Imagine a time when there was no link up between hospital and GP data or even between information stored in individual GP practices. Sound quaint? Yet this was the reality just a few years ago. Even with a joined-up primary and secondary care database, actually interrogating that data, using it to make decisions that would improve care for patients can be incredibly time-consuming, fraught with errors and frustrating.
Through our work on an award-winning project with COPD we have been involved in the development of software which enables quick, easy analysis of where care can be improved, links up care for patients being seen by numerous clinicians and shows us whether the changes we are making have any impact.
Interrogating the data
It started with risk stratification. We had commissioned a provider to give us a system that joined up GP and hospital data for a pilot project in Redbridge. Through this work it became apparent that only one in ten of our high-risk patients were being seen by our long-term conditions team.
What we initially had was linked primary and secondary care data (we later added in local authority and community databases), and a risk stratification tool which automatically flagged up those frequent hospital attenders.
Yet we quickly realised we needed more flexibility. There was much more we needed to be able to do with the data and this was becoming increasingly evident in a project we were working on in conjunction with University College London and Health Innovation Education on tackling COPD.
Our aim was to increase the value of care provided to patients with long-term conditions and COPD – an expensive condition with patients frequently having unplanned admissions to hospital that is also massively under-diagnosed in the population – seemed a good place to start.
Step one in solving a problem – in this case improving a care pathway – is usually to understand what the issues are, where the flaws in the system might be. And we thought we knew what those were.
But actually when we looked at the information – something that at this point we were having to do manually – we found a whole host of problems and using specialist nurses went into practices to audit and confirm what our information suggested.
Improving COPD care
In response to what the data were telling us, we identified a range of training opportunities to address development needs. These included COPD masterclasses for GPs, and a practice nurse mentorship programme focusing on non-diagnosed COPD admissions to secondary care, annual reviews, inhaler technique and spirometry testing.
Every patient subsequently received a COPD action plan providing them with details in six key areas: when they were due an annual review, spirometry testing, stopping smoking, inhaler technique, pulmonary rehabilitation and support with self-management which included providing them with a rescue pack of antibiotics and steroids should they have an exacerbation of their condition.
We implemented the changes in the 188 practices across North East London and at the same time we realised we needed to be able to use the data available to us to monitor whether the programme was working and whether we were having the expected impact. We also needed to be able to map and track groups of patients to quickly identify problems.
Between 2010 and 2012 the interventions we put in place led to:
· A dramatic increase in the number of self-management plans given out – from 171 in the month we started to more than 500 a month by the end.
· A large rise in the number of patients reporting being well-informed about their disease and feeling confident about managing their own condition
· Patients went from being mostly unaware to quite or very aware of the services available to them
· There was measurable improvement on all NICE indicators including spirometry testing, annual review and measures of disease severity.
· We identified and diagnosed hundreds of previously unknown patients
· Fewer emergency admissions
· Achieved overall savings of £600 on average, per practice, per month (with higher drug and equipment costs taken into account) to a total of £2m per year.
The work we did on COPD, for which we recently won an NHS Innovation Award was only possible because we used the data to guide the interventions we put in place and monitor the impact we were having. The whole project was based on the ability to make use of quality data to improve quality of care and cost right down to the individual patient level.
Back in 2009, we did not have the capability to do this – at least not in a way that was easy to achieve. Every time we wanted to find something in the data, to ask a question of the information we had in our primary and secondary care database, we had to go to our provider to ask if they could do it for us. It was incredibly difficult to analyse the data.
So we asked our provider to design a way we could do this analysis ourselves – so that clinicians sat in practices could look at the impact of projects, such as the one on COPD, and see where the flaws were and how changes were leading to improvements. It took them a year but they rose to the challenge and now we have a system called SNOVA which enables us to easily do this in-depth analysis, without a business analyst spending hours working out how to extract the information.
