Covid: Fight uncertainty with data
Collect granular data on incubation period, serial interval and R0. Frame policy with these findings
Arguably the most important scientific statement on the coronavirus disease (Covid-19) was made by Dr Anthony Fauci, director of the United States (US) National Institute of Allergy and Infectious Disease, in his testimony to the US Senate, “But I am very careful, and hopefully humble in knowing that I don’t know everything about this disease.” This is the reality one is confronted with and, therefore, the biggest challenge, political and scientific, is to make decisions in uncertainty. Our best hope perhaps in the short- and mediumrun is that we urgently invest in building our knowledge about the virus, while hoping in the long-run that we will have a vaccine.
In the early days of the pandemic, limited data availability led the first generation epidemiological models to predict that hundreds of millions would be infected with the virus, and millions of lives lost. Fortunately, the reality so far has turned out to be very different. However, the situation is still worrisome. From 657 cases on March 25, India has more than 125,000 cases today. Even more troubling is that active cases continue to grow, despite a significant slowdown since the national lockdown.
To understand the future trajectory of the pandemic and to frame appropriate policy responses, we need granular data based on contact tracing at the level of a city or district to provide information on three important epidemiological parameters. One, incubation period, which is the interval between infection and symptoms. The distribution of this parameter helps the government and experts understand the nature, extent and possible future scenarios of the outbreak. It also informs in the evaluation of the disease-control strategy.
Two, the serial interval, which is the time between the onset of the illness in the primary case (infector) and illness onset in the secondary case (infectee). If the estimated average of the serial interval is shorter than the estimated average incubation period, then presymptomatic transmission is more likely to happen than symptomatic transmission. Research from Japan has indicated that the median serial interval for Covid-19 is 4.1 days, which is less than the mean incubation period of approximately five days. The public policy implication of this is that containment via case isolation might be a challenging task. Containment would, therefore, require to be guided by an aggressive testing and rapid contact tracing strategy.
Three, the basic reproduction ratio (also popularly known as R0), which is the average number of secondary cases per primary case. There has been a great deal of focus on this parameter, because if the R0 is greater than one, then the probability that there will be an outbreak is extremely high.
Given the importance of the parameter R0, one has to exercise great caution in interpreting and estimating it. Most models that estimate this parameter assume that all individuals have a homogenous transmission and constant recovery rate. Therefore, a populationbased R0 is estimated with the implication that if this is greater than one, then outbreaks from a single infected person is highly likely to happen. However, research based on the previous Severe Acute Respiratory Syndrome (Sars) epidemic in 2003 has shown that there is a great deal of variability in individual infectiousness. For example, research on Sars epidemic from Singapore revealed that the majority (approximately 73%) of the cases were mildly infectious; in other words, they had an R0 of less than one, while a small proportion of them (approximately 6%) was highly infectious or “super-spreaders” with an R0 > eight.
The variability of R0 plays an important role in the dynamics of an outbreak. Models that account for individual variability show that even if the population-based R0 is greater than one, an outbreak could still be a lowprobability event. Introducing individual-level variability in the model thus explains why during the Sars epidemic in 2003, several cities did not witness explosive outbreaks despite undetected exposures to infectious cases. In these models, outbreaks are typically caused by super-spreader events (SSEs).
In the Indian context, this might explain why Mumbai is experiencing an explosive outbreak, while many other large, highlydense cities with significant populations dwelling in slums, are not experiencing such an outbreak.
The above point becomes apparent when one compares Kasaragod to Mumbai. On April 2, Kasaragod had 127 confirmed cases, while Mumbai had 185. However, by April 16, there were zero new cases in Kasaragod while Mumbai experienced a devastating outbreak. In late March, the police in Kasaragod, adopted an aggressive contact tracing model, and identified approximately, 20,000 potential “super-spreaders” — these were primary and secondary contacts of those who returned from Gulf countries. A strategy of “triple lock down” was adopted by the police, whereby these potential super-spreaders were put under a more stringent home quarantine compared to the rest of the people in the district.
This prevented an SSE in Kasaragod and minimised the risk of an outbreak. A key implication of this from a policy perspective is that if highly infectious individuals or superspreaders can be predictively identified, we could avert more general lockdowns in the future. Moving forward, armed with more granular data and a better understanding of the Covid-19 virus, we could move away from a policy of general lockdown towards a policy of a smart lockdown.
It is important to remind ourselves that we know very little about the virus. Our best hope, until the vaccine is discovered, is to collect as much granular and disaggregated data as possible on the epidemiological parameters that have been outlined here. This should inform our real-time policy in the collective fight against the Covid-19 virus.