Covid-19 Testing transmission

Mass screening: is this the answer?

In the news again this week is the message that new approaches to testing for Covid-19, particularly using self-administered saliva tests, with just a 15 minute wait for a result, could be the answer to the pandemic.  Indeed, there are many examples across the world where everyone in a region or organisation (such as a workplace or university) has been so tested.  In a previous post on this blog, I mentioned that such a saliva test was just ‘around the corner’.  There are 7 important questions that now need to be addressed when considering the value from such a strategy:

  1. Tests for screening or diagnosis? 
  2. What is screening?
  3. Do screening tests for Covid-19 give the right answer?
  4. What are the causes and consequences of false negative tests?
  5. What are the causes and consequences of false positive tests? 
  6. Do we know what the false negative and false positive rates are for the various Covid-19 tests?
  7. Are there any other aspects of the test to consider?
  1. Tests for screening or diagnosis?
  • Early in the pandemic, tests were developed based on mouth and nose swabs to determine if individuals with typical symptoms of Covid-19 actually had the virus
  • The test was therefore used to diagnose an individual as ‘a case’
  • The positive rate of tests in those with symptoms has varied depending on what symptoms lead to the test and the prevalence of other diseases such as seasonal flu at the time of the test
  • In the early stages of the pandemic, when tests were not widely available, proving the diagnosis in people who were mildly unwell was considered unnecessary as it would not change the advice (stay at home etc)
  • Even though there was still no effective treatment for mild cases, the need for a diagnostic test later became more relevant in a negative sense: ie showing someone was negative allowed that person to resume normal activities
  • Over time it has become apparent from testing random samples of the population that a significant proportion of ‘cases’ of Covid-19 do not have symptoms at the time of a test and might never develop them
Proportion of people with symptoms in those who tested positive
  • In these data from the UK national survey, currently only about a third of those who tested positive had symptoms.  Indeed the prevalence was much lower in the summer, the symptomatic rate was under 20%
  • Having asymptomatic infection may not be a problem for the individual but it is for greater society, given that such individuals can pass the infection on to others
  • Thus, the question is raised as to the value of Covid-19 testing in  asymptomatic people for the purposes of screening

2. What is screening?

  • In a strict epidemiological sense, screening is defined as the application of a test to identify cases that had not yet, or might never, become apparent
  • Screening of individuals is of value only if there is benefit of earlier detection
  • For example, it is no use introducing a cancer screening programme if the people who are detected gain no benefit from the earlier detection 
  • Thus, screening of individuals for a disease is only of value if there is something that can be done to change the outcome, for example earlier treatment.
  • By the same token screening of populations is only of benefit if that leads to an intervention that would reduce the population impact of the disease
  • Thus, screening asymptomatic people in the general population for Covid-19 is only useful if there is something that would be done that would
    • Reduce the spread of the infection
    • Allow sections of society to function normally 
  • The testing on its own is useless unless there are effective measures that can achieve those goals
  • For the rest of this post, I will assume that the test results will lead to successful actions in terms of contact tracing, quarantine etc that will bring about the desired reduction in the rates of infection in the target populations 
  • The rest of this post thus considers how important that the screening test will give the right answer enough of the time

3. Do screening tests for Covid_19 give the right answer?

  • Although this appears a simple question, it is really challenging to answer but could be crucial in determining whether there is any benefit from mass screening
  • No test for Covid-19 is completely accurate all the time!
  • There are several reasons why you may get the wrong result
  • Looking at the diagram above we can see that there after testing, people will fall into one of four groups:
    • Have the infection and the test was positive (the green figure)
    • Have the infection but the test was negative (the black figure)
    • Don’t have the infection but the test was positive (the red figure)
    • Don’t have the infection and the test was negative (the blue figure)

At the end of the blog for those who are interested I show how we calculate the rates of false negatives and false positives

