Implementing Population Health Management (PHM) Across an Integrated Care System in England, Using Action Learning Sets and an Embedded, Formative Process Evaluation

Introduction

In the UK, Europe, and USA, rising demand on health and social care systems combined with financial and workforce pressures [1, 2] has led to increasing calls for a Population Health Management (PHM) approach to redesigning services in an integrated, whole system way. PHM aims to improve patient experience, population health and reduce inequalities by using linked electronic patient records (EPR) from different providers to design more targeted, preventative, and person-centred services [3, 4]. In this sense, PHM brings together two of IFIC’s nine pillars of integrated care; population health and digital solutions.

In England, PHM has been promoted since 2018 [4] and is now embedded in national policies such as Core20PLUS5 health inequalities strategy, Universal Personalised Care, Anticipatory Care and the Community Mental Health Framework. Integrated Care Systems (ICSs) are responsible for implementing PHM at system, place and neighbourhood (Primary Care Network) levels, with growing expectations for primary and out-of-hospital care to play a proactive, preventative role [5, 6].

PHM involves linking and analysing historic health, social care and wider social determinant data at patient and population levels. Patients or communities can be segmented or stratified to identify those ‘at risk’ of poor outcomes or unmet needs [7, 8, 9], informing tailored service responses. These can be tested and evaluated for their impact on health, patient experience, inequalities, workforce capacity and costs [10] – the ‘Quintuple Aim’ [4] – completing the PHM cycle (Figure 1) [5, 11].

Figure 1

Such steps are reflected in NHS England’s model of PHM implementation, summarised as: i) the development of an Integrated dataset; ii) Insight generation; and iii) targeted Innovation development. A fourth step, advocated in two recent literature reviews [12, 13, 14], notes the importance of demonstrating iv) Impact on health, inequalities, and the quality and person-centredness of care. Together, these critical building blocks form what might be called the 4 ‘I’s framework. Figure 2 illustrates this process.

Figure 2

Arguably, PHM is another approach to needs assessment in public health, albeit based on individual rather than area level data. Individual NHS data has been underused historically for improvement purposes [10]. With the current interest in ‘Big Data’ [15] and recent policy, structural and legal changes (e.g. Federated Data Platforms, shared electronic records), PHM offers commissioners and health and social care providers a means of basing local policy on more granular data, supporting NHS policy objectives [16].

However, evidence on how to implement PHM is still emerging. At system level, this largely comes from a dozen studies from the USA, Canada, Singapore and Europe [17, 18], often in the context of insurance-based health care provision [9, 19]. These suggest that implementation, as conceptualised in Figure 2, may be challenging. For example, building integrated infrastructure can present regulatory, technical, methodological and cultural issues [15, 19, 20]. Segmentation and risk stratification may be hindered by a lack of digital and analytical skills [13]. Cross-national implementation studies, including in Greater Manchester, UK, point to the importance of investing in and developing the ‘softer’ aspects of PHM – building a vision and relationships, working collaboratively, paying attention to governance, transparency and issues of trust [21]. Findings include the importance of an enabling learning environment that facilitates the workforce (managers/clinicians), patients and other stakeholders to co-design innovations and support implementation [17, 21]. Insight and innovation development may require skills such as networking, quality improvement, and change management if insights are to be turned into innovations that deliver benefit, and evaluation skills to demonstrate impact [13, 14].

As well as being relatively sparse, UK evidence relating to PHM implementation is based largely around meso-level case studies, often relating to single secondary or primary care provider datasets rather than linked datasets and multidisciplinary teams (functional and clinical integration) [22, 23]. Yet, evidence suggests that, far from being a simple, rational, linear process to implement, PHM is a complex intervention involving macro, meso and micro contexts and mechanisms [17, 18, 19, 21] i.e. a multi-dimensional programme requiring structural (vertical/horizontal), functional, clinical and service integration [24].

Finally, few studies have sought to understand PHM implementation as happening in a real-world setting, within a complex adaptive system. Subtle differences in local context and circumstance challenge positivism claims of causality. Understanding such variations enables a focus on transferability of findings to other areas rather than generalisability [25, 26].

Against this background, this paper presents a pragmatic, participatory, formative process evaluation of an English ICS’s PHM implementation programme across its Primary Care Networks (PCNs). This was commissioned by the ICS in recognition of the need for an adaptive learning approach given the lack of evidence, the complexity of the integration context, with efforts somewhat floundering despite integrated care pioneers in 2015 and the creation of ICSs in 2021 [27] and on-going financial pressures.

