Using the same variables, I get a slowly declining mortality rate currently around 6.3% for the US which includes a two week span between confirmed and death. Your number is significantly lower. Are you offsetting the death tally?
I don’t have the granularity in my data subset to do that.
But looking at world data I see an unmistakable drop in the mortality rate over time.
If I use world completed cases (recovered+deaths) as the denominator the raw world death rate has dropped from 10.9 to 8.7% in just the past month.
If I do the same two week offset you and The Lancet used it’s from 9.5 to 7.8%.
I favor a three week offset that gives the world one month death rate drop of from 8.7% to the current 7.2%.
I find your US only number of 6.3% very believable.