ASSISTANT SECRETARY FOR December 2021 COVID-19 Vaccination Associated with Reductions in COVID-19 Mortality and Morbidity in the United States, and an Approach to Valuing these Benefits Key Points e COVID-19 vaccines have been crucial in mitigating adverse public health and economic impacts of the COVID-19 pandemic, but uptake has varied over numerous domains, including demographic and geographic factors. e Using a regression model and county-level data on vaccination rates, cases, hospitalizations, and deaths, we estimate that COVID-19 vaccination was associated with reductions of 25.32 million cases, 1.38 million hospitalizations, and 213,000 deaths in the United States between December 2020 and July 2021. e Oursensitivity analyses provide a range of estimates for cases (15.70 million to 27.97 million), hospitalizations (1.10 million to 1.86 million), and deaths (178,000 to 422,000). e We value the health benefits of COVID-19 vaccination based on individual willingness to pay (WTP) to reduce one's own morbidity and mortality risks.1 We estimate a total value of COVID-19 risk reductions attributable to vaccination ranging between $1.38 trillion and $4.49 trillion. Using the mid-point value of the willingness to pay, the total health benefits are estimated at $2.95 trillion, including reductions in deaths (mid-point $2.45 trillion; range: $1.15 trillion to $3.74 trillion), hospitalizations (midpoint: $354.14 billion; range: $165.27 billion to $539.08 billion), and cases (midpoint: $139.96 billion; range: $65.32 billion to $213.05 billion). INTRODUCTION On March 11, 2020, the World Health Organization (WHO) declared the novel coronavirus disease 2019 (COVID-19) outbreak a global pandemic and called on countries to take action to 1 In economics, WTP is the maximum value, usually price, an individual is willing to give in exchange for a product or service. In this context, itis the maximum amount of money an individual would willingly exchange to reduce one's risk of getting infected with COVID-19, being hospitalized with COVID-19 ordying of COVID-19, while reducing his or her ability to purchase other things. December 2021 RESEARCH REPORT 1 contain the virus.2 The rapid development of COVID-19 vaccines was seenas crucial to mitigating the public health and economic impacts of the COVID-19 pandemic. By the end of March 2020, with the launch of the first clinical trials in the United States, the race was on to accelerate the development of a COVID-19 vaccine.2;4 On December 11, 2020, the Food and Drug Administration issued an emergency use authorization for the first COVID-19 vaccine in the United States. As of November 2, 2021, three COVID-19 vaccines are now available in the United States. The rapid development of COVID-19 vaccines is the result of various collaborations and massive investments (at least $10 billion) from the federal governmentin research and development, manufacturing capacity, and purchasing.>-&78 The COVID-19 pandemic has had and continues to have widespread impacts on Americans. Data from the Centers for Disease Control and Prevention (CDC) show that between January 21, 2020 and November 2, 2021, there have been over 46 million COVID-19 laboratory-confirmed cases and over 700,000 deaths.2 The number of reported cases, deaths, and hospitalizations peaked on January 12, 2021, when CDC data showed that there were there were 4,103 new COVID-19 deaths, over 123,000 new COVID-19 hospitalizations, and 210,000 new COVID-19 confirmed cases (see Figure 1). 1° 2 AP News. WHO declares coronavirus a pandemic, urges aggressive action. Available at: 52e 12ca90c55b6e0¢398d 134a2cc286e, last accessed October 15, 2021. 3 NIH. NIH clinical trial of investigational vaccine for COVID-19 begins. Available at: httos://www.nih.gov/news- begins#:~: 'text=A%20Phase%201%2 Oclinical%2 0trial, Institute%201 KPWHRI)%20in9620Seatt! , last accessed October 15,2021. 4 GAO. COVID-19: Federal efforts accelerate vaccine and therapeutic development, but more transparency needed on emergency use authorizations. Available at: https://www.gao.gov/products/ga0-21-207, last accessed October 15,2021. 5 The White House. Fact Sheet: Biden Administration announces historic $ 10 billioninvestment to expand access to COVID-19 vaccines and build vaccine confidence in hardest-hit and highest-risk communities. Available at: https://www.whitehouse.gov/briefing-room/statements-releases/2021/03/25/fact-sheet-biden-administration- announces-historic-10-billion-investment-to-expand-access-to-covid-19-vaccines-and-build-vaccine-confidence-in- hardest-hit-and-highest-risk-communities/, last accessed October 15, 2021. 6 Congressional Research Service. Domestic Funding for COVID_19 Vaccines: An Overview. Available at: https://crsreports.congress.gov/product/pdf/IN/IN11556, last accessed October 15, 2021. 7 Health Affairs. It was the government that produced COVID-19 vaccine success. Available at: https://www.healthaffairs.org/do/10.1377/hblog202 105 12.191448 /full/, last accessed October 15, 2021. 8 GAO. Operation Warp Speed: Accelerated COVID-19 vaccine development status and efforts to address manufacturing challenges. Available at: https://www.gao.gov/assets/gao-2 1-319.pdf, last accessed October 15, 2021. 2 CDC. COVID Data Tracker: United State COVID-19 cases, deaths, and laboratory testing (NAATs) by state, territory, and jurisdiction. Available at: https://covid.cdc.gov/covid-data-tracker/H#cases_ deathsper100klast7days, last accessed November 3, 2021. 10 CDC. COVID Data Tracker: United State COVID-19 cases, deaths, and laboratory testing (NAATs) by state, territory, and jurisdiction. Available at: https://covid.cdc.gov/covid-data- tracker/#trends dailycases_currenthospitaladmissions, last accessed October 15, 2021. December 2021 RESEARCH REPORT 2 Figure 1. COVID-19 Cases, Hospitalizations, Deaths, and Vaccination Rates Over Time Panel A. COVID-19 Cases and Vaccination Coverage 2° So I Start of Vaccination Lo So © a 5 oe LO | . OF = 8 - oO 3 a § & 31 © eal Log ES SG - oO ® c B Ss o 2 3 oO oO | | © B as o-+ 7 P'S T T T T January 1, 2021 April 1, 2021 July 1, 2021 October 1, 2021 Date Cases | | Vaccination Coverage Panel B. COVID-19 Hospitalizations and Vaccination Coverage el Start of Vaccination Lo © a Lo © 100 1 40 Hospitalizations (Thousands) 50 Vaccination Coverage (Percent) 20 T T T T January 1, 2021 April 1, 2021 July 1, 2021 October 1, 2021 Date Hospitalizations -_[ Vaccination Coverage December 2021 RESEARCH REPORT 3 Panel C. COVID-19 Deaths and Vaccination Coverage LO _| Start of Vaccination = N co S | Loe Oo 2 = o em & Bo 2, 2 fw Loo E +3 o Oo Sol c os 2 o = a 2 8 Lo as LO + of - Oo T T T T January 1, 2021 April 1, 2021 July 1, 2021 October 1, 2021 Date Deaths Vaccination Coverage Notes: Vaccination data represents the percentage of the 18 years and older population that is fully vaccinated. Data for this metric are not available prior to March 2021, but the red line denotes when COVID-19 vaccinations began in the United States. COVID-19 outcome variables are presented asthe number of new COVID-19 cases, hospitalizations, or deaths in a given week, as reported in the CDC COVID Data Tracker. Source: ASPE analysis of multiple data sources for December 2020-November 2, 2021 From January until July of 2021, new cases, hospitalizations, and deaths declined following the introduction of vaccines. Over that time period, there was significant progress in increasing COVID-19 vaccine availability and COVID-19 vaccination coverage of the population. As of November 2, 2021, 80 percent of the US population ages 18 years and older had received at least one dose of the vaccine (67 percent of the total US population).*4 In July 2021, the highly transmissible Delta variant became the predominant variant in the United States, which subsequently caused a new wave of cases, hospitalizations, and deaths. These new cases, hospitalizations, and deaths were concentrated primarily among the unvaccinated, further highlighting the importance of increased COVID-19 vaccination coverage. '2 11 CDC. COVID Data Tracker: COVID-19 Vaccinations in the United States. Data as of October7, 2021. Available at: ://covid.cdc.gov/covid-data-tracker/Hvaccinations vacc-total-admin-rate-total, last accessed November 2, 2021. 