Understanding the Covid-19 forecasts - Daily Nation

As the Covid-19 pandemic unfolds, epidemiologists and data scientists have produced and published grim estimates of how many people will be infected with the coronavirus or die of it.

They present the data in graphs showing “best-” and “worst-case” scenarios.

They have popularised a major talking point about “flattening the curve” — which means keeping the Covid-19 cases below the level that the health system can handle.

Estimates change as time passes and as more data and other insights become available.

The scale and scope of those numbers and the frequency at which they change make some people distrust them or wish them away because they conjure thoughts of disease and deaths in big numbers.

In crunching the numbers to produce projections, the estimators make assumptions that certain conditions will hold. These conditions are assigned numerical values.

They are fed into a computer software, which runs mathematical formulas and spits out results, which are used to see what the future may look like.

The process of creating different results by playing with different scenarios is what is generally referred to as modelling. It’s merely a way to approximate reality.

It’s a way of producing an early warning system, without which we would be flying blind.

In situations where there is no data or not enough of it, the projections of illnesses, hospitalisations and deaths are wild guesses. It, therefore, follows that the more data available — and especially if it is of high quality — the better the forecasts.

The projections change as mitigation measures shift. If Covid-19 prevention methods are relaxed too early, the estimated number of sick people shoots up, and so are the deaths.

If suspected people are quickly identified, isolated, tested and treated, the estimated number of cases declines.

Similarly, if the estimates are generated assuming a 20-day lockdown, but the lockdown is only implemented for half the time, then the forecast would need to be revised accordingly.

Testing improves the accuracy of viral disease forecasting. A higher number of people tested means better estimates that scientists can generate.

From the testing data, they can infer about the different characteristics of sick people — their location, age, gender, travel patterns, the proportion of health workers among them, and a whole lot more.

Not everyone agrees with models, predictions and estimates. Some term them distasteful death wishes.

In a way, they are right. Predictions are often fraught with mistakes. Models get muddled; they are as good as the data and assumptions they are based on.

Many pundits misjudge the future, but what other tools do they have for forecasting?

Here is my point: we should not fight the coronavirus with one hand tied to our back nor fly blind into the future. We can build scenarios premised on locally generated data.

These scenarios equip decision-makers with much-needed insights for tackling the tidal wave of the coronavirus.