The published literature on COVID now exceeds 211,000 papers, books, and documents, which include: 22,866 observational studies, 19,591 reviews, 1496 meta-analyses and 781 randomized control trials. These publications comprise the backdrop for our research and writing. The project began in the spring of 2020 based on a limited source of cumulative COVID-19 data and has broadened considerably. Here is what we have learned.
Our objectives remain to
- Describe trends and the geographic extent of the pandemic, including associated predictors
- Evaluate the effectiveness of vaccinations and exposure limitations.
- Provide public health perspectives.
“We know that people’s behavior, the mode of transmission, and the virus‘s characteristics all play a role but we don’t have a clear quantitative understanding of how all of these forces interact. With COVID, the biggest wild card has been human behavior.”
Dr. R. Rosenfeld, Head, Department of Machine Learning, Carnegie Mellon
Our findings along with other COVID studies are observational, focusing on numbers rather than people. Since the annual COVID mortality rate is about is only about 1 in 700, achieving reliable statistics by state would require a cohort approaching a million subjects, each of which would have to be tracked over time in order to estimate exposures. Observational studies are thus the only practical option and uncertainties about causality vs. association are inevitable.
COVID-19 trends continue to defy analysis in large part because of unpredictable variants, the latest of which is still unfolding. However, some aspects remain and will continue to dominate:
- Daily cases, deaths, and incidence of long-haul COVID can be reduced more than 10-fold by vaccination, notwithstanding deterioration over time that requires boosting.
- Acceptance of vaccination remains a personal choice - a choice that may be associated with personal characteristics including income, education, and political perspective.
- COVID-19 comprises a major cause of death in the United States and may continue to do so.
Findings In 2021.
January. We showed that COVID-19 cases increased 10-fold, by 30% per month, during the 2020-21 winter; deaths generally followed suit, while case fatality rates (CFRs) decreased up to 10-fold. The Northeast region had shifted from worst to best, so that urban predictor variables like population density were no longer important. Regional rates coalesced in January.
February. The regional trend analyses showed declining cases but not deaths, with steady CFRs. We reported that “No plausible hypotheses have been advanced for the order-of-magnitude increases in cases and deaths since September”, now referred to as the “winter surge”. We compared urban and rural rates and noted the shift towards higher cases, deaths, and CFRs in rural areas.
March. We found that cases and deaths declined while CFRs increased 5-fold. Regional trends, which had ranged 4-fold for cases and deaths now coalesced. We analyzed short-term deaths and found a strong day-of-week effect, probably due to reporting error, but no evidence of important holiday surges.
April. We tried explaining the cyclical behavior of cases in terms of “susceptibles;” predicting an underlying trend of 5000 new cases per million per month. By contrast, the average case rate is now about 8000 cases per million per month - 53 million cases in total. However, after vaccinations got underway and prior to the Delta variant, new cases dropped to levels similar to those at the beginning of the pandemic. CFR’s ranged from about 0.015 in northern regions to 0.05 in the Southwest by April. We showed that Caucasian and mixed-race subjects had far lower COVID death rates than persons of color, and that COVID death rates increased with age at the same rate as non-COVID deaths.
May. COVID rates remained low in May. Comparing states, we reported significant relationships between COVID rates and political preference along with situational factors like household crowding. An increase in Republican voters of 60 percentage points, used as a marker of political perspective, was associated with a doubling of cumulative cases.
June. We revisited our previous consideration of airborne virus transmission, which had been espoused by CDC and the epidemiological community. We estimated ventilation rates and concluded that exposures in a small apartment were likely worse than in subways or aircraft. We also revisited urban-rural differences in more detail and showed that regional COVID rates had continued to coalesce.
July. We did a detailed analysis of vaccination rates and benefits. Daily vaccination rates peaked in April, at about 50% higher in the Northeast than elsewhere. We showed a strong significant decrease in daily state-level cases associated with full vaccination. We estimated unvaccinated case rates to be hundred-folds higher than with full vaccinations. We compared vaccination effects with education, and air pollution concluding that such personal characteristics could also be important. We also showed a negative state-level relationship between voting Republican and cumulative vaccination rates. Interestingly, vaccination rates correlated with COVID rates in 2020 before mass vaccinations began, and vaccinations at this time, also apparently, reduced mortality not associated with COVID. “Could the decision to vaccinate have been more critical than the vaccination itself?”
August. We reported that COVID case rates showed a sharp upturn, followed by death, likely due to the arrival of the Delta variant. Death rates and cases had decreased steadily until July to about 30 per million or 10,000 per day – a level the CDC considered as a “tolerable” endemic. We have not had these low levels since then. CFRs peaked in July, growing six-fold with substantial geographic variability.
September. We examined cyclical variations in daily infection rates and found substantial heterogeneity. State-level vaccination rates predicted both cases and death; and complete vaccination decreased case and death rates about 100-fold, even in the presence of the Delta variant.
October. We compared October’s COVID rates with those of the 2018-19 influenza to obtain a public health perspective. Total COVID-19 and influenza cases were similar at about 30 million and both were controlled by vaccination. Compare to influenza, COVID hospitalizations were 4-fold higher and deaths were 20-fold higher - COVID is clearly the more serious disease. We concluded that 178,000 lives may have been saved by COVID-19 vaccination.
November. We continued examining vaccine effects and found no difference in the real-world effectiveness of Pfizer or Moderna vaccines. We found that COVID vaccinations were associated with reduced non-COVID deaths by 3-fold. We built an empirical mathematical model of the temporal variation of cases that fit the existing data very well but grossly underestimated the current situation. We predicted that full vaccinations for the U.S. might reach 72% in the next year, but with a range of 50-90% among states.
December. Cases began a sharp upward trend at years end, with deaths lagging behind. Regional gradients shifted, with Northwest highest and Northeast lowest. Vaccination rates continued to increase slowly, led by the Northeast. Previous beneficial effects of vaccination had been overshadowed by the severity of the Delta variant. We reported that vaccine effectiveness appears to decrease substantially over time. Long-haul COVID, neglected by the epidemiology community, was inversely associated with vaccination rates and the socioeconomic factors underlying vaccine reluctance or refusal. We estimated trends and the contributions of immunity acquired from previous infection, which we found to be statistically modest.