4. What are the causes and consequences of false negative tests?

  • These are some of the reasons why there might be false negatives from a Covid test:
    • The person actually was infected but was too early to show up in any test
    • The person was infected and was not producing enough material to show up in that test
    • The person was actually infected but there was a problem with the test itself:
      •  A swab test did not collect samples from the correct areas of the mouth and nose
      • A saliva test sample was not collected sufficiently well  
      • Something went wrong with the  storage and processing
      • The test in the lab gave a false reading
      • The threshold for saying a test was positive was too high (this will be discussed further below)
  • There are a number of consequences of the false negative rate in terms of screening, both for the individual and society

5. What are the causes and consequences of false positive tests?

  • There are some reasons why there might be false positive from a Covid test:
    • Contamination of a swab sample by Covid-19 in the lab, on the glove of the handler etc
    • Contamination by Covid-19 of the chemicals and other materials used to do the test
    • Test is positive but due to the presence of a different but harmless corona virus
    • Test is positive but due to the presence of genetic material that was not from Covid-19 
  • There are a number of consequences of the false positive rate in terms of screening both for the individual and society

6. Do we know what the false negative and false positive rates are for the various Covid-19 tests?

  • This is a challenging question as to know when a test is giving a wrong result; it implies that there is a way we can find out what the true result was, other than from the test!
  • In the lab,  a new test is analysed on samples from patients with known virus loads from more stringent tests and who may have obvious clinical infection with Covid-19 – but that might not be so useful in the real world as we want a new test to pick up infections in people with smaller viral loads who may, for example, be asymptomatic
  • Even in the lab with the ideal conditions, both false positive and false negative rates for Covid-19 tests are around 5%
  • We really do not know what these rates are in community -wide settings
  • A paper in the Lancet at the end of September suggested a false negative rate could be as high as 30%
  • Researchers and test developers have compared new saliva tests with  swab tests as in the example below.  But that study can only give the rates of positives and negatives relative to each other
  • Some research has estimated the rate of false negatives by doing repeat tests after a few days, to see if people are positive the second time
  • This may be sensible but of course the person’s infection status might have actually changed in that time
  • So there can only be a best guess as to what are the false negative and false positive rates for any new test

7. Are there any other aspects of the test to consider?

  • One final comment is that in all the lab tests the result is not a clear cut yes/no and a decision is made as to what reading would be considered as positive
  • If that threshold is made much less stringent -‘raise the bar to being positive’, then this would reduce the number of false negatives – but of course raise the number of false positives
  • If that threshold is made much more stringent – ‘raise the bar to being positive’ then this would increase the number of false negatives but of course reduce the number of false positives
  • It is a difficult decision where to ‘put the bar’ and this will depend on the relative harmful consequences of a being a false positive or a false negative


  • Any speedier test that can be applied to populations and that can lead to effective action to reduce transmission of infection is of course welcome
  • Testing alone which does not change behaviour or policy is a waste of time
  • My aim in this post was to emphasise that we should accept there is no perfect test but we do need to understand the imperfections of the tests that are coming into use

Appendix: Calculation of false negative and false positive rates

  • (In the diagrams below I have used the same colouring as in the image above of the different results)
  • For any test we can calculate the percentage of all results that should be positive but were wrongly called negative: this is called the False Negative rate* and is calculated as below (multiplied by 100 to give a percentage):
*Epidemiologists prefer to use the ‘converse’ of the false negative rate by calculating 100%-false negative rate – we call that “sensitivity”
  • This seems to be counterintuitive, as you would think the false negative rate should be the proportion of all negative tests that are false.
  • But the aim is to make sure that as many as possible of the true cases are identified
  • Similarly for any test we can calculate the proportion of all tests that should be  negative but were wrongly called positive: this is called the False Positive rate* and is calculated as below (multiplied by 100 to give  percentage)
*Epidemiologists prefer to use the ‘converse’ of the false positive rate by calculating 100%-false positive rate -we call that “specificity”
  • That also seems to be counterintuitive, as you would think the false positive rate should be the proportion of all positive tests that are false.
  • But the aim is to make sure that as many as possible of those who are not cases cases are also correctly identified

If you would like to receive email notifications of new posts on this blog just click this link and enter your email at the bottom of the home page:

Covid-19 Symptoms Testing

Who should be tested for Covid-19: the role of symptoms

In the news, especially across Europe, is the inability to meet recent growing demand for Covid-19 swab testing.  This threatens return to normal life and impacts on the functioning   of educational, health and care institutions.  Testing authorities are firmly advising only those with symptoms should be tested, but does this make sense? In this post I present an epidemiological analysis to inform thoughts on  this decision.

