AI-guided prediction and nurse-led prevention of patients clinical crises: A powerful solution to the challenges facing the NHS demand equation

In the first of a series of blogs ahead of the NHS ConfedExpo conference, Prof Matthew Cooke, Dr Steven Laitner and Dr Joachim Werr recently wrote about the challenges facing the NHS and described four opportunities for radical change.

In this article, Mark England (CEO, HN) and Graham Prestwich (Patient Director, HN Advisory Board) describe how predictive healthcare can shape the future of healthcare delivery.

Why is Predictive Healthcare a critical part of cost effective population health management?

In much of healthcare the focus is on data, but not necessarily the right data. Activity levels are often given much attention – visits to emergency departments (EDs), number of admissions, counting outpatient attendances. However, to meet increasing demand, particularly of people with new and ongoing unmet needs, we must understand the problem facing the NHS in more detail. This requires a fundamental shift toward person-centred rather than organisation and event-centred thinking.

NHS urgent and emergency care (UEC) performance standards have been declining for the last decade, and for many patients and families this is having a profound impact on their health and quality of life. There is compelling evidence that the harm caused by longer waiting times causes patient harm and leads to worse health outcomes. For every 80 patients who wait over 6 hours there will be one additional avoidable death. What are the root causes of this problem? Is it more activity, workforce challenges, pandemic-deferred presentations? The most consistent root-cause is delayed discharge from EDs where there are insufficient available beds. Leaving hospital processes and community integration aside for a moment and examining individuals alone, there are some interesting patterns emerging. In all healthcare systems examined, just 1% of the population will take up between 53-76% of beds reserved for unplanned care in any one year. In an average hospital catchment this represents 2-3,000 individuals. These people are typically living with multiple morbidities and often reside in more deprived communities.

It’s important to note that the same people are not returning to hospital year-on-year. People enter this cohort for 12 months, 10-12% die in any given year, 13-14% remain using a large amount of bed days, but almost 70% return to average or below levels of care consumption So, If routinely collected data can be used to reliably identify these patients early, it creates the opportunity to address their unresolved issues before they become a medical emergency in their “impactable” window.

Finding this critical segment in time to provide the right support is where predictive analytics supports predictive care. In recent modelling, using routinely available healthcare data, it is possible to predict between 80-83% of all individuals who will become an emergency admission. This is astounding and presents a potential paradigm shift for health and care systems. If we know who will have an avoidable and adverse event such as an emergency admission, then we have a strong case to move away from a reactive care system towards more predictive and preventative care.

What role do patients play in predictive and preventative care?

For society to achieve good health and wellbeing outcomes, there are a number of critically-important activities that need to take place:

  • Timely presentation and information-sharing by patients, and accurate diagnosis by clinicians
  • Effective and affordable treatment, consistently accessible to everyone
  • Timely and ongoing treatment maintenance and reviews
  • Access to the right urgent care services at the right time based on need
  • People adjusting their lifestyle to optimise the benefit of interventions.

This list is not exhaustive but illustrates that for each activity the patient can have a profound positive or negative effect on outcomes depending on how each element of care is managed. In other words, the performance of the patient as a competent project manager can make all the difference. A poorly managed project not only fails to achieve its goals, but it also usually means additional, often expensive, resources and expenditure of additional and extra effort for an overburdened system.

How predictive health and care works in practice – an example

It's worth discussing HN’s recently published study in the British Journal of General Practice.

Applying AI to routinely collected hospital data in York and Scarborough Teaching Hospitals NHS Foundation Trust, HN was able to identify patients at high risk of unplanned care. HN then contacted these patients through their GP for a nurse-led preventive intervention. As the study was randomised, some patients who consented received the nurse-led “extra” intervention, whereas the rest were supported through NHS standard care. All patients followed for at least two years.

Some of the patients were surprised at being proactively identified and approached by the NHS as this was not the way they were used to interacting with healthcare services. After explaining the reason and purpose behind the intervention, none objected to the approach and over 50% onboarded the trial.

This is an example of a data-driven approach leading to care outreach to patients, rather than waiting for the patients to come to A&E.

The findings after 2 years were in several aspects expected as well as unexpected. Comparing contacts with the GP practice in both groups, there was no difference. Both consumed 46 contacts per annum per person of which 25 were face to face. This is almost a weekly contact with primary care on average, and a face to face every fortnight. Interestingly, when looking at who consumed care in both groups, patients aged ≥80 years in the intervention group had a 33% increase in the mean annual rate of primary care events compared with the control arm. And patents <80 years instead reduced their need for primacy care contacts with 10%, offsetting the increased care needs in elderly patients. Importantly, the rates of referrals to secondary care were 26% less in the group receiving nurse-led support compared to those under NHS standard treatment. All changes quoted above were statistically significant (P<0.001).

The predictive and preventive approach significantly shifted primary care resources towards older patients and away from secondary care. When interviewed in a comprehensive PPIE study, many patients expressed their initial surprise that the NHS can find them rather then them needing to seek care, and that they did not object to their data being used for such preventative purpose. You can watch some of the patients share their experiences from the study here.

What does this mean for the NHS?

With the NHS under increasing strain, HN’s approach represents an important opportunity to ease the pressures across primary care services. HN’s practical and scientifically-documented example of predictive and preventive care represents a paradigm shift.

For the NHS, HN’s approach is easy to implement and transformative in nature. HN’s predictive engine uses routine healthcare data and does not require extra data collection. The implementation did not disrupt existing clinical pathways and could be smoothly integrated among existing providers. As nurses worked remotely, the model did not compete for or “cannibalise” local clinical resources. Despite being delivered remotely, the preventive, nurse-led intervention showed an increase in patient engagement, activation and quality of life.

There are close to 2 million GP referrals to consultant-led care every month. With the predictive health and care approach it would be realistic to prevent about 500,000 of these, while delivering on the triple aim of improving health, care experience and reducing costs.

UEC is transient and patients enter and exit their clinical crisis while often needing hospital care for days or even weeks. Today we are missing that many of these clinical crises could be both predicted and prevented, at lower cost and better patient -and system outcomes. Predictive health is one of the most powerful solutions to the challenges facing the NHS demand equation. So the question remains: Why do we continue letting clinical crises happen when we can - to a large extent - predict and prevent them?