One of the most satisfying parts of this project from my own point of view has been the fact that I am extremely close to the technical development and can see the system that we imagined actually coming to life. I can influence the development on a week-by-week basis and have been hugely impressed at the way our technical developers have solved problems, to shape a system with immense clinical utility.
In the past I have programmed computers, and have a real interest in how technology, and computing in particular, has influenced healthcare. I have provided clinical input into many IT projects in my hospital, sometimes leading to real successes, and at other times seeing how technology can in fact disrupt efficient healthcare. However, I am, first and foremost, a doctor, and even as a clinical lead on a hospital IT project, you have only limited influence on how systems are developed and how they subsequently evolve.
In this project however I have been extremely close to the technical aspects. Firstly, we are dealing with a large amount of data. Once we sorted the issues around data access (see previous blogs) we started processing huge numbers of blood test results covering a population of around 500,000 people. Of note, a simple blood test can easily generate 20-30 ‘results’ – a standard liver test has 4-6 results within it, a ‘blood count’ [for anaemia, infection etc] typically has 15 or more. When your GP or hospital doctor requests 3 or 4 tests, in reality 20-30 numbers are produced. In addition, some tests have a result that comes as words or phrases, rather than numbers. Each of these results produces a ‘row’ of data that we need to then process. And it rapidly escalated to about 200 million tests (or in data terms, rows). Whilst this isn’t huge in terms of modern computing power, it easily outstrips what you can manage on your typical personal computer, with a programme such as ‘Excel’. Even if the personal computer programme could handle all that data, the processing would be too slow (some of our estimates ran into weeks to get an answer, and for the most complex it might take the computer years!).
For a project such as ours we needed expertise and powerful computers, and indeed we have both. But even then, cross-referencing all that data (200 million results for 500,000 patients) is a big task. And big tasks take computers a substantial amount of time. So, as we designed the system and worked out how to deliver the project results, we discussed various ways of solving these problems. In part this is just about having sufficient computing power to store and process data. But we also wanted the system to be incredibly responsive. We wanted clinicians to be able to sift through data and generate results (cohorts of patients requiring further potential treatments) in seconds. Being a part of the design of the solution to this problem has been immensely satisfying. When we first tested the system – real data, on real NHS computers – and generated results in less than a second or two it became clear we had made huge progress.
As well as rapid analysis, we need a system that is really simple and intuitive to use. Many clinicians will report that healthcare IT systems are not really intuitive. We need training (which needs to be squeezed into already extremely busy jobs). We forget how to use them if we haven’t been accessing them for a while. We get upgrades that are confusing. And then when we really get to grips with a system, it will of course be taken away, because the technology driving it has become obsolete, or the contracts expire!
We wanted something much simpler for our system. We talked about how we would train clinicians to use it, and soon concluded that if you don’t need to go on a training course to buy something on ‘Amazon’, or to search for a new house or car online, then you shouldn’t need to have training for our system. It should be as intuitive as ‘Prime Location’ or ‘Air BnB’.
Fortunately, our technical experts and developers have been enthused about usability from the start. Together we have developed a ‘front end’ for clinicians to use which is simple and powerful in equal measure. And just recently I have been able to demonstrate an early version to clinical colleagues with the response being “wow, we have been asking for something like this for many years; a system that WE can use to sift through patients results, and find the ones we really need to see”. We are not finished yet, but it is exciting to be part of something that will turn out to be so useful, and help so many people.