Test, trace and isolate – classic infection control measures – have been widely endorsed as the best way to tackle the Covid-19 pandemic, even as the world waits for a vaccine to be developed. These basic public health principles were evolved by epidemiologists. Physician John Snow’s use of maps and records to track the spread of a cholera outbreak back to its source in 1854 in London, provided the foundation of identifying and tracking diseases.
In this interview with Scroll.in, Professor Madhukar Pai, Canada Research Chair in Epidemiology and Global Health at McGill University, Montreal, explains the role of epidemiologists or ‘disease detectives’ in battling Covid-19, why it’s tough for epidemiological disease models to predict the future of the pandemic and the challenge of imperfect data.
Public health is the discipline that addresses health at the population level: preventing diseases, protecting and improving the health of communities. We know that epidemiologists are public health scientists and researchers. What exactly does that role entail, in regular, non-pandemic times?
Epidemiology, no doubt, is a key component of public health, but is also widely used in clinical medicine (where we call it ‘clinical epidemiology’).
Epidemiology is the application of the scientific method to health research (public health or clinical medicine). Using epidemiologic research, we generate evidence to answer questions like what is the cause of a disease, what are the risk factors for getting the disease, what is the prognosis of a disease, and what treatments can prevent or cure the disease.
All of these are evident in the current Covid-19 pandemic. Epidemiologic research has shown us the cause of Covid-19, risk factors for getting the infection or dying from it, case fatality of the infection, prevalence of the infection at the population level, and clinical trials are showing us which treatments work or do not work.
How does the work done by epidemiologists help governments or non-governmental entities plan and implement public health interventions? If you could give us an example that illustrates how work done by epidemiologists influenced or changed the trajectory of major public health interventions.
Historically, disease surveillance and epidemiology played a key role in eradication of smallpox and elimination of polio in most countries. Without epidemiology, we would not have had these successes.
I can give you examples from my own field of tuberculosis. Epidemiological research on TB has shown us who is likely to get TB, who is likely to die of TB, how to diagnose and treat TB. Such research enables agencies such as WHO to make policies and guidelines, which are then implemented by National TB Programs in various countries. Thanks to such efforts, TB mortality has declined over the years, but we still have a long way to go.
AIDS is another great example: clinical trials have provided great evidence on how to prevent HIV transmission and how to effectively treat people with anti-retroviral drugs. This has greatly reduced the burden of HIV/AIDS globally.
In the early days of the Covid-19 pandemic, epidemiologists had said more accurate projections of how this will unfold, how many people will become infected, and how many will die, will be better made when enough reliable scientific data is available. Six months into the pandemic, are we in a position to do that?
Since Covid-19 is a new disease, everyone, including epidemiologists, are learning as we go along. Nobody can afford to be confident that they have the answers. So, early projections or correlations are often proven wrong, but that does not mean researchers were dishonest or that epidemiology is not working. Science is all about improving upon what we know and changing course, as new evidence appears.
For example, early data suggested some promise with hydroxychloroquine for treating Covid-19, but, today, based on randomized trials, we know that HCQ is not an effective therapy. So, many trials have discontinued the HCQ arms. This is how epidemiology and science works. We start with some hypotheses and we then do studies to prove or disprove them.
It took years to prove that smoking causes lung cancer and for anti-tobacco policies to get widely implemented. With Covid-19, we have just had 6 months. So, not all questions are answered at this time. Eventually, with time, they will be.
Governments and policymakers have been relying on disease models to determine when their jurisdictions could possibly run out of hospital beds, ventilators and ICU beds, how much time the administration has to ramp up facilities, or when and in what manner to reopen the economy.
How is disease modelling done? And if the input data is largely the same, how have projections varied between models? For example, in the Indian context, some models claimed that the pandemic in India will be over by May-June.
Yes, lots of confusion around modeling studies. The public and policy makers want certainty, but that is not what models can offer. Models are not designed to predict the future, nor to capture the full complexity of the world. But models can provide a decision-making framework that is systematic (better than just guesswork), data-driven, transparent (all assumptions are openly stated), comprehensible, replicable by others, and able to describe uncertainty.
So, all models need to be publicly shared with the code used for analysis, and with a list of model parameters used and assumptions made. Transparency is key. That has not been the case with many models on Covid-19 in India. If someone shows a model as a powerpoint slide in a media briefing but does not share the full report and code used, that model is useless and cannot be trusted.
Regardless of how good the model is or how good the modeling team is, modeling Covid-19 has been challenging for many reasons:
• New disease – everyone is learning
• R0 is not static – it changes over time
• Testing rates are changing
• It is unclear how long natural immunity will last
• Proportion of population infected is unclear
• Case fatality is highly variable
• We don’t quite know if asymptomatic people infect others
• We don’t quite know the duration of infectiousness
• We don’t quite know if children transmit infection
• We don’t know how many people come into contact with a case
• We don’t know if there are ‘super-spreaders’
• Lockdowns, restrictions & other interventions keep changing
• Human behaviour is not easy to predict
So, all models can be off-base, but they can still be helpful.
