Quote:
Originally Posted by NickMPK
Everyone says this, but it doesn’t seem to be true. I posted this two days ago in the OOT thread:
With the recent spike in cases in the South, it occurred to me to check something out with respect to the common argument that deaths are a lagging indicator of infections/cases. Thus when we see a spike in cases, we shouldn't expect to see a spike in deaths until several weeks later.
This may intuitively feel true, but empirically it has not been true, at least in places that have had major outbreaks. I looked through my data for the peak in the seven-day averages for cases and deaths of the worst hit countries and states. Most hit their peak in both within a few days of each other:
Countries:
Italy: Cases peaked March 26, deaths peaked April 2
Spain: Cases peaked April 1, deaths peaked April 3
France: Cases peaked April 9, deaths peaked April 8
UK: Cases peaked April 14, deaths peaked April 13
Early states:
Washington: Cases peaked April 9, deaths peaked April 10
New York: Cases peaked April 10, deaths peaked April 13
Michigan: Cases peaked April 7, deaths peaked April 16
New Jersey: Cases peaked April 7, deaths peaked April 21 (though there was a long plateau for both)
Connecticut: Cases peaked April 22, deaths peaked April 25
Pennsylvania: Cases peaked April 10, deaths peaked May 5 (this is the biggest exception)
More recently:
Illinois: Cases peaked May 12, deaths peaked May 13
Maryland: Cases peaked May 24, deaths peaked May 1 (?!?, also a long plateau for both)
Virginia: Cases peaked May 31, deaths peaked May 28
I have been tracking 11 states and the US and my 7 day rolling avg data are close to yours in some cases, but I do see lags.
"+rate" below stands for the positive test rate %. This is averaged out for the past 7 days (i.e., total positive cases/total tests)
US 11 day lag: Cases peaked 4/10 (+rate=19.7%), deaths peaked 4/21
NY 4 day lag: Cases peaked 4/10 (+rate=43%), deaths peaked 4/14
MI 10 day lag: Cases peaked 4/3 (+rate =27.7%), deaths peaked 4/13
NJ 10 day lag: Cases peaked 4/7 (+rate=39%), deaths peaked 4/17 then again 5/3
CT 4 day lag: Cases peaked 4/22 (+rate=39%) deaths peaked 4/26
PA 24 day lag: Cases peaked 4/11 (+rate=23.2%), deaths peaked 5/5
AZ 30 day lag: cases peaked 4/9 (+rate=9.8%), deaths peaked 5/8
AZ ??? lag: cases are at all time highs (+rate=17.4%), deaths going up since 6/11
CA 14 day lag: Cases peaked 4/9 (+rate=8.8%), deaths peaked 4/23
CA ??? lag: cases are at all time highs now (+rate=4.6%), deaths are near lows since peak.
FL 30 day lag: cases peaked 4/8 (+rate=11.1%), deaths peaked 5/8
FL ??? lag: cases are at all time highs now (+rate=8.2%), deaths at all time lows since peak
GA 10 day lag: cases peaked 4/13 (+rate=23.3%), deaths peaked 4/23
GA ??? lag: cases are at all time highs (+rate=7.6%), deaths are going up since 6/7
NV 9 day lag: cases peaked 4/4 (+rate=14.7%), deaths peaked 4/13
NV ??? lag: cases are at all time highs now (+rate=4.5%), deaths are at lows since peak
RI 11 day lag: cases peaked 4/26 (+rate=14.4%), deaths peaked 5/7
RI 7 day lag: cases peaked again 5/21 (+rate=12.6), deaths peaked again 5/28
I'm seeing an average of a little less than 13 day lags between case peaks and death peaks.
Of the states that I am looking at that have all time highs in cases recently, 2 of the 5 are experiencing significant increases in deaths. Though both AZ and GA still have fairly low death rates per capita at this point.
I'm also thinking that the European countries you mentioned had alarming death rate %'s early on which is an indicator that their testing capacity was very low. So they may not have come close to capturing their actual peak in cases at the time. edit: you can see this is true for NY and CT which had very high positive test rate %'s at their case peaks.
In looking at this I think we have to consider that deaths are certain statistically and positive cases are somewhat random in that the positive test rates at the time of the peaks may have been abnormally high. I realize that some states may be intentionally under counting their deaths but as long as they are doing it consistently, it shouldn't affect the death peaks.
Last edited by Mr Rick; 06-19-2020 at 11:14 AM.