1.1. What are learning health systems?

A learning health system (LHS) is a way of describing a team, provider or group of providers in the health and care system that, working with a community of stakeholders, has developed the ability to learn from its own delivery of routine care and improve as a result. At its most fundamental, an LHS comprises a set of activities and assets that enable continuous learning and improvement of services.

LHSs are an important method for improving the quality, efficiency and effectiveness of health and care services. But there are many different types of LHS, ranging from the clinical microsystem level to the national level and everything in between. As a result, the term tends to be used in many different ways. We think that this can, on occasion, be a stumbling block to making progress in this field – with people sometimes talking at cross purposes or using terminology in overly restrictive ways.

For that reason, we begin this chapter by developing an analytical framework for understanding the key components of LHSs and for characterising the variety of types that exist.

Common aspects of learning health systems

While there are many different types of LHS, they all have some key factors in common, although important variation can emerge in relation to these factors.

• The provision of services. At the core of an LHS sits a service provider or providers, and the desire to improve service provision and outcomes drives the LHS’s activity. An associated factor is that a key source of data from which an LHS learns is data generated from routine service provision (whether clinical data, operational data, patient-reported data and so on). This is one important thing that makes an LHS different from many types of research or trials relying purely on bespoke data collection. (Another important difference is the continuous, iterative nature of the learning that takes place within LHSs – discussed further below.) The presence of a provider means the type of improvement that LHSs do can be endogenous (driven from within) rather than simply exogenous (externally driven by factors such as policy or regulation).

One possible source of variation among LHSs is therefore the type and sector of the provider in question and the nature of services being delivered (for example, health care, public health, social care or community services). What is being improved will also affect who the ‘service users’ might be on any particular occasion (for example, patients, staff, carers or citizens).

• The learning community and improvement ambition. An LHS is driven by a learning community that has been formed around a common ambition of improving services and outcomes. Not everyone in the learning community will necessarily be involved in every stage of the LHS (for example, patients might be involved in formulating ideas and trialling service changes but not in data analysis; data analysts might be involved in generating learning from data but not implementing service changes; and so on). However, all share and contribute to the common ambition in some way.

Another critical source of LHS variation is therefore the nature of the learning community, and its corresponding improvement ambition:

• They could be place based, and if so they could exist over a range of geographies (for example, based around an individual provider organisation, a health economy or the whole NHS).
• They could be condition based (for example, improving care for people with cystic fibrosis).
• They could be thematic (for example, improving procurement, adopting a particular technology or reducing a particular type of medical error).
• They could combine these properties (for example, improving asthma care for young people in London).

• The learning and improvement cycle. LHSs effect change through iterative learning cycles based on generating and learning from data and formulating and testing service changes. Despite the huge diversity of LHSs, their learning and improvement cycles tend to be based on a common set of stages, illustrated in Figure 1. And at each stage of the cycle, the same types of activities tend to be going on: measuring outcomes, formulating hypotheses, analysing performance, designing improvements, implementing service changes and so on (see Figure 2).,, The cycle is then repeated, allowing each subsequent iteration to test and evaluate the service changes implemented in the previous iteration. These activities are the ‘bread and butter’ of LHSs. And it is by focusing on how to do these activities well, we can support the development of LHSs, whatever their form.

Source: The Health Foundation’s Insight & Analysis Unit

Importantly, if it is the presence of certain activities that constitutes an LHS, then it does not really matter whether the individuals involved think of it as an LHS or not. As we will see with some of the case studies explored in this report, something can be an LHS even if the practitioners involved do not use that terminology or did not set out explicitly to develop an LHS.

Figure 1 is not intended to be a systematic analysis of the learning and improvement cycle, but simply a useful way to think about the constituent activities that go on in an LHS. It broadly corresponds to Charles Friedman’s influential three-stage characterisation of the learning cycle: practice to data; data to knowledge; and knowledge to practice (indicated in the figure).

Source: The Health Foundation’s Insight & Analysis Unit

However, something the classic tripartite LHS schema does not always make explicit is how values and intentions get into the cycle – how questions get asked and priorities for improvement get determined. And this is a more fundamental issue than just the initial selection of an improvement ambition – the need to ask questions, gather and aggregate views, reconcile differences, and make judgements and course corrections is an intrinsic part of the learning process. For that reason, we think that it is important to consider problem definition and solution design as explicit parts of the learning and improvement cycle – signified in Figures 1 and 2 by the actions ‘ask questions’ and ‘identify and agree potential improvements’. Indeed, as we will discuss further later, it is these fundamentally human processes of convening, interacting, deliberating and making decisions that make the social infrastructure of LHSs just as important as their technical infrastructure.

Differences in the way each stage of the learning and improvement cycle happens can be another important source of variation between LHSs. Each stage can differ in scale and intensity (appropriate to the LHS’s goals) – that is, in the depth and granularity of the work going on, the number of people involved, the timescale, the cost and so on. For example, the data analysis involved in a learning and improvement cycle could range from reading a patient feedback form to a lengthy research study involving novel and complex analytics, while the service changes could range from putting up a sign in a waiting room to redesigning a whole health care pathway.

The scale and intensity of different stages of the learning and improvement cycle will therefore greatly affect what the LHS looks like in practice. In particular, the greater the scale or intensity required, the more likely it is that different individuals will lead different aspects of the cycle, or that these stages will happen in different environments. While at the smallest scale, the stages of an LHS could be executed by a single individual in a single environment (for example, a clinician using real-time feedback in a mobile app to optimise their practice), at the other end of the spectrum might be a learning and improvement cycle that involves lengthy research projects, specialist engagement exercises, the lab-based development of new technology or data tools, or the cross-organisational implementation of new clinical pathways.

