Special Report: BI and analytics


Using PACS and RIS data: see what can be done

Imaging systems hold a wealth of data, and suppliers are increasingly working with trusts to make sure they can use it to improve efficiency and patient pathways. Kim Thomas reports.

Along with the picture archiving and communications systems used for storing images, radiology information systems are part of the standard toolkit of hospital radiology departments.

The RIS plays a critical role: handling referrals, making appointments, managing workflow and keeping a record of radiologists’ reports on imaging procedures.

As a result, it holds a rich repository of operational data that can show how efficiently the department is performing – how many films are unreported, for example, or which patients are regularly not attending appointments.

Most PACS and RIS vendors offer business intelligence systems to collate and analyse that operational data, so that trusts can use it to improve performance.


Clemens Janus, general manager, analytics, cardiology and high acuity care at GE Healthcare, says that demand is very high.
“Every single customer is asking for help. The reason is that everyone understands that there are process inefficiencies, a shortage of radiologists, increasing patient numbers, all happening at the same time.”

This is confirmed by one PACS manager who did not want to be named, but whose department is using the HSS CRIS Insight tool: “Currently the NHS is under pressure to deliver diagnostics in a defined time frame, whilst constantly trying to meet an ever increasing demand for these services.
“There is therefore a pressing need for readily available data to ensure that services are monitored closely in terms of both performance, and also planning of the outstanding workload ahead.”
Improving operational efficiency

Janus says there are two main ways in which RIS data is used to improve operational efficiency: waiting time optimisation – looking at how many patients are waiting too long for a CT scan, for example – and productivity.
It’s possible to benchmark within the same hospital, comparing waiting times or output between different teams in the same department, and analysing why some are more efficient than others. As Anjum Ahmed, global marketing manager, enterprise imaging at Agfa, points out, business intelligence tools offer a hard return on investment.

This type of data analysis is usually known as descriptive analytics, but vendors are now seeing demand for tools that can deliver predictive analytics, says Thierry Verstraete, product manager for the business analytics at Carestream.
“Our customers are asking us about making predictions with a certain level of confidence for the number of CT exams or the number of MRI exams in the next three or six months or year.

“Essentially, customers are looking to leverage predictive analytics to do proper capacity planning, to understand when they need to add a modality to the department or at least budget for it, or when to start looking at adding another radiologist to their reading teams.”

Predictive analytics are not an exact science, says Verstraete – they give predictions within a certain range – and have to be used with care, because a significant change, such as a modality breaking down for an extended period of time, will affect the data used as the basis of the prediction. Seasonal fluctuations similarly need to be taken into account.

Customers are also beginning to use radiology data to inform an enterprise-wide view, says Ahmed: “The intelligence you’re driving out of the system should be based on following the patient journey.”

Getting a better view of the enterprise

Jane Rendall, managing director of Sectra, agrees: “On an enterprise level, you need to be able to push the information out of your RIS, out of your PACS, out of whatever other applications you’re using, into a separate database which is accessible remotely from the department.

Then, the chief operating officer or chief information officer can drag data and understand how it’s impacting the whole service.

“One of the things they want to do is to get patients out of bed quicker, so where are the hold-ups? Why has this patient not been transferred out of hospital? Is it because they’re waiting for an ultrasound in radiology?

“When I look at radiology and see that this patient has been waiting for x number of days, why is that? So you begin to get a more holistic view of how the enterprise works.”

Trusts can also work together to use business intelligence for comparison purposes, says Janus: “We can compare the CT team in hospital A, with the CT team in hospital B, and identify what’s going on, and also drill down to identify the bottleneck in the hospital that has longer waiting times.”

Starting to think about demand

At the same time, Ahmed points out, business intelligence provides the opportunity for trusts within a region to manage the care pathway more effectively across their organisations.

An example of what might be possible in future comes from Ireland, where McKesson has rolled out RIS to every hospital in the country, connected by a single central database.

This means that approximately a million HL7 messages come into the database every day, says Ray Cahill, vice-president sales at McKesson.

“Each one of those messages tells a story about a patient journey. If you apply proper BI techniques and methodology on top of that, that supplies fantastic insight.”

That could include, for example, information about how many patients have to wait more than six months for an urgent MRI, making it possible to shift patient care to a different, less busy centre to reduce waiting lists.