One problem is the low pick-up rate

  • In the graph below, I provide the percentage of those who had a test who were positive in the latest full week
  • The UK (depending on how one interprets the results) does badly with less than 1% of those tested being positive
  • The UK government questions the value of such a high proportion with a negative test result
  • We do not know how many of those tested in each country had symptoms, or indeed what symptoms they had.
  • But the data raise the issue as to the relationship between symptoms suggestive of Covid-19 and having a positive swab test

A bit of basic epidemiology!

  • In the table below, it shows that individuals in the population who are tested can be divided into one of four groups based on their (i) presence or absence or symptoms and (ii) whether they are positive or negative on the swab test
  • Government advice is that only those who would be in the yellow or blue boxes would turn up for testing
  • That would mean the people in the pink box would be missed 
  • Also, as those  in the blue box would be tested “unnecessarily”, we should aim to reduce the proportion in that group
  • Therefore, there are the twin challenges of:
    • capturing the ‘pink’ people
    • refining what counts as symptoms to reduce the proportion of ‘blue’ to ‘yellow’ people

Do we know how what proportion of those with Covid-19 don’t have symptoms?

  • Yes we do!
  • This is because (and similar studies have been done in many other countries) a random population has been giving swab tests each week and answering questions on symptoms-thereby avoiding any biases of just looking at people who turn up at test sites 
  • In the UK less than half of this random population of households had symptoms of Covid-19 in the week before or after their positive test 
  • What is interesting is that this proportion had not really changed over the first few months of the pandemic
  • The graph below right the proportion with symptoms has been constantly below 40% (The grey areas represent that there is a range in the likely true result because of the relative small numbers)
  • Testing only symptomatic people will miss a substantial proportion of cases

Can we refine our criteria for those with symptoms to increase the likelihood they have Covid-19?

  • We have a problem with Covid-19 which some viruses do not present
  • This is  child with chicken pox, a disease caused by a virus
  • The rash is distinctive and we don’t need to identify the virus from a swab to confirm the diagnosis
  • This is a person with a viral illness, it could be Covid-19, but it could be influenza, or other minor illness
  • The challenge is that their symptoms from Covid-19 vary between patients and overlap with other viral diseases 

Are there patterns of symptoms that should be used to prioritise for testing?

  • The symptoms that were initially publicised of fever and persistent dry cough clearly overlapped with similar symptoms of other respiratory viral infections.  
  • Although, in some people these symptoms were for sure more severe than say seasonal flu, it was not possible to differentiate.  Given limited testing capacity at that stage, those with mild symptoms were assumed to have the disease and told to isolate, as were their close contacts
  • In what has been a ground breaking study, a team led by Tim Spector from King’s College London*, with a US tech partner, introduced an app for the general public to give daily updates of their symptoms and to note whether they had tested positive for Covid-19 
  • Around 4 million have now signed up in the UK in addition to a sizeable number in the USA
  • What this enabled the researchers to address is how specific were the symptoms of Covid-19
  • Their results are shown as a graph below
  • To  explain, just focusing on the red squares, the square opposite loss of taste and smell is associated with a 6 fold increase.  What that means is that for people with that symptom they were 6 times more likely to report a positive Covid-19 test.  
  • In that sample 2/3 of those with a positive test had this sensory loss
  • Compare that with the very modest increases in fever and cough – of course that doesn’t mean that Covid-19 doesn’t give coughs and fevers.  More that these are very common symptoms and are just as likely to be associated with non-Covid-19 causes eg other viruses such as flu or just ‘having a cough’
  • Almost certainly if there had been more details about the cough and the fever, say how high the temperature etc, then the risk scores would have been higher.
  • But this perhaps explains how the simple public health messages about cough and fever lead to a large number of tests but only a small number of positive results.
  • Just a word of caution about all these results, the percentages and risks will change depending on how common say Covid-19 and seasonal flu are at the time of the research 

*Tim was my first PhD student, but I can’t claim any credit for his later career success!