In summary, variation in the three aspects of LHSs outlined here – the nature of services provided, the nature of the learning community and improvement ambition, and the scale and intensity of each stage of the learning and improvement cycle – makes many different types of LHS possible. Figure 3 illustrates some of this diversity, using a selection of the case studies presented in this report.

Source: The Health Foundation’s Insight & Analysis Unit

These six case studies are presented at the end of Chapter 1.

1.2. The assets underpinning learning health systems

Another approach to thinking about LHSs is to consider the capabilities and infrastructure that typically underpin them – in other words, going beyond thinking about the activities of the learning and improvement cycle to consider the assets on which these activities rely.

These assets, illustrated in Figure 4, include:

• data, data analytics capability and research capability, including skilled researchers and analysts
• technology, including data platforms, tools and systems, as well as an organisation’s wider digital maturity
• learning communities and networks, along with mechanisms, spaces and support for convening, deliberating and sharing knowledge
• improvement capability and culture, along with resources to enable the planning, design and implementation of improvements to care.

 

Source: The Health Foundation’s Insight & Analysis Unit

As the scale of an LHS increases, these assets will tend to become more visible and significant. And, of course, they are not static: they will develop and mature over time with successive iterations of the learning and improvement cycle and successive projects. And there is a need to continually nurture them.

Importantly, a provider or learning community might exhibit some of the components of an LHS but not all of them. And it is worth emphasising that the activities, capabilities and infrastructure required for successful LHSs are valuable in their own right, even when they are not being used as part of a full LHS.

Focusing on these kinds of assets and how they can be successfully developed should therefore be an important aim in itself. In practice, when thinking about how to support the development and evolution of LHSs, rather than trying to create entire LHSs in a single step, it can be more effective to focus on developing one or two components first – for example data analytics or improvement capability. In many cases, it will be about identifying what components are already in place and building on these. LHSs are not ‘all or nothing’ in this respect. Nevertheless, ultimately, it will be by bringing all these different components together and ensuring they are working in partnership that the LHS will become more than the sum of its parts.

Furthermore, there are other important elements of learning infrastructure in health and care that complement the concept of LHSs described here, and on which LHSs rely to connect and share learning (see Box 3 for further discussion).

1.3. How learning health systems can help

Why should providers and learning communities be supported to adopt an LHS approach?

First, LHSs can support the delivery of externally led change through national programmes by providing the means to implement changes, test them and iteratively adapt and refine them, working with the very patients and staff the changes apply to. For example, in England, there are significant opportunities to develop LHS capabilities within integrated care systems as a way of supporting the successful adaptation and embedding of new pathways and models of care. In this guise, LHSs can be thought of as sophisticated ‘implementation mechanisms’ that can help deliver national priorities for service transformation.

But LHSs also matter because they are critical for enabling locally led service change. Many of the challenges in health and care cannot be solved by top-down change programmes alone. And while local systems face common problems and many solutions are generalisable, there are also problems and solutions that are specific to individual contexts, which those closest to them will need to diagnose and solve.

By creating the capability to learn and improve from within, LHSs can turn providers into ‘engines of innovation and improvement’, driving improvement in a way that is not reliant on national initiatives or investment. And over the long term, endogenous, continuous improvement has the potential to achieve more than a series of centrally led improvement initiatives and may in many cases be more effective in achieving sustainable quality and efficiency gains.,

Another reason for the growing currency of LHSs is that they are an important way to capitalise on the increasing availability of data and analytical tools. In short, our ability to learn has never been greater. Crucially, developments in data and data analytics are giving providers themselves the power to gain insights about pressing challenges and how to solve them, reducing the need for external analytic capability. The increasing sophistication of technologies such as artificial intelligence also presents further significant opportunities for data-driven service improvement.

National policy has been slower to focus on LHSs than other drivers of health care improvement (such as targets, incentives and competition). However, interest in LHS approaches has grown over the past decade as the limitations of these more traditional, top-down policy levers have become apparent. In England, for example, the 2013 Berwick review set out a vision for the NHS to become ‘a system devoted to continual learning and improvement of patient care’ and this led to the government proposing that the NHS should become ‘the world’s largest learning organisation’. More recently, the 2021 Integration and Innovation White Paper contained the ambition of ‘accelerating [the system’s] ability to learn, adapt and improve’, while NHS England have argued that integrated care systems should become ‘consciously learning systems’. Meanwhile, Healthcare Improvement Scotland sees the development of ‘human learning systems’ as a key part of its approach to quality management. So there now appears to be acceptance at the national level that building a culture of continuous learning and improvement is essential for improving quality, efficiency and effectiveness.,,

For all these reasons, LHSs are an idea whose time has come. Not only are there increasing opportunities to deploy them and increasing interest from national policymakers, but we will not be able to solve the huge challenges that services are facing adequately unless we fully exploit the potential of providers to learn and improve. Box 4 explores where the greatest potential for further development might lie.

The chapter concludes with six case studies from the UK of LHSs of varying scale and focus. And to gain insights from other countries where LHS approaches are used, Box 5 gives three international examples.

2.1. Introduction

As highlighted in Chapter 1, there is growing interest in learning health systems (LHSs) as a vehicle for improving care quality and service delivery. But given that LHSs are multifaceted and can be highly complex, it can be difficult to know where to begin. So, what should policymakers, and organisational and system leaders, be focusing on to support the development and spread of LHS approaches?

Our research and engagement with expert stakeholders highlighted four key areas where targeted action could help to advance the use of LHS approaches:

• learning from data
• harnessing technology
• nurturing learning communities
• implementing improvements to services.

These four areas, which link to the assets and capabilities underpinning LHSs as described in Chapter 1, form key lines of enquiry in this chapter. Specifically, we discuss each of the four areas in turn, setting out the current context, opportunities and challenges. In each case, we also report the results of our stakeholder survey to identify which of the challenges we describe currently require the most attention.