Some experts have argued against relying on mathematical modelling and instead focussing on dynamic real-time data to make decisions.
It is not one versus the other. We need both approaches. We need good primary data on many of the questions I have listed. Once we have that, we can use them as input parameters to refine our models. So, good modeling requires primary, input data. That is constantly coming in and getting better. Modeling done in July will be better than modeling studies that were done in January or February.
Last week, commenting on a surge in cases in America, Dr Anthony Fauci, director of the US’s National Institute of Allergy and Infectious Diseases, said “America is still knee-deep in the first wave of this”. What is a wave in a pandemic?
The trajectory of an epidemic can be plotted on what we call an ‘epidemic curve’ where time (days) is on the X-axis, and number of daily cases or deaths is on the Y-axis. If the numbers keep climbing, then we are still in the first wave or peak of the epidemic. If the numbers reach a peak, then flatten, and then decline, then the epidemic is subsiding. If the curve dips and then goes up again to produce a second peak, then we call that a ‘second wave.’ Some places did manage to control the first wave (for example, Singapore, Wuhan), but then had a second peak. Case numbers in the US declined, but are now going up again.
In India, we have been talking about whether infections have peaked, if there would be an all-India peak or would it be more localised. How exactly do epidemiologists determine if infections have peaked and does a peak mean good news?
India is yet to peak in the first wave and there are other countries like India (for example, USA, Brazil) where case numbers are increasing every passing day. And there are countries which have successfully ‘bent the curve’ and reduced case numbers to a low level (for example, New Zealand, South Korea, Germany).
If case numbers keep going up, then the peak is yet to come. When case numbers stabilise (that is, roughly similar numbers each passing day), then we say the curve is starting to plateau or flatten. When case numbers decrease every passing day, then we say that the curve is dipping (that the outbreak is getting controlled).
Apart from the number of cases, two other parameters are referred to a lot: doubling time and death rate. Are these the most reliable indicators of how good or bad the situation is?
Every epidemic needs us to use multiple measurements to keep track of: incidence rate, case fatality rate, infection fatality rate, doubling time, excess mortality, etc. I have an entire lecture on this. No single number is enough to capture the complexity of a pandemic.
The biggest challenge, in my view, is the low rates of testing in India. Without adequate testing, we can never get the real scale of the epidemic. If only really sick people get tested for Covid-19, we will always underestimate the numbers. So, if India has reported about 742,000 confirmed Covid-19 cases, we know the real numbers have to be much higher, since most people with symptoms are currently not getting tested, and we also know most people have little or no symptoms.
The virus, contrary to early speculation, has not been killed off in warmer weather. Some epidemiologists (like Jayaprakash Muliyil) now believe that bigger countries will just have to learn to live with the virus until such time that the population acquires herd immunity. Would you agree or disagree with this?
Unless we get a vaccine soon, it is clear that most countries will have to learn to live with the virus. That means we will see a steady number of people getting infected. Thankfully, most people will have mild or no symptoms. The goal is to keep this number at a manageable level, so the health system has the capacity to deal with those who are really sick and need hospital care.
Eventually, if enough people (say about 60%-70%) in a population get infected, then herd immunity starts to play a key role and will protect the vulnerable. But this could take months to happen. A recent study showed even in places (for example, Spain) where lots of cases have occurred, less than 10% of the population has antibodies. So, it might take several months for herd immunity to build up.
Can you tell us how you became an epidemiologist? What was the motivation?
I did my community health training at the Christian Medical College in Vellore. During residency training, I was responsible for community health initiatives in a number of villages in Vellore district, and was involved in running TB and leprosy control programs.
As a resident, I also had to tackle infectious diseases such as malaria and diarrheal diseases. These real-world experiences made me realise the tremendous importance of public health and inspired me to pursue a career in epidemiology. I was lucky to get a fellowship to do my PhD in epidemiology at the University of California, Berkeley, and my PhD thesis was on tuberculosis in India. I am now a full professor of epidemiology at McGill University in Canada, where I teach courses on epidemiology and global health.
You have been actively teaching epidemiology to journalists in India and Africa. Why do journalists need to learn epidemiology?
The Covid-19 crisis has put health reporting on the front pages of all media worldwide, but has also raised big challenges with fast-tracked research that is often contradictory, compromised normal standards of scientific peer-review and dissemination, sensationalised media reporting, and confused policy makers who are making irrational decisions.
All health and science journalists will benefit from understanding the fundamentals of epidemiology, so they can critically read medical research studies and media releases, and report them in an accurate and balanced manner. Journalists, in my view, need to be more skeptical. But many health journalists lack training in epidemiology and public health.
To address this, I offered an online course, in partnership with Suno India, for 70+ journalists in India during June and all course materials and videos are available on my teaching website. The course was very well received. I am now working with several partners to teach an epidemiology course for journalists in Africa next week.