  • There are no easy answers and this blog is not about politics
  • But there are some clear conclusions
  • Testing only those with symptoms will miss a sizeable number of cases
  • Although there is no justification for testing random people (apart from research purposes) there are groups who although asymptomatic should be prioritised for testing including:
    • Contacts of cases
    • Health and care workers
    • Teachers etc
  • Loss of taste and smell is by far the most specific symptom of  Covid-19.  Indeed, given that, and especially in the presence of any other symptom or if a close contact of another person with confirmed Covid-19 , there may be a case for not testing and just assuming the person is positive, and they should isolate.
  • I am drawn to the conclusion that testing people with other symptoms, especially when they’re mild, is not helpful and should, given the constraints of testing, not be a priority compared to the groups listed above 

If you want to be notified of my future posts delivered to your inbox, and are not already subscribed, just enter your email on the form at the bottom of the ‘About’ page of the blog (

Covid-19 Testing

A saliva test for diagnosing Covid-19: is it a ‘game changer’?

Currently the test for the diagnosis of Covid-19 has rested on analysing a ‘scrape’ collecting cells from the mouth and nose for the presence of genetic traces of the virus.  Reports yesterday  suggest that taking a sample of saliva might be a better approach  and certainly more acceptable.  I discuss the science behind this story and the issues involved.

Current standard approach to diagnosis

  • Current approach to diagnosis is based on using a swab to collect a sample of cells from  the back of the throat (‘nasopharynx’) and the lining of the nose
  • The cells at these sites from infected people should contain the virus
  • Although the virus can be grown (cultured) from these cells, it is a much easier job to look for the presence of genetic material from the virus using what is call RTPCR technology
  • The nature of the RTPCR test is that it requires specialist equipment and chemicals and can only be done in specially set up laboratories
  • In theory the laboratory analysis can be completed within an hour; though given the pressure of numbers, the results take much longer to come through

Sample collection using swabs

  • It is unpleasant to provide a sample: it is easy to ‘gag’ on the swab
  • It is not so easy to do oneself and the scrape might not ‘hit the right target’
  • If health care professional staff are used they need to be trained 
  • Such staff taking the sample are putting themselves at risk and facilities for collecting samples need to have stringent infection control 

Have tests on saliva been considered in the past? 

  • There have been a very large number of studies showing that saliva can be used instead of a swab to diagnose other viruses and,  more recently, Covid-19 
  • By the beginning of August over 26 research studies had compared saliva testing to the swab test
  • Overall those studies showed that 91% of people who were positive on the swab test and had symptoms of Covid-19  had a positive saliva test
  • A research paper published last week found higher amounts of virus genetic material in saliva samples than in the swab test. 
  • In that study, compared to the swab test, the saliva samples were positive for more days after diagnosis
Source: NEJM August 28, 2020
DOI: 10.1056/NEJMc2016359
  • Given these results and the easier collection, health authorities in several countries are now suggesting saliva as a replacement for the swab test

Why aren’t saliva tests used up to now?

  • There needed to be a standard way of the patient giving the sample in order to compare the results with the scrape test.
  • It is not that the test can say ‘just spit into this pot’!  
  • For example one cannot eat or drink for 30 minutes prior to the sample
  • There also needed to be a standard way for collecting the sample, and storing it so it can be posted or transported 
  • The RTPCR test that is used for the swab samples also needed to be adjusted for testing saliva 
  • All these issues are now sorted 

But the way that saliva can be tested is also changing!

