched via Google for press mentions of trusts taking over the running of general practices. Finally, we cross-checked the findings of the blinded reviews against study team knowledge of areas where vertical integration is occurring and found that all instances of vertical integration previously known to members of the research team had been identified, plus some further instances.

Stage 1b: validation via primary care general practitioner workforce data and practice websites

A key finding from our phase 1 evaluation2 was that vertical integration was a way of sustaining local primary care, and one form of vertical integration was for a trust to employ GP staff directly. We therefore inspected the GP workforce data published by NHS Digital at 31 March 2021,42 which reports all types of GP employment contracts. Our focus for searching was on GPs categorised as ‘other’ (i.e. not salaried GPs, GP partners, trainees or locum practitioners). A preliminary review of these data identified 243 practices where one or more ‘other’ GPs were employed. We then searched on the practice names to find and review the corresponding individual practice websites for further indications of possible vertical integration with a trust. Practice names were cross-checked against information provided by NHS Digital. 43

Further robustness checks were provided by reviewing all general practices with alternative provider medical services (APMS) contracts on the UK government contract finder website to identify those where acute or community trusts hold the contract. 44 This revealed no further examples of vertical integration for our analysis. We also cross-checked our findings of vertically integrated practices with the Care Quality Commission register of general practices39 filtered by ‘location primary inspection category’ and ‘provider type/sector’.

Stage 2: finalisation of linked dataset

The second main stage in the identification process was to return to the annual reports of those trusts that had been found to be running general practices, to enable identification of practice level details including the individual practice code (the ‘organisation code’ which identifies any NHS organisation uniquely). The practice code is required to be able to link the study outcomes (described in Chapter 4) with the vertically integrated practices. We also needed to identify the practice name and the date that it became vertically integrated into a trust. Practices entering a vertical integration arrangement with a trust are undergoing an organisational change and may consequently have a change in their practice code. We found that there were sometimes practice mergers occurring around the time that practices became part of vertically integrated organisations. To help us track which preintegration practices had merged into which post-integration practices and when that happened, we took note of information about the sites (geographical locations) from which vertically integrated practices were working. A site becoming a location for a practice with a different practice code than previously, is an indication that the practice working from that site has merged with another. All this information gathering entailed hand-searching the annual reports of each of the vertically integrated trusts, starting with the 2020/21 annual report in each case and then working backwards preceding year by preceding year, noting all mentions of general practices and/or practice sites being vertically integrated (and whether they were merged at any point or left the vertically integrated organisation at any date) until the date of the first vertical integration event (the first general practice whose management was taken over by the trust).

Details on models of integration at the practice or trust level are not always stated clearly in trust annual reports, specifically the date when the change happens and what happens to the practice codes at that date. Where practice code, name and date of integration were not mentioned or were otherwise unclear in trust annual reports, the following detailed search process was followed in hierarchical order:

1. Visited individual practice websites and carried out online searching for date of entry into vertical integration; we then visited local and national media sources.

2. Used NHS Digital ‘GP and general practice related data’ sources for two separate years (2022 and 2016; 2022 was chosen as the first year to identify practice codes as it was the most up-to-date version available. We also used an archived version from 2016 as being around the start of vertical integration and would enable us to identify any practice codes which might since have changed due to mergers, closures and so on) to identify the practice code using the data files for both GP practices (epraccur) and branch surgeries (ebranchs). 43,45–47 The datasets include all general practices active in that year. We searched on practice name to match to practice code. Where there was uncertainty, practice address was used for confirmation of a match.

3. Contacted trust or practice contacts, where research team members had those, to ask for practice code and date of entry into vertical integration.

During our preliminary work we identified that horizontal integration (where general practices join with one or more practices) would be a methodological challenge for the quantitative analysis in our study. We therefore undertook a two-step approach to identifying where horizontal general practice mergers had taken place.

In step 1, we used person-level outpatient data from the Hospital Episode Statistics (HES), held under the BRACE data sharing agreement. We used data from the financial year April 2013/14 to the financial year April 2019/20 (i.e. April 2013–March 2020). In each year, we selected practices with at least 1000 patients and then identified practices in the following financial year where over 50% of patients from the previous financial year were now registered. For example, if 100 individual patients with the same practice code in the HES data for 1 year are again treated in hospital the following year but more than 50 of those patients all now have the same practice code, that is different from the practice code they had the previous year, then we may reasonably assume that the first practice has been merged into the second. We also looked in the data set from April 2018/19 for practices where the majority of patients were registered previously and cross-checked the practice codes from the two approaches.

In the second step, we used practice-level data on branch surgeries listed by the NHS Organisation Data Service via NHS Digital40 and matched current practice codes for branch surgeries with historic practice codes for practices at the same postcode (where a postcode was unique by practice) but which no longer had registered patients. This would indicate that a surgery that was previously the base for one practice has become a branch surgery for a practice based elsewhere and thus is likely to have merged with that practice.

Once we had identified the general practice codes before and after vertical integration of all practices integrated vertically with trusts, whether acute or community trusts (by our definition, as described in Chapter 1) we retrieved data on practice characteristics from general practice level data sources provided by NHS Digital at the national level. Further details are provided in the following ‘data and sources’ section. The datasets were extracted as raw CSV files from the primary source and imported into STATA^® Se version 17 (StataCorp LP, College Station, TX, USA). The required variables from each of the data sources were then merged, matching on the practice code. The dataset we had created detailing vertically integrated trusts and their general practices was then merged into the practice characteristics dataset, again using the practice code as the identifier.

Data and sources

We constructed a cross-sectional dataset for March 2021 for all general practices in England and with all the relevant characteristics of the sites from which the vertically integrated practices operate, as follows.

1. Practice code – an organisation code that uniquely identifies each general practice in England.

2. VI – a binary variable which has a value of 1 if it is a vertically integrated NHS trust or general practice. Identified from the search process described earlier.

3. NHS trust type – NHS trusts are publicly owned organisations providing health care to NHS patients and are either NHS foundation trusts or NHS trusts. For the purposes of our work, we refer to them collectively as ‘trusts’. Trusts can take different forms: acute, community, mental health – and any combinations of those – or ambulance trusts. For the purposes of our study, we define an acute hospital trust is one providing acute hospital-based secondary health care, whether solely that or in combination with community and/or mental health services. An acute trust can have more than one acute hospital site within it. In March 2021, there were 230 trusts in England, with 144 of them being acute hospital trusts. 41 For our analysis, we split the trust type into a binary variable with a value of 1 if an acute hospital trust or a combined trust including acute hospital services, and a value of 0 if any other type of trust. For brevity we label this second category as ‘community trusts’, as explained in Chapter 1.

4. Date of vertical integration – This is the date that the first general practice was integrated with that trust. This is retrieved from the primary data source: the trust annual report or the general practice website.

5. Practice list size – This is defined as the number of patients who are registered at a general practice on the first day of each month. These data are nationally available at the practice level via the NHS Digital website. 48 For the purpose of this analysis, the list size data are retrieved for March 2021.