This national analysis of operational data is something that McKesson is now working on with Ireland’s Health Service Executive, along with more advanced analytics, linking the reason for referral to the patient outcome.

Getting into the predictions business

There are two other major trends on the horizon. One, says Verstraete, is towards prescriptive analytics, where data is used to provide guidance on what steps should be taken to achieve particular business goals.

The other is unlocking the valuable information held in trusts’ PACS stores that now date back ten years or more. Roy Kinnear, UK sales director of Intelerad, says a new generation of tools can analyse those images to inform radiologists’ decision-making.

The radiologist would go through the normal reporting workflow, but at the same time algorithms would be applied to thousands of current and prior studies to identify patterns and alert the radiologist to areas of potential interest.

“The algorithm is picking up from the current and historical studies and highlighting where they may be issues,” says Kinnear.

It is still early days for the technology, though Intelerad is in discussions with an NHS trust. But at a time of increased pressure on the NHS, a tool that saves time for the reporting radiologist is likely to have a strong appeal.

Kinnear adds: “It may save 12 or 24 months of treatment or endless callbacks by picking up issues that can be treated earlier in the patient’s pathway.”

Really, really big data sets

In the longer term, Rendall thinks that imaging data could aggregated and analysed across trusts (at, for example, a regional level), providing greater potential for computer-aided diagnosis.

“If you are looking at a chest X-ray and at a specific disease, say pneumonia, you have a huge number of pneumonia patients with a clear diagnosis, clear treatment path and clear imaging.

Now if you got a computer to look at millions of images and say, ‘This is what pneumonia looks like on a chest X-ray,’ maybe in the future, a computer could say, ‘Have you considered pneumonia in this situation?”

The aggregated data could also be analysed to see, for example, “how many people came for a chest X-ray, what sex they were, what age they were, what the outcomes were.”

This, says Rendall, will require vendors to collaborate so that their applications will be able to drill down into relevant imaging data where necessary. The move to local devolution in Greater Manchester, she believes, could provide the ideal vehicle for mining large-scale datasets to better inform clinical pathways.

“The NHS has a golden opportunity to own that data,” Rendall adds. “They need an efficiency-based approach to measure the patient’s entire journey and to understand where the key issues and bottlenecks are and the opportunities to improve.”



Population health management: analysing change

There is growing interest in population health management, or aggregating data from multiple IT systems to change clinical service and financial outcomes. Lyn Whitfield asks why this is happening now, and hears the big driver is policy.

There is a shift underway in healthcare analytics. Certainly, there is still an important role for what might be called traditional business intelligence – surfacing, analysing, and reporting on the data held by specific systems, departments or organisations.

But interest is growing in what is called population health management – aggregating data from multiple systems hosted by different organisations, and using this to inform changes to clinical, service and financial outcomes.

Or, as Graham James, vice president of CACI, puts it: “We are seeing a move towards a wider range of data being collected and used to deliver health and social care in a different way.”

Barry Duke, group analytics product manager at Emis Health, says this has really picked up in the past few months. “Two or three years ago, we were talking to the NHS about this, but the time wasn’t right for a number of reasons. However, this year people have really engaged.”

Getting the data

Why now? Practically, a lot of effort has gone into sorting out the data, technology and information governance required to make a start.

Ashley Woolmore, senior consultant at IMS Consulting Group, says initially a lot of was done trying to tap into and link-up existing data sets, but there are limitations to this approach.

For instance, he points out that while the ONS and NHS Digital have very relevant datasets, combining them is challenging “if you want to look down to patient level, for example down to the level of a patient with, say heart failure.”

Indeed, as people start to live for five or ten years with heart failure or cancer “death (which is only partially available in a dataset such as the Hospital Episode Statistics) might not even be the appropriate outcome measure, particularly where evaluations need to focus on a shorter time horizon.” Sometimes, bespoke data collection is needed.

Yet the traditional way of doing this, through randomised control trials and one-off studies are time consuming and expensive; and might still not provide the kind of trend data needed for service planning.

“What we are interested in doing, and what the NHS is interested in doing is not setting up a seperate study every time [we want to know the impact of a new treatment or intervention],” he says. “We are putting the infrastructure in place to monitor outcomes continuously.”

