This article has been updated.

When it comes to the coronavirus pandemic, President Trump takes wins where he can get them. So, for much of July, he and his administration touted a metric that, they suggested, demonstrated how effective the government’s response had been.

“We’ve done 50 million tests,” Trump said at a briefing July 23. “There’s nobody even close in the whole world. You look at our mortality rate. You look at our death rate. You look at different statistics. We’re doing very well.”

The part about testing isn’t new. It’s central to Trump’s rhetoric, in fact, an effort to shift concern about surging cases into a positive focused on the expansion of testing in the United States. It’s the mortality rate that became a focal point over the past few weeks — similarly because it turned the rising number of new cases into an asset.

In the United States, the pandemic surged twice. The first surge came in the spring, as the novel coronavirus arrived. The second came in June and July, after Trump publicly called for states to scale back efforts to contain the virus’s spread. The result was a period of a few weeks when new cases were climbing rapidly but the death toll was still falling because it takes a while for an infection to lead to a death.

From June 9 to July 6, the seven-day average of new cases in the United States jumped from 20,936 to 50,193. The seven-day average of reported deaths, though, fell from 794 to 463. The rough mortality rate — the number of cases that result in deaths — shifted from 3.8 percent (794 divided by 20,936) to 0.9 percent (463 divided by 50,193). So the White House kept touting the figure.

“When you compare us to other countries on case fatality rate, other industrialized nations, we’re very low and beating most countries, if not all, in Europe,” White House press secretary Kayleigh McEnany insisted on July 13.

But even then, it was obvious that the metric wouldn’t hold. From July 6 to Saturday, the daily average number of cases increased from 50,193 to 62,058, but the average number of new deaths jumped from 463 to 1,137. The mortality rate, then, jumped from 0.9 to 1.8 percent.

July 6 was actually the low point in this metric because it was also the low point in the daily average of new deaths. Since then, the number has begun to climb again, after having fallen fairly steadily since May.

But this metric is also odd. It compares things that aren’t really related at all: the number of deaths reported on a day with the number of new cases tallied, though almost none of those new cases are also new deaths. (The theoretical exception, probably more common a few months ago, is when someone tests positive for the virus shortly before death.) If we compare the number of deaths reported with the number of cases four weeks prior, making an accommodation for the lag between the detection of an infection and the progression of covid-19, the disease the virus causes, the mortality rate has been more stable.

Comparing same-day averages of deaths with cases since June 1 gives a “mortality rate” range from 0.9 percent to 4.6 percent. Comparing deaths with the number of new cases 28 days earlier results in a narrower range, from 2.2 to 4 percent. On average, the lag-adjusted mortality rate since June 1 has been 2.9 percent.

But this, too, isn’t quite right as a metric. What we are curious about is not when deaths were reported but when they occurred. Most states report the number of new deaths as they are discovered, meaning that days- or weeks-old deaths might be added to tallies on any given day. The Centers for Disease Control and Prevention counts weekly deaths based on when the patient actually died, offering a better look at the issue.

Using CDC data on new cases and new deaths, the shift in the mortality rate looks like this.

The dotted line at the end is a function of the fact that the CDC data trickles in more slowly. Dependent on death certificates, more recent weeks show fewer deaths not because there were fewer deaths but because they haven’t necessarily been recorded yet. The last three weeks of data, for example, shows a decline in new deaths, almost certainly a function of hundreds of deaths not being reported to the CDC yet.

What’s more, the CDC graph above doesn’t account for the lag. If we do, the shift in recent weeks has been more modest. On average, the lag-adjusted mortality rate seen in the CDC data from the week ending June 6 to the week ending July 11 was 2.6 percent, a bit under the rate seen when using the daily data.

These are not exceptional numbers. Since the pandemic emerged, the mortality rate in the United States has been about 3.3 percent, comparing all known deaths to all known infections. There are 57 countries with higher overall mortality rates — and 155 countries with lower ones.

What McEnany was talking about, though, was the recent mortality rate. On that metric (deaths vs. cases over the preceding seven days), the United States is currently in the middle of the pack. Fifty-two countries had higher mortality rates over the past week; 78 had lower rates. (An additional 65 had no deaths, and 19 had no cases.)

Comparing deaths to reported cases 28 days prior, the same pattern emerges. Sixty-one countries had higher rates, and 63 had lower ones.

The last time the White House spoke specifically about the overall mortality rate was July 27, in the context of how drug treatments have improved the likelihood of recovery. The last time the White House hyped the mortality rate as a measure of its own success was in that July 23 briefing.

With new cases falling nationally and deaths rising, it seems unlikely that it will be a go-to metric again at any point over the short term.

Update: At his briefing on Monday afternoon, Trump again raised the mortality rate in a specific context.

“As a result of improvements we’ve made in treatment for Americans over the age of 18,” he said, “the mortality rate has declined 85 percent from its April peak. Eighty-five percent decline.”

It’s not clear where this number comes from, though it may be a comparison of the April 20 mortality rate generated by comparing daily deaths to daily cases with the rate in early July. The rate on July 9, for example, was about 86 percent lower than the rate on April 20.

But, again, that’s no longer the case. The drop is now 76 percent.