Search interesting materials

Sunday, April 12, 2020

More testing: From concept to implementation

by Ajay Shah.

There are concerns about quantity and quality in Indian Covid-19 data. In terms of the quantity of testing, India is at 120 tests per million of population, which is among the lowest in the world. Countries like Germany, Italy, South Korea, etc. are at values of about 10,000 tests per million people. In addition, in India, there are concerns about the imprecision in measurement of infected persons and deaths.

As cities and states grapple with the challenge of Covid-19 in the coming year, improvements in testing are being envisioned. Everyone wants more testing. What does it mean to do more testing? In this article, we show the four elements of testing, from a public health point of view. At present, only one of these is in play in India. Public health leaders at the city and state level need to think about this full range of possibilities, and put all of them into motion.

Basic facts about testing


Before we get to the interesting public health and public policy questions, let’s review the science and engineering of testing. For people who are infected with Covid-19, the overwhelming majority recover in about 14 days. Most people experience minor symptoms, do not seek health care, and would not know that the Covid-19 infection took place.

There are two kinds of tests: The PCR test and the antibody test.

The PCR test looks for the virus in your body. It delivers a positive verdict when the virus is present at concentrations above a certain threshold. The absence of the virus can either mean that you are not infected or that you have recovered. Generally, by day 10, the PCR test returns a negative result.

The antibody test looks for the presence of antibodies made by the human body in fighting the virus. These antibodies are generally in place from day 7 and are likely to persist for months. It is possible that while the antibodies are present, you are still communicating the disease, which roughly corresponds to the second week of the progress of the infection.

After day 14, generally the antibody test would yield a positive result and the PCR test would yield a negative result.

After the recovery, immunity from the disease is likely. This paves the way for reopening the country. Ideally, we would like to run large scale antibody tests, find the people who have bounced back from the infection (and might not even have known that they were sick), so these persons can get back to normalcy in their lives. While this is a field of active research, the conclusions are not yet in. It is likely that once you recover, you are immune to the disease, for about 2 to 3 years. But Covid-19 is a new disease, and these are not settled questions.

In terms of operationalising the two tests, the PCR test requires more expensive equipment, the test takes more time, and scaling up to millions of tests is hard given global supply shortages of many of the inputs (reagents, primers, positive controls, extraction kits). PCR testing is unlikely to be available in large quantities in India or other countries for the next month or two, which will limit its usage to relatively low volume applications. The antibody test, though not a diagnostic test for health care settings, is more readily rolled out at scale, and most countries are exploring this from the viewpoint of mass deployment.

There is a neat idea called a “Group test”. Imagine pooling the blood sampled from 10 people and doing one PCR test on that pooled sample. If this came out positive, we would conclude that one or more of these 10 persons is infected. This is particularly useful given the limitations on the number of PCR tests that can be done. At the present moment, the precise protocols for pooled PCR testing for Covid-19 are not yet in place. This is also the subject of active research and we expect this problem will be solved soon.

This shows a testing landscape with two tests (PCR and Antibody) and the possibility of doing a group test with either.

With both classes of tests, there are many vendors with products of varying quality. It would be efficient for India to free ride on the state capacity of advanced countries, and accept any product which is approved by a regulator in a country that is a member of the OECD.

The role of testing in public health and health care


When we are admonished to increase testing, what testing is to be done at scale? It is useful to go to first principles, and think about the objectives of public health. This guides us in designing testing strategies. There are four pathways to testing, organised by the objective of the testing, by the question that is being asked.

  1. Testing in the context of health care: A person shows up in front of a doctor with certain symptoms, and the doctor commissions a test in order to know whether it is a Covid-19 infection. In this case, the question is: Is this person presently infected with Covid-19?
  2. The public health objective of understanding the state of a neighbourhood: When thinking about rules of social isolation in a geographical area, an assessment is required about the state of a neighbourhood. As an example, an airport may be in operation, and we might like to get a daily reading of the state of infection in the airport staff. In these cases, the public health crew cares about the question: Is there an active infection in this group of people?
  3. The public health objective of understanding the progress of the epidemic: The public health team in a city needs to have a situational awareness about what is going on in the city. In this, they would ask the question: What is the overall number of infected and recovered people in my city? How are these numbers moving over time? Each city would like to know: Is the active level of infection likely to rise beyond the available health care in the city in the next few weeks? This forecasting can be assisted by statistical epidemiological models, which can be estimated once this data is observed. In addition, understanding the epidemic curve will help guide decisions about escalation or de-escalation of social distancing measures in the city, and across cities.
  4. Antibody testing for the purpose of restarting the economy: It is likely that there is immunity against infection, for a few years, after a person has recovered from the infection. The entire course of events, from infection to immune system response, can happen without displaying any symptoms. Hence, counting the number of people who sought health care and then recovered is not useful. The antibody test shows whether a person has such immunity. Such persons are likely to be ready to rejoin the economy, and should particularly be brought into front-line roles. From the viewpoint of society and the economy, the key factor to watch for is the fear; Covid-19 is hacking into our minds, much like a terrorist attack. Each of us would like to know: Have I finished with one bout of Covid-19? Once a person tests positive on the antibody test, the terror would subside, one person at a time.

