One of our health care partners, Ascribe, will be joining Microsoft at EHI Live on 5-6 November, 2013. Paul Henderson from Ascribe discusses here what the team will be showing at EHI Live.
According to Gartner, a huge percentage of decisions made at work ignore the large volume of evidence that could have influenced them. This is because information is largely inaccessible. Every organisation seems to have a data warehouse of some sort, which is accessed via reports or KPI-driven analytics, and this is sourced from a number of places, including clinical systems. However there are barriers to connecting people with the evidence they need in order to make good decisions. Many barriers. Too many to discuss in this blog and so I am going to zoom in and look at a couple which might be more addressable than you might think.
Clinicians and patients talk to each other. That is how I tell you what hurts and you tell me what the problem might be. This information might be recorded in case notes. A subset of this information might be recorded in the pick-lists and radio buttons of a clinical system. There is value in the structured data and many organisations benefit from harnessing it and using the information it contains to drive clinical and operational good practice. However the unstructured case notes tend to be less well used. They are used, don’t get me wrong, but they aren’t used as a source for analytics as they are seen as unstructured and therefore not usable in the conventional way that numbers are used.
We all know the volume of this data is growing because, not least, patients with long term conditions are living longer and paper records are being digitised. Many of us have seen this described as the Big Data phenomenon and have marvelled at how the retail sector can target us with things we didn’t know we needed long before we knew we needed them. We should have these sort of options for using health data shouldn’t we? Surely we could improve decision-making if we were able to unlock the value of case notes and deliver that value to clinicians and operational teams in a timely and consumable format?
Ascribe, Microsoft, Intel and most importantly Leeds NHS Teaching Hospitals have been looking at this issue. Together we set up a pilot project to consider how we could aid note taking and clinical acuity at the point of care using a Windows8 mobile tablet. We then pushed those notes through a Natural Language Processing engine running in Microsoft’s Azure platform using a platform called HDInsight and married them back into a locally held data warehouse. The results were amazing. I have been an advocate of NLP for years but have never been able to afford “the computer” required to do it at scale. For this exercise we did it in the Cloud. It cost us a few pounds to do 6 years’ worth of data, overnight. The NLP engine structured the data using a combination of grammar (not drunk being different from drunk for example), clinical terms (such as complaints, like influenza) and clinical abbreviations (such as BIBA which means brought-in-by-ambulance).
Having processed the data we learned some 30 things we hadn’t known previously and which will have an impact on service design and delivery. For example, we could spot early spread of infectious diseases, guide social services to problem spots in our local area, track use of new services and prove-up their business case, see changes in abuse of recreational drugs, improve clinical governance and training, assess risk to patients such as children and many other things that are captured in notes but not always in the structured areas of the clinical system. There was a financial side. A whole bunch of diagnostic tests had been ordered, according to the notes, and probably not charged for, according to the structured data.
I’ll be at EHI talking about the project. Clearly, this is a solution that has not been adopted on a widespread basis and there are many issues to be solved before it can be. However it just may be that the answer to the conundrum of improving the safety and quality of services at a time of financial pressure, lies in understanding behaviour in this way. Taken to its end, this could actually reinvent the notion of healthcare IT. This could be a challenge that will be thrown down to Electronic Medical Records vendors, who may not need radio buttons and pick-lists in future but may derive meaning from words spoken between clinicians and patients. The IT will get out of the way of that relationship, whilst paradoxically improving it.
Paul Henderson will be speaking on “Improving the safety, quality and cost-effectiveness of unscheduled care using Big Data,” at 4pm on Wednesday 6th November (Day 2 of EHI Live).
Further information on healthcare solutions at EHI Live will be profiled in this blog or follow us on Twitter. Are there particular areas you would like to discuss at the show? Contact us to book a meeting.