Benedikt von Thüngen, chief executive and founder of Sanome, argues that many hospital-acquired infections could be avoided altogether through clinical AI. 

Every winter, we talk about pressure on the NHS as if it is a surprise, but it’s not. Bed occupancy regularly exceeds 92%, and hospitals operate with little to no margin for error. When there is no headroom, small failures can compound the situation fast.

Hospital-acquired infections are one of those failures, and they shouldn’t be. Around 8% of inpatients in England acquire a hospital-acquired infection during their stay, rising to roughly 15% in higher-risk cohorts. These infections cost the NHS more than £2.7 billion a year in direct hospital costs alone. These are staggering stats, which highlight the need for change. 

It’s well known that winter exacerbates the impact of hospital-acquired infection. Older patients are sicker, more frequently, and often admitted to the hospital for longer. Wards are full. Staff are overstretched. When signs of infection are missed, the infection quickly and effectively takes hold. In this environment, escalation is fast, and recovery is slow. Beds stay blocked, and patients’ lives are at risk. The sad thing is, it doesn’t have to be that way. 

Why early detection matters

A significant proportion of hospital-acquired infections could be avoided altogether. Estimates suggest up to 55% could be prevented with earlier identification, more timely intervention and better infection control practices. Yet our systems are designed to react, not anticipate.

NEWS2 (the NHS-standard National Early Warning Score v02) has helped standardise how we respond to deterioration, but it only flags risk once physiological decline is already visible. By the time infection risk is flagged, deterioration is often already underway, and the window to prevent escalation has closed. The opportunity to stop the problem before it starts has already been missed. 

Add in winter bugs and increased hospital capacity into the mix, and the conditions that make early detection possible are under great strain. Clinical teams are managing higher patient volumes, increased acuity and limited staffing headroom. 

These delays matter. Patients who deteriorate are more likely to require higher acuity care, remain in hospital longer and occupy beds that could otherwise be turned over. The cumulative effect is reduced flow across the hospital and increased pressure on teams already working at capacity. Data can make all the difference in this situation.

Hospital-acquired infections

The data is already there; we just need to use it properly

Hospitals generate vast volumes of clinical data every day: observations, blood results, medication records, clinical notes. Around 80% of it is unstructured, locked in free text, fragmented across systems and effectively inaccessible in real time. Consequently, an estimated 97% of data from hospitals remains unused. 

No clinician can reliably synthesise all of this across dozens of patients during a winter ward round. That is not a failure of clinical skill. It is a limitation of the system.

This winter has made this limitation painfully clear. Flu admissions are at the highest level ever recorded for this time of year, increasing up to 55% each week so far this winter, with no peak in sight. Thousands of beds are already occupied by patients with respiratory infections, before secondary infections even enter the picture.

When infection risk is missed in this environment, escalation is rapid, recovery is slower, beds stay blocked longer, and pressure compounds.

This is exactly where clinical AI can make all the difference. Not by replacing clinicians, but by doing what humans cannot do at scale: tracking patterns over time, connecting signals across data types, and identifying risk before it becomes clinically obvious.

MEMORI is built with clinicians and shaped by real clinical practices. Designed on the ward, for the ward, it has been co-designed with frontline teams to define locally relevant next best actions, pathways and workflows. Alert thresholds are set with clinicians to reflect different patient populations and risk tolerances, helping reduce alert fatigue rather than contribute to it. Interfaces and summaries have been developed iteratively with users to minimise cognitive load during already pressured shifts.

It continuously analyses data already held within the electronic patient record to identify infection risk up to 72 hours earlier. Not as a black box alert, but as a clear, explainable signal delivered directly into existing clinical workflows.

Earlier insight gives teams time. Time to investigate. Time to intervene. Time to prevent escalation before it becomes another winter bed blocker. Timing changes everything.

Earlier detection prevents bedblocking

When infection risk is identified earlier, clinical actions change. Diagnostics are ordered sooner, treatment starts earlier, patients are isolated when it still prevents infections from spreading, and deterioration is slowed or avoided altogether.

The impact is practical and immediate. There are fewer escalations, lower acuity, shorter lengths of stay; beds turn over faster, and the flow improves.

We have seen this principle before. The TREWS sepsis early warning system showed an 18.7% relative reduction in mortality when alerts were acted on within three hours. Different conditions, with the same lesson. Timely insight saves lives and capacity.

In a winter where flu alone is filling the equivalent of multiple hospital trusts, even small reductions in excess bed days matter and can make a tangible difference. That’s why we need to switch from a reactive to a proactive clinical model. 

Stop reacting and start anticipating

Winter pressures are not an anomaly. They are a condition of the operating environment. The NHS cannot build its way out of this with more beds alone.

Hospital-acquired infections will never be eliminated entirely. But many can be intercepted earlier, before they escalate and consume scarce capacity at the worst possible time of year,

We already have the data. We already have the clinical expertise. What has been missing is the ability to bring them together early enough to change outcomes.

Earlier detection is not a nice-to-have. It is a requirement for a resilient NHS, especially in winters like this one.