6. Contract type – general practice services are contracted by NHS commissioners (until July 2022, these bodies were known as CCGs; since July 2022, NHS commissioners are ICBs) following national guidelines for the practice to provide general medical services for the population registering with them, who are typically resident locally to the general practice’s premises. Most practices are run by at least two, and usually more, GPs alongside other health-care professionals such as practice nurses and physiotherapists who work within the same building. One or more of the GPs may own a stake in the practice business, although practices can also be owned by companies and via other ownership models. 8 Each practice holds an NHS contract to run the practice, of which there are three different types as described in Box 1 in Chapter 1: general medical services (GMS), personal medical services (PMS) and APMS. We obtained the contract type for each practice from NHS Digital ‘NHS Payments to General Practice’ for 2021/22. 49

7. GP workforce – We obtained the number of FTE GPs at the general practice level using NHS Digital data on GP workforce. 42 We also retrieved data on the number of GPs funded by ‘other’ as defined by NHS Digital (i.e. those GPs who are neither salaried, nor partners, trainees or locum practitioners). We did this in case GPs employed by trusts would be recorded in this category.

8. Quality and Outcomes Framework (QOF) score – Quality and outcomes indicators are agreed as part of the NHS GP contract negotiations every year. These indicators have points attached that are awarded according to how well general practices do against them, and so the QOF score is a proxy measure for the quality of the practice. We retrieved data on practice level QOF scores from NHS Digital,50 and specifically whether each practice achieved a QOF score within 25% of the total quality points achievable.

9. Index of Multiple Deprivation (IMD) code – The IMD is the official measure of relative deprivation for small areas in England based on seven different domains to produce an overall relative measure of deprivation. 51 It provides a rank from 1, which is the most deprived area to 32,844, the least deprived area. Using the ranks, decile 1 represents the most deprived 10% of areas nationally and at the other end of the scale, decile 10 represents the least deprived 10% of areas nationally. We report the percentage of practices in the most deprived decile as of March 2021. We obtain IMD codes for the postcode of each general practice in our dataset using the ‘English indices of deprivation’ 2019 online tool. 51

10. Rural or urban classification – We distinguish rural and urban areas, where the classification defines areas as rural if they fall outside of settlements with more than 10,000 resident population. 52 We retrieved data on the rural or urban classifier through the NHS Digital, ‘NHS Payments to General Practice for England dataset 2019/2020’,49 which we then linked to our practice level dataset using the practice code.

Number of vertically integrated general (medical) practitioner practices

As of March 2021, we identified 26 trusts in vertically integrated organisations, running a total of 85 general practices (i.e. with unique practice codes) across a total of 116 general practice sites (as some practices work from two or more locations). The locations of these vertically integrated practices are scattered across England (see Figure 3). There are concentrations of vertically integrated practices in several locations, but with no particular geographical pattern, although we note that there are few instances in London.

FIGURE 3.

Locations of vertically integrated general practices in England, March 2021.

We initially identified 26 vertically integrated trusts from financial year 2020/21 trust annual reports and a further 7 from the 2019/20 reports. Further in-depth checking of these 33 annual reports by a second researcher, confirmed that 7 of the total number of apparently vertically identified trusts from this search process were not in fact cases of vertical integration. The reasons why these trusts were incorrectly identified was most commonly because the trust was either in a joint working arrangement with a general practice or was using primary care premises for some trust activity. The final total number of vertically integrated trusts identified from the annual reports identified in Stage 1a of our analysis was thus 26.

Stage 1b confirmed many of the instances of vertical integration that we found by inspecting trust annual reports and did not reveal any additional instances. The resulting list of vertically integrated trusts and general practices included all of those that had been known to the members of the research team (which includes two researchers who had taken major roles in the phase 1 evaluation of vertical integration) prior to the analysis. This triangulation encourages confidence that the list of 26 vertically integrated trusts with 85 practices operating from 116 sites obtained from the search of trust annual reports is robust and reliable.

During the search process, it became evident that there is not a single criterion for identifying vertically integrated general practices, and that identification can prove complex, for example:

• One general practice we identified is co-owned by both an acute and a community trust and is included in our dataset as owned by the acute trust only.

• Five practices owned by acute hospitals and identified in this process were serving populations of homeless people or provided services for patients who are otherwise unable to access primary care and had been running in a similar structure for over 20 years. These practices are excluded from the analysis in this chapter.

There is also heterogeneity in the ownership arrangements for practices who become vertically integrated; for example, practices may undergo vertical integration at the same time as the acute or community trusts takes on the APMS contract. For some practices within a vertically integrated model, their GMS contracts are retained, although for some they are not. Ownership of practices can be taken on directly by the trusts or in some cases the trust sets up a wholly owned subsidiary company, which then owns the vertically integrated practices within the trust. However, the biggest challenge was in identifying the practice codes and the date of which the vertical integration of a practice took place (as detailed in the methods section).

Our preliminary scoping work revealed that general practices may undergo several stages of reorganisation as part of the vertical integration process, for example initially becoming a vertically integrated practice, and then undergoing horizontal mergers with other vertically integrated practices within the same trust. Although horizontal mergers have been a feature of primary care nationally, it has become evident that they are also a particular feature of vertically integrated general practices. The fact that practice codes within vertically integrated practices are particularly susceptible to change has meant that being able to reliably identify horizontal integration as well as vertical integration of primary care is an important methodological step for this evaluation (Appendix 1 provides further details of the process of identifying horizontal integration and the application to vertical integration of general practices into NHS trusts).

Data

Outcomes

Intervention and control practices

In primary care, vertical integration is one of several new models that have developed and spread in the 2010s. Horizontal merger of primary care (general practice mergers) is a second model, where previously separate practices merge to work together in larger organisational units. Other new models include private ownership of primary care practices, either directly or through umbrella companies, and the use of APMS contracts. To evaluate the impact of vertical integration, and particularly in terms of identifying the most appropriate counterfactual, it is important to be able to distinguish the effects associated with vertical integration from these other changes in primary care in England that have occurred over the last decade. Our findings from WP1, identifying and describing where vertical integration is occurring, highlighted that vertical integration general practices across England are also more likely than other practices to have merged horizontally and are more likely to have APMS contracts.

An impact evaluation tries to understand whether what has happened in the intervention group after the intervention has occurred is different from what would have happened had the intervention not taken place. In randomised studies this is controlled by assigning intervention and control groups at random. However, in real world implementation of health-care service change (including vertical integration) this is not possible. Rather there is sequential roll-out and with selection into the intervention rather than random allocation. Consequently, thought is required to identify an appropriate control. For example, we know from the phase 1 evaluation of vertical integration that practices often struggled with staffing before switching to a vertical integration model,2 and this is the kind of context it is important to try to match on, but that is not possible in practice because information on the extent of difficulties with staffing is not routinely available.

