Two analytic periods were used. First, seasonal estimates used the week of October 1, 2023, through the week of April 21, 2024, to align with burden averted estimates of other respiratory viruses.(11) Second, annual estimates utilized an expanded analytic period of September 24, 2023, through the week of August 11, 2024, since SARS-CoV-2 circulates year round. The dates for the annual estimates were chosen based on ACIP recommendation dates: the week of September 24, 2023, was chosen as the start of the analytic period to allow for a minimum of a week after 2023-2024 COVID-19 vaccinations were recommended(12) and the week of August 11, 2024, was the last full week before 2024-2025 COVID-19 vaccinations were approved by the U.S. Food and Drug Administration.(13)

Population data comes from the U.S. Census 2022 national estimates.(14) Age was categorized into the following five groups: 6 months-4 years; 5-17 years; 18-49 years; 50-64 years; and 65 years and older.

Estimates of the national count of COVID-19-associated hospitalizations were derived based on surveillance data of hospitalizations with confirmed SARS-CoV-2 infection collected in the COVID-19–Associated Hospitalization Surveillance Network (COVID-NET) platform.(15) COVID-NET is an active sentinel surveillance system that produces population-based rates of COVID-19-associated hospitalizations among residents of defined surveillance catchment areas across 98 counties in 12 states. COVID-19-associated hospitalizations are defined as those among persons who have received a positive SARS-CoV-2 reverse transcription–polymerase chain reaction (RT-PCR) or rapid antigen detection test result during or within the 14 days preceding hospitalization. Estimates of national COVID-19-associated hospitalizations were extrapolated from data collected in the COVID-NET sentinel surveillance system using a hierarchical Bayesian modeling approach.(16) This modeling approach incorporated temporal trends in the reported counts as well as unmeasured differences among sites reporting data to the COVID-NET system and variation in SARS-CoV-2 testing frequency and accuracy. In addition, the modeling approach enabled estimation of the uncertainty in the estimated age group-specific hospitalization counts due to extrapolation from the COVID-NET catchment areas to the national population.

Using previously described methods,(15) the prevalence of intensive care unit (ICU) admission and in-hospital death were abstracted from medical records from an age- and site-stratified random sample of hospitalizations. The prevalence of receipt of COVID-19 vaccination was obtained from state immunization information systems. Specific measures included the likelihood that a COVID-19-associated hospitalized person received a 2023-2024 COVID-19 vaccination, Pr⁡VH), the probability of a COVID-19-associated ICU admission given a COVID-19-associated hospitalization, Pr⁡ICUH, and the probability of a COVID-19-associated in-hospital death given a COVID-19-associated hospitalization, Pr⁡IHDH, where Pr(⬚) denotes a probability. For sampled data, unweighted case counts and weighted percentages to better represent the hospitalized population of the catchment area are presented. Monthly estimates were used, except in situations where the weighted sample size was less than 20; in those situations, a quarterly estimate was used instead. For a week split across two months, the estimate from the month with the most days in that week was used.

National, weekly 2023-2024 COVID-19 vaccination coverage estimates were from the National Immunization Survey (NIS) for adults (17) and children.(18) NIS defined vaccination coverage with the 2023-2024 COVID-19 vaccination as receipt of at least one COVID-19 vaccination since September 14, 2023. Thus, COVID-19 vaccination coverage does not take into account the second dose of individuals aged 65 years or older who received a second 2023-2024 COVID-19 vaccine doses after ACIP’s recommendation in February 2024.(19) Analyses of Annual estimates carried forward vaccination coverage estimates from the week of May 5, 2024, for the weeks of May 12, 2024, through August 11, 2024.

To estimate the number of COVID-19-associated hospitalizations averted by 2023-2024 COVID-19 vaccination, we compared the total number of people with a COVID-19-associated hospitalization in the presence of 2023-2024 COVID-19 vaccinations (referred to as “factual”) to the hypothetical total number of people with a COVID-19-associated hospitalization without the presence of 2023-2024 COVID-19 vaccinations (referred to as the “counterfactual”) (Figure 1). In mathematical form, this was represented as (1) CHA=CH^−CH

where H denoted a person with a COVID-19-associated hospitalization in the factual condition, H^ indicated a COVID-19-associated hospitalized person in the counterfactual condition, HA represented the number of COVID-19-associated hospitalizations averted, and C(⬚) was a function for the count of people. We categorized the U.S. population into four mutually exclusive groups where V denoted people with a 2023-2024 COVID-19 vaccination, NV signified people without a 2023-2024 COVID-19 vaccination, H is a person with a COVID-19-associated hospitalization, and NH represented people without a COVID-19-associated hospitalization (Figure S1). We made the following assumptions about each population subset:

