Population Immunity to Pre-Omicron and Omicron Severe Acute Respiratory Syndrome Coronavirus 2 Variants in US States and Counties Through 1 December 2021

Abstract

Background

Both severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and coronavirus disease 2019 (COVID-19) vaccination contribute to population-level immunity against SARS-CoV-2. This study estimated the immunological exposure and effective protection against future SARS-CoV-2 infection in each US state and county over 2020–2021 and how this changed with the introduction of the Omicron variant.

Methods

We used a Bayesian model to synthesize estimates of daily SARS-CoV-2 infections, vaccination data and estimates of the relative rates of vaccination conditional on infection status to estimate the fraction of the population with (1) immunological exposure to SARS-CoV-2 (ever infected with SARS-CoV-2 and/or received ≥1 doses of a COVID-19 vaccine), (2) effective protection against infection, and (3) effective protection against severe disease, for each US state and county from 1 January 2020 to 1 December 2021.

Results

The estimated percentage of the US population with a history of SARS-CoV-2 infection or vaccination as of 1 December 2021 was 88.2% (95% credible interval [CrI], 83.6%–93.5%). Accounting for waning and immune escape, effective protection against the Omicron variant on 1 December 2021 was 21.8% (95% CrI, 20.7%–23.4%) nationally and ranged between 14.4% (13.2%–15.8%; West Virginia) and 26.4% (25.3%–27.8%; Colorado). Effective protection against severe disease from Omicron was 61.2% (95% CrI, 59.1%–64.0%) nationally and ranged between 53.0% (47.3%–60.0%; Vermont) and 65.8% (64.9%–66.7%; Colorado).

Conclusions

While more than four-fifths of the US population had prior immunological exposure to SARS-CoV-2 via vaccination or infection on 1 December 2021, only a fifth of the population was estimated to have effective protection against infection with the immune-evading Omicron variant.

By 1 December 2021, >48 million coronavirus disease 2019 (COVID-19) cases and 780 000 COVID-19­–associated deaths had been reported in the United States [1, 2]. Between 1 December 2021 and 1 February 2022, an additional 26 million cases (35% of cumulative US COVID-19 cases) and 100 000 deaths (11% of all US COVID-19 deaths) were reported [3]. Reducing COVID-19 morbidity and mortality rates depends largely on reaching high levels of population immunity. The emergence of the Omicron variant [4, 5] illustrates the importance of identifying areas of highest vulnerability and underscores how continued viral evolution may reduce effective protection.

The true number of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections that have occurred is unknown. Estimates of the percentage of the US population ever infected vary between 37% and 62% [6–9]. Seroprevalence estimates from a nationwide convenience sample suggested that, as of 21 December 2021, 33.5% of the US population >16 years old had infection-induced SARS-CoV-2 antibodies [10]; a nationwide blood-donor study estimated 28.8% infection-induced seroprevalence for the same period [11]. The level of protection that infection confers and the rate at which protection and seropositivity wane are incompletely understood [12–14].

By 1 December 2021, >240 million US residents (72.9%) had received ≥1 dose of a COVID-19 vaccine [1], and >80 million residents had received both the initial 1- or 2-dose schedule and a booster. Reported efficacy against symptomatic infection for the 3 vaccines available in the United States ranged from 66% (Johnson & Johnson) to 94% (Pfizer and Moderna) in clinical trials [15–17]. Vaccine efficacy against infection was estimated to be lower during the Delta surge compared with earlier waves, and further reductions in efficacy against the Omicron variant have been reported [4, 5]. Declines in vaccine efficacy may reflect both waning immunity and increased immune escape for viral variants. Despite evidence of waning efficacy [18–21], vaccination appears to provide durable protection against severe disease, and boosters partially restore vaccine efficacy [22–24].

Local estimates of population immunity are important for understanding the risks of continued SARS-CoV-2 transmission. State-level estimates of infection- and vaccine-induced SARS-CoV-2 seroprevalence based on blood donation data have been reported [25], with estimates for May 2021 ranging from 63.7% in Mississippi to 91.7% in Connecticut. While these estimates provide a direct measure of seroprevalence in the study populations, they may be affected by systematic differences between blood donors and the general population. Moreover, these data do not provide county-level estimates or account for waning of protection.