12 CDC. Delta variant: What we know about the science. Available at: https://www.cdc.gov/coronavirus/2019- ncov/variants/delta- variant.html?s cid=11609:is%20there%20a%20vaccine%20for%20delta%20variant:sem.ga:p:RG:GM:gen:PTN.Gran ts:FY22, last accessed October 15, 2021. December 2021 RESEARCH REPORT 4 In this study, we examine associations between vaccination rates and COVID-19 cases, hospitalizations, and deaths across counties in the United States from December 2020 to July 2021. Specifically, we use a regression model to estimate the association between increased COVID-19 vaccine coverage and reductions in cases, hospitalizations, and deaths. In addition, we use estimates of willingness to pay (WTP) for reductions in health risks to value the improvement observed in COVID-19 outcomes. In economics, WTP is the maximum value, usually price, an individual is willing to give in exchange for a product or service. In this context, it is the maximum amount of money an individual would willingly exchange to reduce one's risk of getting infected with COVID-19, being hospitalized with COVID-19, or dying of COVID-19, while reducing an individual's ability to purchase other things. This valuation approach is widely used in benefit-cost analysis.13-14 METHODS AND DATA Data We employ multiple databases that capture county-level COVID-19 vaccination rates and COVID-19-related outcomes: cases, hospitalizations, and deaths, as discussed below. The data are aggregated on a weekly basis (from Friday to Thursday) from December 4, 202015 through July 29, 2021; for brevity we referto our period of analysis to be between December 2020 and July 2021. Appendix 1 provides additional information on the frequency and sources of the data. e COVID-19 Vaccine Administration Data: We use county-level data on vaccine doses administered and reported to the CDC.1° The county information represents the county in which the individual resides.?' This variable is defined as the percent of the population ages 18 years and older that is fully vaccinated*® against COVID-19 as of the reported week in a given county. During our review of the data we identified five states (Georgia, Hawaii, Texas, Virginia, and West Virginia) that either did not report county- 13 Office of Management and Budget, Circular A-4, September 17, 2003. Available at: https://obamawhitehouse.archives.gov/omb/circulars_ a004_a-4/, last accessed November 1, 2021. 14 .S. Department of Health and Human Services. Guidelines for Regulatory Impact Analysis. 2016. Available at: aspe.hhs.gov/sites/default/files/private/pdf/242926/HHS RlAGuidance.pdf, last accessed November 1, 2021. 45 Data for COVID-19 vaccinations begin on December 10, 2020. 16 CDC. COVID-19 Vaccinations in the United States, County. Available at: https: //data.cdc.gov/Vaccinations/COVID-19-Vaccinations-in-the-United-States-Coun October 22,2021. 17 CDC. Reporting County-Level COVID-19 Vaccination Data. Available at https://www.cdc.gov/coronavirus/2019- ncov/vaccines/distributing/reporting-counties.html, last accessed October 21, 2021. 18 In this dataset, "fully vaccinated" refers to those who have received the second dose in a two-dose COVID-19 vaccine series or one dose of the single-shot Johnson and Johnson's Janssen COVID-19 vaccine. 8xkx-amgh, last accessed December 2021 RESEARCH REPORT 5 level vaccinations or for which >40% of vaccinations lacked county information. We use state-level vaccination data for counties in these states in order to include them in our analysis.?9 e COVID-19 Cases: We utilize COVID-19 case data collected by Johns Hopkins University's Centerfor Systems Science and Engineering's COVID-19 Data Repository.2°21 The data in our analysis include confirmed, and where reported, probable COVID-19 cases and include all age groups. This variable is defined as new COVID-19 cases per 10,000 population in a given week and county. e COVID-19 Hospitalizations: Facility-level hospital utilization data were obtained from HealthData.gov.22 These data represent facility-level reporting via HHS TeleTracking or reporting provided directly to HHS Protect by state/territorial health departments.23 These data are aggregated on a weekly basis at the county level and represent adult inpatient beds. This variable is defined as total laboratory-confirmed or suspected COVID-19 hospitalizations per 10,000 population in a given week and county. e COVID-19 Deaths: Data on COVID-19 deaths were gathered from Johns Hopkins University's Centerfor Systems Science and Engineering's COVID-19 Data Repository.2*25 These data are aggregated on a weekly basis at the county level and include all age groups. The variable is defined as new COVID-19 deaths per 10,000 population in a given week and county. We encountered data quality issues with the vaccination and outcomes data. These included systematically missing data for certain counties (including counties in California, Utah, Massachusetts, Alaska, and New York), states (Texas, Hawaii) or for certain periods of analysis (Nebraska). It also included data that our analysis suggested may not have been consistently reported or reliable, e.g., counties in Georgia, West Virginia, and Virginia. Addressing these 19 CDC. COVID-19 Vaccinations in the United States, Jurisdiction. Available at: https://data.cdc.gov/Vaccinations/COVID-19-Vaccinations-in-the-United-States-Jurisdi/unsk-b7fc, downloaded on Octobers, 2021. 20 COVID-19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University, https: //github.com/CSSEGISandData/COVID-19, downloaded on August 18, 2021. 21 Dong, E., Du, H., Gardner. (2020). An|Interactive Web-based Dashboard to Track COVID-19 in Real Time. The Lancet, 20(5): 533-534. https://doiorg/10.1016/S1473-3099(20)30120-1 22 COVID-19 Reported Patient Impact and Hospital Capacity by Facility, htips://healthdata.gov/Hospital/COVID-19- Reported Patient-Impact-and-Hospital-Capa/anag-cw7u, downloaded on August 18, 2021. 23 The hospital populationincludes all hospitals registered with Centers for Medicare & Medicaid Services (CMS) as of June 1, 2020. It includes non-CMS hospitals that have reported since July 15, 2020. The data do not include psychiatric, rehabilitation, Indian Health Service (IHS) facilities, U.S. Department of Veterans Affairs (VA) facilities, Defense Health Agency (DHA) facilities, and religious non-medical facilities. 74 COVID-19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University, https: //github.com/CSSEGISandData/COVID-19, downloaded on August 18, 2021. 25 Dong, E., Du, H., Gardner. (2020). AnInteractive Web-based Dashboard to Track COVID-19 in Real Time. The Lancet, 20(5): 533-534. https://doi.org/10.1016/S1473-3099(20)30120-1 December 2021 RESEARCH REPORT 6 issues involved removing data from our analysis or imputing values where appropriate. Appendix 2 provides more information on these issues and how they were addressed. In addition to our COVID-19 metrics, our analysis includes other time-varying information that may be associated with intentions to vaccinate and COVID-19 outcomes of interest. These data include the status of state-level mask mandates or stay-at-home orders, county-specific data on worker mobility, and region-specific information on the prevalence of the Delta variant. More information about these metrics, including data sources, is provided in Appendix 1. Examining Associations between COVID-19 Vaccination and Outcomes We examine the association between vaccination rates and three separate outcomes: COVID- 19 cases, hospitalizations, and deaths. We do this in four steps. First, we use a Poisson regression model to estimate the association between vaccination rates and each of our three outcomes: cases, hospitalizations, or deaths.76 There are five key predictors: county-level COVID-19 vaccination rates in the current week and for each of the four preceding weeks (i.e., four weeks before, three weeks before, two weeks before, and one week before). These measures of vaccination rates are intended to reflect the time that it may take to achieve maximum protection from the vaccine. The model includes variables that can explain changes in vaccination rates and COVID-19 outcomes. Specifically, we include state-level changes in COVID-19 mitigation policies such as mask mandates, county-level mobility patterns, and region-level information on the proportion of cases that are attributed to the Delta variant. Our Poisson model includes county fixed effects to control for differencesin counties' unobserved COVID-19 mitigation efforts and other unmeasured demographic and health-related differences that are stable within counties over the period of analysis. The model also includes quarterly fixed effects to control for nationwide temporal trends and fluctuations in COVID-19 outcomes over the period of study. Second, using this relationship we generate predicted values for each of our outcomes. Third, we setthe vaccination rates for all counties in all weeks equal to zero, and then generate new predicted values, which we referto as our counterfactual values. Lastly, we subtract our predicted values from our counterfactual values. These resulting differences are our estimates of associated reductions in cases, hospitalizations, and deaths. We also conduct additional sensitivity analyses with varying specifications and samples. These and additional information on our empirical approach are discussed in Appendix 2. 