No announcement yet.

What does flattening the COVID-19 curve mean?

  • Filter
  • Time
  • Show
Clear All
new posts

  • What does flattening the COVID-19 curve mean?

    There has been a lot of discussion by government representatives, media, and public health experts about flattening the curve to limit the strain on health care services during the COVID-19 pandemic. The discussion surrounding flattening the curve dates back to about 2007 (link) and the original CDC chart is reproduced below. Additional history about the evolution of this chart to a recent animated gif by can be found at this link.

    CDC curve 2007.png

    While the concept of flattening the curve is simple, the general public is being bombarded with graphical representation of the ever-increasing cumulative number of COVID-19 cases around the world and in their own countries. Here are several immediate examples of tracking sites on the internet with graphs plotting the total cumulative cases from around the world.

    Usually, the main graphic image is a line graph or histogram showing the changing total number of COVID-19 cases by day. Technically, this type of graph is referred to as an ogive, a graph that displays cumulative data usually over a period of time.

    Cumulative Frequency Graph (Ogive)

    An ogive is a graphic presentation of incrementally increasing cumulative data usually over periods of time. This kind of graph is mostly used to depict progress to specific goal or end point. For example, if a company sets a target goal of selling 10,000 units starting in January how low long will it take for sales to reach the 10,000-unit threshold. The graph below shows that monthly sales surpassed the goal of 10,000 units in August.


    Another example might be a charity trying to raise $8000 for new playground equipment for a youth center, as depicted in the graph below. In this graph the x-axis is not time, but the remaining distance to achieve the goal of $8000.

    Charity goal.png

    Now consider an ogive, a cumulative frequency graph, of COVID-19 cases below which is from the JHUCSSSE team, a website that is widely cited for its statistics. It shows the cumulative growth of COVID-19 cases through March 24, 2020. But is such an ogive curve appropriate for a disease outbreak? By using an ogive, do we have some set goal of the number of COVID-19 cases we are purposefully trying to reach? A million cases? 10 million cases? No, of course not.

    Johns Hopkins Cumulative 3 24.png

    At the beginning of a disease outbreak, a cumulative frequency graph can reinforce the public message that the outbreak could get worse. The COVID-19 outbreak did get worse and it evolved into a pandemic. At this stage of the pandemic the cumulative frequency graph of COVID-19 cases has lost it informative powers. Everyone is already aware that cases are continuing to grow. There is no other message that it conveys.

    A distinctive feature of an ogive is that it can never go down. From a pandemic perspective, even after people stop being infected, the graph will never go down nor will it go back to zero. Thus, when public health officials talk about flattening the curve, they are not talking about an ogive of COVID-19 cases that is prominently displayed across the internet. These officials are talking about flattening the epidemic curve.

    Epidemiological Curve

    An epidemiological curve, often referred to as an epidemic curve, is a graphical representation of the frequency of new cases over time based on the date of onset of disease, usually depicted as a histogram. Epidemic curve are always graphed against time. A true epidemic curve is a plot of the number of new daily case based on date of symptom onset. The significant difference between an ogive of COVID-19 cases and an epidemic curve is that only new cases are plotted on a daily basis, not the new cases plus the total of all the cases that occurred before that date.

    Below is an epidemic curve for the United State Center for Disease Control and Prevention (CDC). From the low number of plotted cases in the graph, it should be obvious that the CDC does not have onset dates for the vast majority of US COVID-19 case. The graph notes that the number of daily new cases from the past few days will change because onset of symptoms can occur from a few hours to several days before the case is confirmed and added to the tabulation. In terms creating the graph, onsets dates almost always happen in the past.

    CDC Epidemic curve.png

    The most significant information component of an epidemic graph is that an inflection point in the curve is almost immediately identifiable. The inflection point is the point in time when the number of cases stops increasing and begins to decreases. When public health officials talk about flattening the curve, they are referring to implementing interventions that will reduce the overall shape of the epidemic curve, particularly at the inflection point, by reducing the maximum number of daily cases based on onset of symptoms.

    Comparison of an Ogive with an Epidemic Graph

    The graph below plots both a cumulative frequency of cases against a number new daily cases. Based on the discussions above, it should be clear as noted on the chart that the line graph of the cumulative frequency graph will never go down.

    comparison labelled.png
    Even though the general public does not have access to the symptom onset dates of confirmed cases, a psuedo-epidemic graph can be constructed from available data. Rather than plotting cases by onset dates, a pseudo-epidemic curve can be graphed based on the number of newly reported cases on a given day. Such a graph is a reasonable approximation of the underlying epidemic curve.

    The comparison of the ogive and the epidemic curve presented above is not hypothetical. It is the cumulative daily total of COVID-19 cases plotted against the pseudo-epidemic graph of the daily new COVID-19 cases for the Republic of Korea (South Korea) as shown in the full labelled graph below. What the epidemic graph or curve clearly shows is that about March 2, 2020, South Korea passed an inflection point and the number of COVID-19 cases starts to decline. Through aggressive intervention South Korea was able to flatten the curve during the first few days of March when the number of new cases started to decline. South Korea is currently on its way to economic and social recovery.

    South Korea.png
    While ogives, cumulative frequency graphs, of the COVID-19 pandemic serve to present dazzling graphics - some static and some animated - it is epidemic curves that will shows the true progress of the disease in a particular country or the world as a whole. That is the curve that needs to be flattened.

    Note: The South Korea data in the epidemic graph above was obtained from DS4C: Data Science for COVID-19 in South Korea.

  • #2
    Perhaps this Politico article makes the point better than I did. Two better ways to chart the spread of coronavirus

    Many coronavirus trackers focus on the total number of cases. But viruses spread exponentially, which can quickly balloon the number of infections from a few dozen to hundreds or thousands within days, making it harder to compare countries in different stages of outbreak. Is there a better way? . . .

    The history books will focus on the total number of people infected with COVID-19. But in the midst of the pandemic, we’re often more interested in the number ofnewly infected people — a sign indicating where the crisis is most severe. . .
    Immediate evidence that a county is flattening its epidemic curve is charting the growth, the peak, and the decline in daily new cases, not the cumulative total number of cases by date.