2.2. Learning from data

2.2.1. Key issues and opportunities

The potential of data for learning health systems

Better use of data has huge potential to improve the quality, safety and cost-effectiveness of care and address unwarranted variation, across the whole health service. The data can be quantitative or qualitative and be drawn from many sources. As discussed in Chapter 1, one of the characteristic features of an LHS is that it uses data generated from routine care delivery – whether clinical, operational or patient-reported. However, LHSs are of course not restricted to using these data and can draw on data from a wide range of sources, such as patient surveys, research and clinical trials, and increasingly data from technology such as smartphones and wearable devices (wearables), like smart watches and medical technologies, that are worn by individuals to track, analyse and transmit data.

Within quality improvement work, it is not uncommon to hear people distinguish between ‘data for care’ and ‘data for improvement’. But a more useful approach is to distinguish between the primary purposes of health care data (delivering direct clinical care) and the secondary purposes of such data (such as research, population health management and improvement). Both uses of data support the work of LHSs, in particular in the design of services and helping to build understanding about how they are performing and how they can improve.

As the volume, breadth and quality of data increase, so too do the opportunities for learning and improvement. Wider sources of data, such as data generated by smartphones and wearables, and data on the social determinants of health, are increasingly playing a valuable role by allowing us to build a more holistic picture of our populations. SAIL Databank (case study 7), a Trusted Research Environment that contains data about the whole population of Wales, shows how this wide range of data can be brought together.

Combining and analysing data from different sources can generate new clinical evidence. The Informatics Consult project (case study 8) shows how this can be done. It is particularly important for supporting treatment decisions for patients with complex health needs, where traditional forms of evidence are lacking.

Understanding the health and needs of people and their communities

The move towards better collaboration between health and care providers and commissioners – through, for example, health boards in Scotland and integrated care systems in England – provides an opportunity to improve the health of their populations. By integrating infrastructure, developing standards for interoperability (the ability of two or more systems to exchange information and use that information) and working collaboratively, these organisations can identify where within their populations the greatest needs lie, and work together with stakeholders, including members of the community, to design data-driven interventions to address those needs.

There are particular opportunities for better use of patient-reported data. But while the NHS collects a large amount of patient-reported data, they could be used much more effectively. This is where an LHS approach can help because it centres on the ambition of putting knowledge into practice.

Data collected from communities offer the opportunity to help health care services better understand what matters most to them. This includes national collections such as patient and service user experience surveys and patient-reported outcome measures, as well as local routes, from focus groups to service user surveys. Initiatives such as the Networked Data Lab, which the Health Foundation leads, are linking these datasets together to build a more complete picture of the relationships between the wider determinants of health, health needs, service use, patient pathways and health outcomes.

Data about particular health conditions, including on diagnosis, treatment and outcomes, also offer potential for LHSs – including the possibility of creating LHSs around particular conditions. Clinical registries and clinical audits, for example, allow comparisons between multiple sites, reveal variation in processes and outcomes, and identify where improvements can be made.

Making data easy to understand and actionable

While the growth in the volume of data means there is ever more knowledge that can be generated, the amount of data already exceeds what people working in health care can assimilate. There are other challenges too. Often data are not presented in a useful way, and long time lags (between data collection and use) mean that they can be too old to have relevance to health care decisions today. For example, as patient registries become able to collect data on patients’ priorities in real time, incorporating patient-reported data with clinical data, they could become a critical part of the infrastructure of LHSs where patients, clinicians and scientists work on service improvements together.

So, there is considerable scope to improve both the curation of data (the organisation and management of data from various sources) and the visualisation of the data – along with how and when the data can be accessed, so that health care professionals can see relevant insights in a timely and digestible way to support their decision making. For example, the Clinical Effectiveness Group at Queen Mary University of London (case study 5) has created interactive dashboards showing performance across GP practices in north-east London, allowing for the identification of areas requiring improvement.

Elsewhere, academics at the University of Edinburgh have developed a dashboard to support GPs to improve care for people with asthma across the UK (case study 9). The dashboard gives GPs weekly data on asthma indicators at their practice and shows how the data compare with those for other practices.

Building data analytics capability

Data analytics methods, tools and skills – including novel statistical models and data linkage – are a critical part of LHSs because they generate new knowledge and insights to support the learning and improvement process.

Our research and funding programmes at the Health Foundation show that the data analytics capability that exists in the health and care system is both underdeveloped and underused.

The impact of data analysts could be increased considerably through training and professional development, and through better access to the software tools required to generate insights from data. Specialist analytical expertise can also help to develop data literacy in the wider health care workforce. For example, the Flow Coaching Academy (case study 1), a model centred on a multidisciplinary micro-team approach to improving service delivery, shows how this potential can be exploited. By bringing data analysts into multidisciplinary team meetings known as ‘Big Rooms’, and coaching clinical staff in how to interpret data, the model enables teams to gain a better understanding of how services are performing and, critically, where improvements could be made.

Evaluating improvements

Testing and evaluation – which rely on collecting the right data – are a critical part of an LHS to show whether changes made have led to improvements. This mostly happens as an integral part of the iterative learning process, with the data collected in subsequent learning cycles being used to assess the service improvements implemented in previous cycles.

There are promising ways in which testing and evaluation can be done that marry robustness with timeliness. For example, the Children & Young People’s Health Partnership (case study 6) shows how running different evaluation models simultaneously can effectively support the learning and improvement process. By running randomised control trials at the same time as service evaluations, the partnership can give evidence quickly to clinicians to inform care improvements, and to commissioners and managers who can make rapid decisions about embedding new services into ‘business as usual’, while building a robust evidence base for children’s health care more widely.

2.2.2. Challenges

While some health care services and systems have made great progress in collecting and using data, our research suggests there are a range of challenges that stand in the way of using data to drive continuous learning and improvement in the manner that is required for LHSs. Several of these challenges have been highlighted in the government’s recent strategy Data Saves Lives, which also sets out steps to make progress on many of them.