26 We examine other functional forms, including a linear model, and present these results in Ap pendix 2. December 2021 RESEARCH REPORT 7 Our statistical approach is similar to that of Gupta et al?" with a few deviations to allow for estimates at a smaller geographical scale. Specifically, we use county-level data rather than state-leveldata to enhance the variation in the data and facilitate analysis of the associated reductions at a more disaggregate measure. Secondly, we use the incidence of cases, hospitalizations, or deaths per 10,000 population reported each week, rather than the cumulative measures, to use the variation over time and across counties. Our model includes quarterly fixed effects rather than weekly fixed effects to avoid diluting the variation that we try to use. Finally, we include a set of region- and state-level variables that change over time to capture changes in mitigation policies and the proportion of cases from the Delta variant. Valuing Morbidity and Mortality Risk Reductions In benefit-cost analysis, the value of an improvement in health, such as a reduction in mortality or morbidity risk, is typically based on individual willingness to pay (WTP). This value is "derived from how much money affected individuals would exchange for a risk reduction they expect to experience, giventheir budget constraints and preferences for spending on other goods and services."28 The WTP estimates apply an estimate of the value per statistical life (VSL)22 when valuing expected changes in mortality risks and apply a value per statistical case (VSC) when valuing non-fatal risk reductions (cases, hospitalizations). HHS guidelines recommend VSL values ranging from $5.36 million to $17.51 million. This range of estimates is based ona literature review tailored to the types of risks HHS regulates. ASPE considered several factors specific to COVID-19 risks that may affect the VSL estimates, such as the age profile of fatal cases and the substantial morbidity experienced prior to death, and concluded that the same range of estimates were appropriate for COVID-19.2° We adopt estimates per VSC that range from $2,728 to $8,900 for mild cases, from $6,115 to $19,947 for severe cases, and from 27 Gupta, S., Cantor J., Simon, K., Bento, A., Wing, C.,and Whaley, C. Vaccinations Against COVID-19 May Have Averted Up to 140,00 Deathsin The United States, Health Affairs, 2021 40(9): 1465-1472 https://doi.org/10.1377/hlthaff.2021.00619 28 Robinson, L., Eber, M., Hammitt, J. (2021). Valuing COVID-19 mortality and morbidity risk reductions in U.S. Department of Health and Human Se rvices Regulatory |mpact Analyses. Available at: -covid-19-risk-reductions-hhs-rias, last accessed October 22, 2021. 29 VSLi isnot the value the analyst, the researcher, orthe government places on saving an individual from certain death, but the extent to which individuals are willing to exchange money for small changes in their own risks within a definedtime period. 30 See Table ES.2 (pg. 7) for factorsconsideredin Robinson, L., Eber, M., Hammitt, J. (2021). Valuing COVID-19 mortality and morbidity risk reductions in U.S. Department of Health and Human Services Regulatory Impact Analyses. Available at: https://aspe.hhs.gov/reports/valuing-covid-19-risk-reductions-hhs-rias, last accessed October 22, 2021. https://aspe.hhs.gov/sites/default/files/2021-08/valuing-covid-risks-july-2021.pdf December 2021 RESEARCH REPORT 8 $846,720 to $2.76 million for critical cases from a recent ASPE report on valuing COVID-19 risk reductions.31 To monetize the value in risk reductions in mortality and morbidity we multiply the VSL or VSC with the corresponding number of reductions in mortality or morbidity. Further, we use the mid-point value of the estimated VSL or VSC for our primary analyses but also consider the low and high estimates as part of our sensitivity analysis. Before valuing morbidity risk reductions, we subtract the number of deaths and hospitalizations from the number of cases to avoid double-counting.?2 Similarly, to avoid double-counting hospitalizations, we subtract mortality from hospitalizations that required intensive care unit admission. FINDINGS Estimated Number of Associated Reductions in Cases, Hospitalizations, and Deaths Table 1 shows the total reported number of COVID- 19 cases, hospitalizations, and deaths between Our model estimates that COVID-19 December 2020 and July 2021, compared with our -_-syqccination was associated with model estimates with and without vaccinations. The reductions of 25.32 million cases, first row in Table 1 presents the total number of 1.38 million hospitalizations, and cases, hospitalizations, and deaths during the period 213,000 deaths. of analysis. The second row in Table 1 presents the number of cases, hospitalizations, and deaths estimated by the model based on the observed association between ourthree COVID-19 outcomes, vaccination, and other covariates. The third row shows the estimated total for each COVID-19 outcome as if no vaccines were available, the counterfactual scenario. The results show that without the administration of vaccines, cases may have more than doubled, and hospitalizations and deaths may have increased by two thirds. To obtain an estimate of the total number of associated reductions in COVID-19 outcomes, we calculate the difference between our estimates without vaccines (Table 1, row 3) and with vaccines (Table 1, row 2). 31 Robinson, L., Eber, M., Hammitt, J. (2021). Valuing COVID-19 mortality and morbidity risk reductions in U.S. Department of Health and Human Services Regulatory|mpact Analyses. Available at: https: //aspe.hhs.gov/reports/valuing-covid-19-risk-reductions-hhs-rias, last accessed October 22, 2021. 32 National Center for Health Statistics. In-Hospital Mortality Among Hospital Confirmed COVID-19 Encounters by Week from Selected Hospitals. Available at https: //www.cdc.gov/nchs/covid19/nhcs/hospital-mortality-by- week.htm; last accessed October 30, 2021. December 2021 RESEARCH REPORT 9 Our results suggest that the introduction of COVID-19 vaccinations was associated with reductions of 25.32 million cases, 1.38 million hospitalizations, and 213,000 deaths.33 Table 1. Number of Associated Reductions in COVID-19 Cases, Hospitalizations, and Deaths from December 2020 to July 2021 Cases Hospitalizations Deaths Reported 20,226,104 2,085,410 331,151 Estimated with 20,193,720 2,080,594 330,748 Vaccines No Vaccines 45,518,022 3,464,854 544,097 (Counterfactual) Associated Reductions! 25,324,302 1,384,260 213,349 Minimum2 15,708,113 1,103,825 178,348 Maximum2 27,970,049 1,860,893 421,874 Notes: 1. Estimates from Poisson regression model that includes the following variables: vaccination rate for a given week and for each of the prior four weeks, state-level information about stay at home orders, state-level information on mask mandates, a county-level measure of worker mobility, and proportion of COVID-19 cases from the Delta variant across ten regions in the United States. The regression model also includes county and quarter fixed effects. See Appendix 1 and Appendix 2 for additional information on the model and the data. 2. See Table S-3 and Table S-4 for the specification and samples used to obtain the minimum and maximum number of associated reductions. Source: ASPE analysis of multiple data sources for December 2020-July 2021. We also estimate our model using other functional forms and samples (see Table S-3 and Table S-4). These estimates range from 15.71 million to 27.97 million cases, from 1.10 million to 1.86 million hospitalizations, and from 178,000 to 422,000 deaths. Our preferred specification is a Poisson model with our full set of controls and all states included in our sample. We prefer using a Poisson model because we believe it better captures the nonlinearity of vaccination rates and spread of COVID-19. Figure 2 displays the cumulative number of associated reductions in COVID-19 outcomes over time. The estimated reduction in cases, hospitalizations, and deaths substantially increased week after week (see also Table S-5), with impacts already evident within the first several weeks of the vaccine rollout. 33 We also compared the reported number of cases (Figure S-1), hospitalizations (Figure S-2) and deaths (Figure S- 3) to their corresponding predicted and counterfactual estimates over time. Overall, the trends fall below the reported values and the model tracks the trends more closely for hospitalizations and deaths than cases. December 2021 RESEARCH REPORT 10 Figure 2. Cumulative Number of Associated Reductions in COVID-19 Outcomes Over Time Panel A. Cases Cases (Thousands) 15000 20000 25000 L | l 10000 ! 5000 o- T T T T T December 1, 2020 February 1, 2021 April 1, 2021 June 1, 2021 August 1, 2021 Date Panel B. Hospitalizations and Deaths 1000 1500 i Hospitalizations or Deaths (Thousands) 500 L oO- T T T T T December 1, 2020 February 1, 2021 April 1, 2021 June 1, 2021 August 1, 2021 Date GE Hospitalizations [EE Deaths Notes: See Appendix 1 for additional information on the model and the data. Source: ASPE analysis of multiple data sources for December 2020-July 2021 December 2021 RESEARCHREPORT 11 Associated Reductions in Cases, Hospitalizations and Deaths by State and County In Figure 3 we presentestimates of associated reductions in cases (Panel A), hospitalizations (Panel B), and deaths (PanelC), as well as vaccinations (Panel D) by county. Because county population density can vary widely, we present averted cases, hospitalizations, and deaths in population-adjusted terms (per 10,000 population). Some counties could not be estimated due to missing case, hospitalization, or death data. These counties are shown in gray. In all four panels, counties were divided into quintiles, with yellow representing counties with the lowest associated reductions in cases, hospitalizations, and deaths per 10,000 population, and blue representing counties with the highest rates. These data indicate that COVID-19 vaccinations were associated with significant reductions in cases, hospitalizations, and deaths per 10,000 population in all parts of the country. We also generated state-specific estimates by aggregating the county-level estimates for all 50 states and the District of Columbia. Table 2 shows the estimates of reductions in COVID-19 outcomes for the 25 most populous states (see Table S-6 for the entire list). Figure 3. Number of Associated Reductions in Cases, Hospitalizations, and Deaths per 10,000 Population by County Panel A. Number of Associated Reductions in Cases per 10,000 Population Associated reductions in cases per 10,000 population 0 to 439 439 to 576 576 to 708 708 to 879 879 to 4,024 Missing December 2021 RESEARCHREPORT 12 Panel B. Number of Associated Reductions in Hospitalizations per 10,000 Population Associated reductions in hospitalizations per 10,000 population Oto 17 17 to 27 27 to 38 38 to 56 56 to 3,167 Missing Associated reductions in deaths per 10,000 population 0.00 to 3.69 3.69 to 5.45 5.45 to 7.33 7.33 to 10.08 10.08 to 51.20 Missing December 2021 RESEARCHREPORT 13 Panel D: Percent of the 18+ Population that is Fully Vaccinated, as of July 29, 2021* * Due to data limitations, state-level vaccination rates are shown for GA, HI, TX, VA, and WV Percent of 18+ population " > that is fully vaccinated 11.40 to 38.80 38.80 to 46.20 46.20 to 51.90 . 