Top news, reports and insights for today:
- Daily headline summaries for Thursday:
- Top medical journals the Lancet and New England Journal of Medicine have retracted articles that called into question the effect of experimental anti-malarial drugs hydroxychloroquine based on third-party data source that has now been brought into question (StatNews)
- U.S. COVID-19 cases have been slowly ticking up since Memorial Day (see graph below) (CNBC)
- Reopening resurgence? New COVID-19 cases surging in the South and West, deaths continue in hotspots across all regions
Across the U.S. more than 147,000 new COVID-19 cases were reported in the previous 7 days, a cumulative rise of 9%. Over the last week, an average of 21,095 new daily cases were reported. This marks a flattening of the overall trend which had been falling in previous weeks. In my opinion, we are beginning to see the uptick in cases that had been predicted following state reopening. The top graph below shows percent rise in new cases in each state. We now have 4 states that added more than 25% to their case totals in the last week, there were no such states in the previous two weeks. In the South, notable rises were seen in Alabama (+18%), Arkansas, which along with Arizona leads the nation with 29% increase in new cases, Mississippi (+16%), North Carolina (+25%), South Carolina (+19%), Tennessee (+17%), Texas (+18%), and Virginia (+17%). In that region, only Louisiana saw cases rise less than 10%. In contrast, 9 of 12 states in the Northeast added 10% or fewer; states on the rise were Maryland, Maine and New Hampshire. The midwest, which had been the hot zone in previous weeks, had lower transmission intensity with 3 states higher than 10% more cases (Minnesota, Nebraska and Wisconsin). New cases continue to spike in Alaska (+24%), Arizona (+29%) and Utah (+21%).
Turning to deaths (lower graphic), no state added more than 20% to their total death tally, however because deaths always lag behind cases, this is further evidence that a reopening-related resurgence has begun and that deaths will also turn higher in coming weeks/days. All regions had a mix of results both over and under the 10% threshold. The midwest and South had the higher fraction of states with bigger increases in deaths. States above 15% increase include Arizona, Iowa, Minnesota, Nebraska, Arkansas, Mississippi, North Carolina, Maine and New Hampshire.
The bottom line: While I cannot say for sure, the overall trends this week suggest a reversal of the declining trend in new cases, suggesting the possibility that reopening-related resurgence of cases is beginning to be apparent. Regions of greatest concern are now the South and West.
- Question from an alert reader: what’s the difference between CFR, IFR and crude mortality ratio?
I got a very thoughtful question from a visitor the other day. I thought I would report that here in the hopes it may be helpful to others. Here is the question:
I’m confused by terms IFR and CFR. I like your “crude death rate” but no one else seems to use the term. And my confusion comes you use CFR the way most others use IFR (IFR seems to be defined by most studies as the real infection rate over the real death rate). The CFR seems to be defined, at least in the media, as the confirmed cases over confirmed deaths. I’m not an epi person, so looking for some clarity. Is your use of those terms idiosynchratic? 😉 thanks, JK.
My answer: Thanks so much for your insightful and alert question. Epidemiology has always been what I call a watering hole rather than a formal field. It’s a thought style more than a particular discipline. We are a loose federation of oddballs from medicine, statistics, social science and virology that gather around common problems more than around solidified doctrine. We get used to the idea that important concepts may go by many different names. We acknowledge that sometimes it’s worth fussing about names and sometimes it’s not. In this case, I haven’t yet decided.
The key conceptual challenge underlying this naming issue is simple. When estimating a population risk parameter (the probability of death given disease), we seek an estimated probability or rate where the error in measurement of the numerator and denominator are similar within reason. In this case, the numerator is deaths from COVID-19 and the denominator is all people who have had COVID-19 and were thus at risk for death. With a pathogen that sickens almost every case/infection, such as Ebola or MERS, the distinction between CFR and IFR is rather academic. That’s because the ascertainment rate (the probability that all cases/infections are captured by surveillance) is high, or is at least about the same as the ascertainment of who has died of the disease. However, with a disease in which there are a large number of inapparent or hidden cases/infections, then the error in the numerator and denominator are no longer in the same ball park.
What is paramount is the need to keep two types of estimates in mind and keep them separate. One is for the death rate in the cases we know about given the surveillance we are doing. The other is the “real” death rate among all those who have the disease regardless of our surveillance. Both numbers have value but we can’t make the mistake of assuming one is an estimate of the other. For myself, I find it more useful to distinguish the case fatality rate (deaths among all cases/infections) as against the crude death ratio (cases among those we know about conditional on incomplete testing). I like this language because the two labels make it clear just how different these numbers are and makes clear how inferior the latter is (it’s called “crude” after all). Others distinguish between CFR and IFR under the premise that “cases” are what we know about in hospitals and testing labs and “infections” are the larger domain. Call me old school but all cases are infections and all infections should be cases. The big challenge is finding a label that captures not only the mild cases but the sizable fraction (perhaps 25%) of infected persons who are entirely asymptomatic. If that’s what people mean by “infections” then I am ok with that, but infection seems to presuppose some sign of illness. I taught for 30 years and generally speaking, the term cases has always meant disease events regardless of ascertainment. So, I find the distinction between cases and infections to be insufficiently clear to rely on the CFR/IFR distinction.
The real problem here is that the WHO and CDC have both contributed to the problem by referring to the crude mortality ratio as the CRF. That has been a big mistake in my view. In classic infectious disease epidemiology, the CFR has been the preferred term for what others call IFR. The result is that many people lost confidence in epidemiologists because they knew that the WHO estimate of “CFR” (at 5.4%) was wrong due to the undercounting of infections in the denominator. That’s something that we have yet to recover from. I believe we should reserve the term CFR for the better, more important and reality-based estimate of population risk. But as usual, I am in the minority. I’m happy for the time being to assume that what others call IFR is CFR for me and what others call CFR I will call crude death ratio (CDR). Confused yet? Perhaps this is a debate we should be having. Thanks for your keen insight. I apologize if my answer is not entirely satisfying.