Around the world, countries are beginning to ease the lockdown process. Economies are opening and students are returning to schools and colleges. With global cases varying vastly, scientists and health experts are reluctant to predict the course of the pandemic. Cases in the US are on the rise again, while it is declining in other countries. Within the country, it is high in one state and declining in another.

Are things going to get better or much worse before this is all over? Everyone is looking answer to this question. It seems that health experts and scientists are also not sure where this pandemic is headed. Let’s talk about the factors that are making it hard for scientists to predict where the pandemic is headed next.

Why are health experts unable to predict the future?

Forecasting a pandemic was never an easy task and it is more difficult with coronavirus. Experts and agencies have failed over and over regarding covid predictions. Given the bad track record of pandemic forecasting, it is shocking how the prediction is playing a vital role in the decision-making process by governments around the world.

Britain's chief medical officer had predicted nearly 3000-65000 deaths during the swine flu outbreak in 2009. However, it killed only 457 people. History is filled with such examples. There are several factors that make it difficult to predict the future of coronavirus.

Some of them are poor data input, wrong modeling assumptions, lack of incorporation of epidemiological features, poor past evidence on effects of available interventions, lack of transparency, errors, the study of only one or a few dimensions of the problem at hand, lack of expertise in crucial disciplines, and selective reporting.

Poor data input on the pandemic that goes into theory-based forecasting

Early crucial data input such as fatality rate, infection rate, etc. was changed intentionally or unintentionally. A lot of countries did not have enough manpower, infrastructure, and funds to track the numbers precisely. Numbers may vary unintentionally because of delays in reporting, errors in data, etc. Authorities should invest more in the collection and management of clean data.

Wrong assumptions in the modeling

Many models assume uniformity, i.e. all people having equal chances of mixing and infecting each other. This is an unpractical assumption and, in reality, tremendous heterogeneity of exposures and mixing is likely to be the norm. Scientists should adopt models that make realistic assumptions and change when the evidence says otherwise.

The high sensitivity of estimates

Some models use exponentiated variables. Even a slight error in the variable can result in a major deviation from reality. There is no way to overcome this problem other than acknowledging that estimate can be much larger than it seems.

Lack of transparency

The methods of many models used by policymakers were not disclosed; most models were never formally peer-reviewed, and the vast majority have not appeared in the peer-reviewed report even many months after they shaped major policy actions.

Selective reporting

Cases have been found where governments are misreporting the report to prevent panic or to control the situation going against their favor. This is very hard to counter.

Looking at fewer aspects of the problem

Almost all the models that played role in the decision-making were based on only a few factors like the number of deaths, bed shortage. Models used for decision-making need to take into record the impact on various corners for example other aspects of health care, other diseases, dimensions of the economy, etc.

It's still a very acute and difficult phase of the pandemic. We have to focus on how we get through the next 6-12 months which might be the most difficult and then talk about the longer-term plan on whether it is going to be the elimination or control.

It also depends a lot on the evolution of the virus, the ability of vaccines to keep up with the variants, and the duration of protective immunity of vaccines. We now know that natural and vaccine immunity lasts for a maximum of 8 months. But at this point, it is very hard to predict the course of the pandemic.

What about Herd immunity?

Many previous predictions have failed. But we know that there will be an end to the acute phase of the coronavirus pandemic. When we would have vaccinated at least 30% of the world population we shall see a significant reduction in the death due to coronavirus.

We still don’t know exactly what will be the optimum percentage of vaccination for herd immunity. Experts are hoping it to be around 60% to 70% of the total population. We hopefully should see that virus becomes just like another viral infection like influenza which comes seasonally.

Nevertheless, Covid pandemic prediction is unlikely to be abandoned. We understand that all issues can not be fixed but there are some that can be fixed. Precise modeling of predictive distributions rather than centering to point assessments, considering multiple dimensions of impact, and continuously reappraising models based on their validated performance may help.

So what can be done?

To summarize, here are some preventive measures that we can adopt:

• Invest more in gathering, cleaning, and curating real, unbiased data, and not just theoretical considerations
• Model the entire predictive distribution, with a particular focus on accurately quantifying uncertainty
• Continuously monitor the performance of any model against real practical data and discard models which give the result that deviates from reality.
• Combine the best epidemiological estimates on age structure and comorbidities in the modeling
• Avoid unrealistic assumptions about the benefits of interventions; do not hide model breakdown behind implausible intervention effects
• Improve transparency about the methods
• Share code and data
• Use up-to-date and tools and processes that decrease the potential for error through auditing loops in the software and code
• Encourage interdisciplinarity and ensure that the modelers’ teams are diversified and solidly grounded in terms of subject matter expertise
• Maintain an open-minded approach and accept that most forecasting is exploratory, subjective, and non-pre-registered research
• Beware of inevitable selective reporting bias

To sum up

Despite these obvious failures, epidemic forecasting still continues to thrive, perhaps because wrong predictions are not usually met with serious consequences. In fact, these false predictions may have even been useful.

A wrong, doomsday prediction may spread fears in the people to enhance their personal hygiene. The problem occurs when public leaders take these false/wrong predictions too seriously and take decisions based on them. With COVID-19, wrong predictions can affect the lives of billions of people in terms of the economy and health.