Personalised prevention in England

Bridging the gap between policy rhetoric and practical reality

The NHS in England is under significant strain, with unprecedented waiting lists, staff shortages and burnout, strike action, a post-pandemic backlog and long-standing underinvestment. Preventive action is required not only to tackle current demand on the NHS – by targeting those at highest risk of ill health – but also to improve people’s health and delay their future need for health care. 

In these circumstances, the promise of increasing the effectiveness of prevention through highly individualised interventions is understandably attractive. These personalised – or precision – approaches range from targeted screening based on genetic data to monitoring personalised health information and have been a growing part of political, clinical and public discourse over the past decade. 

But the term ‘personalised’ is vague and could refer to anything from factors such as age and sex to an individual’s gene/environment interactions. For this analysis, we reviewed English health policy documents from 2010 to 2023 to understand what policymakers mean when they talk about personalised prevention, what personalised prevention aims and ambitions have been set out and where more evidence is needed to help achieve them. 

We found that policymakers use the term ‘personalised prevention’ loosely to refer to targeted clinical support based on individual factors such as sociodemographic, clinical, behavioural and genetic information (Box 1). 

Calls for health care professionals to make better use of patient data and associated analytical tools to provide more personalised care have been being made for nearly a decade. However, despite some progress since the Department of Health and Social Care’s 2022 data strategy, inadequate NHS digital maturity – that is, how well digital services meet clinical and patient needs – and data integration remain barriers to delivering the kind of revolution promised politically.

Improvements to outdated IT systems and data digitisation, alongside better join up of data across the care pathway, have the potential to improve health care efficiency. Plans for a federated data platform (FDP) alongside local efforts to create secure data environments for research and development present opportunities to connect disparate datasets across the NHS and test new data-driven tools. Crucially, however, primary care data has so far been excluded from the FDP, and it has been met with resistance from some professional bodies. This is in part due to concerns around how to ensure adequate data governance processes and solve issues of patient consent for use of their clinical data in research. More standardised approaches to health care data access are also needed. 

Improved data and analytical capabilities will also need to be accompanied by workforce training and capacity building to maximise their potential. Further challenges arise from variable public trust in health technology. 

Risk prediction based on genetic information (polygenic risk scores, or PRS) and AI-based medical devices are becoming more widespread and can be used to identify at-risk individuals early on, aid in disease diagnosis and inform treatment. However, the clinical implications of these techniques are so far mixed, with most still under research. 

A range of initiatives are underway in England to accelerate the evaluation and regulation of AI. But it will take time before the NHS can introduce new, validated, population-wide tools. Until then, it must guard against blurring the lines between research and established clinical care. Similarly, biases remain a risk for AI use in front-line health care.

Ultimately, the NHS needs a dedicated strategy for AI to better coordinate the breadth of AI tools being developed. Such a strategy should ensure that risks from biases resulting from underrepresentation are acknowledged and put plans in place to address these in the medium term.

The value-add of PRS is also highly condition dependent. There is some evidence that PRS can help to separate populations into different levels of disease risk based on genetics for areas such as breast cancer, weight gain and coronary artery disease, but this does not necessarily mean improved risk prediction for individuals. Furthermore, evidence of this potential benefit is still mixed or lacking for other major conditions like cardiovascular disease (CVD) more broadly and colorectal cancer. Similarly, the evidence does not yet suggest AI is always superior to traditional clinical risk prediction, though it can aid in some specific areas of disease detection (for example, in detecting CVD through retinal imaging).

The predictive value of PRS and AI also varies by patient demographic. It is well recognised that the whole-genome sequencing studies from which PRS are derived disproportionately represent white European populations and therefore introduce biases that reduce the accuracy of PRS in non-white groups. Given current English demographics, this means these tools may be inappropriate for a fifth of the population, highlighting the ongoing need for better ethnic diversity in genomic data. 

Finally, for most major conditions and in contrast to behavioural, environmental and social factors, genetics contribute only a small amount to absolute risk. Any national personalised prevention programme should not detract from broader interventions and investment into the wider determinants of health. 

This assumption is central to each of the government’s policy ambitions relating to personalised prevention. Many clinical treatments are already personalised based on characteristics such as age, sex and ethnicity – delivering personalisation albeit in a relatively simple sense. Yet the working definition of personalised prevention among English policymakers includes the use of genomics and behavioural data.

There is some evidence for where personalisation using genomics such as PRS may help target screening to higher risk groups, including for breast cancer, but such evidence is mixed or lacking for other conditions and behavioural areas, such as newborn screening, CVD and nutrition. Even where screening with genomic data can help identify increased risk, nuanced communication of results requires training, with concerns around differentiating risk from diagnosis, managing the emotional and clinical impact of uncertain results and whether genetic screening results lead to behaviour change. There is also evidence that people interpret and react to genetic information differently and that clinicians do not always have access to the necessary preventative services for referring patients. These limitations skew the balance of the benefit versus harm of introducing genetics into population screening programmes and challenge agreed-upon principles for screening.

