We dug into the COVID-19 data in Florida. Here’s how to see what’s happening in your state.

In late March, with the number of confirmed coronavirus cases in Florida climbing, my editor and I sat down with the data and tried to make sense of it.

Our goal was to get a clear picture of the epidemic in Florida — and determine where it might be headed. It wasn’t obvious. We knew the overall number of confirmed cases in Florida lagged far behind states like New York and New Jersey. But we also knew it was growing at an alarmingly fast rate.

We decided to plot the number of tests, confirmed cases and deaths for Florida. Then, we looked at the same curves for New York, the epicenter of the outbreak in the United States. We provided plenty of caveats. More on that below.

Our analysis found that the number of cases was doubling every three days and that Florida was on track to have tens of thousands of cases in the coming weeks. We also found that in some ways, the curve in Florida resembled the curve in New York. Since then, we’ve continued tracking the exponential growth and paying close attention to the relationships among testing, confirmed cases and deaths across the state.

An expert who specializes in disease modeling, broke it down in plain English. “We do understand the math and the models well enough to say with great confidence that Florida is going to have a huge public health crisis,” said Thomas Hladish, a University of Florida research scientist who specializes in disease modeling.  “And we are just at the beginning of it right now.”

We could have drawn from an existing model like the tool developed by the Institute for Health Metrics and Evaluation at the University of Washington. But were less interested in looking at projections and more interested in evaluating the curve as it actually existed in Florida.

Here’s how to you can do an analysis like this in your state:

1. Choose your data source. We pulled data from two places: the Florida Department of Health’s coronavirus tracker and the COVID Tracking Project. The numbers aren’t identical, partly because they are updated at different times of the day. We decided to use the COVID Tracking Project data so we could look at multiple states. You might also consider using the county-level data published by The New York Times or the international data published by Johns Hopkins.

2. Think about what the volume of testing means for your data. The coronavirus data available today reflects the number of tests that have been performed. Remember, we’re not looking at how many people have contracted the virus, but rather how many have tested positive for COVID-19.  So far, the number of confirmed novel coronavirus cases has grown in step with the number of tests. As you dig into the numbers, keep in mind the United States was slow to start testing and has lagged behind other countries in tests per capita. Also remember that some states are testing much more than others, making state-to-state comparisons tricky. For example, it is hard to tell if New York has more confirmed coronavirus cases than other populous states because more New Yorkers are infected or because many more New Yorkers have been tested (and thus counted). Both of those things could also be true.

3. Consider using a logarithmic scale to show how quickly exponential curves are changing. A logarithmic scale is a statistical technique that emphasizes the rate of change. It’s easy to use. Rather than having a linear scale on the y-axis, use a scale that increases by a constant factor (for example: 1, 10, 100, 1000, etc.). Some readers told us they found the approach confusing or misleading. But experts told us it was a smart way to examine and visualize the rate of change. In other words, it makes sense for assessing if your curve is flattening or becoming steeper.

4. Provide the context. All of the experts we interviewed emphasized one important point: The data today reflects where the epidemic was three weeks ago. It usually takes a new infection between two and three weeks to show up in the data. That’s because symptoms don’t start right away and people tend to wait another few days to get tested. Adding to the lag, some labs are taking more than a week to return results. This delayed timeline also means the benefits of interventions like restaurant closures and stay-at-home orders won’t appear in the data right away, either.

5. Show your work to experts. We showed the work to three experts prior to publication: the University of Florida research scientist, a former director of epidemiology and disease control for the Miami-Dade County Health Department and a professor of public health sciences at the University of Miami. We wanted to make sure our methodology was sound and get input on what conclusions we could draw from the data. Epidemiologists and health statisticians are clearly busy these days. But I’ve found that many are willing to make time for reporters.