According to COVID act now, cases are now up to 100k a day, from an average of 40k/day in early October. Two weeks ag case numbers were already up by 50%. Meanwhile the change in deaths has been modest, still around 10%. This would require a three week lag for deaths to be proportional to cases. What might be happening?

New Answer
Ask Related Question
New Comment

3 Answers sorted by

I think there are two basic reasons:

  1. There's a lag between infection and test, between test and hospitalization, and between hospitalization and death. So death is the chart that will go up the latest, and there are a lot of long-tail deaths from people who were infected a long time ago, so death rate rises are less "spikey" than infections.
  2. Case fatality rate has trended down in general. Hospitalization-death rate has roughly halfed since the first wave.

Scientists aren't entirely sure why 2. is happening, but there are multiple possible explanations, all of which probably contribute to some degree.

  1. In some places, many particularly vulnerable people are already dead.
  2. The average age of newly infected people has gone down after the first spike, probably for multiple different reasons, including older people being more careful on average. Also, we might be getting slightly better at protecting vulnerable populations. Younger people have a much higher chance of surviving an infection.
  3. Mask wearing decreases the initial viral load that triggers an infection, which causes a less severe, more survivable infection.
  4. Treatments are getting more effective.

Perhaps also higher availability of testing and higher awareness means more people with mild symptoms get tested?

More people with mild symptoms getting tested sounds like a cause. I buy that. Awareness is a good guess but the July hump also had high awareness. Thanks!

The lag was not previously two weeks. I doubt the lag got longer so I suspect there is a real and large CFR reduction.

The four lower points sound plausible and are some good news, finally :)

This is from back in March, but the CDC found the median time from symptom onset to death was 18 days (5 days from symptom onset to pneumonia, 13 days from pneumonia to death). Add at least three days from infection to symptom onset and you have a three week delay.

And again, this was in March. It's possible that treatment improvements have made this delay longer.

There’s also a delay between symptoms and confirmed infections. The July death peak came two weeks after the July confirmed daily infection peak.

Right now two weeks ago we had a 50% CDI increase, but deaths haven’t caught up.

The answer must be a combination of longer time to death and lower IFR (including less at-risk people contracting). I’m hoping it’s more the second but not much update yet.

Or faster testing!


COVID ACT NOW has tremendously different data than many of the state websites. I know for instance that in my county in Georgia, Cherokee, Covid Act Now has 83 cases per day while the GA DPH has 29 per day. That is a huge discrepancy, and is unrelated to the media reports earlier in the year about discrepancies at GA DPH, which have long since been cleaned up. I had a discussion about this on Twitter today with a former soccer teammate of mine who is now an attorney with Covid Act Now, in fact:

He claims that Covid Act Now is scraping data from different sources than the official count, but we don't know what those sources are. So that discrepancy needs to be nailed down before the question can be answered. Covid Act Now may be double counting somewhere, or counting antibody tests.


When you go to GA DPH's website,

..they track new infections by age bracket, which you can compare to the total infection rate graphs. As we now know, IFR varies widely by age bracket. IFR under age 25 is about one third of ordinary influenza, while IFR in all brackets age 65 or older is around 8 times higher than ordinary influenza. Deaths are very age related. The deaths curve for Georgia looks almost exactly like the "60 and older" infection curve, and seems unrelated to the other age bracket curves, which is what we should expect given the IFR by age differences. So the very easy answer is that the only infections that matter for deaths are the elderly, and the current wave of infections is not appreciably affecting the elderly.


A three week lag between reported infection and death seems entirely reasonable given how this disease progresses typically.

New to LessWrong?