COVID-19 Third Wave: Imperfect, Inadequate Data May Mislead Predictions
Availability of data in India is a serious issue. There was huge under-reporting of COVID-19 infections and deaths.
Mathematical models have become very popular for predicting outbreaks during the COVID pandemic.
Many researchers including epidemiologists and public health policymakers adopted this approach, and a number of models were developed, forecasting the trajectory of COVID-19 during the first and second wave. Some were useful, many went wrong.
While a model can do wonders by predicting the future of a disease outbreak; it can also be misleading, if supporting empirical data is scarce and imperfect.
Mathematical or mechanistic models are useful tools to establish certain qualitative aspects of natural phenomenon such as disease outbreak, but they need enough sophisticated data to make detailed predictions about the pandemic.
Every model is built on certain assumptions, and thus includes parameters that may not be observable, or directly measurable. One needs data to estimate (eg. back-calculate) these parameters such as R0 and further predict the trajectory of the pandemic.
Why Predictions From Recent Models Went Wrong?
Predictive models have a mechanistic framework, thus requiring a nice balance between base assumptions and complexities of interactions arising in the phenomenon. A complex model requires very high-resolution data for correct prediction, and a model that is too simple and uses limited available data, might forecast inaccurately.
The nature of spreading phenomenon or transmission of SARs–COV 2 has become very complicated since its discovery in December 2019 in China. The main reason behind this is that, this particular variant of the family of coronavirus was not known to us from before; neither the intensity nor the severity of the disease developed upon infection by this pathogen. The genetic structure, nature of mutation was also unknown.
So naturally, there was no preventive measure such as vaccines to control the spread of the virus. In absence of any measure, the only way to stop the spread of infection was social isolation. However, this was not also a complete success, as lockdown and social distancing had a huge impact on the economy of this country, which in turn, affected human livelihoods, perceptions and attitude.
In other words, human behaviour is the key to success of such social distancing measures controlling infection. Most of the predictive COVID model developed did not consider this important factor.
Only a few considered this, and one of them was the one put forth by Ngonghala, Goel, Kutor and Bhattacharyya in the Journal of Theoretical Biology.
Why do Most Models Ignore Human Behaviour as a Parameter?
To some extent, this model used detailed human activity, social isolation, and testing data, during this lockdown period in India, to predict the rise of cases during first wave in 2020.
A reason for avoiding human behaviour as one of the model ingredients is that it increases the complexity of the framework and includes many parameters. Moreover, there is little access to precise information or systemic data on human activity in India during this pandemic. In addition, social organization – the pattern of relationships between and among individuals and social groups within the Indian population is very complex and not much information is available on this either.
A detailed model might require information of this pattern and its evolution which include characteristics such as spatiotemporal cohesion, sex ratio, age structure, leadership, division of labor, communication systems, and so on.
Availability of data in India is a serious issue. The actual number of cases is always under question, as there was huge under-reporting. Many infections are never captured by a test and there were reports of gaps in numbers of COVID deaths, especially in rural India.
Naturally, misinformation and imperfections of such empirical data are propagated through the model's predictions.
How Can A Model Accurately Predict the Third Wave?
Predicting the third wave would require multifold information such as number of daily new cases, daily testing, right information on recovered population, hospitalisations and deaths.
Modelers need the right statistics on vaccination coverage and effectiveness – detailed information on age-specific vaccination uptake data as well as how long immunity lasted post jab. An improved sampling process of vaccinated groups is required to have better observation on this. All these will help us to correctly estimate herd immunity in population against the upcoming wave.
The size and outbreak severity also depends on mixing patterns of the population during this time. As observed in the past, individuals start treating this deadly infection lightly as and when the number of daily cases starts decreasing. So, information on movement and daily commuting patterns between cities is also necessary to develop correct models and to make precise predictions.
It is also essential to gain a deeper understanding of mutant strains and their pathogenicity by performing genomic sequences of more samples. Without this information, modelers are forced to make subjective choices of their assumptions and parameters, which makes predictions subjective.
Results get influenced by social and class biases and thus, should not be treated as objective or neutral. Imperfect and scanty data is a serious issue in modelling and prediction science. There is no straight forward resolution to this.
But sometimes a combination of mathematical and statistical approaches might be useful to some extent. Different statistical inference methods may help to estimate the parameters more correctly. An interdisciplinary framework of a heuristic approach, transfer and deep learning, and use of synthetic data may help to increase the gains in understanding and improve the prediction.
(Dr Samit Bhattacharyya is Associate Professor in the Department of Mathematics at the School of Natural Sciences, Shiv Nadar University in Delhi. This is an opinion piece and the views expressed are the author’s own. The Quint neither endorses nor is responsible for them.)
Subscribe To Our Daily Newsletter And Get News Delivered Straight To Your Inbox.