The only subset of the population which could have contained COVID-19-associated averted hospitalizations was CV,NH, people with a 2023-2024 COVID-19 vaccination who did not have a COVID-19-associated hospitalization. Persons who did not have a 2023-2024 COVID-19 vaccination and a COVID-19-associated hospitalization, CNV,NH, nor those who already experienced a COVID-19-associated hospitalization, CNV,H and CV,H, could, therefore, have a 2023-2024 COVID-19 vaccination-averted COVID-19-associated hospitalization. Thus, our goal was estimating the subset of CV,NH that would have a COVID-19-associated hospitalization in the comparison condition (Figure S2). We can represent this as (2) CHA=CV,NH∗γ

where γ is a probability denoting the likelihood of a person with a 2023-2024 COVID-19 vaccination and no COVID-19-associated hospitalization would have had a COVID-19-associated hospitalization in the comparison scenario. γ is an unknown parameter and we considered multiple approaches for estimating this parameter, details of which are in the supplementary materials. We assumed that γ=PrH∣V in equation (2) and (3) CHA=CV,NH∗PrH∣V,

because people who obtained or with the intention to obtain a COVID-19 vaccination have been more likely to engage in risk prevention behaviors than people without the intention to obtain a COVID-19 vaccination.(20, 21)

Multiple studies indicated that COVID-19 vaccinated people with immunocompromising conditions are more likely to experience a SARS-CoV-2 or COVID-19 outcome than COVID-19 vaccinated people without immunocompromising conditions.(22–28) Our approach modified formula (3) by incorporating dependence of this probability on immunocompromise status, i.e., (4) CHAIC=CIC,V,NH∗PrH∣IC,V+CNIC,V,NH∗PrH∣NIC,V,

where IC and NIC indicated immunocompromised and not immunocompromised, respectively, and CHAIC were the COVID-19-associated hospitalizations averted by 2023-2024 COVID-19 vaccination accounting for immunocompromise condition (Figure S3). We were unable to find national estimates of PrH∣IC,V and PrH∣NIC,V, meaning we relied on estimates outside the U.S. or on subsets of the U.S. population.(26–28) To account for differences in PrH∣V between data sources, we adapted equation (4) to (5) CHAIC= CIC,V,NH∗PrH∣V∗Pr⁡HA∣ICA,VAPr⁡HA∣VA+CNIC,V,NH∗PrH∣V∗Pr⁡HA∣NICA,VAPr⁡HA∣VA,

where the subscript A represented an alternate data source. Probabilities of COVID-19-associated hospitalization of people with a 2023-2024 COVID-19 vaccination with and without immunocompromising conditions were taken from multiple sources.(26–28) (Table S1) For a data source to be included, the source needed CVA and CHA,VA to estimate Pr⁡HA∣VA, CNICA,VA and CHA,NICA,VA to estimate Pr⁡HA∣NICA,VA, and CICA,VA and CHA,ICA,VA to estimate Pr⁡HA∣ICA,VA. Results presented are from Evans et al. (28) with results from other sources (26, 27) included in the supplementary materials.

The number of COVID-19-associated ICU admissions averted was calculated using the following formula (6) CIA=CHA∗Pr⁡ICUH

where CIA is the number of COVID-19-associated ICU admissions averted and Pr⁡ICUH is the probability of an COVID-19-associated ICU admission given a COVID-19-associated hospitalization. The number of COVID-19-associated in-hospital deaths were estimated similarly, (7) CDA=CHA∗Pr⁡IHDH

where CDA is the number of COVID-19-associated in-hospital deaths averted and Pr⁡IHDH is the probability of an COVID-19-associated in-hospital death given a COVID-19-associated hospitalization.

The supplementary materials also include sensitivity analyses to the parameter values of the averted burden estimates. In sum, burden averted results were sensitive to changes in key parameter values except for the probability of COVID-19 vaccination and, when considering immunocompromising conditions, the probability of immunocompromise (Figures S4, S5). For the former, this is due to the implicit assumption that parameters depend on each other, i.e., changing PrV will affect other parameters, such as PrH and PrH∣V. If changes in the burden averted arose directly from PrV, we might be double counting averted burden due to PrV. Additionally, the direct effect of changes to PrV may be diminished because PrH∣V is used in both the calculation of CV,NH and CHA.