In behavioural interventions, most personalisation is still reliant on low-tech, human-led approaches and based on measures such as height, weight, age and self-reported physical activity and diet. Health technology may evolve towards enabling dynamic interventions based on personal preferences and in response to feedback, but current evidence and technology to realise this level of personalisation requires scrutiny. A review of self-titled precision health behaviour change interventions found no instances of consideration of social/environmental context. For example, a randomised, controlled trial of a mobile app-based weight loss intervention offered personalised step-count goals based simply on an individual’s baseline, and nutrition advice tailored to the population group (new mothers) but not an individual’s personal preferences, financial situation or dietary requirements.

There is, however, some evidence that tailoring pharmaceutical interventions on the basis of genetic information – pharmacogenetics – could be valuable in discrete settings such as oncology, mental illness or statin therapy, but the approach is unlikely to be widely applicable any time soon.

People in England do not currently have access to granular health and risk data. The NHS App allows people to view their personal health records, but there are no immediate plans to communicate health status and risk to app users or provide behaviour change advice. The forthcoming digital NHS Health Check hints at this potential future functionality in the context of CVD.

However, it is well established that information alone is insufficient to drive behaviour change, as it fails to account for the more persuasive physical, social and environmental factors contributing to decision making. For example, introducing feedback letters into England’s National Childhood Measurement Programme to alert parents to their child’s overweight status led to increased awareness but limited behaviour change. Greater exploration of how to enable these personal decision-making factors is needed to understand how access to personal health data can lead to improved health. 

Technology has the potential help more people be actively involved in managing their health and long-term conditions outside of health care settings, which could improve patient empowerment, experience and health outcomes. 

While progress has been made on remote monitoring for some groups of patients on ‘virtual wards’, there is still limited deployment of technology in England to support personal remote monitoring and self-management. For instance, the latest NHS England Digital Roadmap suggests the majority (80%) of products deemed for individual use will not be deployable in the next 3 years. 

With smartphones and wearables more widely accessible, people are increasingly able to monitor metrics such as physical activity, diet and heart rate. These devices are already being used recreationally, in government-led pilots (for example, the Wolverhampton Better Health: Rewards scheme) and in some areas of clinical practice (such as the digital NHS Diabetes Prevention Programme). However, their reliability and validity remain variable, particularly relating to heart rate monitoring and energy expenditure. There are currently also few examples of easily sharing data from wearables into clinical systems, meaning data sharing with a health care professional is dependent on the individual and risks overwhelming clinicians with information they cannot interpret or act upon.

The ease with which self-management tools can be used and the ability of individuals to engage with their own health and disease management also needs to be monitored. For example, offering the NHS Diabetes Prevention Programme digitally led to weight loss equivalent to face-to-face programmes, but digital delivery may be more suitable for some groups than others. A digital NHS Health Check was trialled in Cornwall last year, but evaluation results are yet to be made public.

Health Foundation research suggests that highly activated patients – those with greater inclination and capability for self-management – use less health care. However, most patients are not highly activated, and the assertion that self-management will be capacity releasing as opposed to labour generating through issues such as managing technical issues, quality assuring readings and handling previously unmet need and new induced demand remains largely untested.

One risk to monitor, in line with wider evidence on digital inclusion, is that we would see an initial widening of existing inequalities in those engaging with these technologies favouring the younger, healthier and more affluent, with the potential to create additional NHS demand in the short to medium term.

Previous governments in England have had major ambitions for the role of personalised prevention in shifting the dial in disease prevention and condition management. However, these ambitions have been underpinned by five key assumptions that need more dedicated testing, evaluation and evidence to bridge the gap between rhetoric and today’s reality.

It is important for the new government to recognise that while personalised prevention approaches are likely to be part of the solution for improving health and reducing NHS demand, it will take time for their potential to be realised and to ensure they help improve the health of all. Many of the opportunities for improving personalised prevention, population health and patient empowerment through better data collection and analysis and risk identification – including across health and non-health datasets – need more investment in evaluation and evidence to be well understood.

As part of a more strategic approach to digital, data and AI in the NHS, the new government should therefore prioritise experimentation in areas where there is the strongest evidence of effectiveness while continuing to build capability and invest in robust research and evaluation in other areas.

Ultimately, any investment in personalised prevention and precision medicine must not come at the expense of action on the wider determinants of health. Instead, new approaches to personalised prevention should be just one part of the broader action required to deliver the government’s ambition to improve health and narrow inequalities.