Medical, statistical and policy experts tell us what we need to know about COVID-19 testing, including the laying out of a broader and potentially more useful testing strategy.
We have broken this up under two heads: Basic facts about COVID-19 testing and evolving a strategy for the same in India.
Basic facts about COVID-19 testing:-
Swati Piramal and Mahesh Balsekar spoke to researchers around the world for key insights and clarifications about testing which they put out in an article recently. Here are key takeaways from the article with some additions from other expert sources, and after clarifying with experts:
- Rapid antibody tests aren’t reliable as a diagnostic tool for COVID-19 or for contact tracing (identifying and monitoring ‘contacts’ who have been exposed to infected persons). They have lower specificity and sensitivity than molecular tests like those using Reverse Transcription Polymerase Chain Reaction (RT-PCR). RT-PCR tests directly for the infection, determining whether the subject is infected or not. Antibody tests look for antibodies the body has developed in response to the infection, but developing these antibodies may take from days to weeks. The strength of the antibody depends on the factors like age, nutrition, disease severity and medications or other infections in the patient’s body. In some COVID-19 cases, confirmed by RT-PCR, antibodies have responded weakly or late or not at all. Also, studies suggest that the majority of patients develop antibodies only in the second week after the onset of symptoms. This means that a diagnosis of COVID-19 based on antibody response will often only be possible in the recovery phase, when opportunities for clinical intervention or interrupting the transmission of the disease have passed.
- However, antibody tests may be used for epidemiological and surveillance studies at a community level. Due to concerns about false positives, these tests may perform poorly in areas with very low rates of infection. But, to some extent, they can help understand the spread of infection in hotspots and the probable nature of spread in terms of asymptomatic and symptomatic carriers. In doing so they may help evaluate the impact of protocols and policies, and how to effectively move out of a lockdown or containment scenario.
- All studies must be designed carefully by involving medical and epidemiology professionals (as well as those who are experts in sampling protocol for the Indian population). According to the research team at the Stanford School of Medicine three critical factors must be ensured: effective selection, testing methodology and robust statistical analysis. We will explore these factors below."
- Any selection methodology for testing and analysis for prevalence in the population should ensure the representativeness of actual population demographic and minimise selection bias (selection such that proper randomisation isn’t achieved and the sample isn’t representative of the demographic to be analyzed) as much as possible."
- Testing methodology must ensure high specificity and sensitivity. Eg. Pooled RT-PCR testing: pooling the samples of a group of people to test the ‘pool’, instead of testing individuals, can decrease the number of tests we have to conduct by 70 to 80%; if a pool tests positive then the members of that pool can be tested individually.
- Statistical analysis is necessary at the onset of testing, to determine the sample size - based on demographics and expected test accuracy - as well as to draw conclusions from the study after it has been concluded, and understand the nature of prevalence of the disease.
For more on what researchers around the world told Piramal and Balsekar, read this article.
An appropriate strategy for COVID-19 testing in India:-
India finally has a good supply of COVID-19 testing kits (though quality is a concern) and a lockdown is a prudent time to test as contact between people is minimal. The more we test and the better we test now, the better we will know how - and in which areas - to ease the lockdown.
From Mid-March we have been testing 30 times more people for COVID-19 than we were doing before, taking the number to 30,000 tests daily. But even if we add a million more tests to this number we would still be testing only one person per 1000, and still be a low-test nation. India needs a testing strategy to find infected persons - even if they’re asymptomatic - to be able to treat them and prevent transmission but also to generate data to put in place a good containment strategy, reducing the need for wholesale lockdowns. Testing strategies have to be tailored to the specific needs of each Indian state but there are some common principles that can be adapted.
The Indian Council of Medical Research (ICMR)’s recommended testing strategy allows testing of asymptomatic persons if they are direct and high-risk contacts of a confirmed case. But ICMR also says that 69% of confirmed COVID-19 cases were asymptomatic. On the other hand, a surveillance of Severe Acute Respiratory Infections patients (SARI is defined according to the WHO as an acute respiratory infection with a history of fever or measured fever of ≥ 38 C° and cough, which has occurred within the last 10 days and which requires hospitalization) found only 2% of confirmed COVID-19 cases. So we may miss many cases.
To give you a sense of the seriousness of this, even if only 0.1% of Delhi is COVID-19 positive and asymptomatic, 20,000 infected people in Delhi could be infecting others as the lockdown eases.
Jishnu Das, Neelanjan Sircar and Partha Mukhopadhyay have devised an ‘ICMR-plus’ testing strategy. Here are its key features:
- Test asymptomatic individuals who are potential ‘super-spreaders’, i.e. those susceptible to infection who interact frequently with others during lockdowns, such as health workers, the police, civic workers in essential services and street vendors. Many in such groups were found to be infected only after they became symptomatic. Such groups can be tested at work itself, with samples pooled for those who work together and the usage of Reverse Transcription Polymerase Chain Reaction (RT-PCR) techniques. Their contact points can be traced, right up to where they live. Such potential ‘super-spreaders’ can be checked weekly for symptoms.
