This post is a running account for shorter and more frequent updates to try to capture late breaking observations as events continue to change dramatically on a daily basis. I will continue regular posts (numbered) for weekly summaries and particularly poignant news.
5/19/20
The U.S. Continues to Lag the World
There is very little to cheer about in the U.S. even though it does look like the worst may be behind us. But the damage is done. The U.S. has less than 5% of the world’s population but 30% of its deaths and nearly 50% of the world’s active cases. The Northeast was devastated and is/was the deadliest place in the world. The following table shows that each of the four deadliest Northeast states exceed the per capita deaths of any country in the world.

There must also be a lot of nervousness in our nation’s capital as DC ranks as the 5th deadliest state (if it was a state).
California – The Next Hot Spot
CA has remained relatively unscathed in terms of infections and deaths (89 deaths per million compared to above table), but that is changing fast and unfortunately you cannot keep people at home when the weather is turning so nice. I witnessed in Newport Beach, crowds the size of Memorial Day crowds at the beach this weekend, with few masks and large congregations and parties and the news is it is happening everywhere. Signs of trouble are already in the statistics. Here are hospitalization rates in Orange County and LA. OC was trending down in early April, but has reversed course lately. LA is at a stubborn plateau.

Here are the death rates for OC and LA. Deaths are not coming down very fast and looks like LA may get a rebound. We need to exercise extreme caution if we are to relax social restrictions. Oddly OC is at about 12% of the hospitalizations of LA but only at about 6% of the death rate of LA.

As discussed before, deaths occur about 2.5 weeks after infections and hospitalizations somewhere in between. Interestingly the death statistics do seem to follow the hospitalization statistics by about 1-2 weeks. My gut is that we are going to see a rise in these statistics in the coming weeks, particularly in OC.
Sweden – The Experiment that Failed
Despite what is being reported, the experiment by Sweden to forego social restrictions and rely on the people’s good sense has failed. The per capita number of deaths now ranks as the 6th worst in the world and it is climbing faster than the ones above it as they are very much now into recovery. There is still a lot of denial on this, but the numbers don’t lie. The nation’s top epidemiologist, who advocated for this policy, to his credit, has pretty much been saying lately, “oops!”
Social Easing Considerations
Another way of looking at social easing is if the contagion rate was say R0 = 3 on average (this is not a constant, but dependent on environment, population density, culture, etc.), then to relax social restrictions, we would still need to reduce our typical contact rate by 2/3 to get to R0 =1 and no amplification. Maybe that is not too hard if we just practice better hygiene and behavior, e.g., keep distance where possible, wear a mask when close to people, don’t shake hands or hug, and wash hands and surfaces frequently.
…..
5/2/20
U.S. Still Lags the Rest of the World in Recovery
I try to keep personal opinions out of this blog as I strive for fact-based analysis, but I can’t avoid letting a little leak into these Daily Rumblings, having seen deaths to people close to me. Looking at the statistics every day it is depressing to see the U.S. continue to trail the rest of the world including much less advanced countries than ours. Other than the U.K. we were the last in the world to acknowledge the problem and then dawdled on how to respond such that even the U.K. is now ahead of us in recovery. Sweden oddly is in my opinion severely misguided in their voluntary approach to social isolation and they are now experiencing an epidemic that they seem to be denying. There will be plenty of time when this crisis passes to do a post mortem “lessons learned” analysis, but until then we have to continue to socially isolate and obey reasonable restrictions and ignore the babble that’s coming out of our administration and other states.
So here are one of my favorite plots, that I update weekly, for showing the course of recovery internationally and domestically. They mirror the conclusions from the death rate plots that I post in the Weekly Updates, but I think these plots below are visually clearer.

An indication of recovery is when the data turn downward from the positive linear slope. This indicates when the death rate slows down but not when it reaches a peak, which typically occurs a couple of weeks later (the death rate plots are better for seeing that). The points in the plots are weekly averages so each point is spaced by 1 week. The U.S. appears to have finally turned a corner this week, but we really need at least 2 weeks to confirm this trend. Countries like Italy, Spain, France, and the U.K. have turned this corner 3-5 weeks ago.
The situation in the U.S. states mirrors the lag in the overall U.S. statistics. Thankfully the NY death rate has taken a big drop, but again this is only a one-week data point. NJ is struggling as well as CA, though the latter is at about 1/20 the death rate per capita. WA is recovering and although MI and LA now show a two-week trend, the drop is not very big so the verdict is still out on their recoveries.

