Identifying the key data analytics challenges faced by NHS CIOs will be the focus of an interactive workshop on day two of the Health CIO Summer School, 14 July.
In advance of the workshop Daniel Ray, director of data science at NHS Digital, who will be jointly leading the session, spoke to Digital Health News editor Jon Hoeksma, about what he believes some of the challenges are, and invites NHS IT leaders to share their views using the embedded poll to support the session.
“What we want to try and find out in advance of the workshop are what are the big data analytics challenges that people face, and how can we help support organisations to overcome them,” explains Ray.
Range of analytics maturity across the NHS
“We’ve got quite a diverse range of maturity on analytics across the NHS. Some organisations have got pretty sophisticated capabilities while others have not yet got anything in place on enterprise wide analytics”
For a CIO in a trust what are my main Business Intelligence challenges? asks Ray. “How can I make sure make sure that data is timely, relevant and, most importantly, actionable.”
Simply getting data out of systems remains a challenge
“A challenge that many trusts still encounter is just getting data out of systems, in the right format and that they can sit reliable analytics on top of”, says Ray.
“The key is getting data out of clinical systems. Unfortunately, sometimes people don’t think in procurement about getting data analytics out. One problem is that data is often in a live operational clinical system – so just getting at the data and in a form you can use in a timely reliable way is a challenge. Some people have built data warehouse solutions themselves or procured to manage this.”
Linking different data together
Assuming you can extract data the next challenge is to link different data sets, often from different systems, together to enable business intelligence. This makes data standards and their consistent usage hugely important.
Then this extends to linking different data and taking the next step and linking to outcomes, “There’s very little yet on outcomes on what happened to the patient after they leave hospital. Key focus needs to be on what happens next to the patient once they have had their care and treatment.”
Developing staff analytical skills within the NHS
Turning to analytical skills Ray says there remains a significant skills gap to be bridged. “We need to better understand what are the core skills that CCIOs and CIOs need. It’s partly about being the translator and enabling clinicians to use actionable insight.”
Ray says that ‘actionable insight’ are the operative words. “The priority has to be about enabling staff to make decisions, not just numbers that are interesting.”
Having obtained the data the next challenge is to be able to apply sufficient clinical knowledge to interpret meaning and be able to deliver ‘actionable insight’ to clinicians. “Have you got the clinical knowledge and input to understand what that data means?”
Presenting data in ways that staff can effectively use
And related to interpretation and the application of clinical knowledge are challenges around how analytics data is presented and consumed.
“You can have problems like dashboards that raise more questions than they provide answers – and different staff groups have very different ways of interacting with analytics data.”
He adds: “Building visuals in a way that people can interact with data is much more challenging than it sounds – clinical staff can get fearful. So getting data out that doctors and nurses can actually use is a key challenge. We have to design analytics that are usable.”
To help investigate how to overcome these problems trialing BI tools that enable users to ask natural language queries, rather than requiring them to know a set of complex and often daunting tools.
“It’s about usable BI, we hope it can save a lot of human time. We have a lot of people responding to a lot of routine questions across the NHS,” says Ray.
He adds that for much this new kind of analytics it’s often not the technology that is the problem “but ensuring the right governance is in place to use the data, as technology has advanced so much”, with a lot of IG concerns centered on the potential of patient identifiable data to arise out of small data set queries.
Capturing data at point of care
Other analytics issues include trying to ensure data items are recorded as close to the point of care as possible. “We know that the further away the data item is recorded away from point of care delivery the poorer the quality usually is”. At the point core clinical decisions are made the quality is usually spot on – but get to linkage and coding further down the line and data quality challenges often occur..
Working out who and what to benchmark against
And benchmarking performance data, who and what should you benchmark against?
“The fairest comparison is comparing a unit like-for-like in terms of volume and care delivery and make up of patients,” says Ray.
“If you have a group of hospitals and want to have a conversation about how effective is your service – it’s really important that the case-mix and type of patients is the same. For example, there’s no point in benchmarking a spinal neurosurgical service against cranial practice for like for like comparison.”
Yet another aspect of the need for consistent data sources. “People will select their own dataset, but you need consistent datasets that everyone is using, that represents a single source of the truth.”
The 14 July Data Analytics challenges workshop will be jointly led by Jim Hatton, currently deputy director of informatics at Nottingham and about to join NHS Improvement to run their model hospital informatics programme.
1 July 2017 @ 19:50
My approach is that a DW is vital, however BI needs something a bit different, a bit tidier, a bit more structured, something focused on performance and efficiency, if you move DATA thru’ the pipe line into the SDS using a well designed SDM then an app can crunch millions of records in seconds providing real insight to the decision makers.