LHSs require the availability of high quality, actionable data on a range of issues, including performance, outcomes, experiences and processes. This has been a particular challenge for patient data as progress towards the digitisation of patient records has been slow. Where data are recorded electronically, they do not always meet quality standards, nor are they always useful. Data can be incomplete or captured incorrectly.

Stakeholders we spoke to during our research told us that there are significant issues surrounding data access and sharing, including information governance and data security, which can limit the work of an LHS. In order for a team to be able to access data, a range of approvals are required, which vary in number and complexity due to a lack of standardisation. In addition, the regulation of data often lags the rate of innovation and the data that it produces. Financial costs associated with data access can also hamper the ability to use data to learn and improve. Health and care data must be in a readable, actionable format to have value, which often relies on data system vendors curating and cleaning the information they hold, and they can then charge sizeable fees.

It can be difficult to know what approvals are required to access data and from whom. For example, innovators involved in the RADAR project (case study 10) told us how introducing the MyWay digital health app into some parts of England required data-sharing agreements with each data controller. This included individual GP practices, clinical commissioning groups and Caldicott guardians (who are responsible for protecting the confidentiality of people’s health and care information and making sure it is used properly) compiling population-level data for the area, which took an enormous amount of time and effort to achieve. Conversely, use of the platform in north-west London was far simpler as data access is dealt with at the integrated care system level, where there is a structure for data controllers signing off electronically, and a population-level dataset already exists.

Our experience of funding data analytics projects at the Health Foundation suggests that linking and integrating datasets can also be particularly difficult, not only because of technical challenges, but also because of the way in which increased data linkages can challenge processes of anonymisation. The number and complexity of processes required to link data can be particularly challenging. For example, in the LAUNCHES QI study (linking audit and national datasets in congenital heart services for quality improvement), run by University College London, which sought to improve services for people with congenital heart disease by linking five national datasets, the process took 2.5 years to achieve, requiring nearly 50 documents for the data application processes, which needed to be submitted 162 times in total. Stakeholders told us that there are also challenges with integrating datasets, and insights derived from them, within electronic health records and that integrating analytical tools into electronic health records can be both difficult and very expensive.

System interoperability – the ability of two or more systems to exchange information and use that information – continues to be a challenge in health and social care. With an assortment of different data systems, and a lack of common standards, it can be difficult to gather and aggregate the data needed to drive an LHS. A 2020 National Audit Office report on digital transformation in the NHS in England highlights that while interoperability has been a priority for policymakers since 1998, progress has been slow, and a lack of clarity on standards to encourage new suppliers to enter the NHS could make interoperability harder to achieve.

For health care services to learn about and design the most effective interventions, they require data that go beyond routine clinical data – for example, data on the social determinants of health. Yet access to a wide-enough range of data can be challenging and many types of information are not captured in datasets. For example, routine NHS data do not include data on those who do not access health services, nor do they include information about a patient’s health in between interactions with the health service – both of which could improve understanding of the drivers of ill health.

The disproportionate impact of the COVID-19 pandemic on some groups has rightly highlighted the importance of tackling health inequalities, but to do this requires data that are representative of all patient populations and free from bias, which is often not the case. This can be due, for example, to a failure to capture data on characteristics (such as ethnicity) accurately, a lack of representation of different demographic groups in research or a lack of access to technologies that capture data., Such factors can render datasets biased, which can lead to decisions that either do not bring about improvements to service delivery or – worse – risk adverse outcomes and experiences for some service users. This is a particular concern for data-driven health technologies such as those using artificial intelligence and machine learning.

Understanding Patient Data research shows that most people support sharing patient data for individual care and a high proportion of people support sharing patient data for research where there is public benefit. But making sure that patients and the public, and also the health care workforce, perceive the collection, sharing and use of data to be trustworthy can be a challenge. Transparency and open dialogue with the public are important ways of achieving this but are not always addressed sufficiently.

Data need to be formatted and ‘cleaned’ to allow them to be analysed. This can be laborious and time-consuming work, which often fails to get the attention and resources it requires and is often done on an ad-hoc basis, leading to duplication of data. Both this data curation work and the subsequent data analysis require specialist skills and knowledge, and while there have been moves to build this capability across the health and care system – including through NHS England’s work to develop a National Competency Framework for Data Professionals in Health and Care – more is required. Another related issue is more general health data literacy across the workforce, so that staff, including those in management and leadership positions, can interpret data appropriately.

We asked our survey respondents which of the data challenges highlighted in bold above they considered to be the most significant for adopting LHS approaches; the results are shown in Box 6. (For each part of our survey, we also asked our respondents to tell us whether there were any challenges we might have missed, but the responses suggested that there were no significant omissions.)

2.3. Harnessing technology

2.3.1. Key issues and opportunities

While not a necessary part of an LHS, technology can clearly play a very important enabling role, for example by:

• enhancing data collection and analysis
• accessing high quality information through electronic health records
• helping with clinical decision making
• supporting the design and implementation of improvements.

Increasing the use of digital health technologies is a priority for the UK’s health and care systems to support both service delivery and improvement. For example, NHS England’s service transformation plans are currently seeking to capitalise on the potential of LHSs to generate and embed knowledge and drive improvements in health and care.