We see that there is not one concept in testing. Testing technology (PCR, Antibody, Pooling) are technical tools that are synthesised to answer four categories of questions. Each of these objectives is distinct, requires a different mechanism for implementation, and supports decision making in different ways. Let us dive into each of them.

Pathway 1: Testing in the context of health care, “Is this person infected with Covid-19?


The normal protocols of clinical care will be applied by a doctor, who will trigger a test when certain symptoms come together. The test of choice is the PCR test, because this will report on the presence of the virus in the first week. These results are very interesting for the individual and for the doctor in determining health care for the individual.

The statistics that are produced, out of such testing, are highly sensitive to: (a) Who are the individuals who feel symptoms (most don’t) and who are the individuals that access health care facilities; (b) The protocols and skill of the doctor in deciding to prescribe the test and (c) Capacity constraints in doing the PCR (maybe all the machines in the country max out at N tests/day) and the fact that when there is a backlog, delays in the heat might degrade samples and bias the results in favour of a negative outcome.

A lot of epidemiological research is presently being done using data that is produced from this clinical setting. It is important to be cautious about the extent to which this data can be interpreted. The only thing that we are sure of is that such testing helps in the clinical process. The four kinds of censoring described above (some individuals access health care, the protocols used by the doctor, the limitations of testing, the degradation of samples) are central to the data generating process, and are absent in most models of this data. This limits the usefulness of epidemiological models, when estimated using the existing data, for decision making in public health.

One step away from the clinical setting, and closer to the questions of public health, are PCR tests administered every day to samples of high risk groups (SARI, ILI, health care workers). These could be an early indicator of the spread of the disease and could help containment efforts. Such measurement projects are important and interesting for public health. Fusing this surveillance data into an overall dataset of clinical data, however, induces additional difficulties for epidemiological modelling.

Pathway 2: The public health objective of understanding the state of a neighbourhood, “Is there an active infection in this group of people?


At the level of a city, there is value in containing outbreaks. In an ideal world, we would have a PCR test result for each person for each day, and this would generate perfect information. However, this is not feasible as PCR testing is slow, expensive and does not scale up readily.

Public health staff face questions such as “Has an outbreak begun in X neighbourhood?” “Has the outbreak in X neighbourhood ended?” or “What is the state of health of the airport staff?”. These questions can often be nicely answered by doing a pooled PCR test [link, link]. The results can be used for modifying social distancing and isolation procedures on a day to day basis.

There is value in establishing the institutional infrastructure through which a civil servant is able to ask this question about a neighbourhood, after which a random sample of N persons is taken, and the pooled PCR test is run. Ideally, the turnaround time from decision to result should be about two days.

There is also value in establishing ongoing monitoring of high risk activities, such as health care workers or the airport crew, who should be sampled every day in this fashion, thus inducing the systematic creation of datasets.

Pathway 3: The public health objective of understanding the progress of the epidemic, “What is the overall number of infected and recovered people in my city?”, “How are these numbers moving over time?


The holy grail of this field is panel (i.e. longitudinal) data measurement of persons on infection and presence of antibodies. It would be particularly valuable to observe comprehensive socio-economic information about each individual, over and above these two facts. As an early step towards this objective, a recent study in one town in Germany measured 500 persons and found that 2% were infected and 14% were immune.

Suppose we are sampling 1000 persons, and suppose the true positive rate is 2%, and 2% of 1000 is 20 persons. To fix intuition, on average we would see about 20 persons testing positive, and a few sample realisations are : 24, 16, 25, 22, 27, 19, etc. In this, a 95% confidence interval of the estimated rate runs from 1.1% to 2.9%. This suggests that at rates of about 2%, a sample size of 1000 individuals is quite useful. (Lower rates call for larger samples).

With this dataset is in hand, it becomes possible to estimate epidemiological models, understand R0, anticipate the future course of the disease, watch how modifications to social distancing impact upon R0, etc. This can be particularly useful in anticipating surges in health care requirements, building and dismantling temporary hospitals, etc. These models can help improve decisions about social distancing measures in the city taken by the citizenry and public health authorities.

Pathway 4: Antibody testing for the purpose of restarting the economy, “Have I finished with one bout of Covid-19?


India has young demographics, and many will get the infection and bounce back without noticing it. Individuals should be able to test their own antibody status and put their fears at rest. There is a role for testing at regular intervals for each person, until that person gets the first positive reading.

This is important for restarting the economy. The economy today is hampered by lockdowns and fear. Antibody testing on scale holds the key to ending the fear, one person at a time, and restarting the economy. Persons who have a positive antibody test result should be preferred for front line roles.