51.90 to 58.10 a 58.10 to 99.90 Notes: Estimates are not available for certain counties (shown in gray) due to missing COVID-19 case, hospitalization, or death data. Vaccination data for certain counties in California, Massachusetts, and New Mexico, as well as all counties in Georgia, Hawaii, Texas, Virginia, and West Virginia, were imputed with state-level vaccination rates due to missing data or data quality issues. Vaccination data are shown through July 29, 2021, representing vaccination rates on the final day of the study period. Due to a discontinuation of Nebraska's COVID-19 data dashboard, estimates for the state of Nebraska represent through June 3, 2021. For additional details on how missing or incomplete data were handled in the analysis, see Appendix 2. Source: ASPE analysis of multiple data sources for December 2020-July 2021 Alabama 5,024,279 272,114 18,830 Arizona 7,151,502 859,233 38,499 California 39,538,223 2,566,307 145,545 Colorado 5,773,714 265,805 10,540 Florida 21,538,187 2,215,557 107,370 G ia 10,711,908 995,450 53,286 Illinois 12,812,508 818,561 48,432 Indiana 6,785,528 459,390 27,376 Louisiana 4,657,757 299,756 14,525 Maryland 6,177,224 338,698 29,145 Massachusetts 7,029,917 567,943 21,003 M an 10,077,331 748,899 43,788 Minnesota 5,706,494 357,433 17,417 Missouri 6,154,913 447,357 29,053 3,146 December 2021 RESEARCHREPORT 14 D New Jers 9,288,994 854,690 45,949 6,376 New York 20,201,249 2,387,463 131,120 13,953 North Carolina 10,439,388 767,595 34,810 4,879 Ohio 11,799,448 834,285 50,132 7,742 Pennsylvania 13,002,700 1,059,848 69,080 11,535 South Carolina 5,118,425 395,715 15,746 2,795 Tennessee 6,910,840 280,338 16,458 3,727 Texas 29,145,505 2,263,058 139,812 17,923 Virginia 8,631,393 713,673 36,765 6,350 Washington 7,705,281 459,922 19,965 2,585 Wisconsin 5,893,718 303,834 16,132 2,615 Total United States 331,449,281 25,324,302 1,384,260 213,349 Notes: Results shown for the 25 most populous states. See Table S-5 for the entire list of 50 states and the District of Columbia. Population data from the 2020 Census. Source: ASPE analysis of multiple data sources for December 2020-July 2021 Estimating Associated Reductions in Cases, Hospitalizations and Deaths by Social Vulnerability Index In Table 3 we presentthe estimated reduction in COVID-19 cases, hospitalizations, and deaths based on CDC's Social Vulnerability Index (SVI) fora given county. This is done by summing up the total number of estimated reductions across counties with a given SVI. In this table we also present the percent of the population covered in each SVI category. The SVI summarizes the extent to which a community is socially vulnerable to disaster.?4 The overall SVI for a county is calculated using the American Community Survey (ACS) data across four main themes: (1) socioeconomic status, (2) household composition and disability, (3) minority status and language, and (4) housing type and transportation. SVI values range from 0 (least vulnerable) to 1 (most vulnerable). Counties with very low or low social vulnerability tended to make up a lower proportion of the associated reductions in cases, hospitalizations, and deaths than would be expected based on their population (Table 3). In contrast, counties with high or very high social vulnerability tended to make up a higher proportion associated reductions in cases, hospitalizations, and deaths than would be expected based on their population. Table S-7 presents the results for each of the four themes. 34 Agency for Toxic Substances and Disease Registry, CDC/ATSDR SVI Data and Documentation Download, available at https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html, last accessed October 20, 2021. December 2021 RESEARCHREPORT 15 Table 3. Estimated Associated Reductions in COVID-19 Related Outcomes by Social Vulnerability Index Overall Social See Vulnerability goede Index in the Cases (%) Hospitalizations (%) Deaths (%) error: United gory States (019). 16.19% 3,741,523 | 14.77% | 167,209 12.08% | 28,492 | 13.35% Low 9 0 0, 9 (0.20-0.39} 19.51% 4,775,729 18.86% 235,332 17.00% 36,514 | 17.11% Moderate 0 9 te} Gg (0.40-0.59) 23.38% | 5,969,199 23.57% 340,377 24.59% 47,414 | 22.22% High 25.60% 6,645,468 26.24% 410,025 29.62% 61,773 28.95% (0.60-0.79) ° ° ° ° Very High 15.32% | 4,192,383 | 16.55% 231,318 | 16.71% 39,156 | 18.35% (0.80-1.00) ° ° ° ° Notes: CDC's Social Vulnerability Index (SVI) values range from O (least vulnerable) to 1 (most vulnerable). Table S-7 presents the results for four separate SVI themes. Population data from the 2019 American Community Survey. Source: ASPE analysis of multiple data sources for December 2020-July 2021 Valuing COVID-19 Morbidity and Mortality Risk Reductions As noted in our methods section, to value risk reductions in morbidity and mortality we apply estimates based on the value of individual willingness to pay to avoid a COVID-19 outcome using estimates of the value per statistical life and estimates of the value per statistical case that vary by case severity.?5 Specifically, mortality risk reductions are valued at $11.5 million per death. Morbidity risk reductions are valued based on the severity of the disease: mild cases are valued at $5,846 per case, severe cases that require a regular hospital admission are valued at $13,104 per case, and critical cases that involve intensive care unit (ICU) admissions are valued at $1.8 million percase. Because our estimates of associated reductions in hospitalizations include non-ICU and admissions, we use an estimate of the percent of ICU admissions to estimate the number of associated reductions in ICU admissions. We do this using the percent of ICU admissions among all COVID-19 hospitalizations, which we estimate to be 29 percent using hospitalization data through October 22, 2021.3 The estimated reductions in ICU admissions is 188,086 (29 percent 35 Robinson, L., Eber, M., Hammitt, J. (2021). Valuing COVID-19 mortality and morbidity risk reductions in U.S. Department of Health and Human Services Regulatory Impact Analyses. Available at: https://aspe.hhs.gov/reports/valuing-covid-19-risk-reductions-hhs-rias, last accessed October 2021. 36 HHS Protect Public Data Hub. Hospital Utilization. Data as of October 22, 2021 available at https: //protect- public.hhs.gov/pages/hospital-utilization. Last accessed October 22, 2021. December 2021 RESEARCH REPORT 16 of the 1,384,260 hospitalizations, minus 213,349 deaths to avoid double- We estimate a total value of COVID-19 risk counting).?" We also subtract all reductions attributable to vaccination in the estimated reductions in hospitalizations United States of $2.95 trillion. This valuation from reductions in cases to avoid comes from reductions in deaths ($2.45 trillion), double-counting. Thus, we use hospitalizations ($354.14 billion), and cases 23,940,041 cases (25,324,302 cases ($139.96 billion). minus 1,384,260 hospitalizations), 982,825 non-ICU admissions and 188,086 ICU admissions as the basis to estimate the total health benefits. Applying an estimate of the VSL and estimates of the VSC, we estimate a total value of COVID- 19 risk reductions attributable to vaccination in the United States of $2.95 trillion (Figure 4). This valuation comes from reductions in deaths ($2.45 trillion), hospitalizations ($354.14 billion), and cases ($139.96 billion). The breakdown of these estimates is shown in Table S-8. In Table S-9, we present estimates using low and high values of VSL and VSC. Using those estimates, the value of mortality and morbidity risk reduction ranges from $1.38 trillion to $4.49 trillion. This range of total values includes associated reductions in cases ($65.32 billion to $213.05 billion), hospitalizations ($165.27 billion to $539.08 billion), and deaths ($1.15 trillion to $3.74 trillion). Figure 4. Valuing Associated Reductions in COVID-19 Outcomes in the United States (Sbillions) $354.14, 12% $139.96, N $2,453.81, 83% = Mortality § ™ Morbidity: Cases Morbidity: Hospitalizations Notes: Mortality risk reductions are valued at $11.5 million per death, and morbidity risk reductions are valued based on the severity of the disease: mild cases are valued at $5,846 per case, severe cases that require a regular hospital admission are valued at $13,104 per case, and critical cases that involve intensive care unit (ICU) admissions are valued at $1.8 million per case. The estimated value of associated reductions in hospitalizations includes the sum of ICU admissions ($12.88 billion = 982,825 hospitalizations multiplied by $5,846) and non-ICU admissions ($341.26 billion = 188,086 multiplied by $13,104) where ICU admissions represent 29 percent of the total associated reductions in hospitalizations. Source: ASPE analysis of multiple data sources for December 2020-July 2021 37 This implicitly assumes that all deaths occurred