Enhancing data collection and analysis

Digital technologies can support LHSs by enhancing the way health data are collected, for example through devices such as mobile phones, wearables and sensors. These technologies can enable the real-time collection and analysis of health data that may have previously been prohibitively expensive, intrusive or time consuming to collect or understand. For example, the CFHealthHub (case study 3) shows how data gathered through the routine use of Bluetooth-enabled nebulisers helps clinicians to understand how well existing care pathways are supporting people with cystic fibrosis. By integrating technologies into the existing care pathway, this example shows how data can be used to support system learning and improve personalised support for people with long-term health conditions. The use of digital health technologies such as these may also mean that the data collected is of a higher quality when compared with data inputted into a system manually. But it is important to make sure that those using these technologies are comfortable with using them, capable of using them and motivated to do so.,

In addition to collecting data, digital health technologies also often incorporate additional functionality such as analytics and artificial intelligence (statistical tools and algorithms that enable computers to simulate elements of human behaviour such as learning, reasoning and classification), which seek to maximise the value of the data. Technologies like natural language processing (a branch of artificial intelligence) can make complex data available to LHSs in a timely way by transforming large amounts of qualitative data into an analysable format. For example, it can analyse free text patient feedback, which can then enable a better understanding of what matters most to service users.

Accessing high quality information through electronic health records

Electronic health record systems have the potential to make sure that the right information is available at the right time. As longitudinal records of people’s health information, they are an invaluable resource for front-line teams needing to access an individual’s up-to-date health history and make decisions about their care and treatment. Collectively, they also contain information about the health of populations, which can inform decisions about what services and support are needed, both now and in the future, as well as show whether existing models of care are delivering the desired outcomes. For example, Northern Ireland’s Encompass programme is showing how health and social care records can be integrated and made accessible through a single system so that patients, service users and health and care staff can access timely information.

The information in electronic health records can also be used to help design evidence-based improvements to care. For example, Cambridge University Hospitals’ eHospital programme (case study 11) used insights from its electronic patient records to develop systems and alerts for the early identification of sepsis, which the team estimate led to a 70% increase in patients receiving timely treatment.

Helping with clinical decision making

Electronic health records can be enhanced through the integration of digital tools like clinical decision support systems. These systems can strengthen medical decision making by matching an individual’s electronic health data to a clinical knowledge database to make personalised care recommendations for the clinician’s consideration. This allows clinicians to access and apply continually evolving clinical guidance when it is needed. Clinical decision support systems are an example of how a technology can be used for both delivering care and enabling improvement – for example by incorporating machine-readable guidance that can be automatically updated and improved over time.

Several hospital trusts – such as Great Ormond Street Hospital for Children NHS Foundation Trust, Nottingham University Hospitals NHS Trust, University Hospitals Birmingham NHS Foundation Trust and Liverpool University Hospitals NHS Foundation Trust – have integrated clinical decision support systems into their electronic patient record systems. For example, University Hospitals Birmingham NHS Foundation Trust has used its electronic health record system to improve decision making in prescribing by:

• supporting clinician behaviour change
• minimising unintended drug omissions
• supporting decisions on dosing
• reducing the unnecessary use of blood products.

Clinical decision support systems have major potential for improving services and have recently been identified as a priority in NHS England’s transformation plans. They have been shown to reduce mortality, harm and life-threatening events. But they depend largely on the availability of ‘computer-executable knowledge’, that is, having clinical knowledge available in a format that a computer can apply, which can present a number of professional and technical challenges to implement.

These data tools and clinical decision support systems can also be enhanced through machine learning to support decision making. One example is risk prediction tools, which can support both individual-level care decisions and population-level strategies. For example, Project Breathe (case study 12) applied artificial intelligence to data from self-monitoring devices to predict cystic fibrosis flare-ups 10 days earlier than would have previously been detected.

Artificial intelligence technologies are also being built into clinical decision support systems so that their outputs are based on more complex data analysis, to improve the evidence base for clinical decision making. Artificial intelligence is also being used in platforms such as mobile health apps and risk assessment tools, which can be a way for people to quickly access, view and understand complex information in order to make decisions about their health care. For example, the RADAR project (case study 10) explored whether a digital platform could support people with type 2 diabetes to make decisions about their health by displaying their risk of complications based on their health data. It also allowed clinicians to understand whether and when to intervene earlier to prevent complications and avoid hospitalisation.

Supporting the design and implementation of improvements

Technology can play a role in explicitly supporting the process of designing and implementing improvements. For example, digital communication technologies such as research platforms like Thiscovery (case study 13), collaboration spaces like the FutureNHS collaboration platform and software repositories like GitHub can accelerate the pace at which ideas, evidence, tools and solutions can be shared. And they can bring together stakeholders from across diverse backgrounds and geographies to collaborate, discuss ideas and design improvements.

Technologies can also support implementation by replicating everyday situations and allowing improvements to be tested in safe, while realistic and immersive, environments through simulation technologies such as virtual and augmented reality.

2.3.2. Challenges

Developing, integrating and scaling up technologies within health care is a complex process, and our research has identified several challenges for using technologies to support LHS approaches. These challenges encompass many elements of the health care system, from infrastructure, workforce and skills, to funding, evaluation and trust.

Having adequate, underlying IT infrastructure is essential, in terms of hardware, software and internet access. A recent House of Commons Health and Social Care Committee report stated that providers will be unable to capitalise on the potential of technology while continuing to ‘struggle with basic IT infrastructure’. But despite a focus from national policymakers, digital maturity remains a significant issue, with one in five NHS trusts still relying on paper records.

The most recent data from NHS England’s Digital Maturity Index also show that there remains a significant disparity between organisations, meaning that some will require greater levels of support than others before they can maximise the potential of existing technologies and effectively adopt new ones. And with integration being a vital element of current reforms across the UK, efforts to improve digital systems will need to go beyond acute care to incorporate primary care, community care and social care providers too. Primary care providers have made significant progress on the digitisation of patient records, but wider challenges remain. Community and social care providers continue to struggle with system fragmentation, legacy IT systems and insufficient access to national funding. While the COVID-19 pandemic has accelerated the uptake of technology in out-of-hospital settings, in particular video conferencing, and reduced the number of ‘digitally novice’ organisations, many community and social care providers still require support with fundamental aspects of IT, including cybersecurity and data protection.

