Note: I use IHME to benchmark and compare forecast performance. I use +/-15% of the incremental deaths for the month to define an accurate monthly forecast, and +/-20% for a three-month forecast. Compared to IHME, Briggs-3 performed 120% better on the one month forecast and 125% better on the three-month forecast. For a table of month-by-month forecast vs. actuals, see this google sheet here.
The Briggs-3 model was 100% accurate (+/- 20%) projecting COVID-19 deaths three months into the future, up until August, when the model expected the pandemic to be over in the US. However, the model underestimated how many people (nearly all unvaccinated) would become infected and die of COVID in August through October. The Pandemic of the Unvaccinated revealed new insights on human behavior which inform the forecast going forward.
The Briggs-3 model was designed to make it easy to understand the fundamental drivers that determine COVID-19 deaths. Specifically, the model focuses on: Infections, Infection Fatality and Immunity.
Immunity comes from 1) Vaccinations and 2) Infection Recoveries. It can be affected by variants that can break through immunity.
Infections and Infection Fatality Rates (IFR) are hard to measure directly because not everyone that becomes infected is tested. Part of what makes SARS-CoV-2 particularly dangerous is people can infect others during a pre-symptomatic phase, and some can infect others without ever becoming symptomatic. Still, we can measure trends in infections using confirmed cases and deaths and infer overall infection levels.
It is easier to measure Case Fatality Rates (CFR), which is the percent of confirmed cases that result in deaths. The model infers the overall infection rate based on reported deaths. The model uses confirmed cases to calculate active infections and immunity from infection recoveries. In general CFR and IFR should trend downward as treatments improve but can trend upward when hospitals become overwhelmed.
What the model makes clear is that controlling COVID-19 deaths is as simple as recognizing the gas and brake pedal. The brake pedal for COVID-19 deaths is vaccinations and non-pharmaceutical interventions (NPIs) that slow the spread of infections. The most important NPIs are the use of masks, social distancing, and testing. The gas pedal, accelerating COVID-19 deaths, is more human interactions without NPIs like masks and social distancing, particularly in populations with low vaccination rates, which are more likely to die if infected with SARS-CoV-2. Each person that becomes infected can spread COVID to others, and those with COVID die at a relatively predictable rate. Therefore, each person should aim not to become infected, and if infected, to take every precaution not to infect others. But, if people believe COVID-19 deaths are a hoax, and masks don’t work, and therefore do not modify their behaviors, they put themselves and others they come in contact with at risk.
To assess risk of becoming infected (and thereby potentially infecting others), one can consider the active infections in a given county. I estimate active infections in the US by taking cases reported this week (past 7 days) and multiplying by 10. This roughly accounts for the average amount of time a person is infectious and the likelihood we only catch one in three to four infections with our current testing levels. Taking the total US population of 328 million and dividing by the active infection estimate provides a roughly 1 in 65 estimate for the last week of October, 2021. I would not, for example, join a choir competition right now. In June, the US had cases as low as about 8,000 per week (about 1 in 500 actively infected) but with the removal of masks and insufficient immunity, the fourth wave was seeded. Immunity is higher now, due to more vaccinations and infection recoveries, but not high enough to prevent the spread. I would aim for daily cases of less than 5,000 before discarding masks and social distancing. In the last week of October, the case count was over 70,000. While cases are trending downward, we have a long way to go.
The point of designing my model based on these simple drivers was to remove the black box nature of Machine Learning and statistical models to make it obvious what produces COVID-19 deaths, where risk exists, and the key metrics to end the pandemic. I suspect that everyone reading this analysis feels I am restating the obvious about assessing risk and taking precautions to avoid spreading COVID. The question is, “if it is so obvious, why did the US descend back into the pandemic?”
Why did the US descend back into the pandemic?
The US descended back into the pandemic based on the behavior of the unvaccinated. To blame the recent wave of infections, hospitalizations and deaths on Delta is a mistake. Canada didn’t suffer the same fate, nor did most of Europe, even though Delta represents the vast majority of all cases in both regions. No. The US isn’t suffering from Delta. The US is suffering from the destructive behaviors and beliefs of the unvaccinated.
The destruction in the US amounts to over 140,000 incremental deaths (and counting) and about three quarters of a million hospitalizations. The deaths are among younger people, where vaccination rates are lower. This means more years of life lost. Earlier in the pandemic, when 80% of all deaths were among 65+, the years of life lost per person dying of COVID was about 12 to 13 years. In other words, if a person died of COVID-19 at 70, he or she probably would have lived to be 82 had COVID not taken their life. Now, with a larger proportion of younger people dying, years of life lost will likely double by the end of this year compared to earlier in the pandemic.