in individuals who had been admitted to the ICU. December 2021 RESEARCH REPORT 17 Discussion In this study, we examined the relationship between COVID-19 vaccination rates and associated reductions in COVID-19 cases, hospitalizations, and deaths. Specifically, our model estimates that COVID-19 vaccinations may be associated with 25.32 million fewercases, 1.38 million fewer hospitalizations, and 213,000 fewer deaths from December 2020 to July 2021. Relatively few studies to date have estimated the impact of vaccinations on the full spectrum of COVID-19 outcomes including cases, hospitalizations, and deaths. In a recent analysis that used state-leveldata on vaccination rates and deaths, Gupta and colleagues estimated that vaccination was associated with reductions of about 140,000 deaths from January to May 2021.38 We follow a similar empirical approach as Gupta et al, but use county-level data instead of state-level data and extend the analysis period to July 2021. Using an age-stratified, agent- based national model, Galvani etal (2021) estimated that the United States vaccination program prevented 26 million cases, 1.25 million hospitalizations, and 279,000 deaths by the end of June 2021.29 Galvani et al incorporate several parameters that were not available for our model, such as age-specific risk factors, vaccine efficacy data, and transmission dynamics of variants. In contrast, our model accounts for considerable county-level variation in vaccination rates, which may betteraccount for local trends. Despite these differences, all these models converge on similar national numbers, which supports the validity of our local estimates. In a separate study using individual-level clinical data and local vaccination rates, ASPE found that COVID-19 vaccinations were associated with an estimated reduction of more than 265,000 cases, 107,000 hospitalizations, and 39,000 deaths among Medicare beneficiaries between January and May 2021.9 As the authors noted, these estimates were considered conservative since they did not include the summer increase in both cases and vaccinations, and may not have fully captured some high-risk populations who are underrepresented in the Medicare fee- for-service population. Putting the estimates of that report and this one in context, the estimated reductions in Medicare deaths are a sizable portion of our overall mortality estimates through the end of May 2021 - approximately 22.7 percent of the total - while the Medicare estimates accounted for 13.2 percent of our total estimated hospital reductions and just 2.2 percent of estimated cases. This pattern is consistent with older adults and those with disabilities in Medicare being at highest risk for serious complications and deaths from COVID- 38 Gupta S, Cantor J, Simon KI, Bento Al, Wing C, Whaley CM. Vaccinations Against COVID-19 May Have Averted Up To 140,000 Deaths In The United States. Health Affairs 40(9). https://doi.org/10.1377/hkthaff.2021.00619. 39 Galvani A, Moghadas S., Schneider, E. (2021). Deaths and hospitalizations averted by rapid U.S. vaccination rollout. Available at: https: //www.commonwealthfund.org/publications/issue-briefs/2021/jul/deaths-and- hospitalizations-averted-rapid-us-vaccination-rollout, last accessed September 9, 2021. 40 Samson, LW, Tarazi, W, Orav, EJ, Sheingold, S, De Lew, Nand Sommers, BD. (2021). Associations Between County-level Vaccination Rates and COVID-19 Outcomes Among Medicare Beneficiaries. Washington, DC: Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services. Available at: https://aspe.hhs.gov/reports/covid-19-vaccination-rates-outcomes, last accessed October 15, 2021. December 2021 RESEARCH REPORT 18 19, and therefore disproportionately likely to benefit from vaccinations for those serious outcomes. Rates of COVID-19 cases, hospitalizations, and deaths have varied by race/ethnicity and age throughout the pandemic.*1-42.43 Differences in vaccination rates by race/ethnicity and age," as wellas the increased transmissibility of the Delta variant, have likely changed the landscape of COVID-19 impacts on these groups. We do not have reliable age or race/ethnicity data on vaccination, cases, hospitalizations, or deaths at the county level to estimate our model using these important characteristics. Further, the regression model includes county-fixed effects which removes any county-levelinformation that does not change overtime. Eventhough there is county-level data from the American Community Survey that provides information at the county-level, these data are constant over the period of the analysis and the model drops them from the regression results. Thus, although we are unable to calculate the associated reductions in outcomes by these demographic characteristics directly from the model, evaluation of associated reductions in outcomes at the county-level showed that areas of high or very high social vulnerability tended to make up a larger proportion of each outcome than would be expected based on their population share. This is not unexpected given that areas of high social vulnerability have experienced some of the highest COVID-19 case and death rates.45.46 These results underscore the importance of continued COVID-19 vaccination efforts, particularly in areas where vaccination rates may be low due to access barriers, to mitigate the disproportionate impact of COVID-19 on socially vulnerable communities. Future work should attempt to explore the impact on specific demographic groups, including racial and ethnic minorities, more directly. 41 Simmons A, Chappel A, Kolbe AR, Bush L, and Sommers BD. (2021). Health Disparities by Race and Ethnicity During the COVID-19 Pandemic: Current Evidence and Policy Approaches. Washington, DC: Office of the Assistant Secretary for Planning and Evaluation, United States Department of Health and Human Services. Available at: https://aspe.hhs.gov/sites/default/files/migrated_legacy_files//199516/covid-equity-issue-brief.pdf, last accessed October 15, 2021. 42 CDC. COVID Data Tracker: COVID-19 Weekly Cases and Deaths per 100,000 Population by Age, Race/Ethnicity, and Sex. Available at: https://covid.cdc.gov/covid-data-tracker/#demographicsovertime, last accessed October 15, 2021. 43 COVID-NET. Laboratory-Confirmed COVID-19-Associated Hospitalizations. Available at: https: //gis.cdc.gov/grasp/COVIDNet/COVID19 3.html, last accessed October 15, 2021. * Kolbe, A. (2021). Disparities in COVID-19 Vaccination Rates across Racial and Ethnic Minority Groupsin the United States. Washington, DC: Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services. Available at: httos://aspe.hhs.gov/re ports/disparities-covid-19-vaccination-rates- across-racial-ethnic-minority-groups-united-states, last accessed October 15, 2021. 45 Dasgupta S, Bowen VB, LeidnerA, etal. (2020). Association Between Social Vulnerability anda County's Risk for Becoming a COVID-19 Hotspot - United States, June 1--July 25, 2020. MMWR Morb Mortal Wkly Rep 69: 1535- 1541. http://dx.doi.org/10.15585/mmwr.mm6942a3 46 CDC. COVID Data Tracker: Trends in COVID-19 Cases and Deaths in the United States, by County-level Population Factors. Available at: https://covid.cdc.gov/covid-data-tracker/#pop-factors_7daynewdeaths, last accessed October15, 2021. December 2021 RESEARCHREPORT 19 Limitations Our empirical approach includes several limitations. First, we are unable to fully account for potential confounders in our model, which limits our ability to infer causality. While we include county and quarter fixed effects in conjunction with controls such as mask mandates, worker mobility and the proportion of cases due to the Delta variant, we are unable to control for other factors such as time-varying county-specific information that could be correlated with both vaccination and our outcomes of interest. For instance, it is possible that changes in vaccination rates were correlated with changes in preferences to engage in other COVID-19 mitigation behaviors. If vaccination rates rose in counties where unobserved COVID-19 mitigation behaviors simultaneously strengthened, an estimate of the causal effect of vaccination rates may be biased upward. Moreover, it is possible that changes in COVID-19 cases, hospitalizations, or deaths prompted changes in vaccination rates, which may also bias an estimate of the causal effect of vaccination. Second, we use county-level rather than individual-level data. Consequently, our ecological model captures potential vaccination spillovers, but is not able to examine how demographic and other individual-specific characteristics may be associated with vaccination and our outcomes of interest. Our model also does not distinguish COVID-19 outcomes by age group. During the study period, December 2020 to July 2021, COVID-19 vaccines were available to individuals aged 12 and over;*" however, due to data limitations, our variable of interest was the vaccination rate of the adult (18+) population. We assume that the vaccination rate of adults is associated with reductions in COVID-19 outcomes across all age groups. Third, our counterfactual estimates do not account fora number of potential factors. For instance, in a counterfactual environment in which COVID-19 vaccines do not exist, other interventions may have been enacted to prevent disease spread. Similarly, we are unable to predict the potential economic consequences of such an environment. Furthermore, our model does not explicitly account for natural