Data driven matching (such as propensity score matching) and synthetic control approaches are two approaches to addressing this issue. They try to find the most similar set of controls for an intervention group that cannot be randomised, and for which no obvious control can be identified, and do this by matching the trends in the outcomes of interest in the time period before an intervention (in this study, the switch over of a general practice to a vertically integrated approach). Using synthetic controls67 can give different results compared with analyses of the same intervention using alternative (difference in difference) approaches. 68,69 Synthetic control approaches are designed to better match pre-intervention trends and have been found to be less biased in simulation analyses when some of the assumptions about changes in both the intervention group and the control group over time before the intervention takes place are not met. 70 A synthetic control approach was used in the analysis of vertical integration between an acute hospital trust and general practices in Wolverhampton, England. 34

We concluded that the use of matched control practices (such as propensity score matching or use of generalised synthetic controls) would not be feasible for this evaluation of the impact of vertical integration nationally, as these approaches depend on matching stable practices over time. By contrast, vertical integration practices are more likely than other practices to merge horizontally, or to switch to APMS contacts, and we wanted to be able to distinguish between all these changes explicitly in our analyses. We used publicly available annually reported data from NHS Digital to characterise changes in general practices’ contract types over time, taking the date of the first change during the study period as the date of the intervention. We identified general practice horizontal mergers for both vertically and non-vertically integrated practices through a two-stage approach, tracing the general practices of registered patients attending hospital outpatient services, where the practice codes of those patients subsequently disappeared. We additionally looked at the postcodes of practices that had closed, exploring whether their postcodes were linked to practices which are still open as branch surgeries, and identified these also as mergers (see Report Supplementary Material 1).

Hence, in all analyses we consider four different groups of general practices:

• all vertical integration practices and additionally, separately:

  • practices vertically integrated with acute trusts

  • practices vertically integrated with community trusts

• non-vertical integration practices that have undergone a horizontal merger

• non-vertical integration practices that have changed to an APMS contract

• GMS practices that have not changed contract, undergone a horizontal merger, or vertically integrated.

We therefore carried out three separate analyses for each outcome, describing the impact of vertical integration, of horizontal merger (without becoming vertically integrated and not switching to APMS contracts), and of switching to APMS contracts (but not undergoing horizontal merger or vertical integration) compared with a random sample of general practices with GMS contracts that did not change or merge at all during the study period using a difference in difference framework, adjusting for differences in pre-intervention trends as part of the modelling framework. Terminology varies slightly across disciplines: in our analysis the pre- and post-intervention slopes are modelled (rather than the assumption of parallel trends being tested) and this is sometimes alternatively known as a controlled interrupted time series approach.

We included an additional sensitivity analysis considering only vertical integration practices and compared the impact of the introduction of vertical integration with acute hospital trusts on the one hand with the introduction of vertical integration with community trusts on the other. This, by design, matches preintervention period in practices undergoing vertical integration, as only vertical integration practices are included, but allows the specific impact of integration with acute hospitals to be tested, compared with practices that have undergone vertical integration but with community trusts rather than acute hospital trusts.

The geographical distribution of included intervention and control practices are presented in Figure 5.

FIGURE 5.

All included vertical integration practices (X) and APMS controls (orange), stable GMS practices (light blue) and horizontally merged practices (green).

Multimorbidity

Analysis framework

In this section, we describe our preliminary analysis work, the main analysis framework for all outcomes, and the analysis details separately for GPPS and HES.

Included practices

Building on the work from WP1, we present the numbers in England of vertical integration practices, horizontally merged practices and practices that switch to an APMS contract in Table 8; the final numbers of practices included in our analyses are then shown in Table 9. After excluding practices with shorter follow up than the required 2 years before and after the date of vertical integration, we are left in the final analysis with 59 practices (becoming 42 merged practices by February 2020) that had switched to a vertical integration approach; 33 vertically integrated with acute hospital trusts and 26 with community trusts. Correspondingly, we included 42 practices that switched to an APMS contract and 964 practices that merged horizontally (becoming 331 merged practices by the end of the study).

Analysis 1 – characteristics of patients at vertically integrated, horizontally merged, alternative provider medical services and control practices

In analysis 1, we describe the characteristics of patients at vertically integrated and non-vertically integrated practices using survey data from GPPS. In the same way as for our findings in WP1, patients at vertical integration practices (particularly practices vertically integrated with community trusts) are more likely to be living in the most deprived parts of the country. Levels of multimorbidity are similar across all intervention and control practices (see Table 10).

Analysis 2 – evaluation outcomes during the year April 2014–March 2015 (pre-intervention in all practices)

In our second set of analyses, we described the patient experience and outcomes from the use of secondary care services during the pre-intervention period in all groups of practices (see Table 11). Patient experience outcomes (particularly continuity) were lowest in practices that went on to become vertically integrated with community trusts or switched to APMS contracts, and highest in control and horizontally merged practices. A&E attendance rates were highest in practices that went on to be vertically integrated with community trusts but similar across all other groups of practices. Outpatient attendance rates and inpatient admission rates were lowest in practices that switched to APMS contracts but were similar across all other types of practices.

Analysis 3 – unadjusted changes over time before and after the intervention in practices that vertically integrate, practices that merge horizontally, and practices that switch to alternative provider medical services contracts

For the unadjusted (descriptive) analyses of patient experience in the 2 years before and after the date of the general practice changes in, outcomes declined for all groups in a manner that was consistent with national trends. 76 In general, consistent with the pre-intervention period, patient experience was lowest both before and after the change, in those practices that switched to APMS contracts (see Figures 6–11). No outcome showed a large step change after the vertical integration, horizontal merger or switch to an APMS contract by the practice.

FIGURE 6.

Helpfulness of receptionists.

FIGURE 7.

Ability to get through on the phone.

FIGURE 8.

Time taken to see or speak to a GP.

FIGURE 9.

Overall experience of access.

FIGURE 10.

Continuity.

FIGURE 11.

Overall satisfaction.

For the seven secondary care utilisation outcomes (see Figures 12–18) differences between the three intervention groups were least for length of stay in hospital (see Figure 18) and greatest for rates of emergency inpatient admissions (see Figure 16) and A&E attendances (see Figure 12). For no outcome was there a ‘step change’ after the date of the intervention. Secondary care utilisation before and after the intervention was highest among vertical integration practices and lowest among practices that underwent horizontal mergers.

FIGURE 12.

Accident and emergency (department) attendances.

FIGURE 13.

Outpatient attendances.

FIGURE 14.

Inpatient admissions.

FIGURE 15.

Inpatient admissions for ACSC.

FIGURE 16.

Emergency inpatient admissions.

FIGURE 17.

Emergency inpatient readmissions.

FIGURE 18.

Length of stay.

Analysis 4 – impact of vertical integration, horizontal practice mergers and switching to an alternative provider medical services contract on patient experience in primary care and on the use of secondary care

The results of these analyses are presented in Table 12 (patient experience) and Table 13 (use of secondary care services).

