The coronavirus pandemic in the United States is heading in two different directions at the same time. The number of new cases is spiking, passing a seven-day average of 50,000 daily cases for the first time on Monday. And yet the seven-day average of deaths each day is at its lowest point since the end of March, when there were only 3,700 deaths from the virus in total.

You’ve by now seen graphs like the ones below showing the daily change in new cases and deaths. It makes the point starkly: Cases are rocketing up, while deaths keep sliding lower.

President Trump would like us to focus only on that second graph, repeatedly emphasizing that the number of deaths is declining and repeatedly — and falsely — claiming that the mortality rate from the virus in the United States is the lowest in the world.

But the two are obviously linked. People die of covid-19 after contracting the novel coronavirus that causes it. Sometimes confirmation of infection comes only as the patient is on the brink of death, but usually patients represented in the graph on the right are also represented in the graph on the left. That is, someone is confirmed as having been infected with the virus before they are confirmed as having died of it.

This is the central question of the moment, one explored by The Washington Post’s Paige Winfield Cunningham on Monday: What’s the gap between those two confirmations? How long does it take for someone to go from being confirmed to have contracted the virus to their unfortunately succumbing to it? It’s a question that is important for an obvious human reason, but it’s important for a political one, too. Experts expect the death toll to head back up given the increase in confirmed cases. They just don’t know when or how much.

In an informative thread on Twitter, epidemiologist Ellie Murray noted that our ability to answer those questions is hampered in part by the broad expansion of testing. More testing has allowed doctors to identify infections much earlier in the process — meaning a longer period of time from infection to death. Murray calls this the “lag time bias,” and explains it with a simple diagram.

At the apex of the daily death toll in April, New York was the epicenter of the pandemic. Testing was limited; those confirmed to have contracted the virus were mostly those with severe symptoms who presented themselves at hospitals. In other words, they were patients who were most at risk of death. The gap between confirmation of infection and death was fairly short. Expanded testing means catching people before symptoms are similarly severe (though they may get to that point). The result? A bigger gap between an increase in cases and one in deaths.

One way of looking at how quickly each metric is shifting is to look at the time it takes to hit certain benchmarks. Let’s say, for example, that we want to see how long it took for the United States to hit 200,000 and then 400,000 and then 600,000 cases (and so on) or 10,000, 20,000 and 30,000 deaths.

The result is another way of depicting the graph above. On April 1, we passed 200,000 cases, 77 days after the first known case in the United States. We added another 200,000 cases seven days later. Another 200,000 six days after that.

That was the pattern for a while, until new case totals slowed down in May. But in late June and July, they sped back up. We hit the most recent 200,000-case benchmark only four days after the preceding one. (The gray bar at the end of the chart shows that we’re still between benchmarks.)

The pattern for deaths is different. It took a long time to hit 10,000 deaths, but then we passed similar benchmarks over and over in short order. Then, the pace began to slow. The most recent benchmark was passed more than two weeks after the preceding one.

To compare this pattern to data at the state level, we can adjust for population. If we look at how quickly the United States added 100 cases for every 100,000 residents or 10 deaths, the pattern looks like this.

Fairly similar, though with fewer benchmarks to pass. We can also get a sense for the current value of each metric by looking at these graphs: The United States appears to have seen between 30 and 40 deaths per 100,000 residents, given that there are three complete sections of that line of the chart. That is in fact correct; the current toll is 38.7 deaths per 100,000 residents.

Compare that with the state of New York.

Here we see that both the per capita case total and death toll have been much higher than the nation’s overall. There are far more sections represented on the graph, with benchmarks being passed day after day.

This graph also allows us to see the lag between the two metrics. The number of cases surged in late March; deaths surged in early April. Since, both have slowed way down.

In New Jersey, we see a similar pattern, though with a more modest expansion of the death toll.

We also see a statistical glitch on that graph. The state decided late last month to add suspected deaths to the total, which is why there’s a relatively short section of the chart in a period when deaths weren’t climbing.

Other states that aren’t often talked about as hot spots are revealed to have been more challenging than expected. In Rhode Island, for example, there was a surge in mid-April that led to a smaller surge in deaths. Again, note the lag time between cases and deaths.

(The state’s location between New York City and Boston contributed to its case totals.)

In Louisiana, a state hit hard fairly early, the pattern is similar. An increase in cases and, soon after, an increase in deaths.

Compare those stats with the states at the epicenter of the recent surge in new cases.

In Florida, for example, case totals accrued slowly — until June.

The death toll, happily, hasn’t kept pace.

In Texas, a less-dramatic increase in per capita cases. Despite the recent increase, the state hasn’t yet passed the 100-deaths-per-100,000-residents mark.

In Arizona, a more worrisome pattern. The state’s been adding cases at a near-New-Yorkian pace and is past 20 deaths per 100,000 residents. Those deaths are still accruing slowly, but they’re accruing.

Ideally, the reason that coronavirus-linked deaths are increasing so slowly isn’t just that the lag time has been extended. Hopefully it’s much more a function of improved treatments, ample hospital space and more infections among less vulnerable populations — all of which seem to be the case.

At some point, though, experts say, they think deaths will start heading back up. An enormous amount, psychological and political, rides on the pace at which they do so and the total that results.