Mark Twain talked about “lies, damned lies, and statistics.” Lying with statistics isn’t always intentional; sometimes a writer doesn’t understand their meaning or applicability.
Recently I saw a news article on the COVID-19 infection rate in northern New Hampshire. It reported a higher positivity rate than in previous weeks. That doesn’t tell us anything about whether infections are increasing or decreasing, though. If the positivity rate goes up, it could mean one of several things:
- More people are infected.
- More of the people who are infected are getting tested.
- Fewer people who aren’t infected are getting tested.
These are all real possibilities, and more than one of them could be at work. With no change in the infection rate, people’s behavior can drive the positivity rate up or down. If people get smarter about whether they need a test, that will drive the rate up even without any increase in infections. The number of infections reported relative to the population is a more useful statistic.
The positivity rate is a useful number for judging and adjusting testing strategies. A high rate suggests more testing needs to be done. A low one suggests resources are being wasted, and that the tests should be either better targeted or reduced.
Curiously, the WHO has used the positivity rate as a criterion for recommending re-opening. I don’t know why they took this approach, but I have a guess. States that push everyone to get tested will have a lower positivity rate than ones where people get tested only when they’re seriously ill. This has little to do with how much of the population is infected. The WHO may have been trying to promote widespread testing, knowing that states that implemented it would look better regardless of per-population infection rates. The thrust of the article is to promote more testing, which is consistent with that aim. If so, it’s not an honest approach. It could even be a dangerous one.
Likewise, the fatality rate is important but can be misused. It’s the ratio of people who die of the disease to people who are tested positive. It can go up without an increase in deaths if more people are tested before they die. It could also go up because the virus is getting deadlier, or because a greater proportion of those getting infected is at high risk of death. Testing more people who are infected with mild symptoms or none will drive the fatality rate down without saving any lives. The more important number is how many people relative to the population are dying.
When someone hands you a statistic, think about whether it’s the best one for the purpose and whether biases might play a role in using it.
Update, Nov. 4: I just came across another misuse of COVID statistics. A Boston.com article says: “There were 4,732,126 people vaccinated [in Massachusetts] as of Oct. 30, meaning 1.1% have reported a breakthrough case of COVID-19. The rate has been steadily increasing — it was 0.23% on Aug. 7.” Since that rate is based on the cumulative total of breakthrough cases, it’s virtually impossible for it to do anything but increase steadily. But news sites always need ways to scare us.