Sometimes you have to keep banging away about something. For some time I have been complaining that the government’s system for providing tests for Covid-19, and then for its approach to contact tracing, suffers from a fundamental flaw of process design. I see this being occasionally mentioned by others, but the idea hasn’t caught on. So I will say it again.
This is relevant because the testing regime seems to be in a state of complete dysfunction. The government is not being transparent about what is going wrong, a an issue which is not unrelated, so I’m having to join some dots, based on a flood of anecdotal evidence from people at different levels in the system that have popped up on the news. The system has been overwhelmed by a surge in demand. Whether or not this should have been foreseen is one question, but taking a step back and looking at the outcome prompts another. This excess demand seems to have caused the whole system to fail, so that while testing capacity is very high (the government claims it is higher than in most other countries), all, or most, of the tests are taking far too long to return results, which completely undermines their usefulness. I have heard experts suggest that if results take longer than 48 hours to be returned, then they are of little practical use. That sounds about right. Results seem to be taking much longer than this in the official system, or at least that which serves most users (I think hospitals are linked to a different one, which might be working a bit better). The problem seems to be at “Lighthouse” labs where samples are analysed. The government suggests that this is just a numbers game: these labs have a capacity and demand is in excess of it, leading to delays – which is perfectly plausible explanation and doubtless at least part of the problem. There are other stories of staffing issues as these labs are losing temporary workers as the university terms start, and finding them hard to replace.
How to manage this? The first response is to stop people taking tests through the booking system, by telling them that they are unavailable, or only available hundreds of miles away. One story is that the only way that people living in the London suburb of Twickenham can get a test locally, rather than one in Aberdeen (in the north east of Scotland), is to say that they are living in Aberdeen. This is causing an immense amount of distress, which is feeding back in complaints to MPs. The government is now trying to impose some form of prioritisation on tests to give this more rationality. But that will be hard going, with goodwill in short supply. There is a least one new Lighthouse lab in the pipeline, and the government doubtless is placing its hopes on this. Alas any relief is unlikely to last for long. The whole thing gets much worse when the need for contact tracing is brought into the picture, where similar problems are emerging, though not, excess demand. The whole damn system is flawed.
What’s gone wrong? The designers of the government system are bewitched by the idea of scale economies. The unit cost of a large scale system operating at full capacity is generally very low. And because covid tests are basically quite standard, at first sight the building of such high volume facilities looks like a sensible way forward. When the government suggested that the system would be “world-beating”, this is doubtless what lay behind it, as many countries have gone for a much more artisanal approach. But that is only one aspect of process design. The problem comes from how you manage the whole process from end to end (i.e. from the moment somebody decides that they need a test to the moment they get the result). The more steps there are in the process, and the more disconnected their management, the less efficient the whole becomes. This can seem quite paradoxical. Each part of the system can seem to be operating well, but the whole can be dysfunctional, and doesn’t seem to be anybody’s fault. The problem is compounded by the the government’s preference for the use of mass-sampling facilities. People are sent to drive-in centres that are able to process large volumes. But these are often idle and simply make the whole process more disconnected. The symptoms of such a disconnected “silo” based process design are very familiar. Bottlenecks, queues, delays, lost files, and all the while managers working frantically hard at their own little section in the knowledge the the problems are all somebody else’s fault. And managers blaming users for making unreasonable demands. There is so much at stake in the overall design that nobody dares point out that it might be better to scrap the whole thing and start again. Instead they work on fixes that ameliorate the worst problems but make the whole process more cumbersome. In this case designing systems to prioritise demand.
What’s the alternative? It is to create local facilities that do the whole job end to end, or as much as possible. Best of all is if the testing function can be integrated with a similarly localised system of contact tracing, all accountable to a local director of public health, part of local government. Where possible staff should be able to cover multiple jobs, rather than specialising in just small parts of it. This is more or less how it works in Germany, among other places. Such a system cannot solve all problems. It may not be able to overcome shortages in critical supplies (reagents for tests, and so on), though managers are more empowered to find work-arounds. How would it cope with excess demand, as is happening at the moment? It is superior in several ways. Firstly because managers are likely to have a better overview of the whole system, problems are more likely to be foreseen. Secondly bottlenecks are more easily fixed. It is easier to recruit two or three extra staff in a local centre than a couple of hundred in a centralised facility. And where there are problems, they will not bring the entire country to a halt. And finally communication with the end user is likely to be far superior, as they are much closer to a knowledgeable, human interface.
These principles have been well-understood since the 1990s (when I used them to reorganise processing operations that I was managing). Alas the government, and those it appoints as advisers, are far too wedded to the imperial silo-based model and seem incapable of understanding that they are dealing with poor systems design rather than a few teething difficulties. Doubtless the silo approach works well in some contexts. But not here. But quite why the lessons of the 1990s are so widely forgotten in 2020 remains something of a mystery to me – my guess is that managers and politicians have been distracted and beguiled by new technology.