The fortnightly newsletter that unpacks system leaders’ priorities for digital technology and the impact they are having on delivering health services. Contact Nicholas Carding in confidence here.

The government has made it clear it will not end the lockdown until it is satisfied the NHS can cope with demand. But how will the NHS know when it is in such a position, and what are the tools at its disposal to achieve this? 

One key factor is demand modelling. This is not a new concept, but it is something the NHS does not always manage to do well.   

The key to acute hospitals regaining some degree of normality in the next few months lies — to some extent — with managers being able to keep intensive care capacity ahead of demand.

Therefore, having the ability to accurately forecast ICU demand is something managers are yearning for.

Excitingly, the NHS may be closer to this than many realise.

Predicting the pressures 

Last week, it emerged that NHS Digital and Public Health England are piloting a system that uses machine learning to help predict the upcoming demand for ICU care.

Developed by NHS Digital and the University of Cambridge using PHE data, the system is being piloted at four hospitals in England, after which there will be tweaks and fine-tuning and — hopefully — full national deployment.

The Download has been told the system can determine what the likely ICU demand will be in both two and seven days’ time — information which will be gold dust for hospital managers.

Crucially, this will allow hospitals to obtain the right level of ventilators, oxygen, drugs, and other equipment when needed.

Mihaela van der Schaar, the Cambridge academic who has led the system’s development, believes hospitals will be able to say — with a “high level” of confidence — that 30 of 40 ICU beds will be occupied next week.

Professor van der Schaar and her team are also hoping the system can be used to predict a patient’s length of stay in hospital, which will help trusts enormously with things like freeing up more capacity and mutual aid.

It all relies on good data, however. The data that currently informs the system is PHE’s “covid-19 hospitalisation in England surveillance system”, but NHS Digital hopes to integrate more data to improve the accuracy further.

The end result is hoped to be a system that not only provides vital statistics about covid-19 patients admitted to hospitals, but also the forecasting element and a “simulation environment” that allows managers to test the effect of different scenarios, such as changes in the profiles of patients admitted and the subsequent impact on required resources.

Time for a rethink?

The work between NHS Digital, PHE and University of Cambridge is just one of many examples of how having good data and the ability to use it for modelling and analysis will be essential to help the NHS limit the impact of covid-19.

The need for more development in this area was neatly summarised by former NHS England deputy chief executive Matthew Swindells in a webinar hosted by advisory group Public Policy Projects last week.

Mr Swindells, whose previous role at NHSE included overseeing large parts of the NHS’ tech agenda, called for a “profound rethink” about data use within hospitals and their supply chains for vital equipment.

“We should be able to identify problems in advance and buy ourselves days’ and weeks’ worth of warning,” he said.

“From infection rates, we can predict admission rates, from admission rates we can predict length of stay, from length of stay we can predict bed numbers.

“From the path of disease, we can predict ventilator and ICU need, equipment need, medicines need, and disposables need.

“From delivery lead times to a venue of care, we should know what storage is needed locally, and, from supplier lead times, we should know what we hold in our warehouses.”