We utilized a modified formula from the ACIP benefit-risk assessment framework for COVID-19 vaccine policy decisions (7) to calculate COVID-19-associated burden averted estimates. The original formula calculated the COVID-19-associated hospitalizations prevented as the product of the COVID-19 hospitalization rates, analytic time horizon, the number of COVID-19 vaccination doses administered, and an estimate of the VE against COVID-19 hospitalization. We modified this formula to (8) CHAB=PrH∗CV∗VE

where CV is the number of COVID-19 vaccinated people and VE was an estimate of the vaccine effectiveness against COVID-19-associated hospitalization of the 2023-2024 COVID-19 vaccination compared to people without a 2023-2024 COVID-19 vaccination. Results from equation (8) were compared to results from equation (3). Estimates of VE against COVID-19-associated hospitalization of the 2023-2024 COVID-19 vaccination compared to people without a 2023-2024 COVID-19 vaccination were taken from the VISION platform.(29)

Monte Carlo simulations were used to generate the number of COVID-19 hospitalizations, 2023-2024 COVID-19 vaccination coverage, the probability that a COVID-19-associated hospitalized person received a 2023-2024 COVID-19 vaccination, the probability of a COVID-19-associated ICU admission given a COVID-19-associated hospitalization, the probability of a COVID-19-associated in-hospital death given a COVID-19-associated hospitalization, and VE against COVID-19-associated hospitalization. Data were generated by age group and week for COVID-19 hospitalizations and 2023-2024 COVID-19 vaccination coverage, monthly for all probabilities, and seasonal for VE. Data were generated from multivariate standard Gaussian distributions based on estimated means and standard deviations. A correlation matrix was created to incorporate potential correlations within and between data elements (Table S2). 10,000 Monte Carlo simulations were generated. Full details about the estimation of averted COVID-19-associated burden due to 2023-2024 COVID-19 vaccinations including the assumed correlations among modeling elements and logical values used for truncation are presented in the supplementary materials.

For each simulation, the estimated number of COVID-19-associated hospitalizations was converted to PrH by dividing by the census population. Then, using the calculated PrH and the simulated values of PrV and Pr⁡VH), PrH∣V was calculated using Bayes Theorem. For each simulation, CV,NH was calculated by the product of the population size, PrV, and 1−PrH and CHA from the formula above. Age group estimates were created by summing across all months for an age group by simulation and then computing the median of 10,000 simulations. Uncertainty intervals were the 2.5^th and 97.5^th quantiles of the 10,000 simulations. Overall estimates were computed by summing over all months and age groups by simulation and are summarized by the median and 2.5^th and 97.5^th quantiles.

Additional sensitivity analyses for averted COVID-19-associated hospitalizations, ICU admissions, and in-hospital deaths were performed. For analyses without accounting for immunocompromising conditions, we varied the correlation structure, correlation parameter, and the likelihood that people with a 2023-2024 COVID-19 vaccination and no COVID-19-associated hospitalization would experience a COVID-19-associated hospitalization in the counterfactual condition. Other parameters varied included the probability of a COVID-19-associated hospitalization in the counterfactual (see supplemental materials for details). For analyses that accounted for immunocompromising conditions, we varied the correlation structure, correlation parameter(s), and the probability of COVID-19-associated hospitalization of people with a 2023-2024 COVID-19-associated vaccination with and without immunocompromising conditions.(26–28)

Seasonal COVID-19 hospitalizations averted by 2023-2024 COVID-19 vaccination without accounting for immunocompromising conditions was 68,315 (95% uncertainty interval [UI]: 42,831 – 97,984) or 98 COVID-19-associated hospitalizations averted per 100,000 2023-2024 COVID-19 vaccinated people (95% UI 61 - 141) (Figure 2 panel A, Table S3). The annual estimate of hospitalizations averted was 107,197 (95% UI 80,692 – 137,643) (Figure 2 panel B, Table S4).

The seasonal COVID-19-associated hospitalizations averted by 2023-2024 COVID-19 vaccination was 724 (95% UI 177 – 1,527) for people aged 6 months to 4 years, which was approximately 71 (95% UI 17 - 156) COVID-19-associated hospitalizations averted per 100,000 2023-2024 COVID-19 vaccinated people aged 6 months to 4 years. COVID-19-associated hospitalizations averted by 2023-2024 COVID-19 vaccination was 453 (95% UI 44 – 1,917) for people aged 5 to 17 years, 2,467 (95% UI 992 - 4,197) for people aged 18-49 years, 6,770 (95% UI 2,564 - 11,761) for people aged 50-64 years, and 57,665 (95% UI 35,442 - 84,006) for people aged 65 years and older. COVID-19-associated hospitalizations averted per 100,000 people with a 2023-2024 COVID-19 vaccination was lower for people aged 6 months to 4 years compared to all age groups except people aged 65 years and older.

