Dr Marc Farr on health data analytics

The founder of Beautiful Information has seen big changes in the NHS’ ability to collect data and in its willingness to act on it. He tells Lyn Whitfield that more changes are coming, as AI comes on stream.

Beautiful Information is an unusual organisation; a business-NHS hybrid. Managers and information specialists may have come across it through its work with NHS Improvement to automate the daily Situation Reports that trusts send to NHS England.

The SitReps keep the commissioning board up to speed about the impact of winter – and now summer – pressures on the NHS, by detailing A&E closures, ambulance diverts and the like.

When the automated returns were announced last June, NHS Improvement estimated they would save 36,000 staff hours that would otherwise go on collecting and submitting information manually.

Yet the programme is just a small part of Beautiful Information’s work. It’s core job is to work with local trusts on a range of information products that it can offer to the wider health and care system.

Live updates and analytical dashboards

“Beautiful Information came about because Monitor came to see what we were doing, and then asked us to go into other trusts and help them,” says Marc Farr, the director of information at East Kent Hospitals University NHS Foundation Trust, who is also the founder of Beautiful Information.

“Eventually, our local academic health science network suggested that we should set up as a company that is half private and half owned by the NHS. So, we are organised like a business, with a formal   board and all the governance that entails.

“That allows us to be entrepreneurial, but the equity and the profits go back to the NHS and benefit the NHS. We use our founder trusts [East Kent Hospitals and Ashford St Peters NHS Foundation Trust] as a research lab. They get the benefit of anything we build, and after that we can provide it across the country.”

Beautiful Information has two main types of product: mobile live aggregate status updates that can be used in A&E and ward settings, and dashboards that enable that data to be analysed in more detail.

So, for example, it has an Operational Control Centre app that provides real-time information about demand and bed capacity that can be displayed across a trust or health economy.

And it has an Activity and Finance Tracker that plots activity against financial information, so budget holders can analyse the efficiency of their services and make better decisions about them. Other trackers focus on RTT activity, cancer, maternity and pharmacy.

Getting the data

Farr used to work for Dr Foster; the company that grew out of the Sunday Times’ decision to publish the first ever Hospital Guide back in 2001, with comparative mortality figures based on the work of Professor Brian Jarman.

He worked on some of the software that the company went on to build, before joining East Kent Hospitals, where he focused initially on data quality. Back in the day, just getting data out of proprietary systems could be a challenge.

Now, an increasing number of hospital systems have application programming interfaces or APIs that can be used to export information into web-based applications, using HL7 messaging. But, Farr stresses, the data still has to be high quality.

For example, an app like the Operational Control Centre won’t work if it doesn’t have up-to-date information; which it won’t if medical staff are used to jotting down notes and only updating the systems that feed it after the event.

“When we implement these systems, trusts often need to do some organisational development, to get people out of the habit of handwriting notes, and into the habit of working on electronic whiteboards,” Farr says.

A demand for better data at all levels of the organisation makes it easier to drive this kind of culture change. “Even seven years ago [when Farr joined East Kent Hospitals] there was a culture of reports to the board,” he says.

“It would meet once a month and be told ‘we have done x hip replacements this month and been paid this much cash for doing it’.

“If the target was for the hospital to do 110 hips, and it only did 100 hips, there wasn’t much anybody could do. Now, the management team can look at theatre performance three days in, and if activity is falling behind they can do something; they might schedule a Saturday clinic to get back on track.

“Using data has become normal. People don’t expect Google, or the BBC, to be a month out of data; and they don’t expect the information they use at work to be a month out of date, either.”

Asking new questions

Farr is constantly pushing his teams to move along the “data, information, intelligence, action” continuum. At the moment, Beautiful Information apps work in real time (they can tell a trust how busy it is) and look back at the past (they can tell a trust how much it cost to cover a spike in activity).

But Farr wants trusts to start looking ahead (at when a service is likely to be busy, and at how it might adjust resources to cope). “We are trying to shift from counting how many ambulances are turning up at A&E to predicting whether it is going to be a good day in A&E,” he says.

“I have tried to build an analyst team that goes off and talks to people and asks questions like: ‘what makes a good day in A&E?’ Is it a day that Whitfield and Farr are on duty? And if so, what’s different when Smith and Jones are on duty?”

Prediction should get easier as artificial intelligence, or AI, starts to mature. “We are very interested in AI, in the sense of sophisticated analysis of data,” Farr says. “If you have a full electronic patient record, then it should be possible to get a robot to trawl its [correctly clinical coded] information and tip you off about trends or issues with a patient.

“For example, if it sees references to diabetes and to an amputation, then it should be able to work out that the amputation was a complication of diabetes. Then, it should be able to alert staff that this patient needs to be supported appropriately.”

The AI analytics revolution

Farr has yet another role, as digital lead for Kent and Medway sustainability and transformation partnership; the body responsible for local delivery of the Five Year Forward View’s agenda of integrated health and care services that offer treatment and support as “close to home” as possible.

Using data to analyse population needs and plan new services will be vital to success, and the STP is already working on elements of what is becoming known as population health management.

“By linking data sets, we can do population projections and risk management [working out which patients may be at risk of a hospital admission or other adverse outcome, so services can be put in place to try and prevent this],” Farr says. “We can also do strategic planning; we can start to look at big questions, such as how many frailty beds we will need in 2025, or where to put a new major stroke unit.”

This is another area in which Farr thinks AI will have a major role to play. For example, he says, analysis of very large data sets may show risk criteria for patients who are not targeted by existing public health programmes. Or reveal ways of making treatment more effective.

“A good example might be stroke,” he says. “We know that only 50% of the patients who have a stroke take the tablets that are prescribed for them. If we could work out who doesn’t take the tablets, we could offer them a different treatment or more support.

“It might not take very sophisticated analysis. It might be we soon found out that everybody who didn’t take the tablets was old and frail, or young and female. In reality, it is likely to be a combination of a lot of factors that we probably haven’t thought about – but if you can design an algorithm to find them, then you can design a programme to address them.”

Think big

Over the time that Farr has worked with and for the NHS, some important building blocks have been put in place. Data used to be lost in notes is becoming electronic. More open and interoperable systems are creating new datasets.

AI is about to make analysing those datasets much easier. Which means that new questions can be asked. “We can start to say things like: ‘Everybody called Lyn has had this happen to them’; which was just not possible in the past,” Farr says.

“What we need now is for people to say: ‘If we know this something happens to people called Lyn, then what are we going to do about it? What we need is for people to think big about what can be done.”