…..
4/23/20
U.S. Lags the Rest of the World in Recovery
The U.S. is behind all other major hotbed countries regarding signs of recovering from COVID-19. Even the U.K., the last of the countries to implement social distancing, is showing clear signs of getting over the death rate peak. One can look at the death rate plots in yesterday’s post “12. Weekly Update: All Populations are in Recovery” to see the progression from death rate acceleration to deceleration (getting over the peak) to see how much the U.S. is lagging. However, a very sensitive way to also see this is from the log-log plot of death rate vs. cumulative deaths below.

Deviations from the positive linear slope is an indication of progressing past the exponential growth rate. So, the point of departure is actually about half way up the death rate curve, which is about 2 weeks before the peak. The points in the plot above are weekly averages so each point is spaced by 1 week. This shows the U.S. is no less than 2 weeks behind the rest of the world in its progress toward recovery.
This situation in the U.S. states is also not encouraging and mirrors the lag in the overall U.S. statistics compared internationally. The log-log plot for U.S. states below shows encouraging signs for WA, and maybe for LA and MI. But for NY and NJ we may have prematurely called a peak in yesterday’s post. They do not look like they are recovering yet.

…..
4/18/20
A curve ball everyday
Let’s start internationally. Italy and Spain are clearly on the downward slope of the death rate curve and therefore also the prevalence (cases) curve. However, whereas Spain continues down a relatively symmetric path as the way up, Italy seems to have hit a plateau on the way down experience 500-600 deaths/day for the last 5 days (still down from a peak of 1000). If this trend continues up to our weekly update cutoff of Mondays then we will revise up the total forecasted deaths, but probably by no more than 20%. France and the U.K seem to have reached a peak, but we will not know that for sure until we see some downward motion.
Domestically, things are bleaker. NY continues to have a climbing death rate and our call on 4/14 that it had reached a peak appears to be premature. By comparison the UW IHME model claimed it was 4 days past it’s peak so that attests to how difficult it is to make a definitive claim. This will most likely push the total death count in NY to greater than 20,000. The CA death rate has risen to about 90 per day. Though a comparatively low number it is still climbing. Still we think it will be by a large margin a much less affected U.S. state when measured by per capita.
When deaths are not deaths: Not so accurate a measure
Our model premise is based on the assumption that there is no more accurate outcome measure than deaths. Well in the last couple of days we have learned that death counts have been understated for two key reasons: (i) The reported numbers are dependent on how diligent healthcare facilities report deaths and whereas hospitals are pretty reliable and also the primary location for deaths, it is now coming out that nursing homes, assisted living, and other less emergency based facilities have been slow to report this date. (ii) Reported deaths are for confirmed COVID-19 patients and those about to die in hospitals are usually diagnosed, however, it is now acknowledged that many deaths can be attributed to COVID-19 that were not diagnosed. This has caused corrections to data either officially as decreed by the CDC on April 15 or informally as has been done in France and China. In our next weekly update we will describe how we handle these one time surges in reported data.
Can current antiviral drugs work
Everyone is hoping for the magic bullet. Our President has harped on repurposed drugs as the savior and just around the corner. Dr. Fauci has presented a more measured and truthful response. Still there is some hope that a currently approved drug for other viral diseases might at least offer some reduction in death risk for the most afflicted patients. Of the promising candidates that are in expedited clinical trials, Gilead’s Remedivir, developed for Ebola virus, has reported unofficial results from the University of Chicago that suggests for this limited study a marked reduction in deaths in the most severe patients who were given the drug. There is not much more one can say until the statistics are better understood.
Reported confirmed cases way undercounting
In just the last few days we have had the following reports:
- China has restated the number of deaths in Wuhan from 2,579 to 3,869 attributed to unaccounted for deaths at home. That this increase is precisely 50.0% raises suspicions about the truthfulness of this report. Based on demands for funeral homes and cremations, some accounts estimate the death count to be closer to 40,000. Regardless of the absolute numbers, I believe they have significantly reduced the death rate so that the curve if not the amplitude is somewhat believable.
