Covid-19: Testing, Testing

Research nurse from the NIHR Clinical Research Facility processing patient samples using SAMBA machines at Addenbrooke’s Hospital in Cambridge.      Credit: Cambridge University Hospital

As I write this post the UK has been in Covid-19 virus ‘lockdown’ for over four weeks. My age puts me rather closer to the danger-zone than I would like, but WFH does reduce my chances of infection, even if that means going crazy with the near-saturation coverage of the subject on all media. We are bombarded daily with tables, graphs and forecasts, the latter seeming to depend more on assumptions than hard data. Forecasts come from mathematical models running on powerful supercomputers. In order to produce daily estimates of infection numbers and spread, the models constantly compare their current numbers with new, measured data, refining the variables so as to produce a better result the next day. And there lies the problem: relatively little data has been gathered through testing, so we have no idea what proportion of the population remains uninfected and still vulnerable, and how many have been infected and recovered without hospitalisation. Or indeed the number of ‘Typhoid Marys’ who are infectious but show no symptoms.

The Weather Forecast

Without regular inputs of accurate data, computer models will always produce unreliable outputs following the ancient maxim from the dawn of electronic computation: GIGO or Garbage-In, Garbage-Out. This applies universally and is independent of both processor power and the sophistication of the model. Sure, you need a supercomputer with petaflops throughput to process a vast amount of data in a short space of time, but if that data is dodgy, GIGO still defines the outcome. This leaves scientists with no choice but to make assumptions based on their experience. It’s why the forecasts from Covid-19 spread-models developed by scientists at Oxford University and Imperial College London are so very different.

Weather forecasts used to be very inaccurate, frequently not even close. Now, thanks to world-wide data gathering, fast communication and satellite observation, even very local forecasts can be relied upon to give you not only correct future conditions, rain, sun, temperature, etc, but accurate timing of the events. If only the same could be said for the spread of viruses across the globe.

Testing, testing

At the moment, a test consists of taking a swab/blood sample from a subject and then putting it through a complex laboratory process taking many hours to get a result. The recent introduction of new technology has enabled the processing time to be cut to 90 minutes. Nevertheless, you can see why the ideal situation of an entire population being checked every day, producing masses of accurate data for a forecast model is simply not going to happen. It ought to be possible to use random sampling like opinion pollsters, but maintaining the essential randomness, that is avoiding introducing bias is quite difficult to do. A particular problem is finding that some people in the randomly selected set don’t want to be tested for some reason. Replacing them with willing subjects can distort the data.

There are new tools for diagnosing covid-19 and they inevitably involve Artificial Intelligence. One such uses AI to search for tell-tale signs of the disease in chest X-rays. At the moment it provides a very useful ‘second opinion’ for doctors, but wouldn’t be much use for widespread data-gathering.

New Technology

Citizen Science offers new hope in the struggle to gather the vast amounts of data required for a forecast model to operate. Smartphone apps for corvid-19 contact-tracing are appearing across the world, some ‘official’, others not. Generally, they require the user to register with a website and provide some personal information including any symptoms that might suggest they are infected with the virus. As they walk around, if another registered user with symptoms comes close, a warning is issued to both them and the central server. They will then be required to self-isolate and book a full corvid-19 test online. Such a system does - at the moment – rely on sufficiently large numbers of people registering voluntarily to make it useful. In some countries where this form of digital contact-tracing has been piloted, the take-up has been poor. Privacy concerns dominate because such a scheme can be set up to provide continuous surveillance for purposes other than health monitoring.

In the UK, a ‘benign’ COVID Symptom Tracker app is supported by King's College London, Guys and St Thomas’ Hospitals working in partnership with ZOE Global Ltd – a health science company. The user just provides a daily report on their state of health, even if they show no symptoms.

The official UK NHS contact-tracing app is scheduled for release soon and this relies on Smartphone Bluetooth LE communication to detect other registered users in close proximity.

How Epidemics End

Epidemics always end in one of two ways:

• The virus is so aggressive it quickly kills its host, often before it manages to infect another. Such outbreaks can be relatively short-lived when all the available hosts have died. For obvious reasons, such viruses tend to mutate and become less virulent.
• Viruses like covid-19, while definitely not benign, only end up killing the old and sick hosts with compromised immune systems. All others will recover as their immune systems get ‘trained’ to recognise the invader and eventually destroy it. Before that happens though, the host can infect others and so spread the disease. Eventually, most of the potential hosts in a group become immune, usually a minimum of 60% and the virus dies out when all still-infected hosts become surrounded by immune ones – so called ’Herd Immunity’.

Herd immunity is achieved in one of two ways:

• By the natural process described above. It’s slow and most of the sick and elderly will die.
• Artificially by a campaign of vaccination. This can be quick and saves lives. But only if a vaccine is available, obviously.

I’ve not mentioned the use of antiviral drugs. That’s because, unlike antibiotics for bacteria, they do not destroy the virus, merely disrupting reproduction allowing the host to survive for as long as the drug is taken.

So where are we now?

At the time of writing this post, the UK government is wrestling with a hard decision that cannot be put off much longer: when and how to release the lockdown restrictions. It all comes down to the numbers, numbers we don’t have. At the beginning of the lockdown Oxford researchers believed that a large proportion of the population were already infected; Imperial assumed a much smaller fraction. If Oxford are right then lifting the lockdown will not cause a large increase in infections because herd immunity may soon be achieved. On the other hand, a low level of immunity will leave the population vulnerable to a ‘second wave’ of infection. What to do? If only we had the data….and a vaccine.

The Future

Two things are certain: there will be more pandemics in the future, and economies, including that of the UK, will not be able to survive another lockdown like this one. Most experts agree that early detection of disease carriers with contact-tracing and isolation rigorously applied will be essential to avoid the chaos of covid-19. Tracking the spread using smartphone apps may be the answer. Apple and Google have devised a competitor to the NHS app which they claim has fewer privacy issues. But it may well be that governments will insist on all phones having the tracking programs built-in and outside the control of the user. Perhaps that’s the price for the ‘New Normal’, free from lockdowns and economic crises.

If you're stuck for something to do, follow my posts on Twitter. I link to interesting articles on new electronics and related technologies, retweeting posts I spot about robots, space exploration and other issues.