The programme was based on a previous national NHS England PHM development programme (provided by a USA-based commercial partner, Optum) to four of its 31 PCNs and one of its four locality care partnerships (LCPs) in the ICS during 2020–21. As an early adopter, acceptance onto this programme required a degree of infrastructure development and data linkage. This study evaluated the ICS’s revised, shortened version of Optum’s PHM programme. It had a similar format and structure, but comprised less Action Learning Sets (ALSs) and was delivered by Health Innovation Southwest (HIN SW) over 20 weeks across all its localities and PCNs between April 2022–24.

The programme aimed to provide formative learning throughout the two-year period, as well as potential summative learning from innovations implemented in this period. As ultimately no innovations were delivered, this paper, therefore, reports on the formative learning, addressing the research question: What factors facilitate or hinder PHM roll-out at PCN level across the ICS?

Methodology

Data analysis

The study undertook a formative analysis of why the programme had or had not achieved its inputs, activities, outputs and outcomes comparing separate implementation, progress and contextual factors across the four localities. Qualitative data was coded in real time inductively and deductively at the study’s interim and final reporting stages (informed by a blended coding framework based on principles identified in a Dutch implementation study [19] and adapted the 4 I’s framework shown above [4, 6, 13, 14]) by one researcher (TB) and grouped into basic themes [34]. Coding categories were checked by the other researchers (JE & FG) and dissonances discussed and resolved. Codes were grouped into themes and refined through discussion and on-going reflection with the PHM delivery group. Analysis was conducted in Excel and NVIVO software. Themes were used to reflect on, refine and adapt the 4 ‘I’s framework to produce a revised conceptual model of implementation [35] (Figure 3).

Figure 3

Results

Table 1 shows just over a third of study participants were from primary care (including GPs), with VCSE, acute providers, public health and NHS commissioners making up most of the remainder.

There were 21 ALSs over 9 months (plus 4 supplementary sessions) with three localities only reaching 5 of the 6 programmed workshops (localities B, C and D). ALS workshops involved 198 attendees (on one or more occasions), varying between 40 to 68 attendees across localities. Overall, two-thirds of attendees (134/198) were not primary care stakeholders. Locality B had the lowest number of attendees and proportion of GPs attending (Table 2). No locality involved patients in ALS workshops, although co-design was being actively considered in locality A.

By the end of the study, six project ideas were identified across PCNs, of which four were worked up from a partially completed logic model into a proposed innovation (localities C and D). One innovation was on the verge of piloting (locality A) (Table 3). Locality B did not reach a point of designing an implementable intervention. Three innovations focused on secondary prevention, two on service efficiency and one on primary prevention. Although no innovations were piloted, the logic model work estimated that the interventions could have reached up to 630 patients and one of the service efficiency innovations ~5% of PCN patients.

The partial implementation of the PHM programme and limited outcomes were explained by several interacting factors, operating at the macro (system context), meso (organisational) and micro (individual and process) levels. These themes surfaced at all stages of implementation – infrastructure development, insight generation and innovation.

This ultimately affected programme impact, leading to a revision of the 4I’s implementation framework as shown in Figure 3. ‘Softer’ enabling process factors were also identified and incorporated: coordination, flexibility, time, inclusiveness and trust.

Discussion

This detailed process evaluation of PHM implementation within one English ICS highlights the complex interplay of contextual, organisational and individual factors shaping the translation of a national policy goal into practical reality. These factors interacted in a non-linear fashion across the stages of PHM implementation – Infrastructure, Insight, and Innovation – across system, place, and primary care levels [38]. Each stage was also sensitive to how the PHM programme was managed, facilitated and resourced in terms of coordination, inclusiveness, flexibility, timing and trust. Our locality comparison also showed local context matters, justifying a bespoke and pragmatic approach taken by the PHM programme, the ALS facilitators and the evaluation.