- At home deaths in NY are also now reported to be significantly undercounted and the official count was adjusted by nearly 4,000 deaths on 4/17/20 to correct for this and this now accounts for about 25% of the total death count in NY.
- France made a similar one-day correction on 4/4/20 add about 1,500 to its previous count of 4,500 (now at 19,000, 4/18/20).
Accuracy of COVID-19 diagnostic tests
There have been reports of patients who have had the virus who get re-infected, which has baffled the medical profession. But there may be another explanation, that they never had it in the first place.
The accuracy of a diagnostic test is typically measured by sensitivity, which is the percentage of positive samples the register as positive (true positives and the misses are called false negatives) and specificity, which is the percentage of negative samples that register as positive (false positives vs. true negatives). No test is 100%, but the standard reverse transcription polymerase chain reaction (RT-PCR) is reported to have sensitivity near 100% and specificity of about 96%. This sounds pretty good and it is, however, here is how it plays out in real life. Let’s say you take 100 random people and that 5 have the virus. Then the PCR test will probably detect all 5 positives. However, of the 95 negative patients and a 4% false positive rate there is likely to be 4 negative patients who are diagnosed as positive. Well that is now 9 positive readings of which only 5 are real. So one has to suspect that a significant number of confirmed cases are not true positives.
Now based on current protocols of testing only those most suspected to be infected in which case about 25% register positive, the false positive contribution will be about 16% (4%/25%). However, as we expand testing to more of the general public, this ratio will change dramatically and we need to be prepared for how to handle that situation. I have no suggestions to offer at this time other than to be aware of this error when this data is used for real purposes.
…..
4/11/20
Response time is everything!
U.S. just surpassed Italy today and Spain three days ago in deaths becoming the world’s most afflicted country and with a death rate still growing at an alarming rate.
I try not to get on my soap box, but quick recognition of a problem and immediate implementation of solutions is absolutely crucial to controlling an epidemic. I have shown previously (Post #7: Weekly Update: Grim News), and I’m not by any means alone in this assessment, that every week of dithering will cost a factor of 2-4 in total deaths. That is the perniciousness of exponential growth.
There is now strong evidence (post-crisis analyses should make this irrefutable evidence) that quick action to implement intervention such as social distancing and extensive testing can considerably curb the outbreak and death rate (China, S. Korea, Taiwan, Singapore). Conversely the evidence shows that countries that delay a response suffer costly multiplications of death rate. Examples of late comers are the U.S., France, and U.K, each first nationally sounding the alarm on 3/16, 3/16 and 3/28, respectively, and they are posed to be the three deadliest countries in the world. The U.S. alone, however, didn’t implement any federal actions to recluse, that was left to the states. And again, a good example of the necessity for quick action. CA enacted a stay-at-home edict on 3/18, NY not until 3/23. CA currently has 14 deaths per million people, NY has over 400! So the U.S., the most advanced nation in the world, will be ignominiously remembered for its lack of responsiveness and preparedness costing tens of thousands of American lives. Even worse is that this was willful. This country has 5% of the world’s population and 20% of the COVID-19 deaths.
As far as detecting a rolling over to a peak in infections (new cases or incidence), hospitalization rates are a good indicator, but whereas death rates lag incidence by about 2.5 weeks, hospitalization probably lags by 1-2 weeks since the virus will be coursing through someone before they are serious enough to require hospitalization. So, I still think the gold standard for knowing when you hit a peak is deaths.
The question when is it safe to relax social distancing? That will be addressed in a new blog in a few days. But suffice it to say that life can never return to normal until we have a vaccine or when upwards of 50% of the population has gotten the disease and has developed immunizing antibodies. Otherwise we will be just back to where we started.
…..
4/4/20
Italy and Spain reaching peak deaths. U.S. still accelerating.
Below are latest death rate plots for our coverage of hot-spot countries. There appears to be a real rolling over in the rate for Italy, that we perceived a couple of days ago, which allowed us to rank it yellow (red, yellow, green ratings) and now looks to be reaching a peak. If so our death projections in Post #8 may come down by a factor of 2. Spain also may be further up the death rate curve than thought two days ago. Unfortunately the U.S. France, and U.K are still in exponential growth so still low on the death rate curve.