  • This week the results of a tie up were announced between a UK biotechnology company MicroGEM and the University of Southampton in the UK
  • MicroGEM produced a test, that does not need a laboratory to do the analysis and can provide the user with the result in 15 minutes
  • My guess is that the kit might look like a pregnancy test and could be just as easy to use: anyone should be able to test themselves and get an answer
  • Obviously, it needs to be robustly evaluated but early results look promising
  • The US government have given MicroGEM $41m to expedite development and the means of widespread production of  the test
  • The test will also identify those with influenza (important when people have typical symptoms of a viral infection)

Point of care (POC) testing

  • POC testing has been the goal since the start of the pandemic
  • It means that patients, health care professional staff and groups such as school children, employees etc can self-test regularly and those infected can be identified very quickly
  • I had previously suggested in my post (10 August)  about returning to school that a test such as this could be the  strategy needed to reassure parents, students and teachers
  • A POC test aiming to identify proteins produced  by the virus had been proposed earlier this year but there were too many false negative results and this approach had not been developed further


  • This could be a ‘game changer’
  • It is not a substitute for social distancing and other preventive measures but it will make it much easier to track the disease and control local outbreaks
  • This technology could be used for diagnosing flu and other common  virus outbreaks in the future
Covid-19 Testing

CoVid-19: Update on Recent Information

[This was circulated on 11 March 2020. Although much is still probably relevant, at that time encouraging infection amongst younger people was being discussed to achieve herd immunity. Also, interestingly, aerosol transmission was ruled out – which given my recent post was probably premature]

What is the test being used to detect  CoVid19 infection?

The test used in the UK and elsewhere in Europe is to take a swab from the back of the throat and to determine if there are fragments of viral genetic material (RNA) present which are distinctive for Covid19.  Thus, the test determines the virus is present and not necessarily proves active infection.  This test should be able to determine the different strains of virus as it mutates in different populations. 

In China, they have also been using blood tests to determine if there are antibodies to CoVid19, which prove an antibody response to a viral infection

Can the test be positive in people who do not have the infection?

This question can also be rephrased as “are there people who carry the virus and remain well but could pass on the virus to others (asymptomatic carriers)?”

There is limited data on this question as far as I am aware.  Reports from China have identified  asymptomatic carriers who are family members or other close contacts of infected individuals who test positive for viral RNA.  These individuals are then shedding the virus without themselves ever displaying signs of illness.  Interestingly, the reports thus far suggest that while many of these may not have the illness when tested, when carefully followed up, do show signs and symptoms up to 28 days afterwards. By contrast a very few still have active virus after a month but are not ill.

I am not aware of any random population studies in the areas affected eg Wuhan to see how many people with apparently no contact with an infected individual are harbouring the virus in their throat, with no illness. 

What is the incubation period?

This has now been relatively well characterised and widely known. The period between being exposed to the virus and developing symptoms does vary but the median time is about 5 days (ie 50% of cases have developed by that time interval)  and the large majority by 7 days.  In all, 98% of those going on to have the disease  have developed it by 14 days after exposure. The 28 days reported in family members above is extreme.

If someone is infected, how long are they infectious for after the symptoms start?

That is a different question to the one of incubation but important of course.  The question is: assuming people have recovered from the illness, eg cough, cold and fever subsided, do they still pose a risk to others?

The Chinese data suggest that the time of maximum amount of virus shedding , ie the passing of the virus in droplets from the respiratory tract-nose and mouth, is in the early part of the infection.  As the days go on the amount of virus declines.  What researchers have done was to try and grow the virus from nasal and throat swabs at different time points after the illness starts. Basically by 8 days it has proved very difficult to grow the virus in anyone.  Their conclusion is to assume by 10 days after symptoms have started, assuming the symptoms have stopped, people are not infectious.  

However, the Chinese have also shown that fragments of the viral RNA can persist for days, perhaps for weeks after the symptoms have subsided They cannot grow the virus in culture and their  conclusion though is that these genetic fragments, of themselves do not indicate that the person is still infectious.  Indeed, given it has been difficult to grow the virus after 8-10 days, the persistence of the viral fragments may be of little relevance.  More research will follow for sure.

How do the authorities know how many cases of coronavirus there are?

