## Reading coronavirus: pandemic ABCs

Author: Alex Weinreb
05.04.2020

We are being deluged by daily updates from the Ministry of Health, news media, data aggregation services, all showing us a range of coronavirus statistics. Here we provide a brief guide—and some warnings—to help readers interpret these numbers

**There is inconsistency across sites, so pay close attention to categories.**The number that we hear most frequently from the Ministry of Health in Israel, and therefore in the press, is the “total number of confirmed cases.” This is simply a running total of all confirmed cases. It includes “active cases”—that is, people who are known to be infected—*and*people who have recovered or died. If we want to understand trends in infection, this “total number of confirmed cases” is not always the best category because it can. Using it to understand coronavirus trends is like using a running total of spending to evaluate trends in household consumption. The best these types of measures can do is flatten out. For some uses, it may be more advisable to use the number of “active cases,” which is the total confirmed cases minus the number of recoveries and deaths.*never come down***The “total number of confirmed cases” is basically a measure of clinically confirmed infection.**As such, it is a simple function of: (a) how many people have been tested; and (b) testing protocols, that is, the rules about who gets tested. This is particularly problematic for international comparisons since infection rates will likely be lower in countries with lower rates of testing, or that allow people to buy a test (this is the case in parts of the US, and means that testing rates will likely be higher for the wealthy than for the poor). But this is also a problem when trying to understand trends within a country. In Israel, for example, testing increased from 500-700 per day in the second week of March, to more than 6,500 per day in the final days of the month. In this case, it is important to look at the percentage of tests that are positive, the number that are repeat tests (e.g., to check if a person is still infected), while also being aware of changes in testing protocol.

- If we want to know whether we’re slowing the spread of coronavirus (i.e., have “flattened the curve”), we need to
**focus on the rate of change in the number infected**, not the absolute number of new infections. For example, the 712 new infections on April 1 represented a 13.9% increase. The 726 new infections 6 days earlier (March 26) represented a 31.6% increase. So even if a headline tells us “record number of infections,” look at the rate of increase.

- The flipside of the last point: if we want to know what the impact of infection is going to be on the medical system (or on any other sector in society), looking at the rate of infection is not sufficient. We also have to
**look at the absolute number.**Flattening the rate of infection earlier in the epidemic—e.g., at a level where we get 700 new cases per day—places much less stress on the system than if we flatten it at 1400 new cases per day.

**We can use ratios between numbers to generate useful indicators of the epidemic’s trajectory.**[1] One example is the number who’ve recovered relative to the number who’ve died. This is a simple measure of how well a system is able to channel known infections at some prior time toward recovery as opposed to toward death. Another useful ratio is the number of people on respirators at a given point in time relative to the number of confirmed cases 10 days earlier. Again, this tells us something about changes in the proportion of cases that appear to end in the most serious clinical condition. Note that these ratios are not frequently presented in media. You’ll have to calculate them yourselves (or rely on us—see our coronavirus blog).———————————————————————————————————————————

[1] Underlying these ratios is a simple mathematical identity. The number of people *S* infected with coronavirus on any given day *t* is equal to the number of people who were infected at some point in the past, *I*, minus those who have recovered in the interim, *R*, and those who have died, *D*, as in: *S _{t}* =

*I*(

_{t-n, t}–*R*)

_{t-n, t}+ D_{t-n, t}By moving various components in this identity between the left- and right-side, especially if we pay careful attention to time (*t*), we can identify turns in the trajectory.