Note: Following the publication of this article, the PDF of the paper was removed by the journal without notice from the URL at which it had been originally available. There is now only a 404 notice at the link, indicating the paper can no longer be found there. The paper is available at a different URL now, here.
In the last two weeks, two studies pertaining to COVID-19 – one about the use of hydroxychloroquine and the other about the effects of blood pressure drugs on the disease – have been retracted by the prominent medical journals that published them. Both studies were marred by deep ethical issues inherent in the data the authors used to arrive at their respective conclusions, and the incidents reminded the world that even as scientists and officials negotiate speed versus accuracy, ethical considerations are not up for debate.
But in India, the absence of ethics in the government’s COVID-19 decision-making remains a fundamental problem. As the latest symptom, two researchers from the Directorate General of Health Services (DGHS), an office within the health ministry that “renders technical advice on all medical and public health matters and in the implementation of various health schemes”, have published a paper with dubious claims that belie the importance of their stations.
The authors of the paper, Anil Kumar and Rupali Roy, are both editors of the journal in which the paper was published, Epidemiology International. This is a conflict of interest. But at the end of the text, next to the words ‘conflicts of interest’, the authors have written ‘none’. This also raises questions about the ‘separation of powers’ between the journal’s reviewers and editors. If such separation doesn’t exist, the study’s conclusions could be compromised as well.
Within the DGHS, Kumar is deputy director-general of public health, and Roy, the deputy assistant director-general of leprosy. According to an order dated April 7, Roy has been deputed to “look after work related to the coronavirus”.
Second, according to the document itself, it was submitted to the journal for consideration on May 22, 2020, but accepted for publication a year earlier, on June 2, 2019. This may well have been a typo. But such errors are quite hard to find in legitimate journals, suggesting a lack of attention to detail.
Fourth, the migrant labourers’ crisis, born after the Centre failed to plan for them after the lockdown first began on March 24, has progressed by now to the reverse migration ‘stage’. Many North Indian states are now contending with the consequences of lakhs of workers returning home from cities. However, the paper’s modelling exercise hasn’t accounted for this important social phenomenon. The authors have simply included a caveat in the concluding portion: that “natural calamity, unpredictable population movement and important national or international events may have significant influence” on the model’s predictions.
Fifth, despite the DGHS’s position within the health ministry, the paper’s authors have used “secondary data” from the website Worldometers.info. Why not use the health ministry’s own, more detailed data, not to speak of official data from elsewhere – all readily accessible – given major discrepancies in the Worldometer case count for some countries (e.g. Spain)?
Sixth, the authors advance a simple model to be “used at the national, state and district levels” without explaining how the model can be both specific and wide-ranging at once. In addition, Gautam Menon, a professor of physics and biology at Ashoka University, wrote in The Wire Science on April 4, “Local models are more useful than global models. A model that shows state-wise behaviour is better than a model that purports to be India-wide. It is indeed more rational to think of different policies at the level of individual states or districts than to demand a single nationwide policy be applied uniformly across India, mindless of local circumstances.”
Low quality of research
By 10 pm on June 7, at least a dozen news websites had published articles syndicated from news agencies. Each article focused on the model’s prediction: that India’s COVID-19 epidemic would end around mid-September. However, the paper’s own wording is neither consistent nor directly supports this conclusion.
First, the authors write that the paper’s result are indicative, not prescriptive: “Application of mathematical modelling and statistical analyses about coronavirus is to indicate what will be the situation if [there are] different circumstances. These models are not envisioned to predict fixed outcomes.” Then, in another part, they predict that in mid-September, the per-day “number of infected will be equal to the number of removed patients”. This implies a flattening of the cumulative case-load curve, not the end of the epidemic, although it is also a fixed outcome.
But then they caution that this prediction should be taken with a pinch of salt because the model they have used is stochastic. In stochastic models, the values of one or more of the input parameters change randomly. However, their analysis itself is not stochastic, predicting fixed outcomes that cannot vary randomly. Finally, the authors write: “This a very good model to support analysis and interpretation of state/district data (whenever numbers of cases are high) and it will also help in relevant decision-making in control activities of COVID 19 pandemic.”
In effect, the authors claim their model is stochastic but make fixed predictions, and use vague language but say their work is “very good”.
Irrespective of the paper’s conclusions, these flaws only highlight the low bar of research and evidence government officials have accepted in the last few months to draft policies that affect millions, including healthcare workers. The most prominent offender has been the leadership of the Indian Council of Medical Research. It has been unusually fond of hydroxychloroquine as a pre-exposure prophylactic against COVID-19. To date, there exists no evidence in support of this mode’s efficacy and safety – and ample evidence that suggests the drug’s use won’t help.
This said, the health ministry hasn’t been exempt. It permitted the use of hydroxychloroquine without requisite clinical trials as required by law. The NITI Aayog also joined in, announcing at one point that India’s COVID-19 epidemic would end in mid-May. Together with numerous other issues with the government’s response, there is now significant uncertainty over the scale and effects of the coronavirus’s spread in India.
As it happens, the conclusions of the new study are suspect as well. According to the authors Kumar and Roy, more people will recover from COVID-19 in India per day than will be infected sometime around mid-September. So that’s when, they say, India’s coronavirus epidemic will ‘end’. This focus of the paper mirrors a claim that NITI Aayog CEO Amitabh Kant has oft-repeated in the last few weeks as well: that India’s supposedly superior response to the coronavirus outbreak is borne out by its increasing recovery rate. (Factcheck: it’s not.)
Beyond the confused wording itself, Giridara Gopal Parameswaran, an epidemiologist working in India, questioned the paper’s methods in a brace of tweets. His argument was that the model Kumar and Roy had used to arrive at their results was designed to study malaria. However, malaria’s characteristics as a disease (vectors, transmission, infection rates, etc.) are quite different from those of COVID-19. Menon, who has been leading a large-scale modelling effort, also tweeted, “The methods are unclear and the prediction involves a questionable extrapolation to four months on.”
As such, if the health ministry formulates any policies based on the new paper, it is likely to join the growing pile of claims, notices, statements and other papers that have been the foundation for the Indian government’s more questionable choices. Ultimately, the paper undermines the usefulness of epidemiological modelling and technical expertise itself.
An email to Rupali Roy, listed as the paper’s corresponding author, requesting her comments on these issues went unanswered. This article will be updated as and when she replies.
Note: This article was edited at 9:46 am on June 8, 2020, to clarify the conflict of interest.