Analysis 5 – comparison of the impact of vertical integration with acute hospital trusts compared with vertical integration with community trusts, for both patient experience and secondary care use

We present our findings from analysis 5 in Table 14 for both patient experience and secondary care use. We find no evidence of a differential impact of vertical integration with an acute hospital trust compared with vertical integration with a community trust for any of the six patient experience outcomes.

We find that the rate of A&E attendances differentially falls by 3% more in vertical integration with acute hospital trusts (IRR 0.97, 95% CI 0.95 to 0.99; p = 0.01) compared with vertical integration with community trusts, after the practices switch to vertical integration. Conversely, the rate of inpatient admissions increases by 4% more in vertical integration with acute hospital trusts (IRR 1.04, 95% CI 1.01 to 1.06, p = 0.0095) compared with vertical integration with community trusts (see Table 14).

Although we find no differential impact between the two types of trusts at the date of vertical integration with respect to outpatient attendances, we note that in practices that vertically integrate with acute hospitals the post-integration rate of outpatient attendances does increase over the next two years at a considerably faster rate than in practices that integrate with community trusts (IRR 1.13, 95% CI 1.03 to 1.24, per year; p = 0.013; see Table 14).

Analysis 6 – impact of vertical integration for people living with multiple long-term health conditions

Overall, before a practice vertically integrates, we found that people living with multiple long-term health conditions, relative to other patients, reported better experience of the helpfulness of general practice receptionists, ability to get through on the telephone and continuity, but with poorer experiences overall of access to primary care, and with mixed (although small) differences for time taken to see or speak to a GP and for overall satisfaction (see Table 15).

We find that when practices vertically integrate with trusts any differential impact for people living with multiple long-term health conditions compared with people living with a single condition or no conditions, is either not significant or of very small magnitude for the measures of patient experience.

Sensitivity analysis

In our sensitivity analysis considering only practices from Wolverhampton, our results were broadly consistent with the main analysis including all vertical integration practices, albeit with smaller effect sizes overall and often lacking statistical significance, reflecting the smaller sample size. We note that when considering vertical integration practices in Wolverhampton alone we do not see a statistically significant overall impact of vertical integration on the rate of A&E attendances, outpatient attendances, inpatient admissions or length of stay (see Table 17). However, in contrast with the analysis including all vertical integration practices (see Table 13), the impact of vertical integration on the rate of emergency readmissions is a larger reduction than for the main analysis: a 12% reduction in Wolverhampton (IRR 0.88) compared with a 5% reduction (IRR 0.95) nationally.

Summary of findings

In this quantitative analysis, we explore the impact on patient experience and secondary care use of a practice becoming vertically integrated compared with a control group of practices that have not undergone organisational change.

We contrast this impact with the impact of horizontal mergers between general practices and with the impact of switching to APMS contracts not associated with vertical integration. We additionally compare practices that vertically integrate with acute trusts to practices that vertically integrate with community trusts.

We found that the impact of vertical integration on patient experience was small, except for the effect on continuity where there was a 6.6% points worsening of patient experience. Horizontal mergers have minimal impact on patient experience, while switching to an APMS contract was associated with slightly worsening patient experience for continuity, access and satisfaction (i.e. for three of the six included outcomes). Patients of practices switching to APMS contracts reported the worst (albeit still small) declines in their experience in all categories, compared with vertically integrated or horizontally integrated practices.

Regarding use of secondary care services, we found that vertical integration was associated with a small reduction in rates of A&E attendances (−2%), outpatient attendances (−1%), emergency admissions (−3%) and emergency readmissions (−5%) after it was introduced, with no impact on length of stay, overall inpatient admissions or inpatient admissions for ACSC. The falls in A&E and outpatient attendance rates appeared to be temporary, as the rates resumed growing at faster rates than for practices that held GMS contracts throughout the study period. We additionally found small reductions in secondary care use after non-vertically integrated practices underwent horizontal mergers for A&E attendances, outpatient attendances, emergency inpatient admissions and ACSC admissions; with no impact on readmissions. When practices switch to APMS we found no impact on secondary care use except for an increase in inpatient admissions. For none of the interventions was there any impact on length of stay.

We found little difference between the impact of vertical integration with acute hospital trusts and vertical integration with community trusts, although we note that in the 2 years following vertical integration, outpatient attendances increased faster in practices that had merged with an acute hospital trust. Differential pre- and post-intervention trends varied between outcomes but, in general, differential yearly changes in use were small and not in consistent directions. We found no evidence that the impact of vertical integration was differential for people living with multiple long-term health conditions compared with people with a single or no long-term health conditions.

How findings relate to previous research

Our findings are broadly consistent with an earlier statistical evaluation of vertical integration in Wolverhampton;34 and we extend their work by considering a larger number of vertical integration sites, longer follow up, and additional outcomes, both patient experience and secondary care use. When considering only outcomes used in this earlier evaluation, we found overall (small) reductions in inpatient admissions, emergency readmissions and A&E attendances when practices changed to a vertical integration approach. This impact on A&E attendances was not found by the earlier Wolverhampton analysis. Our sensitivity analysis considering Wolverhampton practices alone suggests that this is a feature of vertical integration in Wolverhampton rather than of methodological differences between the two evaluations, as we similarly did not find an impact on A&E attendances in these Wolverhampton practices when considered in the sensitivity analysis.

This highlights that the impact of vertical integration is likely to be heterogeneous across the country. Two early evaluations of APMS contracts found that practices performed more poorly on measures of clinical quality and patient experience compared with GMS and PMS practices. 77,78 Previous research suggests that continuity may reduce when general practices undergo horizontal mergers,79 and although we did find a small reduction, we note that this was not as large as the reduction in continuity associated with vertical integration or when practices switch to an APMS contract.

Research questions

Our research attempts to address gaps in the literature regarding how service delivery and coordination changes following the implementation of vertical integration and the patient experience of vertical integration, with a particular focus on whether patients with multiple long-term conditions are affected differently from other patients. This WP aimed to answer the following research questions:

1. 3: What impact is vertical integration having on the patient journey with regard to access to and overall experience of care?

   1. 3.1: How do models of vertical integration support patient transitions from primary care to acute care?

   2. 3.2: How do patients experience services, more commonly found in secondary care, within a vertically integrated general practice setting?

   3. 3.3: How does the experience differ for people with multiple long-term conditions compared with other patients?

Design

Our general approach to answering the research questions was a cross-comparative case study qualitative evaluation comprising: (1) focus groups or one-to-one interviews in each case study site, to which we invited key service managers and clinicians from the acute hospital and general practices; (2) one-to-one interviews capturing the views of patients from integrated and non-integrated general practices to understand their experiences of vertical integration models; and (3) refining and validating an overall theory of change for this model of vertical integration, starting from that emerging from our phase 1 study,2 following feedback from case study sites.

