Mortality halved for older males receiving AI screening combined with personalised coaching

The findings, published in the Emergency Medicine Journal¹, stem from a randomised controlled trial (RCT) conducted by HN (Health Navigator Ltd) in the NHS, with advice and guidance from the Nuffield Trust. 

The national study was conducted across eight NHS hospital trusts and included 1,767 patients identified at high risk of disease progression and emergency care, using a specialised AI algorithm on their routine health data. 

Researchers then provided these patients with either standard NHS care or remote, telephone-based preventative clinical coaching, led by nurses. Follow-up checks over two years revealed this method – AI screening followed by coaching – reduced deaths amongst elderly male patients, showing 46% fewer deaths for those aged over 75. 

In this group, for every 1,000 patients, the control group reported 280 deaths whereas the intervention group reported 152 deaths, a reduction with 128 deaths. These results were statistically significant.   

For this cohort, the results demonstrated a substantial positive impact, with one additional life saved for every 8 elderly males receiving the intervention over the period of the two year RCT. Compared to pharmaceutical prevention methods such as statins, which avoid 1 heart attack per 60 treated patients over 5 years and 1 stroke per 268 treated patients over 5 years, the impact for these older males was much more pronounced.

The study does not explicitly explain why older males responded well to the intervention. Further work is needed to understand why this group had a marked decrease in mortality whereas other groups had decreased hospital attendances but no change in mortality. However, one potential factor is that elderly males tend to underreport health issues to doctors. An outreach model that proactively identifies high-risk older males through AI screening and offers preventative care may effectively overcome this disclosure reluctance. By reaching out to elderly males with interventions irrespective of self-reported complaints, their health risks can be managed earlier than otherwise possible.

Peter Elcock, a diabetes patient who benefitted from the intervention said: “By flagging me as high-risk early on, the AI technology made it possible for my coach to take immediate preventative actions tailored to my specific needs. With the help of a coach, I started eating right, got essential tests booked and could plan and prepare for a healthier future. The coaching was a tremendous difference for both my mental and physical health. I firmly believe this programme needs to be expanded so its life-changing benefits can reach the patients who need it most before they fall through the cracks.”

Dr Joachim Werr, Founder and Executive Chair of HN said: “This is one of the most robust research studies undertaken in the NHS looking at AI models supporting frontline care, and we are proud of our collaboration with the Nuffield Trust and the NHS. Through our many years of research, we have now developed a novel clinically proven AI model for predictive and preventive care for patients who need it the most. Applying this technology on routinely collected healthcare data provides a paradigm shift for predicting and preventing adverse patient events, harm and deaths.”

Tom Lovegrove-Bacon, Senior Strategic Development Manager at East Kent Hospitals University NHS Foundation Trust (EKHUFT), one of the contributing NHS partners, said: "EKHUFT has been involved in this pioneering study since 2017 and it’s great to see the findings come to fruition. Early analysis at EKHUFT revealed improved survival trends and we published data showing the intervention's potential cost-effectiveness while driving better clinical outcomes. With these clinically validated results now published, the potential of proactive care models is clear to see."

This peer reviewed study provides initial evidence that AI prediction tools combined with personalised coaching is an effective method in preventing the escalation of emerging health risks.

A companion paper published in the British Journal of General Practice²last year by the same group of authors found the AI and coaching model reduced hospital resource utilisation and did not increase primary care workload. On the contrary, nurse-led coaching was found to contribute to a reduction in hospital referrals. 

While further research on improved survival is still needed, these results highlight the potential for AI screening tools in combination with coaching to extend limited care resources through data-driven risk prioritisation. This could pioneer a new era of preventative, personalised and patient-centric population health management.

References

1. Emergency Medicine Journal: ‘Impact on all-cause mortality of a case prediction and prevention intervention designed to reduce secondary care utilisation: findings from a randomised controlled trial’ Bull, L., Arendarczyk, B., Nguyen, A., Werr, J., Lovegrove-Bacon, T., Stone, M., Sherlaw-Johnson, C., (2023): https://emj.bmj.com/content/early/2023/10/11/emermed-2022-212908?rss=1
2. British Journal of General Practice: ‘The impact on primary care of a case-management intervention for reducing emergency attendance: randomised control trial’ Cohen, J. N., Nguyen, A., Rafiq, M., & Taylor, P. (2022): https://bjgp.org/content/72/723/e755