Securing funding for the development and deployment of digital health technologies also remains a challenge for NHS organisations looking to innovate, adopt and spread technology. Recent Health Foundation analysis suggested that the funding landscape is complex and skewed towards early-stage innovation – implying there would be benefits to more support for testing, adoption and spread. Challenges also exist for companies looking to finance innovative ideas. The high-risk nature of innovation, coupled with the long timeframes required for development and the need to deliver a return on investment, mean that many innovations succumb to the ‘valley of death’ between research and commercialisation, with half of digital health start-ups failing within their first 2 years.

Critical to financing innovations, as well as for their wider adoption and scaling, is generating sufficient evidence to show their safety and efficacy in real-world settings. But as discussed earlier in this report, generating this evidence is costly and complex, particularly for digital health technologies that develop at speed, and can be high-risk. Conventional methodologies, such as ‘gold standard’ randomised control trials, are rarely affordable for the health care technology market, which is dominated by small and medium-sized enterprises. They can also be limited in their ability to account for the complexity of many digital health interventions, whose contexts can vary considerably and continually evolve. Lengthy trials may also be unsuited to the rapid pace of digital innovation, with technologies likely to have been updated or even superseded by the time the trial is concluded. Prolonged, single evaluation approaches rarely align with the ‘fail fast, fail often’ mantra of many technology innovators. Regulators such as the National Institute for Health and Clinical Excellence (NICE) also often prefer a spread of evidence from smaller economic, usability and implementation studies. However, the sheer number of technologies available, and the pace at which they develop, also mean that regulators are limited in their capacity to evaluate the ‘tsunami’ of new digital innovation.

Effective development, evaluation and implementation often requires health and care services, innovators and technology companies to work together effectively and there are challenges here too. For the health service, this includes the need to better articulate where technology can add the most value and to spot opportunities for innovation, while for industry this includes closer working with the health service to collaboratively identify and develop suitable solutions that meet NHS needs.

But developing productive relationships between service providers and innovators can be challenging. With the NHS being a national public service, relationships between health care and industry can be complex, and issues such as intellectual property, contracting and data access can slow or even completely stall progress. High levels of system fragmentation and a complex local and national innovation landscape also mean that innovators can be unclear as to who to contact and work with.

Making sure that technology and data tools are implemented effectively and integrating them successfully into workflows are also significant challenges. It is rarely enough to simply add new technologies into existing ways of working and expect transformation. Instead, effective integration requires sufficient capacity to collaboratively redesign roles, processes and care pathways to derive the benefits of the new functionalities that technology offers – in other words, focusing on the whole change and not just the technology.

Central to effective implementation is making sure there are the necessary levels of technological skill and literacy among both staff and service users. Recent analysis shows that a fifth of the UK population do not have even the most essential digital skills, creating challenges both for the effective uptake of new technologies and for making sure that the ‘digital divide’ does not widen existing inequalities. For example, low levels of digital skills, along with poor internet access and language barriers, can mean people remain excluded from services like online appointment booking and remote appointments.

Also fundamental is building public and staff trust in the use of new technologies. After the COVID-19 pandemic hit, new technologies were rolled out with impressive speed, but Health Foundation research has shown that not everyone is yet sold on technology-enabled care. A significant minority (36%) of UK adults surveyed during the first wave of the pandemic thought that while greater use of technology made sense during the pandemic, it was not something ‘for the long term’.

We asked our survey respondents which of the technology challenges highlighted in bold above they considered to be the most significant for adopting LHS approaches; the results are shown in Box 7.

 

2.4. Nurturing learning communities

2.4.1. Key issues and opportunities

The kind of learning and improvement that LHSs undertake typically needs input from many people. This requires a community of people with different roles and backgrounds who are committed to taking part in the process (a ‘learning community’), along with an ability to bring them together, reconcile differing views and take decisions. Such learning communities can include ‘communities of practice’, improvement collaboratives and professional and patient networks. They are groups of motivated people – such as clinicians, patients, carers, researchers, analysts and improvement teams – who come together around a common goal to identify problems and collaborate on designing and implementing solutions and improvements.

Especially because of this, when thinking about and developing LHSs, it is important to not only focus on the technical aspects such as data and technology but also consider the social and relational aspects of learning and improvement. Discussions of LHSs sometimes imply that iterative learning cycles have a kind of automaticity to them – turning practice into data, data into knowledge, knowledge into practice and so on – but as our description of the learning cycle in Figure 1 (Chapter 1) illustrates, sitting at the heart of each of these stages are social processes of deliberation, problem definition, priority setting and decision making. How well people collaborate at these critical stages of the learning cycle will significantly influence how well the LHS works and how successful the resulting changes will be. The stakeholders we interviewed and the literature we reviewed highlighted how building strong learning communities makes it more likely that improvement work will lead to positive, sustainable change. For example, an evaluation of the HipQIP project (case study 15), an improvement collaborative to improve hip fracture care, found that the learning environment created by bringing people together around a common purpose played an important role in the project’s success.

Bringing diverse perspectives to the improvement process

Healthy learning communities typically have a variety of characteristics. They incorporate diverse roles and perspectives. As well as offering opportunities for professionals from a range of backgrounds to collaborate on change, they can also offer a way for patients, service users and the wider public to engage in the learning and improvement process and shape services according to what works for them. This makes it more likely that changes will achieve the desired outcomes. For example, the National Children and Young People’s Diabetes Network worked with children and families from 173 paediatric diabetes units across England who were able to shape services by informing priorities for change, contributing to improvement initiatives and accessing peer support and information. The stakeholders we interviewed also highlighted how the inclusion of diverse voices from a variety of backgrounds can challenge assumptions and help address issues of inclusion, power and paternalism in care. Here, learning communities can benefit from the involvement of voluntary and community organisations because of their ability to harness knowledge and skills in local communities, as the Health Foundation’s Common Ambition programme shows.