Cloud changes the BI weather

On the technology front, Duke says the big change has been the arrival of public cloud services, such as Microsoft Azure, that dramatically cut the hardware and software costs of storing and crunching data.

“The cloud is just the most cost effective way of doing this,” he says. “You can pay for what you use, and if something does not work out, you can just switch it off.”

The NHS has been slower to start using the cloud for analytics services than other organisations, partly because of concerns about security and information governance. But the bodies tasked with advising the health service on these issues seem to have decided that this is the way to go.

For example, Duke says, the NHS Digital has said it is acceptable to store patient identifiable data in a public cloud, subject to risk assessment and as long as it does not leave the EEA (or, if it originates from named central bodies, England). Microsoft and Google are both investing in UK data centres.

Meanwhile, Dame Fiona Caldicott’s latest report on information governance gives a significant boost to central data collection and certain uses of information, including risk stratification, despite the new patient opt-outs offered for regulation and research.

Vanguards lead the way

Even so, the big driver is policy. The NHS has a long history of experimenting with integrated working and a shorter one of using data-driven techniques such as risk management to try and divert patients from hospital by providing more support in the community.

But the Five Year Forward View plan to try and close a £30 billion gap between funding, rising costs and demand (and particularly demand) that could otherwise reach £30 billion by 2020-21 takes this kind of thinking to a new level.

It puts a big emphasis on new models of working which reach out across both health and social care; where, as James points out, councils have an even more pressing need to find efficiencies, and are further along the road of changing both their own structures and the channels through which they deal with users.

The Forward View’s ideas are now being tried out by 29 vanguards around the country. The acute-led, or ‘primary and acute care systems’ vanguards have helped to pull new players into the analytics market.

The Salford Together vanguard is using the population health management system from Allscripts, its acute electronic patient record provider, while the Wirral Partners vanguard has brought Cerner’s population health management to the UK on a similar basis.

The established analytics providers are also involved with the vanguards. CACI is working with the Better Care Together vanguard that is focused on the Morecambe Bay Health Community. It is looking to create a system that will “take responsibility for the whole health and social care needs of the population, within a single budget.”

Information wise, “it is really focused on a big data solution that pulls together GP data, hospital data, social care data, and possibly, in the future, information from not for profit providers,” says James. “There is a lot of interest in getting information from wearables, as well.”

Emis is involved with the Fylde Coast Local Health Economy, one of the ‘multi-speciality community provider’ vanguards, that wants to create ‘wrap around’ care for individuals, using community teams. (IMS Health is in advanced discussions with a vanguard, focused on improving outcomes and shifting services for a specific disease, although it can’t yet say which one).

The challenge: the money

The flip side of policy driving interest in population health management is that the financial crisis that has inspired that policy puts challenges in the way of investing in it, and in the kind of service change that it is meant to inform.

“Technically, this is very doable. It has to be pulled together in a secure way. You have to consider the information governance. There are issues with collecting information from wearables. But none of that is undoable, and it cannot be a barrier,” says James.

“What we need is for the money to go with this change. We need to get the money out of fixed assets, like hospitals, and into the community. We may need new CQUINs to support this. Or, maybe, NHS England needs to look at some transition funding.”

NHS England is surely well aware of this issue. Several vanguards, including Morecambe, are effectively collapsing some of the barriers between commissioning and providing that make it hard to shift money between provider organisations.

This year’s operational planning and contracting guidance says some CQUIN money will be available for taking part in sustainability and transformation plan work; while signalling a bigger shift towards ‘control targets’ for STP areas that NHS England chief executive Simon Stevens has said could mean an end to payment by results.

Even so, with the acute sector finishing the last financial year £2.4 billion in deficit, trusts being asked to accept ‘control targets’ to get this down to £250 million this year, and tight controls imposed on access to the relatively small sustainability and transformation fund, transition funding looks unlikely.

Still, Woolmore also feels this is the big challenge. “The only way the health sector generates income is through activity. There is a dynamic to increase activity and unless we tackle that, and allow financial rewards for stuff we do that is not payment by results, we will not get through this crunch,” he says.

“For me, the essence of population health management is reallocation; the question is whether we can do something better by putting resources into different treatments and services.

“Potentially, the only way to do that is to offer some protection to the acute sector for a period of time. We need to buffer change by providing financial continuity; but that is not where we are at the moment.”