Signs of falling Covid-19 numbers may be deceptive, warn experts

7-day rolling average of daily cases down from 95,000 in mid-Sept to below 85,000 now: Report

The seven-day rolling average of daily Covid-19 cases in India has declined from the peak of around 95,000 in mid-September to below 85,000 towards the end of the month, according to data maintained by COVID19India, an independent data repository.
 

This gives a sense that the epidemic is coming under “control” in India.
 

The same period shows another positive trend: The rate at which new infections spread has also declined below 1, implying that 100 infected people now infect less than 100 new uninfected ones. This is so if one goes by the parameter R — the dynamic reproductive rate — formulated by University of Chicago-led researchers. However, data across most affected states shows that this positive turn has happened when testing has been slowing down in the country.
 

Experts, too, insist that policymakers observe caution as short-term indications on improvement could be deceptive.
 

At the national level, more than 800 people per million population were being tested daily till about a few days ago. The rate has now gone below that mark. The adjoining chart shows how the decline in rate of persons turning out as Covid-19 positive was associated with a decline in testing. The blip at the end shows how the positivity rate went down as more tests were conducted on September 24.

 

Among leading states, Karnataka is on a similar track. The southern state is showing a rise in test positivity rates hand-in-hand with a decline in testing. Maharashtra, the most affected state, is improving in terms of falling positivity rate, but with falling tests per million, too. Tamil Nadu and Andhra Pradesh, on the other hand, are showing a secular decline in test positivity rates along with improved testing.
 

Delhi’s increased testing and falling positivity are also contributing to taming India’s national average into the “improving” territory, where R is going below 1.
 

This divergence among states, and the improvement on an average, brings up a questions: How much should we look into this esoteric number R? There are a few key limitations to the conclusions that can be drawn from this data.
 

Satej Soman, a lead researcher in the UChicago-led group, said that a current estimate of R may undercount the actual spread of the infection as there is delay in data availability.
 

“The reproductive rate R going below 1 for a few days is not necessarily a cause for celebration. Conversely, a single day above 1 is not always a cause for concern. These numbers are driven by random processes. What we should want to see is extended periods of time with R below 1,” he told Business Standard.
 

According to their model, Adaptive Control, 100 infected people are infecting about 90 new people since September 19. The infection spread rate has largely been above 100 new-for-100 infected till now.
 

Now, the R0 (R naught), for SARS-CoV2 is estimated to be 2.5, according to a recent Lancet paper. This means that put in a horde of uninfected — thus susceptible — people, 100 infected would infect 250 uninfected ones. But the one currently under consideration is effective R, that changes over time as the virus propagates among the human population.

“In epidemiology, when the effective R equals 1, the disease has reached its peak of the epidemic stage. An R value consistently below 1 would suggest that the epidemic is no longer an epidemic, but an ‘endemic’: The disease is on a decline,” said Jayaprakash Muliyil, who chairs the scientific advisory committee of the National Institute of Epidemiology.

The situation on ground, however, is not in sync with this observation.
 

Eminent virologist T Jacob John, said: “When testing is not uniform, we cannot accurately decipher the effective R value. The efficacy of value calculated from available data is limited to understanding the epidemiology, but not for policy formulation.”
 

But he also said his own assessment suggests that India may have just turned the corner, and is on the descending limb of the epidemic curve.
 

Further, research has shown that the R estimate may have a limited practical value outside the population from which the data has originated. Differences in data collection and surveillance can also have an impact, according to a study entitled ‘Unraveling R0: Considerations for Public Health Applications’ from authors Benjamin Ridenhour, Jessica M Kowalik and David K Shay published in the American Journal of Public Health.
 

“In isolation, R0 is a suboptimal gauge of infectious disease dynamics across populations; other disease parameters may provide more useful information,” the study noted.
 

A February 2020 study also advised caution on the use of the metric.
 

“Although the concept of R0 is very intuitive, its calculation is based on complex models and may lead to misinterpretations.

This is especially for what concerns the real weight that R0 has on the spreading of an infectious disease,” Giulio Viceconte and Nicola Petrosillo wrote in the paper, ‘Covid-19 R0: Magic number or conundrum?’.