For the current study, we used state- and county-level modeled estimates of cumulative SARS-CoV-2 infections and reported coverage for initial and booster vaccination [6, 26]. We estimated the joint distribution of prior SARS-CoV-2 infection and vaccination from survey data [27]. Using these inputs in a Bayesian analytic framework, we estimated the population with SARS-CoV-2 immunological exposure (ever infected or vaccinated) for each US state and county through 1 December 2021. Incorporating evidence on the time course of natural and vaccine-induced immunity, we estimated effective population immunity against infection and against severe disease over time, as well as effective protection against the Omicron variant, accounting for immune escape.

METHODS

Data

Infections

We extracted time series estimates of SARS-CoV-2 infections from a statistical model [6] that synthesizes reported data on COVID-19 cases and deaths [3, 28], accounting for both underascertainment and time lags. We imputed missing cases and deaths data for Nebraska counties after 30 June 2021 (see Supplementary Methods). We estimated cumulative infections for each US state and 3137 counties from the first reported case date until 1 December 2021. Across all states and counties, dates for the first reported case ranged between 16 December 2019 and 12 November 2020 (interquartile range, 20 February to 7 March 2020). We excluded 6 counties owing to missing or insufficient data.

Vaccinations

We extracted estimates from a repository reporting weekly county-level vaccination coverage based on Centers for Disease Control and Prevention (CDC)–reported data, adjusted for known biases and incompleteness in several states [29, 30]. We imputed missing data and smoothed the weekly time series [31] into a daily time series of residents having received ≥1 vaccine dose (see Supplementary Methods). We summed these counts for all counties within each state to produce state-level estimates. We extracted daily state- and county-level booster coverage data from CDC reports [26]. County-level booster coverage reporting started on 16 December 2021. Booster coverages before this date were imputed proportional to the corresponding state coverage using the ratio of county to state booster coverage on 16 December 2021.

Co-occurrence of Infection and Vaccination

The Census Bureau’s Household Pulse Survey collects data on COVID-19–relevant beliefs and behaviors at 2-weekly intervals, for individuals ≥18 years old [27]. We extracted data from 2 2 February to 30 August 2021, to estimate the joint distribution of infection and vaccination among survey respondents. We extracted the following variables: had COVID (yes/no; whether a respondent has received a positive COVID-19 diagnosis), received vaccine (yes/no; whether a respondent has received ≥1 dose of a COVID-19 vaccine), state, and week. Responses other than yes or no (eg, unknown) were excluded (2.3% of respondents).

Estimation

For each location, we computed the percentage immunologically exposed, defined as the percentage of the population with a prior SARS-CoV-2 infection, ≥1 dose of a COVID-19 vaccine, or both. We calculated values separately for individuals aged <12 or ≥12 years old, owing to differences in vaccine eligibility for these groups over the study period.

Immunological Exposure for the Population Aged ≥12 Years

Using the Household Pulse data, we fit a logistic regression model to estimate the association between self-reported vaccination status and prior COVID-19 diagnosis. We operationalized this relationship as the odds ratio for reported vaccination, comparing individuals reporting a prior COVID-19 diagnosis to those reporting no prior diagnosis (see Supplementary Methods). Using these regression results, we created state-specific prior distributions for the odds ratio of vaccination given prior infection status, for individuals aged ≥12 years (Supplementary Table 1). This approach assumes that the odds ratio for vaccination among those with a prior undiagnosed infection is the same as for those with a prior diagnosed infection. We validated this relationship using data from the Axios-Ipsos Coronavirus Tracker [32]. We calculated the joint probability of being vaccinated or infected as the sum of the marginal probabilities for prior infection and prior vaccination, minus the probability of being both infected and vaccinated, to avoid double counting (see Supplementary Methods).

Immunological Exposure for the Population <12 Years Old

For the population <12 years old, the percentage immunologically exposed was assumed equal to the estimated percentage ever infected, as this age group was not eligible for vaccination during most of the study period. We assumed that infection prevalence in this age group was equal to that in the overall population. We combined immunity estimates for both age groups (<12 and ≥12 years) in a weighted sum to obtain the percentage immunologically exposed in the full population. We validated our results by comparing them with published population immunity estimates based on laboratory data from a blood donor sample [25].

