What is the COVID Curve Tracker?

The COVID Curve Tracker project is led by Professor Tom Pike of Imperial College London in collaboration with Dr. Vikas Saini of the Lown Institute. The COVID Curve Tracker examines the daily percentage change in the death rate for selected countries and creates a unique curve for each country to predict peak and total deaths. 

The work is based on an observation Professor Pike made (during his own coronavirus-related isolation) while exploring public COVID-19 data: that the growth in rate of deaths from confirmed COVID-19 cases in Italy seemed to be the same pattern seen in China. Subsequent exploration of other countries’ death rates indicated that this may be a more generalized phenomenon and could serve to provide useful information about the pandemic. 

More recently, the analysis has been adjusted to allow each country to follow its own curve based on reported death rates. However, many of the country’s curves are similar to each other, indicating that there may be a common path of the epidemic across regions.

How do you create the projections for peak and total deaths?

We create a curve for each country or state curve on their percentage change in death rates. This curve allows us to project peak and total death rates if the numbers follow the curve. We look carefully at how well the curve fits the data. The better the fit, the more confident we can be that the numbers will stay close to our curve. We quantify this as “confidence limits” that we also show in our plots. The tighter the confidence limits, the closer we expect the numbers to follow the curve. We show three different confidence limits as shaded areas about the curve, with the broadest ones giving a 95% chance that the numbers will be closer to the curve. Developing reliable confidence limits has been a large part of this analysis.

Why are peak deaths projected for my country less than the number of deaths today?

All the analysis is based on a trailing weekly average, so the date of the peak is the date of the last day of the week with the most deaths. The data available today is giving us the average from seven days ago to the yesterday, inclusive. Because we are averaging, the highest weekly total, divided by the days on the week, will usually be less than the peak number of daily deaths.

Why isn’t my country/state on the tracker yet?

We are continuing to add countries and states each week. To avoid giving projections based on too little data, we are adding new countries and states once they have reported more than 500 deaths in total. Even with higher death counts, problems with the data can prevent us finding a reliable curve fit. We don’t plot for those countries and states.

What information has the Curve Tracker uncovered so far?

When we track the percentage change in death rates, we see a similar pattern occur in multiple countries. For most countries, we find that once the rate of change in deaths falls below 20%, their death rate curves take on a similar, predictable shape. As the pandemic has progressed, we can identify that some countries are falling faster down the rate curve, while others seeing their rates fall more slowly. This gives us information on how effective social distancing might be. We believe that tracking these curves will help monitor the course of this epidemic and allow some benchmarks for policy decisions.

What makes the COVID Curve Tracker different from other models? 

Most other reports that have emerged in the past few weeks are based on the creation of complex models that take into account multiple variables like infection rate, fatality rate, hospital capacity, etc., using the best available estimates of these variables to explore the outcomes. Our models are different because they are derived directly from the data of reported deaths in order to predict future deaths. We are also finding that we have narrower confidence limits – our curves fit the data better than existing, more complex, models.