Coronavirus disease 2019 prevention, control and modelling metrics.

Dear Editor,

Apropos various forays into the esoteric world of epidemiological modelling in recent MJAFI publications. With reference to the article on Epidemiological Modelling of COVID 19 scenarios, we would like to recommend that the term ‘i-metric’ may be a more suited term instead of ‘q-metric’ as the authors are referring to ‘isolation’ of ‘infected’ people. In the social media infodemic that has characterized the current times, the distinction between isolation and quarantine has become blurred, with most non–public health personnel using it loosely and interchangeably. Isolation is for cases, and quarantine is for healthy contacts of those who have been detected to be infected.

The article by Chatterjee et al. provides a mathematical analysis of the options available for controlling coronavirus disease 2019 (COVID-19), which is most welcome at this juncture. The recommendations are relevant, practical and implementable.

Our concerns are the limitations that have not been mentioned by the authors, besides certain seemingly minor calculation errors that influence the overall outcomes, interpretations and recommendations. The analysis necessarily is based on several assumptions; however, it is not clear how substantially the modelled outputs may have changed if input parameters had been changed within the range of their observed or estimated values. It is possible that the proposed ‘q’ values may have had a very wide and potentially unusable confidence interval (CI) if such an analysis had been performed. It is pertinent to keep in mind that an individual becomes infectious for an unknown number of days before becoming symptomatic or testing positive for the coronavirus. Hence, even if we isolate a confirmed case, that individual would have potentially spread infection to others already before being isolated. Hence, the conclusion that by achieving q50 or q90, there will be a few thousand cases would be incorrect as the authors have assumed that these q50 will not cause even one secondary infection, which may not be an accurate representation of the ground realities.

In ibid article, the infection fatality rate is based on data from China, with evidence collected up to February 11, 2020. These data are also based on repatriation flights in the initial time frame of the outbreak in China. Extrapolating these data to calculate or develop models for Indian settings will bring forth errors and will not be a representative picture to rely on for decision-making.

Supplemental information (Table 2a) has data estimating the infection fatality rate for India; however, there is error in this data set as 95% CI does not include point estimates. The multiplication factors bring out errors in calculations from 49,842 cases per 100,000 to cases per 10 million. Table 2b has presented the percentage of people hospitalized in India. In many countries, not all of those infected were hospitalized; however, in India, this hospitalization policy has been constantly evolving at the national level, with different grades of implementation at the state level. Initially, all positive cases once diagnosed by reverse transcriptase - polymerase chain reaction (RT-PCR) were being admitted to COVID-19 care centres or designated hospitals irrespective of age and not the percentage of cases that the authors have brought out, based on their utilization of data of European hospitalizations. Extrapolating data of other geographies with differing population age pyramids and variable policies for admission/discharge of cases to the Indian setting will not be representative of what is to be expected in India and thence will not serve as a reliable guide for policymakers to take decisions.

The authors use an SEIR model, rather than the SEIRUS model, which has only rarely been reported in use with COVID-19–related modelling., The earlier published analyses either accept that reinfection is unlikely for 4–12 months or state that the chances of reinfection appear to be uncertain and possibly unlikely. Given the reality that reinfection is unlikely in an early time frame (based on the observation that otherwise we would have had large numbers of reinfections being reported worldwide), it would have been of interest to see an analysis of ring-fencing symptomatic people with test-converted (recovered) individuals, rather than a full isolation, and for those exposed to infected cases, a similar modified quarantine. The impact of this analysis on disease spread, balanced with availability of resources, would be of interest as an alternative approach. This definitely appears to be more feasible than locking down a country of the size and complexity of India.

Some of the other assumptions that appear rather simplistic, illustratively, are the personnel requirement for management of ventilators. As the situation is evolving, it appears that ventilation may not be the panacea for all ills and that it may not require only intensivists to handle severe cases. Assuming only 5000 professionals are available to handle ventilatory requirements may be fallacious, when these machines have programmed algorithms and other available doctors can easily be trained for a crisis response.

We are highlighting that epidemiological modelling is just one of the tools that are available now with widespread access to computing software; however, they cannot be the keystone for decision makers in the realm of public health, who must rely on pragmatic interpretation of the surfeit of inputs available to them. The mantra for the modern times is recommended to be: ‘Model with abandon, but interpret with sentient discrimination’.

Disclosure of competing interest

The authors have none to declare.