Regarding infrastructure, our findings echo those of previous UK-centric reviews [13, 14] and the wider international literature [17, 18, 19, 21], which emphasise that successful PHM implementation requires leaders to articulate a strong, strategic vision for PHM, supported by long-term investment in developing, maintaining and utilising linked datasets effectively [20]. Robust and transparent governance structures around linked datasets and data protection are essential to avoid continual renegotiations that stall implementation. Relational work is needed to build trust with primary care providers and other stakeholders to address concerns about data safety and confidentiality. These are critical issues in this and other sectors, especially insurance-based systems sensitive to financial confidentiality [39], and are vital to ensure comprehensive dataset coverage. Given the significant capacity impact on BI staff who had to spend time renegotiating and reorganising contracts and provider data flows, a clearer national strategy on data governance would be valuable.

Extrinsic motivation in the form of funding proved an effective tool for supporting engagement, particularly for PCNs facing capacity constraints. Our comparative analysis showed that progression to insight generation was influenced by existing collaborative capital, the quality of relationships, local priorities and the intrinsic motivation of stakeholders. Thus, investment in local partnerships and better (structural) integration of local interests with ICS priorities could strengthen implementation by aligning resources and facilitating spread. This needs to be accompanied by ‘soft’ management skills and a flexible process that supports ongoing learning as these were important to managing different interests across PCNs, maintaining stakeholder engagement and creating momentum towards innovation development. This is consistent with factors and principles identified as key to successful implementation elsewhere [17, 18, 19].

Most PCNs will be using integrated datasets developed by ICSs, as few will have capability or capacity to link practice data themselves. As we have found this currently generally covers health and social care data. Producing more nuanced and holistic insights into patient needs, a key rationale for PHM [40], will require engaging stakeholders from non-statutory and VCSE sectors as well as patients themselves. Effectively leveraging their knowledge and experience to develop integrated clinical pathways requires dedicated resources and development of co-production skills among facilitators, coordinators, and stakeholders.

The importance of externally resourcing pathway co-design, supporting dashboard access and interpretation, providing primary care workforce training and facilitating patient and public involvement must not be understated. This has been noted in other UK PHM programmes, for example, coronary vascular disease and mental illness risk stratification tools developed for some London PCNs and an acute mental health trust [23]. Although innovations in our study did not reach these stages, the London experience showed that a lack of dedicated implementation support and financing were significant barriers to delivering tailored service pathways. Demonstrating impact on patient health outcomes is also challenging [23]. New PHM care pathways will require thorough assessment to ensure they meet PHM goals (and do not inadvertently widen inequalities), providing necessary evidence for business cases supporting wider adoption [20, 21].

These findings suggest the need for a more dynamic model of PHM implementation than NHS England’s linear ‘4I’s’ framework [12, 13, 14]. Based on a learning approach, van Ede et al.’s (2023) implementation model incorporates not only integrated infrastructure and data analysis but also a domain for community and workforce co-design, alongside an emergent implementation strategy grounded in learning cycles [9, 17]. Understanding how to align and integrate infrastructure and technology with effective processes and stakeholder engagement will be critical to successful implementation [38]. In this context, PHM should be recognised as multi-level, multi-dimensional integration programme, requiring cultural change and distributive leadership to foster local ownership and action [17]. Leaders must provide vision and adaptive strategy, backed by resources targeted at multiple system levels, to transform organisational and individual behaviours, that is also sensitive to local circumstances [17, 18]. Providing ‘social proof’ of PHM successes, for example, through case studies and building a learning community across organisational, clinical, and community levels, based on advocacy champions and continuous improvement, will further support effective spread [18].

Conclusions

Implementing PHM is a national priority for ICSs in England, supporting policies aimed at delivering more preventative, person-centred care and addressing health inequalities. Despite incentives and resources and using a flexible approach to engaging PCNs and stakeholders, raising awareness and analysing locally linked data, the programme evaluated in this paper did not achieve its aims – no locality developed and implemented an innovation within a 2-year period. System pressures, weak integration with ICS priorities, and reduced capacity affected engagement, contributed to delays and led to a loss of momentum.

Given the intention of better harnessing shared data and predictive analytics to support the shift from sickness to prevention, the lessons of this study have wider relevance to the NHS 10-Year Health Plan. Without addressing system integration, IT infrastructure and IG challenges as well as the ‘human factors’ that affect implementation, the potential to create an NHS that is fit for the future may not be realised.

Reviewers

Dr Dan Alton, Chief Clinical Information Officer, Buckinghamshire, Oxfordshire and Berkshire West NHS Integrated Care Board, Oxford, UK.

Dr. Annefrans van Ede, Health Campus The Hague, Public Health and Primary Care department, Leiden University Medical Center, Netherlands.