immunity following COVID-19 infection or for the degree of effectiveness of COVID-19 vaccines; instead, these factors are reflected in the observed data we applied in the model. However, our calculation of the counterfactual "no vaccination" scenario might not fully account for the impact of changes in natural immunity in the population. Fourth, data quality issues may have biased results for some counties. Due to missing or underreported county-level vaccination data in five states (Georgia, Hawaii, Texas, Virginia, and West Virginia), we impute county-level vaccination data using state-level vaccination data. This eliminates within-state variation in vaccine coverage, which results in an averaging of associated reductions in outcomes within these states. By attributing vaccination coverage that is higher or lower than the actual coverage in a given county, we likely over or under-estimate 47 On May 10, 2021, FDA authorized the use of a COVID-19 vaccine for emergency use in adolescents 12 through 15 years of age. On October 29, 2021, FDA authorized a COVID-19 vaccine for emergency use in children aged5 to 11. December 2021 RESEARCH REPORT 20 associated reductions in outcomes and therefore the county-level estimates for these states should be interpreted with caution. Missing case, hospitalization, or death data in some counties likely resulted in an underestimation of national numbers of associated reductions in the outcomes examined. While our review of the data did not reveal that these missing data could be identified systematically, measurementerror due to missing data may impact our ability to fully capture the extent of COVID-19 spread and hence our estimates. Additionally, our data cannot account for COVID-19 cases that were not reported or identified, such as those in asymptomatic individuals. To address these issues with missing or underreported data, we employ a number of sensitivity analyses. These results appear in the Sensitivity Analysis section in the Appendix. The results of these analyses suggest that our national-level estimates are robust to these data limitations. Finally, our data and approach do not permit us to incorporate time invariant information, such as the demographic composition of a county that would enable us to examine our estimated associations by groups of interest, e.g., age, race/ethnicity. Relatedly, our estimates of economic health benefits assume that the willingness to pay to reduce morbidity and mortality risks are the same across the population. The literature suggests that they may vary by individual and risk characteristics but there is no consensus on appropriate estimates to use generally and in the context of COVID-19. Future research should explore alternate data sources and methods to examine these questions. Conclusions Our county-level regression-based estimates suggest that COVID-19 vaccinations were associated with reductions of approximately 25.32 million cases, 1.38 million hospitalizations, and 213,000 deaths in the United States from December 2020 and through the summer of 2021. With 60.1 percent of Americans ages 18 years and older fully vaccinated at the end of the study period, these results emphasize the importance of continued outreach to unvaccinated individuals and communities with low vaccination rates. In terms of willingness to pay for morbidity and mortality risk reductions, these translate to total estimated benefits of $2.95 trillion, which far outweighs the estimated cost of federal investmentin vaccine development and distribution, at least $10 billion. Our results emphasize the importance of vaccination for reducing the spread of COVID-19 and saving lives. Moreover, they highlight the incredibly high societal rate of return on vaccine investment in conjunction with vaccine uptake. December 2021 RESEARCHREPORT 21 Appendix 1: Data Sources Table S-1. Description of Variables and Data Sources Used in ASPE's Analysis Data Description Level/Frequency Source Percent of people 18+ who arefully © County/Daily* COVID-19 Vaccinations in the United vaccinated (have second dose of a States, County (CDC two-dose vaccine or one dose of a single-dose vaccine) based on the jurisdiction and county where recipient lives Confirmed and probable (where County/Daily* COVID-19 Data Repository by the reported) COVID-19 cases for all Center for Science and Engineering ages (CSSE) at Johns Hopkins University Average number of patients Facility/Weekly COVID-19 Reported Patient Impact currently hospitalized in an adult and Hospital Capacity by Facility inpatient bed who have laboratory- (Department of Health and Human confirmed or suspected COVID-19, Services) including those in observation beds reported during the 7-day period Confirmed and probable (where County/Daily* COVID-19 Data Repository by the reported) COVID-19 deaths for all Center for Science and Engineering ages (CSSE) at Johns Hopkins University Stay at Home Order is Active State/Daily* COVID-19 State Policy US Database (yes/no) Face Mask Mandate in Public Spaces State/Daily* COVID-19 State Policy US Database is Active (yes/no) Mobility Trends for Places of Work County/Daily* COVID-19 Community Mobility Reports (Google) Proportion of Delta Variant Causing Region/Weekly CDC's National SARS-CoV-2 Genomic Cases Surveillance Program Note: * Denotes frequency of the data available in the original source files; where appropriate data were aggregated to capture measures ona weekly basis orto match the last day of a week. December 2021 RESEARCHREPORT 22 Appendix 2: Methodology Regression Model We use a Poisson regression model to estimate the number of associated reductions in cases, hospitalizations, and deaths due to vaccination. Our baseline model is Equation (1) where y,, denotes one of three measures (m = cases, hospitalizations, or deaths) per 10,000 population in county c during week f. Each measure mis the number of new cases, hospitalizations, or deaths at weekt in county c. vax_rate,, is the county-level vaccination rate (percent of population 18 years and over fully vaccinated) in county c at week t. To account for the lag between vaccination and COVID-19 outcomes (i.e., the current rate of COVID-19 outcomes being driven by vaccinations administered in previous weeks), our model incorporates four lags of vaccination rates in each county; these are denoted as vax_rate,, 1 through vax_rate,,_4.48 Xis avector of region- and state-specific variables that vary over time: whether a state has a stay-at-home order active at week t, whether a state has a face mask mandate in public spaces in effect at week t, county-specific mobility trends for places of work at week t, and the proportion of the Delta variant causing cases in regions of the US at week t. t denotes quarterly fixed effects. 0, includes county-specific fixed effects.*9 The standard errors are clustered at the county-level. (1) ye = exp (a+ bo* vax_rate,,+ b,* vax_rate,,4+b2* vax_rate,,.+ b3* vax_rate,;3+b,* vax_rate,,,4+td*X+tT+0,+ ect) Estimating Associated Reductions in Cases, Hospitalizations and Deaths Using estimates from Equation (1) we predict the number of cases, hospitalizations, and deaths for each county in each week. We call these our predicted values (ym). We then set vaccination rates equal to zero for every county in every week in our sample and generate a new set of counterfactual predicted values that estimate our outcomes in the absence of vaccination. The difference between these counterfactual predicted values and our original predicted values are our estimates of associated reductions in cases, hospitalizations and deaths in a given county and week due to vaccination. We then sum these counts over all counties and weeks to arrive at a total number of associated reductions in cases, hospitalizations, and deaths for all 50 states and the District of Columbia over our entire study period. Results of the Model In Table S-2 we present the exponentiated coefficients of the Poisson model, or the incidence rate ratio (IRR), in Equation (1). An IRR of 0.91 in the first column of Table S-3 suggests that, holding all other variables constant, on average a one unit increase in vaccination coverage is associated witha decrease in the rate ratio of cases by a factor of 0.91. Our model suggests a similar association for hospitalizations and deaths. Because our model includes both the contemporaneous vaccination rate and its four lags, to estimate the association with vaccination and its lags we would multiply the corresponding coefficients. For instance, when using cases as our outcome, this would be 0.955 (0.955 = 0.911*1.040*1.028*1.012*0.969). Furthermore, we test for the joint significance of our 48 Our choice of four lags follows the approach of Gupta etal. (2021). 