SNOVA is not a database in itself, but a programme that sits to the side of the database and enables us to challenge it. We can use that to do all analysis and it is massively time saving. A clinician can just type in the query – such as how many of their COPD patients have had annual reviews and how this compares to other practices, or how much they are saving in expenditure on COPD, and gets a quick answer in return.
Cinicians can see the web-based system in their practices – they don’t need us to produce spreadsheets and can take cohorts for their practice and look at numbers compared with other populations.
Prior to getting this joined up system up and running, we had to send someone into the actual practice to run a query and bring the information back. Over the past four years the analyses we have needed to do have become ever more complex and the need for SNOVA has never been greater.
One of its most important features, is what it allows us to do with monitoring projects as we go along. With COPD, dashboard reporting allowed nurses and GPs to see the impact of the interventions and by implementing the project in a staggered way and using one borough as a control, we were able to see what was having an effect and how.
What we have now essentially is a database that through role-based access enables people to look at different elements of care. It enables clinicians and commissioners to review data across pathways to identify where there are shortfalls and calculate where significant savings and improvements could be made.
What started as a risk stratification tool, then a tool for analysis also became a tool for financial oversight, including analysis of budgets, routine challenges and improving data quality – something we continue to work on.
And there are many other areas where we have used more sophisticated informatics to guide our clinical practice.
Integrated care is a good example. Local hospitals, GPs and social care organisations can now share data which is analysed to produce profiles of each ‘at risk’ local person. Those patients can be offered integrated packages of care along a single care pathway avoiding duplication and the risk of vulnerable patients falling between the cracks.
Central to this is a shared ‘virtual’ care plan that all those involved in treating the patient can access – a scheme we have now been developing for about a year and is now ready for prototyping. . Those at high risk of hospital admission are identified before being assessed by the clinician to see if they would benefit being looked after by the multidisciplinary teamwhich includes a GP, community matron and social worker.
We have recently started to do some work on dementia after noting a disparity between the number of patients recorded as having dementia and the number of patients with no recorded diagnosis in primary care but who are on dementia medication or who have been in hospital with dementia. We are looking to see if they are not known to the practice or are actually coded improperly. So using data has driven clinicians to make changes in the populations they look after.
The system essentially enables us to use a wide trawling net to identify innovation projects and work out which will provide significant benefit before rolling out further. We know their likely impact before they are implemented on a large scale. And anything unlikely to work can be scrapped.
The main barrier we have come up against time and time again is around information governance. Role-based access to patient identifiable information is key to determining who can access data, and when and how. GPs and other clinicians, CCGs, social services, public health all have very different requirements. And that has been a big problem for us. How to you do that without breaking information governance rules?
Knowing who is directly responsible for caring for the patient is simple but there is a grey area involving indirect care that has been the biggest challenge. It took two and a half years to get right. We have developed controls to allow patient consent, recorded in the GP clinical system to enable specific individuals to see just that patient record.
Joining up all the different clinical systems was also fraught with problems. We had to develop a system to collect data from each of these clinical systems and transform it to a common format. For example different practices were using different codes. And you would be amazed how many patients in different practices are using the same NHS number – it is a lot more common than you would think. If you get it wrong, patients can be given new conditions as well as all the ones they actually do have.
It took four years and an amazingly diverse set of skills to bring this together – from finance-based intelligence to public health and when we were a PCT is was easy as we had all those skills to hand
Public health is a very big user of the system and they have a massive need for the dataset we have developed. At their request we have been able to automate the NHS Health Check process, coordinating all the manual steps from identifying patients to payment. This has encouraged GPs to do more of this work and we believe it would work well for other public health type programmes such as immunisation.
What started as a easier way of us doing statistical analysis will become more and more a clinical tool, as a dashboard available at the point of care to track patients and identify issues with care pathways as well as comparisons with neighbouring practices and CCGs.
We continue to trawl through the data to identify areas where we can improve care and save cost, implementing small pilot projects before assessing to see if wider roll out would have an impact. Hopefully we will be able to emulate the success we had with this approach in COPD in other long-term conditions.