The U.S. is lagging most of the rest of the world perhaps due to being late to declare COVID-19 a national emergency. New York and California are still in exponential growth, though per capita NY is still about 20x higher in total deaths and death rate. Washington state is appearing to be reaching a peak.

That’s it for today. Serious social isolation needs to be practiced.
…..
3/28/20
U.S. News
We are tracking NY, CA, and WA, but may add to that soon. Below are plots of the daily death counts for these states and the following are observations:
- It appeared for a few days that WA was containing the contagion but on 3/27 there was a surge.
- NY and CA also look to be leveling off, but we will need at least a couple of more days to see if the death rate is slowing; however, they are not going down yet either.
Another way to look at the data is per capita. The table below reveals how horrendous NY is at this time especially compared to CA that is also considered a hot spot, but not nearly so when viewed per population.
- Per capital NY has 15x more deaths and 25x greater daily death rate than CA. Washington has a very high cumulative deaths per capita but its deaths per day is subsiding.
- Not shown in the plots or tables, we estimate the number of true cases today are about 1 in 240 people in NY, 1 in 4000 in CA and 1 in 1300 in WA. The numbers will be less if social distancing has been working since inception about 1-2 weeks ago.
- On the whole the U.S. cumulative deaths now exceed 1,500 though flattening out at about 200-250 deaths per day! We’ll be anticipating a reduction in the next few days due to social distancing (fingers crossed!).
Regarding other states:
- Louisiana has catapulted to #3 in the nation for total deaths (119 yesterday and 137 today and counting) and 1st in death rate (as of yesterday), but NJ has now just overtaken them today to #3 (140 deaths and counting) and is now comparable to the top death rates (about 25% increase per day). Louisiana’s sudden emergence may be due to a surge in cases arising from Mardi Gras that is now evolving to either recovery or death.
- Other new trouble spots are MI (111 deaths), FL (54 deaths) and IL (47 deaths).
International News
Below are plots of daily death counts for several countries that we are tracking and the following are observations:
- Spain has nearly overtaken Italy for number of deaths/day although both countries are still experiencing alarming death rates.
- The press sometimes reports news based on two-day trends. This leads to false conclusions, which is dangerous misinformation and what I try to avoid here. The reasons why short term trends can be false include:
- Official numbers are often miscounted when counting live and it sometimes takes a couple of days for the numbers to settle down. I see this in the gold standard daily reports published by the WHO. For some reason they lag quite a bit on U.S. deaths.
- Often there are two sets of daily numbers, those collected on Greenwich Mean Time (GMT) and others collected in local time zones. For example, a GMT reaches midnight when say California is at 4 pm. That means by the GMT standard, the subsequent CA count will be included in the next GMT day. I will update with whatever data is most current, but I then reconcile to the WHO GMT data when it comes out the next day.
- The U.S. appears to be trending toward a flattening out of the death rate. Let’s follow this for a couple of more days to see if it persists.
- Italy looked to be flattening too but then saw a surge on 3/27. Too soon to tell what the trend is.
- Iran seems to have flattened their death rate. Let’s now hope to see a reduction in daily rates that trends toward near zero
- The U.K. also gave a false hope for a few days, but then surged upward on 3/26 and 3/27.
I like the following representation of cumulative deaths plotted on a log scale. I have presented similar plots, but this puts it on a scale starting at 10 deaths for each country so the days since do not occur on the same calendar days, but is a good way to see whether a country’s death rate presages a previous country’s agony. For example, although Spain lags Italy in deaths because its contagion started about 11 days later, the death rate appears to be greater than it was in Italy for the same time after inception. Spain is in Italy-scale serious trouble.
We still do not see too many bright spots around the world, but most countries are still within 2 weeks of initiating strict measures, so if these are effective we may not see them in the death statistics for another week or so.