Selection of case study sites

We recruited case studies at three vertical integration sites. We focused on sites in England to correspond with the scope of the quantitative analysis (WPs 2 and 3). The research team already had (through BRACE) access to the requisite NHS activity data for England but was unable to acquire corresponding data for Northern Ireland, Scotland or Wales within the timeframe of this rapid evaluation. We returned to the two phase 1 case study sites in England (Urbanville and Greenvale) and recruited one additional vertical integration site in England to undertake further qualitative data collection and analysis. During WP1 (see Chapter 3), the research team identified NHS trusts in England where vertical integration between an acute hospital and general practices was already being delivered. Members of the study team undertook a purposive approach to identify the third case study site, with the aim of ensuring variation in our sample in terms of: (1) geographical location; (2) population served; (3) legal and governance arrangements; and (4) time elapsed since vertical integration was introduced. Members of the study team (JS and MS) emailed to vertical integration organisational strategic or clinical leads (or their equivalent) at two potential case study sites, and following initial informal conversations with both regarding participation, one site agreed to take part. One site declined to participate because they were in the process of addressing pertinent local challenges about the contractual and ownership model of vertical integration.

All three case study sites sent formal confirmation of participation letters/emails. Table 18 summarises key characteristics of the three sites as of September 2022.

Interviewing staff across case study sites

We planned to complete focus groups and/or interviews at each case study site, inviting key strategic and service level managers and clinicians from both primary and secondary care. The aim was to understand the model of vertical integration in operation and whether/how it has evolved since it was first established, and what (if any) elements of the current model are expected to improve the care and support of people with multiple long-term conditions. The evaluation team planned to complete one to two focus groups and six to eight interviews at each case study site or until data saturation (meaning that, once data were triangulated, no new emerging information was being discovered during data analysis, but only data that confirmed existing themes and conclusions). The point of saturation was agreed by discussion with all members of the full research team. Fieldwork was completed in parallel across all three case study sites (May–September 2022) by four members of the research team with experience of undertaking interviews and qualitative data analysis (IL, GM, led by MS). MS was responsible for all communication at two sites and, with IL, undertook data collection there, while GM was responsible for communication and data collection at the third site. As part of bimonthly team meetings, early findings were shared by IL, GM, MS and FW to provide the principal investigator (JS) and quantitative team members (CD, CS) the opportunity to embed themselves appropriately within the data collection process to better inform data analysis and subsequent synthesis.

Participants for interview were purposively sampled and were approached through each case study site’s gatekeeper via email invitation, including a participant information sheet (see Project website document). 83 A gatekeeper was defined as a person based at our case study sites who could act as an intermediary between a researcher and potential participants with the authority to deny or grant permission for access to potential research participants. 84 Informants included directors (clinical and non-clinical) and other NHS managerial staff (related to strategy, delivery of health-care services, financial and governance-related management); GPs, PCN leads and other primary care staff who had implemented the vertical integration model in each area.

Individuals participated in a focus group or one-to-one semistructured interview, completed at their place of employment, via telephone or Microsoft Teams® (Microsoft Corporation, Redmond, WA, USA) or Zoom (Zoom Video Communications, Inc., San Jose, CaA, USA). Each participant was provided with the participant information sheet at least 48 hours in advance to enable them to make an informed decision regarding whether to participate or not. Interviewees had the opportunity to ask questions about the study and/or wider BRACE-related work. Participants signed a consent form (see Project website document) prior to participating in the interview, including whether they consented to the audio recording of the interview. Participants were informed that they were entitled to withdraw from the study at any time and were given information about how to find out more about the study and how to raise any concerns about its conduct.

In total, the study team approached 26 potential participants to complete an interview across all 3 sites with 22 agreeing to take part (9 participants took part in 3 focus groups, 13 completed one-to-one interviews). Salient characteristics of staff interviewed across the three case study sites are provided in Table 19.

A topic guide was developed and used by researchers as an aide memoire during the interviews (see Project website document). The main themes the topic guide covered were: understanding the rationale behind the implementation of vertical integration (or whether the drivers have changed since we last collected data in the two areas involved in the phase 1 study); understanding the experiences of staff involved with the delivery of this model; what outcomes this model is expected to deliver in the short, medium and long term and what data are collected to monitor them.

Interviews were audio recorded (subject to consent being given), transcribed verbatim by a professional transcription service, anonymised and stored securely in compliance with the General Data Protection Regulation 2018 and the Data Protection Act 2018.

Interviewing patients with multiple long-term conditions across case study sites

The evaluation team planned to complete up to six to eight interviews with patients living with multiple long-term conditions across each case study site or until data saturation (N = 24). We used the definition of ‘multimorbidity’ from the NICE (2016), which defines multimorbidity as ‘the presence of two or more long-term health conditions’, which NICE states can include:

• defined physical and mental health conditions such as diabetes or schizophrenia

• ongoing conditions such as learning disability

• symptom complexes such as frailty or chronic pain

• sensory impairment such as sight or hearing loss

• alcohol and substance misuse. 35

To identify patients with multiple long-term conditions who could be invited to be interviewed, we asked gatekeepers to facilitate communication between members of the study team (IL, GM and MS) and leads at the PCNs that the vertically integrated practices at each site are part of (Urbanville = 1 PCN, Greenvale = 10, Townshire = 2). We sought to engage with members from patient participation groups (PPGs) of vertically integrated practices and of practices not part of a vertically integrated arrangement. The rationale for working with PCNs was to work with a group of practices that may include those that are integrated and not integrated as part of local vertical integrated arrangements, and hence may expedite access for researchers into case study sites. Where PCNs were entirely created from integrated practices, then the study team asked the case study site gatekeeper to facilitate communication with local neighbouring PCNs to identify suitable patients for interview.

Study team members asked gatekeepers to identify suitable patients from PPGs for interview, using the following inclusion criteria:

• 18 years or older

• living with at least two long-term health conditions (in accordance with NICE guidance)

• accessed primary and/or secondary health-care services within the past 12 months

• not on a palliative care pathway

• capable of giving informed consent

• not prevented from being able to take part in an interview due to their ill-health.

To support uptake at Urbanville, MS gave a presentation of the study at an online monthly PPG meeting (September 2022) while the site gatekeeper sent an invitation to all patients registered to receive their PPG newsletter. At Greenvale, IL and MS gave a presentation of the study to all PPG leads for practices that were part of the integrated arrangement local to the area (online, August 2022) with follow-up emails sent by the site gatekeeper. At Townshire, GM worked closely with the site gatekeeper to identify suitable members for interview. Study team members purposively recruited patients based on a convenience sampling strategy due to a low response rate.

Each patient was emailed a participant information sheet (see Project website document) prior to commencing the interview and given at least 48 hours to decide whether they would participate. Prior to commencing the interview, interviewees were given the opportunity to ask questions about the study and about what taking part would involve (either at the end of online recruitment meetings or direct with a named site-specific study team member at any time during the duration of the study). Participants signed a consent form (electronic via email) or provided verbal consent before participating in the interview, including whether they consented to the recording of the interview. Participants were permitted to withdraw from the study at any time without having to offer a reason for doing so, though none did so, and they were given information about how to find out more about the study or to raise concerns about its conduct.

