Top news, reports and insights for today:
- Crude fatality ratios remain widely and mysteriously divergent
One of the big mysteries of this epidemic has been the tremendous variation in the crude fatality ratio across nations. The crude fatality ratio is simply the number of COVID-19 deaths divided by the number of confirmed cases. As I have said in this blog, this number is a poor estimate of the true case-fatality rate because the denominator (confirmed cases) is such a bad estimate of the true number of cases. That is because so many of those infected have mild or no symptoms and are not tested, and because the rate of false-negative tests is disturbingly high. The graph below shows the current best numbers on variability in crude fatality ratio across 60 countries (calculations are my own using Worldometer data). Italy’s rate has climbed from 11.6% to 12.5% amid a surge in deaths, even as new cases rose by the lowest number (3,600) in almost three weeks. Among these 60 countries, the crude death rate is higher than the WHO estimated CFR of 3.4% (green reference line) in one third of cases. The range is staggeringly wide. The top tiers include both countries with advanced health systems (Italy, UK, Netherlands, Spain and France) and low-resource nations (Algeria, Morocco and Iran). The same is true of countries in the lower tier (Norway, Finland, Australia and New Zealand vs. Bahrain, Qatar, South Africa and Latvia).
What does it mean? It’s not yet clear to me why the spread is so large. Much of it doubtless has to do with differences in the way testing is being done in each country. It does tell us that the crude death rate is probably itself not a good metric to compare the magnitude of the epidemic across nations. Countries at the high end are very likely testing selectively (only the sickest patients are being tested). But testing is surely not the only story these data are telling. It should be a high priority to try to understand and explain why some countries are keeping crude death rates so much lower than others.
- South Korea has been a model, now it offers a warning about how the curve doesn’t always stay flat
South Korea has been on many people’s radar since reporting its first case on January 20. Despite an initial wave of intense and rapid spread, South Korea used early and extensive testing, digital contact tracing and cluster identification to contain the outbreak and flatten its curve. Now, NBC News is reporting that the country is seeing what may be a second wave of new infections after social distancing measures have been relaxed in recent weeks. According to Worldometer, South Korea has 10,284 cases, placing it 17th among the world’s nations, with just 186 deaths. The crude fatality ratio remains low in South Korea (1.8% deaths among confirmed cases). Last Monday, 125 new cases were reported, up from a low of 64 per day in mid-March. COVID-19 deaths are also on the rise.
Why is this important? It’s too soon to say for sure that South Korea is experiencing a second wave, but a recent uptick in new cases points to the reality that even a country that has been successful in containing the outbreak and flattening its curve must remain vigilant. COVID-19 is likely to come in multiple waves over the coming months as nations struggle to balance outbreak control measures and a return to normal life.
- COVID-19 deaths are being undercounted in the U.S. and elsewhere
An article in today’s Washington Post by Emma Brown, Beth Reinhard and Aaron C. Davis makes the case that COVID-19 deaths are being undercounted. It’s not just that accurate counting is challenging and often lower priority than taking care of very sick patients. It’s also a function of politics and procedures. The CDC has only been counting deaths of patients who have had a confirmed positive laboratory test for coronavirus. That means patients who die while their sample sits in a lab waiting to be processed, won’t be counted as a COVID-19 death. Last week in California, doctors were still waiting for test results on 67% of hospitalized COVID-19 patients. Patients who died before test results arrived won’t be counted using the CDC’s rules. Many reports describe large numbers of critically ill patients in hard-hit New York city with unknown test results. In that city, it takes a week to get results due to backlogs at the major commercial testing labs. Across the nation, those who die and are suspected of having COVID are not being tested due to scarcity of testing materials. Patients who die in nursing homes and prisons may not be tested to avoid liability and scrutiny. Another factor is the poor sensitivity of current testing, which results in as many as 30% false-negatives. During an epidemic, accurately counting deaths due to the disease is always problematic. After-the-fact research on the 2009 H1N1 “swine” flu showed that pandemic probably caused 15 times as many deaths as were officially recorded based on lab-confirmed cases. As Harvard epidemiologist Marc Litsitch has pointed out, many patients are dying with COVID-19, but not of COVID-19. Because those with chronic conditions like COPD, heart disease and diabetes are at higher risk of dying anyway, it’s hard to know what role COVID-19 may have played.
Why this matters? Keeping track of deaths is a vitally important part of our work as disease detectives. Deaths will be a better indicator of when the disease peaks compared to cases. But it’s impossible to escape the fact that tabulating deaths is a very challenging and politicized enterprise. We may not have a true picture of the death toll for years after the crisis has subsided. The COVID-19 epidemic, when combined with the political reality of an election year, creates strong incentives to undercount. We see evidence of politics impacting the tabulation of the disease’s toll in China, Russia and other nations. That should not stop us from working hard to keep careful records so we will know with confidence when one wave peaks and another begins.
The bottom line: The CDC should report separate numbers of deaths among confirmed and suspected cases and not allow politics and the testing backlog to cloud the picture.