IIT-B scientist predicts Covid-19 toll may peak in hotspots in 2.5 months but others disagree

A disease prediction model proposed by a Nobel laureate and tailored for India by an IIT Bombay professor suggests that the peak of Covid-19 death count in Indian hotspot states would be over in another 2-2.5 months barring Karnataka where the toll has begun to rise.

Several Indian researchers tracking the pandemic using their own models, however, didn't give credence to such a prediction citing technical flaws associated with the model.

Proposed by 2013 Nobel laureate Michael Levitt, the scheme relies on a simple mathematical formula, which has been applied to India by Bhaskaran Raman, a computer science professor at Indian Institute of Technology, Bombay.

Follow live updates on the coronavirus here

Going by the prediction, Mumbai will witness the peak in Covid-19 deaths in another two weeks whereas for Delhi, it is still two and half months away. For Chennai city, the peak would come after a month.

Looking at the states, the same model suggests a peak in the death toll in Tamil Nadu in six months whereas for Maharashtra, it would come after two more months. Gujarat has already achieved it but the pandemic pain for Karnataka has just begun. For Kerala, the data is too small to make any meaningful prediction.

“The model is childishly simple and hence blindingly brilliant. It has been known since March-April when Michael Levitt proposed this metric based on his analysis of Hubei data, and later for other countries’ data,” Raman said in a presentation. The model can be used to predict both Covid-19 deaths and cases.

Several researchers disagree with the formula. "This method does not take the disease dynamics and biological characteristics of the outbreak into account and linearly extrapolates the growth rate observed over a finite period, which is bound to deviate later on," Tanmay Mitra, a scientist at Helmholtz Centre for Infection Research, Germany told DH.

"In case of India, this method is bound to fail as we still have a time-dependent reproduction number (Rt) with a value substantially more than 1 (about 1.4 for India as a country around the second week of July) and we are still continuing on a rising phase of the pandemic."

"The problem in Levitt's method is that it linearly extrapolates the growth rate observed over a finite period for future times, with no physical/epidemiological basis. We need much better evidence for reliability of this simple extrapolation," noted Ramachandran Shankar, a retired professor at the Institute of Mathematical Sciences, Chennai who provided some of the data for the study.

Mitra, however, conceded that the metric might be useful to anticipate the peak of the reported daily deaths, owing to the fact that the growth curve of reported deaths were less exponential than that of an infection curve.

“The analysis is not based on any epidemiological model but on a curve fitting of the available data. So it would "explain" the past data well. In predicting when the pandemic will end, the team makes the assumption that the future behavior of disease spread will be determined by its past behaviour. This may or may not be the case,” observed Dibyendu Nandi, a professor at Indian Institute of Science, Education and Research, Kolkata.

“This may match the future behavior of some states but fail to do so for others. It will depend on whether or not the epidemic dynamics, specifically containment measures, remain the same or change from the fitted behaviour,” he added.