The main themes covered by the topic guide for the patient interviews were: recent experiences of accessing care within their local health-care system including both primary and second care; coordination of care; communication and care planning between clinicians and the patient; speed of access to care; range of clinicians and allied health professionals to whom patients had access; barriers to and facilitators of accessing care (accounting for the COVID-19 pandemic); and perceived quality of care and care-related decision making by patients.

In total, the study team approached 24 potential patients to complete an interview across all three sites with 14 agreeing to take part (Urbanville = 8, Greenvale = 2, Townshire = 4). Despite a low response, the patients interviewed represented a varied sample of high and low users of primary and secondary care, gender, ethnicity, age and range of long-term conditions (see Table 20). The team also interviewed two participants who were both patients living with multiple chronic conditions and carers. No patients from non-integrated practices were interviewed from across the three sites. There were challenges in gatekeepers encouraging non-integrated practices to take part due to pressures faced by primary care staff during the period of data collection. Study team members (IL, GM and MS) completed semistructured interviews either by online (Zoom, Microsoft Teams) or by telephone. Interviews lasted between 20 and 75 minutes.

Analysis

Findings

Summary of findings

Analysis of interviews with staff and patients showed that vertical integration remains one model of integrated care that can help general practices remain open. Vertical integration between acute trusts and general practices is supported by the introduction of novel ways of clinically integrating care across primary, community and secondary care to meet the needs of all patients locally, but specifically those at risk, have complex, and/or multiple long-term conditions. However, there are some caveats. First, health service improvements driven by local acute trusts in collaboration with clinicians in general practice focus on health service changes that can improve population health for all patients locally and may use vertically integrated practices to test services before encouraging wider local roll out to non-vertically integrated practices. Notably, health service improvements are introduced following a significant period of cultural interchange (i.e. colleagues from acute and primary care understanding ways of working, challenges to delivering provision and aligning back office functions). Second, health service improvements were not only between primary and acute care, but also primary and community care, with clinicians from community and acute care becoming more embedded in providing services in general practice. Finally, following the pandemic and the introduction of new health services, patients with multiple long-term conditions encounter significant ‘navigation work’ concerned with choosing and accessing health provision with diminishing continuity of care with the onset and scaling up of MDT working; however, this holds true for both vertically integrated and non-vertically integrated general practices nationally.

How findings relate to previous research

There is growing interest in understanding how vertically integrated models of care can meet population and workforce needs in primary care from across the UK. Findings from this chapter extend the literature on understanding the drivers for vertical integration and the impact on the patient journey regarding their overall experience of treatment between primary, community and secondary care. Shortell et al. 97 identified several factors that need to be considered before commencing vertical integration. These included (but are not limited to): the need to establish trust among clinicians and institutions and consensus on practice and care delivery guidelines. Further, Shortell’s more recent work on accountable care organisations98 has detailed the importance of vertical integration to be physician led given their perceived legitimacy among clinicians and knowledge of local patient needs. Across our three sites, we found considerable engagement between GPs and secondary care clinicians and managers in attempt to understand how vertical integration could support and alleviate pressures faced in primary care, especially in relation to addressing workforce challenges and delivering services within community settings. The acute trusts in our study attempted to establish an almost bottom-up approach to engaging with primary care clinicians and attempt to understand organisational practices, historical challenges to providing care, and culture; nevertheless, tensions remained. Interviews with members of staff revealed how long the process of relationship building and cultural understanding has taken and how it entailed differing components of integration;99 we relate such findings to Gittell et al. ’s relational coordination theory and its associated dimensions. 88 Thus, akin to our phase 1 findings, organisational and functional integration takes a considerable length of time, whereby clinical integration (e.g. introduction of a range of services across primary, community and secondary care) has only recently been attainable.

Robinson and Casalino100 go further, highlighting that vertical integration focuses on integration between acute care and larger primary care entities (rather than solo physician or small physician practices) because of advantages such as increased efficiency in services for hospitalised patients, reduction in duplication of services, improved transitions across care settings, as well as facilitating cooperation where financial incentives across settings are not aligned. Similarly in our work, we found examples where vertically integrated primary care services were enhanced through greater horizontal integration of primary care. All examples of vertical integration between acute trusts and general practices include a degree of horizontal integration (either through practices merging or working as part of a network). Such a model can provide opportunities to redesign patient pathways to deliver care more efficiently.

Data from our qualitative study shows that vertical integration provides opportunity to embed clinicians from secondary and community care into primary care settings and also target patients at greatest need as well as those living with multiple long-term conditions, predominantly with the introduction of MDT working. Yet, it remains difficult to decipher whether changes to patient services resulted directly from vertical integration arrangements or due to increased horizontal collaboration amongst general practices, which also comes with additional resource (e.g. NHS DES contact as part of PCNs). Although, ‘innovation projects’ as described in this chapter demonstrate the greater level of flexibility and autonomy vertically integrated trusts have to deliver new services that are more responsive to local population needs and are not nationally prescribed.

Lastly, we build on Turnbull et al.’s82 application of navigation work encountered by patients and their sense making strategies and help-seeking behaviours and applied their understanding on interpreting patient journeys as part of vertically integrated models of care. We found that patients continued to face challenges accessing care and showed a preference to seek clinicians who understood their illness narratives. Although, much of the navigation work completed by patients was not specific to vertically integrated models of care but how the primary care landscape has changed since the COVID-19 pandemic.

Research question 1

1. 1: How many general practices have already vertically integrated with NHS organisations running acute hospitals in England; when did the integration between general practices and acute hospitals take place; and what are the characteristics (in terms of geographical location, patient demographics and practice size/workforce) of those practices where vertical integration has taken place?

Identifying all instances of vertical integration, where a trust runs one or more general practices, proved challenging. There was no single dataset that could be interrogated to determine this, so we created one, drawn mainly from the annual reports and accounts published by each trust individually. We have established that, at the end of March 2021, 26 trusts (15 acute hospital trusts and 11 community trusts), which is roughly 1 in 10 trusts in England, were each running one or more general practices, 85 in total, which is equal to somewhat over 1% of all general practices with significant variation in how many general practices each trust has vertically integrated. The number of vertically integrated general practices in England has been increasing annually since 2015 in a steady trickle: vertical integration is an established approach to organising care in some places, but it is far from being common. The number of practices run by each trust ranges from just a single practice in several cases, to as many as 12 practices in the largest vertically integrated group.