Waning of Protection

Protection conferred by natural infection and vaccination declines over time [24, 33–35]. Findings of prior studies suggest that antibody titers decay rapidly in the 3 months after infection and more gradually thereafter [14, 36]. Neutralizing antibody activity has been observed up to 8 months after symptom onset [12], and simulation studies suggest that titers wane below 1:20 (often used to infer 50% protection) for the majority of previously infected individuals by 341 days after symptom onset [14]. Antibody titers in vaccinated individuals are believed to wane at similar rates [13], although vaccine efficacy against symptomatic SARS-CoV-2 infection has been shown to remain robust in the first 6 months after inoculation [20, 21].

Based on studies of antibody titers, clinical trials, and vaccine effectiveness studies, we formulated 3 simplified waning scenarios, designed to capture major uncertainties in waning rates (Supplementary Figure 1). For the main analysis (base-case scenario), we assumed that infection or vaccination each initially confer 80% protection against infection that declines to 25% by 12 months after exposure, and protection against severe disease starts at 95% and declines to 85% after 12 months. For individuals both infected and vaccinated, we assumed constant protection of 90% against infection and 95% against severe disease. (See Supplementary Methods for optimistic and pessimistic scenarios used in sensitivity analyses.) We assumed that booster uptake was randomly distributed in the eligible (fully vaccinated) population and that receiving a booster restored immunity to original (prewaning) levels and subsequently waned following the “both infected and vaccinated” curve [23, 24, 37].

Immune Escape Under the Omicron Variant

Early evidence indicates the Omicron variant can escape immunity acquired by immunization or infection with earlier variants [4, 38]. Available evidence suggests that the immune escape may range between a 30-fold drop [38] to halving of the protection of earlier variants, with greater protection retained by boosted individuals [22]. Protection against severe disease appears more robust [39]. We translated this evidence into high, medium (used in the main analysis), and low immune-escape scenarios, to capture a simplified yet plausible range of scenarios. In the medium escape scenario, protection against infection was reduced by 70% (40% for those who received a booster) and protection against severe disease was reduced by 20% (15% for those who received a booster) compared with immunity against pre-Omicron variants. (See Supplementary Table 2 for low and high escape scenarios.)

Model Implementation

We executed the analysis with R software [40] and the rstan package [41]. (https://github.com/covidestim/covidestim/tree/immunity-waning). For state-level results, we report uncertainty using equal-tailed 95% credible intervals (95% CrIs). We calculated national estimates and conservative uncertainty intervals by summing state-level estimates and upper and lower bounds of state-level intervals. County-level estimates were produced using an optimization routine [6] that produces point estimates without uncertainty intervals. In summarizing county-level results, we excluded counties with a population <1000 (0.9% of all counties).

RESULTS

By 1 December 2021, 59.2% (95% CrI, 46.9%–75.6%) of the US population was estimated to have been infected with SARS-CoV-2, with state-level estimates ranging from 24.0% (16.0%–40.3%; Hawaii) to 78.5% (68.7%–88.6%; New Mexico). County-level estimates ranged from 9.0% (San Juan County, Washington) to 91.3% (San Juan County, New Mexico). The percentage of the US population that received ≥1 COVID-19 vaccine dose was estimated to be 65.2%. State-level coverage varied between 32.0% (West Virginia) and 82.8% (New Hampshire), and county-level coverage between 13.3% (Morgan County, West Virginia) and 89.9% (Pitkin County, Colorado).

Based on the results of the Household Pulse Survey, individuals reporting a prior COVID-19 diagnosis were substantially less likely to report being vaccinated. The odds ratio of vaccination among individuals with a prior COVID-19 diagnosis (compared with no prior diagnosis) varied from 0.40 (95% CrI, .36–.44) in Florida to 0.58 (.53–.63) in Texas, with a national mean of 0.52 (.50–.55).

Immunological Exposure

The national estimate for the population immunologically exposed was 88.2% (95% CrI, 83.6%–93.5%). State-level estimates ranged from 76.9% (95% CrI, 67.6%–87.6%; West Virginia) to 94.4% (91.2%–97.3%; New Mexico) (Table 1). Across counties, the percentage immunologically exposed ranged from 42.4% (Sioux County, Nebraska) to 98.3% (San Juan County, New Mexico; interquartile range 80.2–87.3%) (Figure 1 and Supplementary Figure 2).