Effective learning communities also have trusting relationships and a culture that supports an open and honest learning environment, with the psychological safety for people to raise concerns and try out new ideas and approaches.

Mechanisms for collaboration

Learning communities also need the ability and the mechanisms to bring people together to deliberate, to aggregate or reconcile differing views and to make decisions – whether in person or online. For example, the Health Foundation’s Q Lab UK (case study 14) works with a broad group of people with interest and expertise in a particular topic, including clinicians, people with lived experience as patients, carers and service users, researchers, commissioners and national system leaders. Through collaborative research activities, such as surveys, workshops and interviews, the Lab generates intelligence and shares the findings with teams testing service changes to support improvement efforts.

Meanwhile, the growth of online research platforms such as Thiscovery (case study 13), developed by The Healthcare Improvement Studies Institute, and VOICE, led by the University of Newcastle, allows a wide range of stakeholders to take part in improvement work over the phone, on their computer or face to face.

2.4.2. Challenges

Developing and sustaining inclusive learning communities can be a significant undertaking, and our research found several challenges that can arise in the process.

First, when setting up an LHS, there may be an initial need either to create a new network or learning community or to harness existing ones. The work involved in doing so should not be underestimated. There can be a belief that learning communities are naturally occurring, where interested parties will gravitate together to learn and collaborate. But in many cases, the formation of a learning community requires considerable bespoke design, nurturing and facilitation, not least to bring together a broad range of stakeholders.

Perhaps partly as a consequence of underestimating the work involved in forming and maintaining learning communities, there can be a lack of infrastructure and funding for convening people, to enable discussion and collaboration. For example, there can be difficulties in accessing suitable physical space for face-to-face collaboration, while variation in access to suitable software and hardware can make online collaboration difficult. The stakeholders we interviewed also highlighted how securing backfill to free up clinician time for improvement (getting cover for their role while they are engaged in such work) can be exceptionally difficult, but is nevertheless highly important. This can be especially challenging when there is a desire to collaborate in large groups or beyond single teams or organisations, where many participants may need such support.

Even when the right mechanisms and resources are available, clinicians, data analysts, researchers, quality improvers and other professionals still need to find the time and make the effort to be involved in improvement work. With the health service facing chronic workforce shortages and some of the most severe pressures in its history,, this can be difficult. For LHSs operating at a large scale or over long time periods, maintaining engagement and the dynamic nature of the learning and improvement cycle can be even more challenging.

The time and effort involved can be a particular barrier for patient and public participation; without adequate financial compensation or resources, contributions are often restricted to those who can afford to take part. Some service users also face other barriers, such as ill health or difficulties with communication, which make the burden of participation even greater. Without support to participate, there is a risk that those involved in learning communities are the most motivated and engaged users, rather than a broader, more representative group.

Another challenge is that cultural barriers can make it hard for people to come together around a common set of goals. While it is important to make sure that learning communities are diverse and representative, in some cases, diversity in professional and personal backgrounds, or in organisational approaches and culture, can mean that each stakeholder sees the issues differently, making consensus hard to achieve. In cases like this, the work involved in successfully facilitating the learning community will be greater. Language matters too: when bringing groups with varied expertise together, it is important to make sure that the language used is accessible and jargon free.

Sometimes initiatives can run into resistance to change. For example, overstretched staff already grappling with multiple proposals for reform may see the work of an LHS as ‘just another initiative’ – unless the rationale for change is clearly set out and the work is aligned with broader organisational strategies. On other occasions, staff or service users may be sceptical about proposed changes, or have legitimate quality or safety concerns, especially if they have not been involved early enough in discussions or in formulating the proposals.

Another barrier highlighted through our stakeholder discussions is a lack of clarity about, or competing perceptions of, what an LHS is and what it is trying to achieve. As we discussed in Chapter 1, there is no single definition of an LHS, and discussions can be highly theoretical, making the ideas difficult to apply in the real world. There is also a wide range of models of LHSs, making it difficult for individuals, organisations and systems to decide on the best approach.

Importantly, many of the issues discussed above highlight how successful learning communities – and successful LHSs – require effective leadership. In particular, effective leadership is a critical facilitator for:

• aligning the activities of the LHS with wider operational and strategic priorities
• securing resources and organisational support
• gaining the buy-in of staff and teams across an organisation
• creating a healthy improvement culture.

In this context, effective leadership includes behaviours such as being ‘problem framers’ rather than ‘problem solvers’, listening and observing, and stimulating reflection and critical thinking within teams. (While a detailed exploration of leadership is beyond the scope of this report, it is worth emphasising that this includes clinical leadership, which can have a significant impact on organisational performance.)

We asked our respondents which of the learning community challenges highlighted in bold above they considered to be the most significant for adopting LHS approaches; the results are shown in Box 8.

2.5. Implementing improvements to services

2.5.1. Key issues and opportunities

While gathering data and drawing knowledge and insights from them are key components of any LHS, the goal is to put this into practice and make improvements to care quality and safety. This involves understanding how to identify, design and implement changes to service delivery, and then evaluating the impact of those changes to see whether they have had the desired effect.

Quality improvement skills and methods

Quality improvement – the systematic and coordinated use of specific methods and tools, with the aim of bringing about a measurable improvement in quality – can play an important role in supporting the design and implementation of service changes. Quality improvement tools include ‘driver diagrams’ to understand cause and effect, and methods for improvement such as ‘Lean’ principles or experience-based co-design. Quality improvement approaches sit well with LHSs as they are similarly based on an iterative approach to testing – for example through ‘plan–do–study–act’ cycles.