The geographical spread of vertical integration is broad. It occurs across the whole of England, in both urban and rural geographies. Measures of population deprivation are similar in vertically integrated areas to the average in the rest of the country (a higher percentage of practices vertically integrated with community trusts are in deprived areas compared with acute trusts). However, practices that vertically integrate are below average size in terms of numbers of registered patients and FTE GPs. Practices that vertically integrate are also, on average, doing less well than other practices in terms of QOF scores. Our qualitative findings corroborate this position. Townshire’s and Urbanville’s approaches to vertical integration were predominantly to sustain primary care close to where patients live but also improve key service provision performance indicators (across primary, community and secondary care). These findings are consistent with our findings in the phase 1 study, and internationally, that practices that enter vertical integration include some that are struggling. 101 In short, practices that have vertically integrated with a trust do not appear to be ‘typical’ general practices.

Research question 2

1. 2: What impact is vertical integration having on secondary care (outpatient attendances, A&E attendances, all inpatient admissions, emergency inpatient admissions, inpatient admissions for ACSC, bed days, readmission within 30 days of discharge)? Does this impact differ for people with multiple long-term conditions compared with other patients without long-term conditions or living with a single condition?

Our findings relating to this question are largely, but not solely, from the quantitative analyses we undertook of HES data. We restricted those analyses to the instances of vertical integration that were supported by reliable, monthly, hospital activity data for 2 years before the date of vertical integration and 2 years afterwards. We compared patient use of hospital services for these practices with use by patients of ‘stable’ practices that had held GMS contracts throughout the period. We thereby identified some statistically significant, but modest in magnitude, impacts of vertical integration on use of hospital services.

Our interviews did not reveal any perceptions of notable changes in rates at which patients of vertically integrated practices accessed hospital services. Our quantitative analyses however revealed modest, statistically significant falls in rates of A&E attendances, outpatient attendances, emergency inpatient admissions and emergency readmissions, at the time of vertical integration, and no increases in any broad category of hospital activity. But, in respect of A&E attendances and outpatient attendances, vertically integrating practices had faster growth in hospital activity than for ‘stable’ practices, both before and after the date of vertical integration. A consequence is that the modest reductions in A&E and outpatient activity at the time of vertical integration are temporary; soon erased by increased activity subsequently. For emergency inpatient admissions and readmissions, vertically integrating practices’ trends in hospital activity were not significantly different from those of other practices, so the reductions in emergency admission and readmission rates may persist (at least for the 2 years post-vertical integration that our analyses covered).

We have paid particular attention to analysing whether the impact of vertical integration on the use of hospital services differs for people with multiple long-term conditions compared with patients with one or no long-term conditions. We have found no evidence of any difference in impact on that part of the patient population. Nevertheless, evidence from interviews with staff and patients from across our three case study sites shows acute trusts are prioritising improved coordination of care for those living with multiple long-term conditions. Improved coordination of care includes an increased presence of consultants in general practice, MDT working and improved access to screening services available in community settings. Such service level improvements may also have some bearing on hospital utilisation in the medium to long term. We have compared our results with the only previously published quantitative analysis of the impact of vertical integration on patients’ use of hospital services. Our data include those analysed by Yu et al. 202034 for the effect of vertical integration in one location: Wolverhampton. However, we have taken a different statistical approach and have more data, from the whole country. We are reporting a total effect across all vertical integration sites where there are reliable data for 2 years before and 2 years after the date of vertical integration. In that context, our findings can be seen as broadly consistent with those of Yu et al. 34 They report hospital cost savings resulting from reductions in emergency inpatient admissions and readmission. We have additionally found, when looking beyond Wolverhampton, small reductions in A&E attendances and outpatient attendances.

Our analysis suggests that at the time of vertical integration there are small but significant downward shifts in some hospital activity for the patients of the integrating practices. These reductions appear to persist for vertically integrated practices relative to other general practices in the following 2 years in respect of emergency admissions and readmissions. However, at the vertically integrating practices the rates at which patients were attending A&E and outpatients departments of hospitals were growing faster than at stable practices before the date of vertical integration, and they resumed this relatively faster rate of growth after vertical integration. Thus, the saving in amount of hospital A&E and outpatient activity is temporary. Figure 20 provides a schematic illustration of this effect.

FIGURE 20.

Conceptual depiction of the pattern of A&E attendance rates and outpatient attendance rates for patients of practices that vertically integrate with trusts.

Financial implications following the implementation of vertical integration

The combined effect of the reductions in hospital activity that we have found at the point of vertical integration are summarised in Table 21. The table shows the approximate activity impact and consequent financial impact on the vertically integrated trusts in our analysis. The financial estimates cannot be extrapolated to all general practices; our evaluation has made clear that practices that vertically integrate are not typical or average, so it cannot be assumed that the effects seen there would be replicated widely were vertical integration to be implemented across the country.

We estimate any potential cost reductions in secondary care utilisation resulting from the introduction of vertical integration, using the unit cost per activity provided by the NHS National Schedule of References Costs for the year 2019/20102 and the Unit Costs of Health and Social Care 2020. 103 We use the IRR for the step change in secondary care use for vertically integrated practices compared with ‘stable’ GMS practices, to calculate the change in our chosen secondary care outcomes. This then enabled us to multiply the unit cost for each outcome measure by any statistically significant change at the point where vertical integration is introduced. The results of these calculations are reported in Table 21, which is for all secondary care outcome measures where the step change compared with ‘stable’ vertical integration practices was statistically significant, namely for A&E attendances, outpatient attendances, emergency inpatient admissions and inpatient readmissions.

We find that at the point of vertical integration, there is a reduction of 2526 A&E attendances per year among the patients of vertically integrating practices. This results in a cost saving of £460,000, based on a unit cost of £182 per A&E attendance. 102 There is also a reduction of 4873 outpatient appointments equating to a cost saving of £660,000, based on a unit cost of £135 per outpatient attendance (the weighted average cost of all outpatient attendances as reported in Curtis and Burns 2020). 103 Emergency inpatient admissions drop by 1007, equating to a cost saving of £610,000 if these were short-stay emergency admissions (unit cost of £602 per admission) or £3,400,000 (unit cost of £3366 per admission) if these were long-stay emergency admissions. 103 Likewise, inpatient readmissions drop by 355, equating to a cost saving of £210,000 if they were short-stay readmissions and £1,200,000 if they were long stay readmissions. 103 In total, for the 39 vertically integrating practices we were able to include in this part of the quantitative analysis, the estimated annualised hospital cost saving from the initial reductions in hospital activity seen for the patients of those practices amounts to £1.9–6.7 million.

Research question 3

1. 3: What impact is vertical integration having on the patient journey with regard to access to, and overall experience of, care? How does the experience differ for people with multiple long-term conditions compared with those living with no or one long-term condition?

It is important to note first that the vertically integrated practices stayed open and continued to provide primary care to their local populations. Our evaluation is unable to determine how many of those practices would have closed had they not integrated with a trust, but our phase 1 study2 showed that it is highly likely that at least some practices would have ceased to operate. The consequence of that would have been the patients of the affected practices being redistributed to the lists of other practices, presumably less convenient (on average) for patients and adding to the demand pressures and need for recruitment at the receiving practices.