I think you’ll find New Jersey is part of the New York epidemic. I suspect areas where people travel into NYC from New Jersey likely have the cases.
We are all overdosing on COVID news, data, and stories. Can you stand one more blog post?
It is not fashionable to inject a little perspective in this topic. Media whose income depends substantially on advertising must emphasize emotional rather than rational stories. Bucking this trend, I would like to add a little perspective by looking at annual deaths in the US that are not reported with such alarm.
This season the CDC estimates that, as of mid-March, between 29,000 and 59,000 have died this season due to influenza illnesses. (https://www.health.com/condition/cold-flu-sinus/how-many-people-die-of-the-flu-every-year). The midpoint of this incredibly imprecise number is 44,000.
Currently the annual traffic death rate in the US is about 37,000. https://en.wikipedia.org/wiki/Motor_vehicle_fatality_rate_in_U.S._by_year
The peak was in 1972 at 54,589. (It took 15 years to reach a total 58,000 US deaths in the Vietnam War.)
Last year there were 161,000 avoidable deaths in US hospitals. This number actually represents a decline in deaths from 2016 when Johns Hopkins conducted a similar analysis in which 206,000 preventable deaths occurred. https://www.modernhealthcare.com/safety-quality/161000-avoidable-deaths-occur-hospitals-annually-leapfrog-group-finds
If you’re counting, this is approximately a quarter of a million deaths last year, which occur more or less every year. Most of these deaths are avoidable.
Let’s look at the COVID19.
The only hard data is the pure death number. Almost death percentages are vastly overstated due to undercounting the infection numbers. Unless there is 100% sampling there are likely errors due to sampling bias. There is some good testing data (good = 100% of population is tested). Example: a March 25th Wall Street Journal article authored by two Stanford professors of medicine notes that 100% of the population of Padua Italy (3300 people) were test on March 3 and using this data calculated a death rate of 0.06%, vastly different from the 8% cited. How do we get our head around the >100X difference? There are many confounding variables.*
So will the COVID19 deaths this year reach the 250,000 deaths we overlook every year? No one knows but I would not be surprised. I would be surprised it if greatly exceeded this number. My point is that emotional rather than rational response is likely creating a US that will not recover, ever.
*The deaths are greatly skewed to at-risk individuals above the age of 65. At risk are individuals with pulmonary ailments (asthma, emphysema, COPD), diabetes, high blood pressure, heart disease, etc. The US is especially vulnerable here with a 37% obesity rate which leads to many of these conditions. The death rate varies greatly with country demographics. China’s 65+ population is 12%, Italy’s is almost twice that at 23%. The USA’s is between these two at 17%. The survival rate of these at-risk individuals is a strong function of whether there are advanced ventilators available to save their lives. These ventilators are generally only available in ICU beds. The USA has the highest rate of ICU ventilators in the world at about 100 per 100,000 population aged 65+; Italy has about one-quarter this value, see “The Economist,” March 21-27, 2020 p 57. (The March 28th WSJ has a county by county map of the US statistic if you are interest in your local area.)
Thanks Ed – I was going to do a blog looking at flu statistics as a comparison. I will make two points here: (1) Perhaps with social isolation we will reduce flu disease and maybe have a much lower number of flu deaths thereby offsetting COVID-19 deaths. (2) Hard to predict when deaths will stop, but I would assume that when we hit the peak of deaths/day and it starts to turn over, then we will be at 50% of total deaths. Half on the way up and half on the way down assuming the curve is symmetric. – Jack
I agree the Death rate vs. Cumulative Death log-log plot is the informative for assessing a peak. I wish other sources used this approach.
I owe this plot to a great video that you should watch: https://www.youtube.com/watch?v=54XLXg4fYsc, though it plots cases, and not deaths, which is a less precise measure. By plotting death rate vs. cumulative deaths on a log-log plot one can immediately see deviations from exponential growth representing the start to recovery.