Effective Protection

Accounting for waning of immunity, the percentage of the US population with effective protection against infection with pre-Omicron variants increased from 14.7% (95% CrI, 11.1%–19.8%) on 1 January 2021 to 54.1% (50.5%–59.2%) by 1 December 2021. On 1 December 2021, effective protection against infection with the Omicron variant was estimated to be 21.8% (95% CrI, 20.7%–23.4%). The percentage of the population with effective protection against severe disease was estimated to be 74.1% (95% CrI, 71.4%–77.6%) for pre-Omicron variants and 61.2% (59.1%–64.0%) for the Omicron variant.

Effective protection against infection with Omicron varied across states between 14.4% (95% CrI, 13.2%–15.8%; West Virginia) and 26.4% (25.3%–27.8%; Colorado). Effective protection against severe disease ranged between 53.0% (95% CrI, 47.3%–60.0%; West Virginia) and 65.8% (64.9%–66.7%; Colorado) (Table 1).

Figure 2 shows how state-level percentages of those immunologically exposed, effectively protected against infection, and effectively protected against severe disease have evolved over the course of the epidemic and with the introduction of Omicron. For counties, the percentage of the population with effective protection against infection and severe disease caused by pre-Omicron variants, respectively, varied between 26.6% and 48.4% (Cameron Parish, Louisiana) and 72.5% and 77.4% (Fairfax City, Virginia; Figure 3). Estimates of effective protection against infection and severe disease, respectively, from the Omicron variant varied between 8.4% and 26.4% (McPherson County, Nebraska), and 33.1% and 71.3% (Mineral County, Colorado) (Supplementary Figure 3).

Relative Contributions of Prior Infection and Vaccination

On 1 December 2021, 23.0% (95% CrI, 18.4%–28.3%) of the US population was estimated to have been infected but not vaccinated, 29.0% (18.0%–36.8%) was estimated to have been vaccinated but not infected, and 36.2% (28.4%–47.2%) was estimated to have been both vaccinated and infected. Relative contributions of vaccination and prior infection varied widely across states and counties and over time (Supplementary Figures 4 and 5). Figure 4 highlights regional patterns in the different pathways to overall immunity on 1 December 2021. The state-population weighted regional averages of the population immunologically exposed were 89.1% in the West (20.4% only infected and 30.8% only vaccinated), 84.7% in the Midwest (24.8% and 30.0%, respectively), 88.3% in the South (27.5% and 24.4%), and 91.0% in the Northeast (14.5% and 35.5%).

Validation

We reestimated the odds ratio of vaccination given prior infection using independent survey data, allowing biweekly national estimates between January and June 2021 [42] (Supplementary Methods). For the 13 survey waves included in this period, the odds ratio of vaccination given prior infection varied between 0.35 and 0.98, and the mean odds ratio was 0.51 (95% CrI, .44–.59), similar to the estimated national value in the main analysis. We compared our estimates of the percentage of the population immunologically exposed with blood donor seroprevalence estimates (Supplementary Figure 6). Our estimates of the percentage immunologically exposed were generally lower than seroprevalence estimates from Jones et al [25].

Sensitivity Analyses

We conducted additional analyses evaluating the sensitivity of our effectively protected estimates to waning assumptions, with pessimistic and optimistic scenarios (Supplementary Methods and Supplementary Figure 1), in combination with the Omicron immune-escape scenarios (Supplementary Table 2). In these analyses, national estimates of effective protection against infection from pre-Omicron variants ranged between 47.1% and 64.3%, and protection against infection with Omicron between 12.3% and 28.4% (Supplementary Table 3). Effective protection against severe disease from pre-Omicron variants ranged between 67.5% and 79.0%, and protection against severe disease from the Omicron variant between 57.4% and 61.7% (Supplementary Table 4).

DISCUSSION

We analyzed the joint distribution of COVID-19 vaccination and prior SARS-CoV-2 infection in each US state and county since the beginning of the COVID-19 epidemic and estimated how population immunity changed over this period. By 1 December 2021, more than three-quarters of the US population had prior immunological exposure to SARS-CoV-2 via vaccination or infection; half of the population retained effective protection against infection with previously circulating variants, while only a fifth of the population had effective protection against infection with the Omicron variant.