We have used data from the national GPPS to evaluate quantitatively the impact of vertical integration on the patient experience of primary care and we have conducted interviews with 14 patients and 22 staff to obtain qualitative data. The analysis of GPPS data looked at six measures related to patient experience, all of which are measured on a 0–100 scale with a higher score equating to a better patient experience: helpfulness of receptionists; ability to get through on the telephone; time taken to see or speak to a GP; overall experience of access; continuity; and overall satisfaction. The interviews were able to explore the views of patients and staff in a broader way, allowing respondents to raise issues they considered noteworthy.

The analysis of GPPS data found that patient experience of care was worsening gradually over the study period for vertically integrated practices but that this was in line with national trends. We found no statistically significant step change in patient experience before and after practices switch to a vertically integrated model compared with control practices for five of six of the patient experience outcomes (those concerning access and overall satisfaction). We did find a statistically significant though modest worsening relative to ‘stable’ practices in continuity of GP care for patients (ability to make an appointment with a preferred GP) after practices became vertically integrated. Our qualitative findings show that continuity of care in vertically integrated practices remains a challenge. For instance, patients living with multiple long-term conditions are likely to receive care from a range of allied health professionals working as part of a MDT and this may be despite patient preferences to see a preferred GP.

The interviews we conducted revealed that vertical integration is associated with some service innovations and improved clinical integration, including examples where vertically integrated practices had pioneered initiatives and thereby encouraged other practices in the area to do likewise. We found that, where this was happening, vertical integration facilitated service improvements for patients largely because of a bottom-up, collaborative approach to redesigning patient care between colleagues across primary, community and hospital care. It is unclear whether the innovations could be directly attributed to vertical integration, but circumstantial evidence shows vertical integration as at least driving and supporting innovation in these cases. However, we also found at one of the three case study locations examples of where changes in service provision sought by vertically integrated practices were opposed by others locally.

Despite evidence that integrating health and social care improves access and patient experience for people with complex needs, they continue to experience a lack of coordination across teams and wider community resources. 104 We identified two related but distinct elements of vertical integration which appears to be necessary for innovations in health-care service delivery to thrive: relational coordination and service coordination towards delivery of care. We found that vertical integration is highly relational and the coordination within and across organisations is facilitated by: (1) shared goals, (2) shared knowledge or understanding, (3) mutual respect and (4) increasing opportunities for communication. Despite concerted efforts to establish closer working relationships across primary, community and secondary as part of vertically integrated models, cultural differences remained and hinder the extent of collaborative working that was possible. Yet we heard reports of increased shared understanding between secondary care and primary care staff of each other’s operational intricacies, which helped to determine expectations of what colleagues could achieve as part of vertically integrated model of care. On both a macro and micro level, aspects of relational coordination can contribute to an environment where health-care service provision can deliver timely and appropriate care.

Further still, improvements in relational coordination can lead to better co-ordinated care (e.g. bringing consultants into general practice, the use of MDTs and application of digital tools). Such examples that suggest coordination of care in vertically integrated models of care can be proactive,82 to treat patients in primary care and away from emergency secondary care.

Informed by Corbin and Strauss’ concept of the ‘work’, our approach to understanding patient navigation builds on existing literature which suggests that help-seeking behaviour is a social process with exogenous factors (external to the patient) and influenced by health-care professionals and models of health-care provision. 105 Although we found examples where patients’ ‘work’ was increased (e.g. patients using digital tools for remote consultations), we also identified examples where patients’ navigation work was reduced at the intermediary level due to improved access for appointments in general practice and improved transitions between primary and secondary care. The overall picture of the impact that vertical integration is having on patient access to, and overall experience of care, relative to the national picture across all practices, is that it is associated with a slight worsening in continuity but with no impact (for better or worse) on access to primary care or satisfaction with primary care generally. As a result, those considering introducing vertically integrated models of care need to ensure any changes in health-care provision is underpinned by the importance of continuity and the need to prioritise or incentivise this alongside other developments in health care.

We specifically analysed in the quantitative work whether patients with multiple long-term conditions were affected differently by vertical integration compared with other patients. All 14 of the patients we interviewed had multiple long-term conditions. It would not have been practical within the constraints of a rapid evaluation (a point reinforced by the difficulties we encountered in recruiting patients, as discussed earlier) to interview the much larger number of patients that would have been needed to undertake a comparative qualitative study. The interviews revealed examples of some service innovations in vertically integrated practices that were focused particularly on supporting people with multiple long-term conditions. It is unclear the extent to which these would have happened in the absence of vertical integration. For example, in Greenvale innovative support for patients with multiple long-term conditions had been a focus of a ‘new model of care’ vanguard initiative that was implemented 1 year before the vertical integration started and continued alongside it. However, as our phase 1 study highlighted, sustaining practices via vertical integration at least provides the opportunity for service innovation to happen, including innovations particularly aimed at patients with multiple long-term-term conditions. Yet, akin to existing literature, evidence on the impact of models of integrated care on cost and health-care use remains unclear. 106

Our quantitative analysis of GPPS data showed that when practices vertically integrate with trusts any differential impact for people living with multiple long-term health conditions compared with people living with a single condition or no conditions, is either not significant or of very small magnitude for the measures of patient experience.

Site characteristics

For quantitative analysis, we decided to compare vertical integration practices with: (1) practices that have horizontally merged with other practices but have not merged with hospitals; (2) practices that have changed the type of contract they hold (GMS, PMS, APMS) and those who have not; (3) and practices integrated with acute hospitals compared with those integrated with community trusts.

For the case studies, we interviewed staff and patients across three sites selected purposively (see Chapter 5) from across England from both rural and urban settings, representing a range of vertically integrated models, as well as treating diverse populations with varying needs (see Table 20, Chapter 5).

Strategies to improve representation

Within our evaluation the study team used a range of strategies to promote inclusivity. The team offered interviewees different modes of interview (telephone, Microsoft Teams or Zoom) and different times depending on participants’ preferences and to ensure accessibility for as many people as possible given workforce constraints.

Participant characteristics

Primary, community and secondary care staff taking part in interviews varied in their professional role, clinical and non-clinical skill set, location, duration in position, and the nature of their responsibilities with respect to their vertically integrated arrangement. The study team did not collect data relating to staff personal demographics (e.g. gender, age, ethnicity) as we focused on professional characteristics. Patients recruited from general practice PPGs varied with respect to age, gender, ethnicity and type of multiple long-term conditions they were currently living with at the time of the interview (see Table 20, Chapter 5).

Research topics relating to equality, diversity and inclusion

Our evaluation had a clear focus on inclusivity in relation to capturing the experience of those serving and living with multiple long-term conditions, a group which in the past has been identified as a priority when designing and coordinating care across the NHS in a range of settings. In addition, we recommend that further research with patients to understand experiences of receiving care as part of a vertically integrated model.

Reflections on research team and wider involvement