49 We use quarterlyfixed effects as weekly and monthly fixed effects excessively eliminated usefulvariation. December 2021 RESEARCHREPORT 23 contemporaneous measure of vaccination and its four lags. We find that for each outcome they are jointly significant at the 1 percent level. Table S-2. Estimated Incidence Rate Ratio of the Poisson Model Used to Estimate Associated Reductions in Cases, Hospitalizations and Deaths Cases per 10,000 Hospitalizationsper | Deaths per 10,000 Population 10,000 Population Population Vaccine Doses per 10,000 0.911*** 0.940*** 0.905*** Population (0.00556) (0.00617) (0.00847) Vaccine Doses per 10,000 1.040*** 1.027*** 1.075*** Population 1 Week Prior (0.0105) (0.00815) (0.0173) Vaccine Doses per 10,000 1.028** 1.002 0.975 Population 2 Weeks Prior (0.0114) (0.00765) (0.0166) Vaccine Doses per 10,000 1.012 1.020*** 0.958** Population 3 Weeks Prior (0.0111) (0.00705) (0.0177) Vaccine Doses per 10,000 0.969*** 0.991 1.081*** Population 4 Weeks Prior (0.00592) (0.00697) (0.0123) Stay at Home Order 2.238*** 1.556*** 0.806** (0.152) (0.0995) (0.0744) Mask Mandate 1.584*** 1.281*** 1.729*** (0.0369) (0.0377) (0.0604) Workplace Mobility 1.004*** 0.997*** 1.005*** (0.000420) (0.000564) (0.000935) Delta Variant 2.786*** 0.546** 0.0235*** (0.544) (0.136) (0.00871) Number of Observations 104,150 80,252 102,483 Notes: All specifications are estimated using a Poisson model, and include county and quarterly fixed effects. Each observation is at the county-week level. Standard errors appear in parentheses and are clustered at the county level. * p-value <0.10; ** p- value <0.05; *** p-value <0.01. Source: ASPE analysis of multiple data sources for December 2020-July 2021. Addressing Data Quality Issues We adopt a number of approaches to deal with data quality issues. First, interms of our vaccination data obtained from CDC, Texas and Hawaii do not report vaccination data at the county-level. We also find that over 40 percent of vaccinations in Georgia, Virginia, and West Virginia have no associated county data. Consequently, for all five of these states we use CDC state-level vaccination data instead of December 2021 RESEARCH REPORT 24 county-level vaccination data. Additionally, certain counties in California and Massachusetts do not report county-level vaccination data. Therefore, we drop these counties from our analysis.°°54 There are also several states or counties with missing data in the case and death data obtained from the Center for Science and Engineering at Johns Hopkins University. First, Nebraska stopped reporting COVID-19 case and death data on June 3, 2021.52 Therefore, we treat all case and death data after June 3, 2021 in Nebraska as missing, so estimates of associated reductions in outcomes in Nebraska should be considered underestimates. Second, anumber of counties in Utah report their case and death data in combination with other counties. Since we are unable to identify the individual contributions of each county, we treat case and death data in these counties as missing.>? We alsotreat as missing four Alaska counties that combine their case and death data when using cases and deaths as outcomes; these counties are considered too small to report their data separately. °4 Finally, case and death data for Dukes and Nantucket counties in Massachusetts are not reported, and so these counties are dropped from our sample when using cases or deaths as outcomes.>> Moreover, our hospitalization data in their original form include missing data. Some facilities had missing data for some weeks during our period of analysis. We replace these missing values by interpolating within facility using the number of hospitalizations available for the nearest known date. We also recode some hospitalization values. Specifically, to protect the privacy of individuals, any facility reporting fewer than 4 hospitalizations is originally top-coded with a value of -99999. We replace these top-coded values with a value of 4. Lastly, we then sum total hospitalizations by week and county. We drop from our analysis any county that does not have hospitalization data available for any week during our sample when using hospitalizations as an outcome. We also modify our workplace mobility variable to account for missing data. This variable is the change in visits to workplaces (using aggregated data from Google users) in a given day compared to the median value of visits to workplaces for that same day of the week during the five-week period January 3, 2020 to February 6, 2020. Since this variable is at the daily-county level, we average it to the weekly-county level. Moreover, data is missing for some counties. We replace these missing data with the weekly-state average. Sensitivity Analysis 50 For California, these counties include Alpine, Inyo, Mariposa, Modoc, Mono, Plumas, Sierra, and Trinity. For Massachusetts, these counties include Barnstable, Dukes, and Nantucket. 51 Additionally, McKinley County, New Mexico, overre ports its vaccination coverage, and so is dropped from our sample. 52 Between July 1, 2021 and September 21, 2021, thereis no county-level data available due to a temporary discontinuation of the Nebraska COVID-19 Dashboard. See https://nebraska.tv/news/a-new-nebraska-covid- dashboard-is-up-and-available. 53 These counties in Utah include Beaver, Box Elder, Cache, Carbon, Daggett, Duchesne, Emery, Garfield, Grand, Iron, Juab, Kane, Millard, Morgan, Piute, Rich, Sanpete, Sevier, Uintah, Washington, Wayne, and Weber. 54 These counties include Bristol Bay, Lake and Peninsula, Yakutat, and Hoonah-Angoon. 55 Additionally, New YorkCity does not report probable deaths by county, so these are notincluded in our analysis. December 2021 RESEARCHREPORT 25 We examine whether our estimates of associated reductions in cases, hospitalizations and deaths are robust to the choice of specification in Equation (1) and changes in our sample. The results of this sensitivity analysis are presented in Table S-3. In general, we find that our baseline estimates are robust to these variations. Sensitivity Analysis (1) is our baseline specification as presented above in Equation (1). We showit here for ease of comparison. In Sensitivity Analysis (2), we re-estimate our model dropping the worker mobility variable as a control. We do this to ensure that imputation of missing values does not substantially bias our results. Our robustness check shows that it does not as estimated associated reductions in cases, hospitalizations and deaths do not substantially change. In Sensitivity Analysis (3), we exclude all region-, state-, and county-level controls. Our estimates increase, suggesting that omission of these controls may bias our results upward. In Sensitivity Analysis (4), we exclude Hawaii and Texas to compare our estimates with estimates that others have published that similarly exclude these states. Our estimates decrease, whichis expected given that Texas is one of the most populous states. Similarly, in Sensitivity Analysis (5) we exclude a group of five states (Georgia, Virginia, West Virginia, Texas and Hawaii) as vaccination data for these states were imputed. Our estimates decrease by between 10 percent and 23 percent relative to our baseline estimates. Again, this result is expected given that these states account for a notable portion of the United States population. In Sensitivity Analysis (6), we use a linear regression model instead of a Poisson model. Relative to our baseline scenario, our estimates substantially decrease interms of cases and hospitalizations, but increase in terms of deaths. Finally, in Sensitivity Analysis (7) we use a linear regression model with quadratics in contemporaneous vaccination rates and their lags. Compared to the baseline model, associated reductions in cases decrease by approximately 1 percent, but there is a substantial increase in associated reductions in hospitalizations (32 percent) and deaths (106 percent). We prefer using a Poisson model because we believe it better captures the nonlinearity of vaccination rates andthe outcomes of interest Table S- 3. Sensitivity Analysis: Estimated Associated Reductions in Cases, Hospitalizations, and Deaths SiN UALS Cases Hospitalizations DY 1 dK) (1) Modeluses anon- linear Poisson regression, baseline model 25,324,302 1,384,260 213,349 (2) Modeluses a non- linear Poisson 24,680,971 1,414,165 205,174 regression, excluding worker mobility (3) Modeluses anon- linear Poisson regression, excluding all region-, state- and county-level controls 27,970,049 1,640,618 266,997 December 2021 RESEARCH REPORT 26 Table S- 3. Sensitivity Analysis: Estimated Associated Reductions in Cases, Hospitalizations, and Deaths SEN UNAS Cases Hospitalizations Deaths (4) Medeluses anon- linear Poisson regression, excluding thestates of Texas and Hawaii 22,122,403 1,228,794 184,798 (5) Modeluses anon- linear Poisson regression, excluding 18,924,595 1,103,825 183,808 the states of Georgia, Hawaii, Texas, Virginia and West Virginia (6) Modeluses a linear functional form 15,708,113 1,235,956 247,280 (ordinary least squares regression) (7) Modeluses ordinary least squares where 24,482,573 1,860,893 421,874 vaccination rates are modeled as quadratics Notes: Unless noted, all specifications include county and quarterly fixed effects. Standard errors are clustered at the county level. Source: ASPE analysis of multiple data sources for December 2020-July 2021 As noted above, we find that county-level vaccination rates are underreported for some states. Therefore, as an additional sensitivity analysis, we explore whether other sources of state-level measurement error may bias our estimates of associated reductions. Todo so, we estimate the total number of associated reductions in cases, hospitalizations, and deaths after dropping one state at a time from our sample. We repeat this estimation procedure for all 50 states andthe District of Columbia. The results are presented in Table S-4. Changes in total associated reductions due to dropping a single state stem from two different factors. The first is that when a given state is dropped from our sample, total counts may decrease due to subtraction