We are bombarded by numbers of new cases every day in the media: how are those numbers gathered?  For the moment, the numbers are purely based on the reports submitted to the public health departments from the testing labs  of the number of tests which have proven positive.  For countries that have been very slow of the mark in testing (eg USA) their numbers are a woeful under-estimate.  If, and when, in UK people self-isolate at home with mild symptoms and are not tested, the exact numbers will not be known.  I assume that there may be sample surveys of some primary care practices to identify the numbers to make an estimate.  Thus, although we should expect the numbers to rise in the next days/weeks, these numbers will under-estimate the true overall number of new cases but over-estimate the proportion of the more serve cases.

Are we now clearer about the main route of transmission?

CoVid-19 in almost every case has  spread from droplets by being in close (ie 2 metres proximity) to an infected person.  There is no evidence that the virus hangs around in the air (aerosol).  So, for example it has been suggested that on  a plane, if there is an infected person, it is only those within 2 rows who are at risk!

What is the evidence that it can be spread from hard surfaces?

In addition to droplets, the media have been keen to emphasise that where such droplets land on a hard surface they can persist.  What is clearly not possible is to prove that any individual case was picked up from this route say from a particular door handle touched a week ago.  Testing hard surfaces for presence of the virus is not done routinely.  So the hard surface route being a source for any case or cases will I assume always remain an unproven possibility. The concern about the survival of virus on hard surfaces  came from work on SARS in the laboratory which showed that on hard shiny surfaces, eg metal in the laboratory, virus could be detected sometimes up to  couple of weeks after exposing the surface to the virus.  Experts thus believe that the same is likely to be true for CoVid-19.  Alcohol, dilute bleach and hydrogen peroxide all very effective at sterilising hard surfaces of the virus

When will this epidemic end?

This is the really interesting question and there are lots of unknowns.  There are a number of possibilities:

  1. Successful containment:  The virus can only be spread for person to person (let’s leave the issue of animal transmission) so in theory if all infected people are totally isolated then the virus could have ‘nowhere to go’.  This appears to be the explanation of China’s apparent success and it is an interesting experiment to see if Italy also succeeds in their drastic measures.  However, this might be a short lasting success,  if  infections persisted in other countries and then got re-introduced into a country where they thought it had been contained and people started mixing again, the epidemic could start over.
  2. Naturally die out of the virus in warmer weather One suggestion widely mentioned in the media is that (as with ‘seasonal flu’), come the warm, drier weather, the virus will struggle  to survive and will then naturally die out. Epidemiologists also suggest that in summer we are less indoors and less in close contacts with other.  SARS did disappear in a few months.  The summer reduction theory is accepted by many although it has been suggested that CoVid19 could be with is for years and have outbreaks as with seasonal flu.  Although with increasing natural immunity and a weaker virus, see below this is likely to be less serious.
  3. The virus will become weaker This is an interesting theory which suggests that as the epidemic proceeds the virus becomes weaker.  In scientific terms the theory is that viruses mutate continuously during an epidemic. Further as the virus mutates, it becomes less virulent (causing less harm).  The reason put forward is that for the virus to survive it needs to be passed on from human to human.  Thus, those humans who have the more serious disease and are isolated in hospitals or die, are less likely to pass their version virus on, and there will be selective transmission of those mutations associated with milder disease.  This is speculative but accords with data from other viral outbreaks
  4. Vaccine:  A successful vaccine would easily terminate the epidemic but as I stated earlier, that even though clinical trials have started, it won’t be available for widespread us, I suspect until 2021
  5. Herd immunity:  From discussion with colleagues, but not seen in writing, is that we should be prepared to allow (say) most ‘younger’ healthy people go about normal business, catch a mild disease and then become immune.  Once the proportion of the population level with natural immunity reaches a certain level, then the non-immune people in the population would be too far spaced for an epidemic to take hold.  Intriguing  idea, but probably not politically acceptable and, or too risky as cannot assume only the healthy will contact the infection and,  in terms of the numbers of complications, health services would be unable to cope 
  • Can you get it from your pet?

To finish on  a lighter note, there has been a single case reported from China of transmission from patient to their dog.  Most commentators have dismissed the significance of this