After accounting for immunocompromising conditions, estimated COVID-19 hospitalizations averted by 2023-2024 COVID-19 vaccination was 58,540 (95% UI 36,669 - 83,987) (Table S5). This was approximately 14% fewer COVID-19-associated hospitalizations averted compared to the results that did not account for people with immunocompromising conditions. Age group estimates accounting for immunocompromising conditions ranged from 11% to 17% fewer COVID-19-associated hospitalizations averted compared to results that did not account for people with immunocompromising conditions. Similar differences were found for annual estimates after accounting for immunocompromising conditions (Table S6).

An estimated 13,108 (95% UI 4,459 – 25,018) seasonal COVID-19-associated ICU admissions were averted by 2023-2024 COVID-19 vaccination without accounting for immunocompromising conditions (Figure 3 panel A, Table S7). This translated to approximately 19 COVID-19-associated ICU admissions averted per 100,000 2023-2024 COVID-19 vaccinated people (95% UI 6 - 36). Our annual estimate was 18,292 COVID-19-associated ICU admissions averted (95% UI 10,062 – 28,436) (Figure 3 panel B, Table S8).

Age-based, seasonal estimates were 175 (95% UI 38 - 418) COVID-19-associated ICU admissions averted for people aged 6 months to 4 years, 93 (95% UI 3 - 392) for people aged 5 to 17 years, 307 (95% UI 39 - 776) for people aged 18 to 49 years, 1,499 (95% UI 321 – 3,503) for people aged 50 to 64 years, and 10,878 (95% UI 3,104 – 21,591) for people aged 65 years and older (Table S7). Seasonal COVID-19-associated ICU admissions averted by 2023-2024 COVID-19 vaccination accounting for immunocompromise status was 11,237 (95% UI 3,815 – 21,461) (Table S9) and annual ICU admissions averted were 15,673 (95% UI 8,629 - 24,386) (Table S10).

Without accounting for immunocompromising conditions, a seasonal estimate of 5,301 (95% UI 101 – 14,230) COVID-19-associated in-hospital deaths were averted by 2023-2024 COVID-19 vaccination (Figure 4 panel A, Table S11); annually, that estimate increases to 6,749 in-hospital deaths (95% UI 2,077 - 13,557) (Figure 4 panel B, Table S12). Seasonally, COVID-19-associated in-hospital deaths averted were 2 (95% UI 0 - 17) for people aged 6 months to 4 years, 6 (95% UI 0 - 59) for people aged 5 to 17 years, 36 (95% UI 0 - 185) for people aged 18 to 49 years, 412 (95% UI 0 - 1,447) for people aged 50 to 64 years, and 4,779 (95% UI 0 - 13,132) for people aged 65 years and older (Table S11). After accounting for immunocompromising conditions, the model estimated 4,538 (95% UI 88 – 12,166) COVID-19-associated in-hospital deaths were averted by 2023-2024 COVID-19 vaccination (Table S13) or 5,774 (95% UI 1,776 - 11,613) COVID-19-associated in-hospital deaths averted annually (Table S14).

Point estimates for annual COVID-19-associated hospitalization, ICU admissions, and in-hospital deaths using the conditional probability multiplier model were lower than estimates generated from the formula adapted from the ACIP benefit-risk assessment framework (107,197, 95% UI 80,692 - 137,643, versus 116,026, 95% UI 93,253 - 141,078 when using the benefit-risk formula), but the uncertainty intervals overlapped between the two methods (Table S15).

In formula (2), the number of hospitalizations averted is the product of the number of people with a 2023-2024 COVID-19 vaccination who were not hospitalized and a probability. For our main analyses, we chose that probability to be the probability that people with a 2023-2024 COVID-19 vaccination were hospitalized. In sensitivity analyses, we found the choice of that probability affected the averted COVID-19-associated hospitalization burden (Table S16), with most probabilities resulting in a higher averted COVID-19-associated hospitalization burden compared to the benefit-risk assessment framework formula. Otherwise, the correlation structure and assumed correlation values (ρ) had minimal impact on the point estimates of COVID-19-associated hospitalizations, ICU admissions, and in-hospital deaths averted (Tables S17–S22). For most probability values, the uncertainty interval width increased with increasing correlation for both the autoregressive and compound symmetric correlation structures. When accounting for immunocompromising conditions, there were large differences in the point estimates when using different data sources (Tables S18, S20, S22) with the values from Evans et al. (28) providing the highest averted burden of COVID-19-associated hospitalizations, followed by Suleyman et al. (27), and Smits et al. (26) with the lowest, approximately 45% less than Evans et al. (28)