Training and coaching can help develop the use of quality improvement in practice and maximise its value. For instance, PINCER (case study 2), a pharmacist-led intervention to identify potentially harmful prescribing in general practice, teaches pharmacists how to use quality improvement tools and gives them the opportunity to apply these tools in general practices to help reduce prescribing errors. It includes action learning sets for pharmacists to discuss their experience with their peers and get advice on how to hone their quality improvement skills.

Beyond developing improvement skills, there are a range of other important ways in which the implementation of service changes requires support, which have been discussed in previous sections, such as providing clinical backfill or providing assistance with data analytics and evaluation.

Taking an organisation- or system-wide approach to improvement

Quality improvement tools and methods are not enough in themselves to drive improvement, however. A positive workplace culture that is conducive to improvement also matters, as does the presence of an integrated strategy for improvement at the organisation or system level. It can be difficult to implement and sustain improvements if they are not aligned to wider organisational or system priorities, or in the absence of visible, long-term support from senior leaders who are willing to give front-line teams the space and permission to do things differently.

The relationship between quality improvement and quality planning, control and assurance is also relevant here. Until recently, many organisations have treated these four elements of the ‘quality management system’ as separate entities, each with their own governance processes and strategic goals – a practice that has led to disconnected and sometimes conflicting aims and activities. This is now beginning to change. For example, organisations such as East London NHS Foundation Trust are choosing to take a single approach to quality that encompasses the full quality management cycle. The Trust’s approach is based on engaging staff and the local community at the planning stage to identify where improvement work should be targeted over a five-year period, using quality improvement methods to implement changes, and then making sure improvements are maintained.

Learning across boundaries and improving collaboratively

Learning from peers – particularly those who are facing the same challenges – can be an effective way to build solutions for improving health care services and address unwarranted variation, for example through models such as breakthrough collaboratives and action learning sets. For example, HipQIP (case study 15), an initiative to improve outcomes for people with hip fractures, shows how the collaborative model can drive improvement. Teams from across the UK collaborated to design changes to hip fracture care, test them in their respective organisations and share their results, supported by guidance in quality improvement. But while these kinds of models can be effective, their success depends on the context they are deployed in and whether they are sufficiently resourced.

2.5.2. Challenges

Our experience of funding innovation and improvement projects at the Health Foundation over the past two decades has shown that implementing change in health care services can be a challenging and complex endeavour, requiring specific capabilities and resources.

One key challenge can be a lack of skills and knowledge required to effectively implement improvements to care. Health Foundation research has found that these capabilities are essential for high-performing health care organisations, and they were recognised in the NHS Long Term Plan as an important area for development., A related challenge is how quality improvement is taught to staff.,, Sometimes the focus is primarily on the technical aspects of improvement, neglecting important relational aspects, such as how to build a team, motivate others and secure support. And some quality improvement programmes do not give staff enough opportunities to implement their skills.

Also a challenge is a lack of capacity – the time and space needed for people to effectively implement improvements to care. The capacity required to plan, set up, execute, evaluate and refine service changes is often underestimated and can be very difficult to find when services are understaffed and facing significant demand pressures. And both building capability and creating capacity require resourcing, so a lack of financial and other support to effectively implement improvements to care is another related challenge.

Earlier on we highlighted how the culture of an organisation – the shared ways of thinking, feeling and behaving – is a critical ingredient for an effective LHS. The experience of NHS Nightingale Hospital London (case study 4), which was set up temporarily to create sufficient intensive care capacity to treat patients with COVID-19, shows how an effective LHS relies on the right culture and behaviours among staff at every level – not only among senior leaders. As Nelson and others put it, an LHS is more than simply ‘technical infrastructure’; strong values and a commitment to collective learning need to govern it. Consequently, the absence of a culture of learning and improvement within health and care services can be a barrier to developing LHSs. Research suggests that the prevailing culture in some services is too often defensive, inward looking and hierarchical – characteristics that not only inhibit the improvement of quality but also can have an adverse impact on it.,,

Given the relatively underdeveloped nature of LHSs in the UK, it is fair to say that, while there are many successful examples of improvement going on, few of them take place within an LHS. For instance, they may be time-limited projects or projects not necessarily taking a systematic approach to analysing and drawing knowledge from data. And where actions to improve service delivery do occur within an LHS, there may not be explicit acknowledgement that an LHS approach is being taken. Indeed, many of the teams we interviewed for this report whose work exemplified LHSs were not aware that their projects were taking LHS approaches. For these reasons, while this report documents a series of case studies, there remains a lack of evidence or practical examples of successful LHSs, showing what impact they have had and why. Other issues contribute to this evidence gap too. It is often the case that improvements are incremental and accrue over time. The stakeholders we interviewed told us that there can also be a reluctance to learn from failure, so where attempts to build LHSs have failed, they are less likely to be reported. In addition, our interviewees also indicated that traditional evaluation models such as randomised control trials do not lend themselves to LHSs.

Another challenge for implementing improvements to services is the gap between health care research and practice. Research can provide a critical input into the work of an LHS, and an important objective will be to quickly translate research findings into improvements in practice. But research does not always have a clear practical application and the results do not always reach health care professionals. Even when they do, significant work may be required to interpret and implement the findings. Our case study of PReCePT2 (case study 16), which aims to increase the uptake of magnesium sulphate to reduce brain injury in preterm babies, shows that even though there was compelling evidence to support the use of magnesium sulphate, it took several years of systematic quality improvement work and national support to embed it into widespread practice. Thiscovery (case study 13), an online collaborative research platform developed by The Healthcare Improvement Studies Institute, shows one way in which this gap can be bridged – by bringing together researchers and practitioners and making sure that research projects are designed to have a practical application from the outset.

We asked our survey respondents which of the improvement challenges highlighted in bold above they considered to be the most significant for adopting LHS approaches; the results are shown in Box 9.