At the same time, the US should have exited the pandemic months ago, with an economy pivoting back toward services and inflation pressures easing. The pandemic caused Americans to change purchasing patterns and purchase more goods and fewer services. This creates supply chain challenges. Why would anyone who cares about the US economy want to prolong the elevated purchase of goods over services, increase inflationary pressures and transfer more wealth to China (The #1 source of goods for the US)? At a time when Americans should be done with the pandemic, and planning a family holiday, instead families are bidding prices up to buy goods that are stuck in containers that may or may not arrive in time for Christmas. Why is the US still in the throes of the pandemic?
It is important to document how the US prolonged the pandemic by six months, causing destruction to the economy and to lives, so that we have clarity on what we might do to avoid similar destruction in the future. We can gain insight by studying the difference in positive tests among vaccinated and unvaccinated from February 2021 to the end of June 2021. Working with Brown University, Harvard, the Nevada State Public Health Lab and Washoe County Health District, we tracked every case over these months. You can read the MedRx write-up here. To quote my August Summary of this data:
Unvaccinated are responsible for spreading Delta. In partnership with Brown University School of Public Health, Harvard, The Nevada State Public Health Lab and Washoe County Health District, I studied every case of COVID in Washoe County over a six-month period (from February 1 to July 29th). What we found was 94.5% of cases, 96.0% of hospitalizations and 100% of deaths were among the unvaccinated. The population we studied mirrored the US national average, with 51% of the population in this county “Fully Vaccinated” by the end of the study period. The ratio of confirmed cases among Fully Vaccinated was 1 in 709, while the ratio for unvaccinated was 1 in 40. The implication of this disparity between unvaccinated and vaccinated is that If everyone was vaccinated, the R-naught would have been well below 1, and the virus would have burned itself out.
The Lancet recently published findings that showed those unvaccinated were 52% more likely to spread the virus compared to someone that is vaccinated and experiences a breakthrough infection. The CDC reports that those unvaccinated are 6.1x more likely to test positive and 11.3x more likely to die of COVID-19.
In sum, but for the behavior of the unvaccinated in the US, we would be out of the pandemic. Global supply chains would have had a chance to recover as spending would have shifted back toward services like vacations, theatres, etc. The trade deficit with China would have narrowed, rather than expanded. Inflation would be less. Medical screenings would be returning to normal. In other words, we’d be benefiting from the end of the pandemic. In terms of COVID-19 deaths per million per week, the US would be looking a lot more like Canada or the major economies of Europe. In the last week of October, France had three deaths per million. Canada had four. Germany seven. The US had twenty-four deaths per million -- 600% higher than our neighbours to the north. That’s not to say Europe and Canada will not experience rising COVID-19 deaths this fall and winter -- they will, per our forecast. But they won’t experience nearly the same level of deaths per million and economic destruction the US will experience.
Why is the US going in a very different direction from our friends in Canada and Europe? I’ve been modeling vaccination adoption since last year and generating a monthly vaccination analysis for Ad Council and other interested parties since January 2021. The model predicts vaccination rates at the county and state level by factoring COVID-19 deaths per capita, race, religion, politics, and several other potential factors. Over time, politics went from a non-issue to the #1 factor influencing incremental vaccinations. As a registered non-partisan, and a former registered Republican, I wish I didn’t have to bring politics into this analysis. But politics are so clearly the driving force leading to destruction of lives and the economy, it wouldn’t be intellectually honest not to report the findings.
This isn’t the first time that politics changed the trajectory of the pandemic. In September 2020, when Trump was denying the gravity of COVID and discouraging testing, mask wearing, social distancing as he held indoor rallies, portraying an image that COVID was a non-issue (even though he knew the deadly consequences based on January recording with Bob Woodward), I raised my projection of COVID-19 deaths for the end of 2020 from 300,000 to 350,000, even though the cases were flat to declining at that time. That forecast proved to be nearly exactly right -- the US crossed 350,000 on January 2nd.
Politics are once again leading to more deaths. I’ve been tracking vaccination rates by county, and the factors influencing vaccination since my report in May. Last month, the difference in vaccination growth rate between the counties where Trump won by a margin of 2:1 or more and counties where Biden won by a margin of 2:1 or more was 82%. 82% difference! In addition to my October report, here is another researcher’s chart and analysis of the significant political divide.