- Demarcate areas where you expect a high risk of transmission and/or high vulnerability. Eg. densely populated cramped settlements, areas with larger numbers of elderly people. In these areas, choose people in an explicit statistically structured (a two-tiered structure: areas with high risk and, in them, people with high risk; so as to maximise the probability of finding someone who's infected as well as maximise the efficiency with which you can use any testing protocol) randomised manner and, again, test them using RT-PCR methods. The chances of finding infected persons can improve by first using local information - eg. people reporting symptoms of influenza like illnesses - and then drawing randomly from the area’s voter list. Information on the intensity of contact with others should be collected from those tested, along with data on occupation, age, gender and any previous illnesses.
- States should anonymise and release this data along with associated test results. Besides the test results, the data would comprise demographic data, eg. related to age, gender, occupation, biologically relevant data, eg. medical history of illnesses, weight, as well as other data: how mobile people have been and whether they belong to particular essential services groups. The randomisation (choosing experimental participants randomly, to eliminate bias) will allow us to analyse, in real time, the possible causes of transmission, even in places not directly tested. For eg. if there are 20,000 localities in a state and only 5000 have been sampled, but if we know, in detail, the characteristics of these 5000 localities, we can say predict that certain localities - localities with slums is an example - will have a higher infection and mortality rate. Information from the initial stages of testing and collecting data must be used to improve sampling designs after each round of tests. As data feeds into the system a better predictive model can be built, which will influence sampling decisions. For instance, how risky an area is will determine whether we choose to sample it. Crowd-sourcing this analysis - by inviting inputs from the public, but especially experts who are not necessarily already working with governments - will improve the speed and quality of analysis and allow for not just a better epidemiological understanding but also an improved policy response.
To know more about the ‘ICMR plus’ strategy and the reasoning behind it, please read this article.
Swati Piramal is Vice-Chairperson Piramal Enterprises; Mahesh Balsekar is a senior paediatrician; Both of them are a part of Mahacovid, a voluntary informal group including foundations, management consulting firms, companies, public officials and experts in public health and data analytics who are working on the ground in Mumbai, Maharashtra, to help fight the Covid 19 pandemic. Jishnu Das teaches at Georgetown University; Neelanjan Sircar teaches at Ashoka University. Partha Mukhopadhyay is a senior fellow at the Centre for Policy Research (which Das and Sircar are affiliated with as well).
Special thanks to Neelanjan Sircar for his help in clarifying these recommendations for us.
NOTE: ICMR just issued a revised/updated testing strategy ("version 5") today, on May 18th (or the day this article was published).
It prescribes testing the following:
"1. All symptomatic (ILI symptoms) individuals with history of international travel in the last 14 days.
2. All symptomatic (ILI symptoms) contacts of laboratory confirmed cases.
3. All symptomatic (ILI symptoms) healthcare workers / frontline workers involved in containment and mitigation of COVID-19
4. All patients of Severe Acute Respiratory Infection (SARI).
5. Asymptomatic direct and high-risk contact of a confirmed case to be tested once between day 5 and day 10 of coming into contact.
6. All symptomatic ILI within hotspots and containment zones.
7. All hospitalised patients who develop ILI symptoms.
8. All symptomatic ILI among returnees and migrants within 7 days of illness.
9. No emergency procedure (including deliveries) should be delayed for lack of test. However, sample can be sent for testing if indicated as above (1-8) simultaneously."
An NB ('Nota Bene' or Note Well) adds:
- ILI case is defined as one with acute respiratory infection with fever ≥ 38 C° and cough.
- SARI case is defined as one with acute respiratory infection with fever ≥ 38 C° and cough and requiring hospitalisation (which has occurred within the last 10 days).
- All testing in the above categories is recommended by real time RT-PCR test only."
However, this revised ICMR testing strategy, unfortunately, does not address most of the key issues raised by Das, Sircar and Mukhopadhyay. Neither does it take into account the broader approach of the 'ICMR plus' strategy recommended by them with regard to the collection and analysis of data. Ideally, an understanding of who must be tested, and when, must be grounded in science and the existing ICMR data. While there are significant challenges in aggregating this data, failure to provide such documentation may lead to increased burden on an already limited supply of tests.
#COVID19 #Medical #statistical #SwatiPiramal #MaheshBalsekar #RapidAntibodyTests #diagnostictool #ContactTracing #specificity #sensitivity #moleculartests #ReverseTranscriptionPolymeraseChainReaction #RTPCR #infection #antibody #nutrition #diseaseseverity #clinicalintervention #transmission #epidemiological #surveillance #community #falsepositives #hotspots #asymptomatic #symptomatic #protocols #policies #lockdown #containment #epidemiology #StanfordSchoolofMedicine #effectiveselection #testingmethodology #robuststatisticalanalysis #representativeness #randomization #demographic #pooledRTPCR #statisticalanalysis #accuracy #testingstrategy #testingkits #lowtestnation #IndianCouncilofMedicalResearch #ICMR #SARI #SevereAcuteRespiratoryInfections #WHO #JishnuDas #NeelanjanSircar #ParthaMukhopadhyay #ICMRPlus #superspreaders #statisticallystructured #data