The current study has several limitations. First, we chose to model separately the populations <12 and ≥12 years old separately. On 1 November 2021, children 5–11 years old became eligible for vaccination; however, we did not account explicitly for the initial vaccination scale-up through 1 December 2021 in this group. For other ages, we used vaccination coverage data from Tiu et al and Merritt et al [29, 30], which endeavor to address known biases in CDC vaccination data. We further adjusted these data to assure that no greater than 100% of the ≥12-year-old population could have been vaccinated by the end of our study period. Furthermore, we assumed the cumulative infections to be proportional between the <12- and ≥12-year-old populations. While some indicators suggest lower cumulative infections among children, other evidence shows the opposite pattern, which contradicts this and suggests a higher seroprevalence for children compared to adults [43–46].

Second, to estimate effective protection, we made assumptions about how natural and vaccine-induced immunity wanes over time. Despite accumulating evidence, these assumptions are still uncertain. In sensitivity analyses, we examined additional waning scenarios, providing a range of plausible values for the level of effective protection. We did not account for differences in waning for the pre-Omicron SARS-CoV-2 variants. Our assumptions regarding the immune escape of the Omicron variant are preliminary. The presented range of plausible scenarios demonstrates that SARS-CoV-2 variants that evade immune protection may spread widely despite high prevalence of prior infection and vaccine coverage.

Third, the model for infections assumes individuals can only be infected once, so possible reinfections and breakthrough infections among vaccinated individuals are not accounted for in our estimates of immunity. For this reason, and because reinfections and breakthrough infections with Omicron are common, we used estimates of infections only up until 1 December 2021.

Fourth, we estimated the relationship between prior infection and vaccination status using survey data that have been criticized for nonrepresentativeness [32]. While this relationship was confirmed in independent survey data validated against external benchmarks [32], it is still possible that reporting biases could have distorted this relationship. If there is greater overlap between vaccinated and previously infected populations, then overall population immunity will be lower than estimated in our analyses. Finally, we assumed that booster uptake was randomly distributed among the eligible (vaccinated) population.

Existing and new SARS-CoV-2 variants will likely continue circulating, since neither natural infection nor offers permanent immunity against infection. Recent CDC recommendations for local COVID-19 monitoring focus on hospitalizations per capita [47]. However, monitoring community outbreaks through signals such as testing volume and surveillance of wastewater data remains important [48, 49]. Estimates of effective protection against infection and severe disease in the population presented in this study provide valuable insight into how to assess local (counties and states) risks in the United States.

In conclusion, as of 1 December 12021, the fraction of the US population that had ever been infected with SARS-CoV-2 and/or received ≥1 dose of a COVID-19 vaccine varied between counties and states. Accounting for waning of population immunity, effective protection against infection by pre-Omicron variants in US states was 27.6%–40.4% lower than the percentage immunologically exposed. Introduction and takeover of the Omicron variant reduced effective protection against infection by another 26.2%–37.0% across US states.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author Contributions. T. C., J. A. S., and N. A. M. and conceived and supervised the project. T. C., J. A. S., and N. A. M. acquired funding. F. K. wrote the model code, drafted the original manuscript, and visualized the results. F. K. and M. R. curated the data and executed the analysis. All authors contributed to the development of the methods and reviewed and edited the original manuscript.

Disclaimer. The findings, conclusions, and views expressed are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC), Council of State and Territorial Epidemiologists, or the National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Financial support. This work was supported by the Centers for Disease Control and Prevention (contract 200-2016-91779; grant 75D30121F0003 to N. A. M.), including through the Council of State and Territorial Epidemiologists (support to F. K. and N. A. S.; grants NU38OT000297-03 to T. C. and N. A. M. and NU38OT000297-02 to J. A. S.); the National Institute of Allergy and Infectious Diseases (support to F. K. and grants R01 AI112438 to T. C., R01 AI137093 to V. E. P., and R01 AI146555-01A1 to N. A. M.); and the National Institute on Drug Abuse (grant 3R37DA01561217S1 to J. A. S.).

References

Author notes