Developing learning health system capabilities in the UK National Health Service - using asthma as an exemplar

Abstract

Asthma is a common condition responsible for substantial morbidity and societal impacts, much of which are potentially preventable. This condition, therefore, serve as useful exemplar to investigate the opportunities for developing health infrastructures that allow continuous cycles of improvement. The United Kingdom’s (UK’s) mature electronic health records infrastructure offers considerable opportunity to improve outcomes for people with asthma through the creation of a Learning Health System (LHS). The core idea in an LHS is to use the continuous streams of data emerging as a by-product of clinical care to inform and support policy decisions, organisation and delivery of healthcare and the personalisation of care. This was demonstrated during the recent COVID-19 pandemic when data from across the health ecosystem were harvested and analysed in near real-time to inform decision relating to public policy, public health and clinical care. In this thesis, I reflect on the progress made to advance capabilities to building LHS in the NHS for asthma and summarise outstanding challenges by critically considering eight papers that I first-authored between November 2014 and June 2024.

I begin this thesis with a brief overview of why asthma was used as an exemplar and how LHS has the potential to curb adverse outcomes of asthma and improve health in people with asthma. The concept of Learning Healthcare System had emanated from healthcare. Recognising that health is also shaped by wider determinants such as housing conditions and air pollution, Prof Aziz Sheikh proposed the term Learning Health System to encompass factors beyond healthcare alone. In this thesis, I adopt the definition of LHS as proposed by Prof Aziz Sheikh. Then I proceed to relevant epidemiological and health services research studies and LHS literature that predates and informs my own research and publications. Towards creating an LHS for asthma, I am guided by the four priority areas identified by Health Foundation and Health Data Research UK: i) learning from data, ii) harnessing technology, iii) learning communities and iv) implementing and evaluating improvements to services.

Next, I focus on my research, describing the studies I was involved in. In that chapter I describe three epidemiological and health services research studies and a qualitative study on health information systems I was involved in, which underpinned my submitted publications. For these four studies, I describe my involvement in the study, background to the study, the aims and objectives, design and methods used, implication of the findings and dissemination of findings.

My first paper provided the first evaluation of primary care clinical codes on asthma and allergies as was used in two million GP consultations in Scotland. My second paper is a protocol, which, for the first time, listed the routine and administrative datasets across the four nations of UK, for finding epidemiology and burden of asthma and the methods used to analyse these. In my third paper, I used those datasets to report the first comprehensive asthma burden estimates in the UK. In the absence of linked and longitudinal data, my fourth paper used cross-sectional data to create, for the first time, a population pyramid of asthma morbidity in Scotland. I found that a small proportion of people with asthma make disproportionate use of hospital-based asthma services. The fifth paper described, for the first time, profile of those admitted in paediatric intensive care in England for asthma, with patterns of healthcare utilisation and outcomes.

These exercises involved working with different data custodians across geography, data harmonisation issues and significant time lags between care provision and access to the data. Thus, to explore the barriers and facilitators in utilising electronic health records in near-real time across information systems, my sixth paper was a qualitative study based on interviews of stakeholders and data users in Scotland, which has rich data assets. The next study focused on creating LHS capability and investigating asthma modifiable factors so that targeted interventions could be provided in near-real time. The seventh paper was one of the first publications that analysed asthma modifiable factors that might explain the reduced risk of asthma hospitalisations during the COVID-19 pandemic. This PhD culminated in the development and deployment of a near-real time asthma dashboard designed for use in UK primary care. Using routinely collected data from the Oxford-RCGP RSC, the tool provides weekly feedback on asthma epidemiology, modifiable risk factors, and exacerbations at practice level, benchmarked against network averages. The dashboard operationalises key features of an LHS by enabling a timely, data-informed decision-making tool to support local quality improvement.

This is followed by my reflections in the next chapter, on the findings of my submitted papers, noting how individually and collectively, they contributed to the priority areas for creating an LHS. They have been presented under: i) learning from data: clinical codes, compiling and linking datasets for comprehensive understanding of the asthma burden, focusing on people with high health needs, children admitted to paediatric intensive care, potentially modifiable risk factors for asthma exacerbations, asthma dashboard as a quality improvement tool in asthma care, data fitness for research, data lifecycle approach, accessing data via single trusted research environment with metadata; ii) harnessing technology: overcoming technological roadblocks in real-time data, data integration across systems, modernising infrastructure, call for tech-savvy workforce, data virtualisation, asthma dashboard as an integrated visualisation; iii) learning communities: building collaboration to overcome data challenges, collaboration at national level; and iv) implementing and evaluating improvements to services through real-time monitoring.

After describing my contributions, I attempt an interpretation of my work in the context of the wider literature on asthma and LHS and identify alignments and distinctions. I discuss that by focusing on the role of LHS, my papers and review of the literature offer insights into how LHS capability could be built in the NHS to improve asthma outcomes. I assess the strengths and limitations of my published work and implication of my findings for the existing literature on LHS. I conclude, that to improve asthma outcomes, an LHS approach is a potential solution. My contribution spans i) ‘practice to data’ cycle of LHS by listing the clinical codes and datasets on asthma in the UK nations, ii) ‘data to knowledge’ cycle by providing a) comprehensive estimates of asthma epidemiology, healthcare utilisation and outcomes at every UK-nation level and UK-wide, which can be used for baseline estimates for future studies, b) identification of potentially modifiable factors for asthma exacerbations, c) knowledge on barriers and facilitators that can enable real-time access to data; and iii) beginning of the ‘knowledge to practice’ cycle by deploying an asthma dashboard in RCGP RSC, which now needs promotion for regular usage and evaluation.

My reflections on my research journey involved navigating complex health data ecosystems to investigate how routine data can inform better asthma care. My early studies revealed limitations in coding and data quality, while later work identified modifiable risks and inequalities in asthma outcomes. The development of the asthma dashboard marked a turning point, shifting from knowledge generation to deployment in practice. Key lessons included the importance of data curation, stakeholder co-design, and the structural challenges of sustaining learning systems in real-world practice.

My call to action is to realise the potential of a national LHS for asthma - the NHS must move from data collection to action. This involves embedding real-time feedback tools like the Oxford-RCGP RSC asthma dashboard into clinical workflows, improving coding for treatable traits, investing in local quality improvement, and fostering collaborative learning communities. The UK's world-class data infrastructure can, and should, be harnessed to reduce variation and improve outcomes in asthma care.