Covid-19 has imposed large welfare costs upon humans who have been forced to be away from their loved ones. Widespread antibody testing will help the lucky ones to resume desired human interactions.

Elements of implementation


When we think about these four questions at the level of India, implementations appear implausibly difficult. If we think of creating a panel dataset of 1000 people in Bombay, it is much more tractable when compared with trying to create an all-India panel dataset.

The presence of four pathways to measurement, in any one city or district, creates one immediate advantage: it is then more likely that errors in any one element of the measurement strategy will be detected, and feedback loops established for remedying them.

Some ideas for implementation are sketched below.

  1. The first stage lies in establishment of a mechanism to take in data from all over the country, on all the pathways, and make it available as a unified repository to the public. A coalition of researchers should establish the data standards for all data coming into one shared public facility with data, where all incoming data is instantly released into the public domain. The governance of this effort should combine experts on public health, information systems, and civil liberties. Trust in this data will be enhanced if there is a lack of government control of this repository.
  2. There should be published protocols that determine when a doctor asks for a Covid-19 test as part of Pathway 1. There should be a single definitive source where these protocols, and all changes in these protocols, are made visible. When the doctor writes a prescription, this would generally be filled by a lab. The lab should get paid by the state in exchange for submission of a few (anonymised) facts about the individual, back into a public data facility.
  3. For Pathway 2 (pooled PCR testing in a neighbourhood), a problem that is faced is the stigma and fear that goes with Covid-19 in India today. The city/district government will need a contract with a lab to do the sample collection. Civil servants may need to accompany the staff of the lab in order to make the citizenry comfortable with what is being done. In time, there will be much comfort when people realise that no individual is being identified in the group testing. The contracting framework is required through which a civil servant commissions a test, and in a day or two the answer is obtained. Similarly, the contracting framework is required through which (say) a data point is obtained every day about a random sample of the airport staff. All the data generated here should be anonymised and go to the public data facility.
  4. For Pathway 3 (panel measurement), survey organisations partnering with testing labs are required to meet households and obtain samples. There is an important barrier in this, owing to the stigma and fear surrounding Covid-19 in India today, and these problems will need to be overcome. All the data generated here should be anonymised and go to the public data facility.
  5. For Pathway 4 (antibody testing for an individual), there is a role for a testing voucher through which each person can get tested every x weeks until a test shows positive. This would kick off a decentralised mechanism where individuals would step forward and get tested. Private labs would be required, as part of the voucher arrangement, to electronically submit anonymised data to the public data facility.

At present in India, most of the work in testing is on Pathway 1. In addition, most of the work in this is being done by government labs. Most of the testing capacity in India is in the private sector, so these pathways need to establish incentive-compatible PPP arrangements through which work is done in private firms with public funding, with release of anonymised data into the public domain. Perhaps a nice split is to have the public sector continue to play an important role in Pathway 1, and establish the additional three pathways in private labs.

There is a lot of concern about the practical problems of organising production at private labs. The private sector is best equipped to understand and solve complex problems of supply chains and organisation. The PPP contracts, that a city or a district gives to multiple labs, should establish frameworks for payment for tests, and also embed real options whereby the private firms will be paid for a certain floor level of testing even if the order flow does not materialise. This will create incentives for private laboratories to build their organisational capabilities and solve problems of production. Private firms will solve problems of organising production better than governments will.

All four lines of work will require time and effort in implementation. The Covid-19 epidemic is a problem that will play out in India over a year or two. It is efficient for each city or district to embark on a three month journey to establish these information systems, so as to exert a beneficial impact upon decision making thereafter.

Conclusion


Everyone agrees that more testing is required. But testing is a means to an end. The purpose of testing is to improve situational awareness. At present, we are in a state of high uncertainty; we do not know what is going on, and this induces greater fear and hampers decision making by private persons and by policy makers. The mere intensification of the existing approach to testing (i.e. testing in a health care context) does not address key objectives in public health and in restarting the economy.

To test is human, to create datasets divine. We should shift focus from the words "number of tests" to the intellectual clarity around the four kinds of datasets, each of which has their role in the overall problem.

Fighting an epidemic is inherently a decentralised problem. This is a battle that is played out at the field level. It is more useful to think about the problem of measurement at the city or district level. These four questions will have to be faced by the public health leadership in each city and each district. As an example, the leadership of Pune should ask themselves: How do we organise these four lines of work? This is far more feasible when compared with solving these problems for 3.3 million square kilometres.



I thank Prakash Hebalkar, Manoj Mohanan, Nandu Saravade, Pradnya Saravade and Suja Thomas for useful discussions.

No comments:

Post a comment

Please note: Comments are moderated. Only civilised conversation is permitted on this blog. Criticism is perfectly okay; uncivilised language is not. We delete any comment which is spam, has personal attacks against anyone, or uses foul language. We delete any comment which does not contribute to the intellectual discussion about the blog article in question.

LaTeX mathematics works. This means that if you want to say $10 you have to say \$10.