As a hospitalist, I have been interested in the CDC’s Covid-19 forecasts for hospitalizations. There are multiple entities that provide analysis for the CDC. These are very reputable institutions and they use a variety of modeling assumptions. These researchers include Columbia University, Johns Hopkins University, Georgia Institute of technology College of Computing, Los Alamos National laboratory, UCLA, US Army Engineer Research and Development Center, and The Institute of Health Metrics and Evaluation. The following is a quote from the CDC Website.
“This week, two national forecasts predict a likely increase in the number of new hospitalizations per day over the next four weeks, two forecasts predict a likely decline, and three forecasts are either uncertain about the direction of the trend or predict stable numbers. For September 14, the forecasts estimate 2,000 to 10,000 new COVID-19 hospitalizations per day.”
This statement by the CDC suggests that over the next 4 weeks, hospitalizations for Covid-19 related illness may increase, decrease or stay the same based upon different modeling analyses. Modeling has a role to play but it is not always helpful.
The next two graphs represent information obtained from “Worldometer” which is an independent organization in the United States that analyzes, validates and aggregates data from many sources including the CDC and Johns Hopkins Resource Center. It is run by an international team of developers and researchers. I am a strong proponent of a rational, public health policy that seeks to achieve community immunity in a manner that causes the least amount of total harm to the public. Historically, this has usually required obtaining some form of community or “herd” immunity. While modeling can be helpful at times, predictive modeling should never take precedence over actual evidence and research. The following two graphs comparing Oregon and Sweden are important illustrations.
The graph above represents the actual and projected number of Covid-19 related deaths in Oregon as of 8/27/20.
Current Covid-19 related deaths in Oregon: 433.
Oregon Modeling forecasts for December 1, 2020
2,408 deaths assuming Oregon continues the same public health policy
2,520 deaths assuming Oregon eases policy restrictions
This graph represents the actual and projected number of Covid-19 related deaths in Sweden as of 8/27/20.
Current Covid-19 related deaths: 5,817
Sweden’s Modeling death forecasts for December 1, 2020
6,320 deaths assuming Sweden continues the same public health policy.
Modelers suggest that there would only be 5,958 total deaths if Sweden were to implement universal masking (95%) of the public. This is unlikely to occur since Sweden does not require masks for the general public and their policy has allowed for a targeted community immunity approach.
If Sweden and Oregon continue their current public health policy approaches, the modelers suggest that the number of deaths in Sweden will only increase by 8.6% by December 1st as compared to an increase of 456% for Oregon.
Why is there such a large difference in forecasted COVID deaths? Sweden Covid-19 deaths are stable because they used a targeted public health approach and have reached community immunity. They have not closed schools, shuttered business or indiscriminately masked the general public. However, Swedish public health officials have warned the world to avoid their mistake of allowing the virus to get into large nursing home facilities. They have since adjusted their resources to protect the elderly and vulnerable. States that protect this demographic will have the best death rates per capita.
In the U.S., 2.1 million live in nursing homes or residential care facilities, representing 0.6% of the U.S. population. This demographic represents 45% of all the deaths from COVID-19 and over 50% in some countries like Sweden. I have advocated for a targeted pandemic approach since March; hoping that we could avoid an economic, health and social catastrophe. The idea was to allow viral exposure in the healthy and young individuals which would ultimately reduce transmissibility to the vulnerable. Unfortunately, public health officials misrepresented this targeted pandemic approach with a straw-man argument. They suggested that this approach would just let the virus “rip” through the population and based upon flawed modeling assumptions, cause an unacceptable number of deaths. This public health opinion spread fear and excluded this rational, time-honored option.
A recent egregious example of misrepresentation of public health data was recently demonstrated by the physician head of the Kansas Department of Health and Environment. He highlighted the following graph to emphasize the importance of wearing masks in the general public.
Kansas allows counties to determine their preferred pandemic mitigation strategies– some counties mandated masks and others did not. This lead physician claimed that the group of counties that chose to follow a mask mandate, depicted by the yellow line, “is winning the battle”. He said, “all of the improvement in the case development comes from those counties wearing masks”. Now, look closely at the graph above. There is a secondary “Y” axis with different scales on the right. The “Y” axis scale to the right represents the blue line depicting the counties who did not require masks. This leads to a misleading impression that there is a dramatic fall in COVID-19 cases depicted by the yellow line representing the counties that mandated masks.
In response to this misrepresentation of data, the Kansas Policy Institute filed an Open Records Act and requested the names of counties and calculations used to produce the original publicly facing chart. Their revised depiction of the data has a single “Y” axis with the same scale and expanded the timeline of the “X” axis. They included the same data as the first graph with these adjustments and completely different findings were revealed. Daily COVID-19 cases on August 3rd were 77% higher per capita in counties that followed the mask mandate. Any honest scientist would accept that there are many confounding factors in this kind of data such as population density, test selection criteria and testing capacity that make the comparisons of counties of very little value. So, why would the lead public health official misrepresent the data in this way?
Lastly, I am an ardent proponent of safe vaccines. Some have advocated for a “vaccine or bust” approach. It is crucial that we not skip safety steps to implement a vaccine program. This will take time to accomplish well. Sweden has avoided the “bust” scenario by securing community immunity and avoided severe disruptions to their society. Oregon should also adopt a targeted pandemic approach. We have regrettably already busted the economy, educational opportunities and negatively impacted the health of our citizens. We must do better.
John Powell M.D.