of this given state's associated reductions. For instance, if California is dropped from our sample, then the associated reduction in counts experienced by California would not be included in the total estimate of associated reductions, resulting ina lower estimate. The second factor is that dropping a single state from our sample may change the coefficients estimated for Equation (1). Depending on how the coefficients change, this could result in total associated reduction in counts either increasing or decreasing. In general, Table S-4 reinforces the validity of our estimation strategy. Dropping single states from our estimation procedure does not substantially affect our estimates ina way that suggests systematic state-level measurement error substantially biases our results. For instance, the largest changes in our December 2021 RESEARCHREPORT 27 estimates occur when California, New York, and Texas are dropped. This is not surprising given that these are some of the most populous states. Table S-4. Sensitivity Analysis: Estimated Associated Reductions in Cases, Hospitalizations, and Deaths when Individual States Are Excluded from the Model State Excluded from Model Cases Hospitalizations Deaths Alabama 25,203,952 1,367,314 206,301 Alaska 26,175,702 1,395,567 213,448 Arizona 24,253,682 1,337,347 204,955 Arkansas 24,025,278 1,329,052 208,866 California 22,873,150 1,232,484 178,348 Colorado 26,749,732 1,387,504 217,437 Connecticut 24,913,960 1,360,013 209,846 Delaware 25,222,954 1,377,075 212,897 District of Columbia 25,278,368 1,379,730 213,011 Florida 23,211,188 1,280,078 202,513 Georgia 23,679,080 1,313,613 200,112 Hawaii 25,318,648 1,383,579 213,075 Idaho 25,208,208 1,388,236 213,478 Illinois 24,190,600 1,328,859 205,960 Indiana 24,285,804 1,337,558 207,050 lowa 25,576,350 1,409,115 212,643 Kansas 24,705,598 1,340,085 207,099 Kentucky 24,429,132 1,338,620 228,510 Louisiana 23,933,082 1,357,844 210,054 Maine 25,595,048 1,403,134 213,727 Maryland 24,938,428 1,356,241 212,449 Massachusetts 24,680,350 1,360,558 208,053 Michigan 25,148,604 1,367,892 208,582 Minnesota 26,154,054 1,419,625 212,395 Mississippi 24,793,728 1,374,047 209,426 Missouri 27,140,506 1,330,992 210,666 Montana 25,778,300 1,427,077 212,443 Nebraska 25,389,704 1,400,922 214,134 Nevada 25,142,670 1,357,929 210,911 New Hampshire 25,273,890 1,392,562 213,741 New Jersey 24,450,516 1,329,106 207,884 New Mexico 25,427,418 1,394,053 213,569 New York 22,429,736 1,214,244 193,555 North Carolina 24,298,872 1,345,468 209,225 North Dakota 25,777,476 1,390,282 213,189 Ohio 24,096,162 1,317,400 199,804 Oklahoma 24,285,342 1,347,530 203,415 Oregon 25,545,790 1,401, 748 212,488 December 2021 RESEARCHREPORT 28 Penns nia Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin i 23,865,562 25,218,534 24,805,162 25,679,484 25,064,466 22,128,476 25,142,820 25,274,648 23,922,878 25,367,514 24,726,330 24,871,552 25,238,750 1,335,479 1,378,526 1,356,176 1,394,796 1,343,071 1,229,519 1,374,274 1,383,552 1,347,313 1,389,425 1,362,139 1,396,817 1,373,265 194,925 211,949 209,530 218,504 209,846 185,056 213,088 213,276 225,824 212,892 210,570 212,011 210,629 None (Baseline Model 25,324,302 1,384,260 213,349 Notes: Unless noted, all specifications include county and quarterly fixed effects. Standard errors are clustered at the county level. Source: ASPE analysis of multiple data sources for December 2020-July 2021 December 2021 RESEARCHREPORT 29 Appendix 3. Supplementary Analysis Number of Cases (Thousands) Figure S-1. Reported, Predicted, and Counterfactual Cases 1000 1500 2000 L ! 500 ! o- a T T T January 1,2020 April 1, 2020 July 1, 2020 T T T T October 1, 2020 January 1,2021 April 1, 2021 July 1, 2021 Date eo = Reported Cases ~~~ ='+ Predicted Cases Counterfactual Cases Source: ASPE analysis of multiple data sources for December 2020-July 2021. December 2021 RESEARCH REPORT 30 Figure S-2. Reported, Predicted, and Counterfactual Hospitalizations 100000 150000 | Number of Hospitalizations 50000 1 o-4 T T T T T T T August 1, 2020 October 1, 2020 December 1, 2020 February 1,2021 April 1, 2021 June 1, 2021 August 1, 2021 Date =o Reported Hospitalizations sss se:=)+ Predicted Hospitalizations Counterfactual Hospitalizations Source: ASPE analysis of multiple data sources for December 2020-July 2021. December 2021 RESEARCH REPORT 31 Figure S-3. Reported, Predicted, and Counterfactual Deaths 10000 15000 20000 25000 \ ! ! ! Number of Deaths 5000 | / a =--- T T T T T T T January 1, 2020 April 1, 2020 July 1, 2020 October 1,2020 January 1, 2021 April 1, 2021 July 1, 2021 0 Date me Reported Deaths ~~ ===> Predicted Deaths Counterfactual Deaths Source: ASPE analysis of multiple data sources for December 2020-July 2021. December 2021 RESEARCH REPORT 32 Table S-5. Estimated Associated Reductions in Cases, Hospitalizations, and Deaths, and COVID-19 Vaccine Doses Administered Over Time Cumulative OUT VIEL) Cumulative Cumulative Reductions in Reductions in Reductions in Vaccine Doses e-TY-15 Hospitalizations Deaths Administered 12/10/2020 127,342 12/17/2020 23,765 1,254 383 1,697,791 12/24/2020 59,170 3,222 845 3,738,130 12/31/2020 98,567 5,894 1,447 6,555,680 1/7/2021 130,879 8,316 2,338 11,961,374 1/14/2021 212,317 13,566 4,041 17,546,374 1/21/2021 316,823 20,196 5,889 26,193,682 1/28/2021 448,510 29,015 8,658 35,203,710 2/4/2021 626,624 40,653 12,686 46,390,270 2/11/2021 878,844 56,943 17,858 57,737,767 2/18/2021 1,182,231 78,011 24,036 68,274,117 2/25/2021 1,548,012 101,987 31,727 82,572,848 3/4/2021 1,973,596 129,606 40,850 98,203,893 3/11/2021 2,436,892 159,773 50,404 115,730,008 3/18/2021 2,973,430 195,148 61,159 133,305,295 3/25/2021 3,555,190 233,122 72,467 153,631,404 4/1/2021 4,301,527 279,524 81,471 174,879,716 4/8/2021 5,105,811 331,459 91,019 198,317,040 4/15/2021 6,000,820 389,536 101,477 218,947,643 4/22/2021 6,926,951 450,042 112,370 237,360,493 4/29/2021 7,879,947 513,093 123,640 251,973,752 5/6/2021 8,855,912 577,339 135,118 266,596,486 5/13/2021 9,857,257 644,115 146,533 279,397,250 5/20/2021 10,829,007 709,744 157,312 290,724,607 5/27/2021 11,809,535 775,738 167,930 297,720,928 6/3/2021 12,736,600 838,010 175,736 305,687,618 6/10/2021 13,733,079 898,494 182,841 314,969,386 6/17/2021 14,750,723 958,227 188,828 320,687,205 6/24/2021 15,803,732 1,016,019 193,462 328,152,304 7/1/2021 17,591,712 1,092,115 198,549 332,345,797 7/8/2021 19,411,702 1,167,239 202,601 336,054,953 7/15/2021 21,339,670 1,240,231 206,448 339,763,765 7/22/2021 23,321,174 1,312,192 209,964 344,071,595 7/29/2021 25,324,302 1,384,260 213,349 348,966,419 Source: ASPE analysis of multiple data sources for December 2020-July 2021 December 2021 RESEARCHREPORT 33 Table S-6. Estimated Associated Reductions Cases, Hospitalizations, and Deaths by State Population Cases' Hospitalizations Deaths Alabama 5,024,279 272,114 18,830 3,300 Alaska 733,391 72,399 1,434 227 Arizona 7,151,502 859,233 38,499 7,836 Arkansas 3,011,524 196,800 10,993 1,602 California 39,538,223 2,566,307 145,545 33,484 Colorado 5,773,714 265,805 10,540 1,316 Connecticut 3,605,944 392,837 18,215 2,663 Delaware 989,948 88,527 4,426 558 District of Columbia 689,545 40,400 4,132 308 Florida 21,538,187 2,215,557 107,370 12,522 Georgia 10,711,908 995,450 53,286 7,963 Hawaii 1,455,271 46,410 2,770 283 Idaho 1,839,106 130,567 5,304 764 Illinois 12,812,508 818,561 48,432 6,951 Indiana 6,785,528 459,390 27,376 4,008 lowa 3,190,369 174,335 10,179 1,809 Kansas 2,937,880 181,324 10,518 1,830 Kentucky 4,505,836 353,443 19,963 3,352 Louisiana 4,657,757 299,756 14,525 2,225 Maine 1,362,359 77,246 4,607 490 Maryland 6,177,224 338,698 29,145 3,564 Massachusetts 7,029,917 567,943 21,003 5,073 Michigan 10,077,331 748,899 43,788 7,455 Minnesota 5,706,494 357,433 17,417 2,685 Mississippi 2,961,279 226,837 12,869 2,101 Missouri 6,154,913 447,357 29,053 3,146 Montana 1,084,225 59,678 3,331 532 Nebraska 1,961,504 58,120 7,213 629 Nevada 3,104,614 277,252 17,720 2,332 New Hampshire 1,377,529 51,627 2,127 217 New Jersey 9,288,994 854,690 45,949 6,376 New Mexico 2,117,522 146,434 7,030 1,720 New York 20,201,249 2,387,463 131,120 13,953 North Carolina 10,439,388 767,595 34,810 4,879 North Dakota 779,094 38,159 3,213 311 Ohio 11,799,448 834,285 50,132 7,742 Oklahoma 3,959,353 399,325 19,143 3,849 Oregon 4,237,256 187,332 9,485 1,362 Pennsylvania 13,002,700 1,059,848 69,080 11,535 Rhode Island 1,097,379 122,016 5,691 1,004 South Carolina 5,118,425 395,715 15,746 2,795 South Dakota 886,667 59,947 3,972 634 Tennessee 6,910,840 280,338 16,458 3,727 December 2021 RESEARCHREPORT 34 p Texas 29,145,505 2,263,058 139,812 17,923 Utah 3,271,616 174,502 6,562 507 Vermont 643,077 23,344 996 114 Virginia 8,631,393 713,673 36,765 6,350 Washington 7,705,281 459,922 19,965 2,585 West Virginia 1,793,716 180,819 9,827 1,880 Wisconsin 5,893,718 303,834 16,132 2,615 i 576,851 31,700 1,760 262 Total United States 331,449,281 25,324,302 1,384,260 213,349 Source: Population estimates from the 2020 Census. ASPE analysis of multiple data sources for December 2020-July 2021 December 2021 RESEARCHREPORT 35 Table S-7. Estimated Associated Reductions in COVID-19 Related Outcomes by Social Vulnerability Index Social heron Population Cases (%) Hospitalizations (%) Deaths (%) Category Theme 1: Socioeconomic (0019). 26.47% | 6,221,288 | 25.57% | 299,953 | 21.67% 45,237 | 21.20% Lexi 24.43% | 6,189,630 | 24.44% | 327,638 | 23.67% 47,226 | 22.14% (0.20-0.39) 0.40.0'59) 23.67% | 6,315,098 | 24.94% | 398,169 | 28.76%| 51,819 | 24.29% ee 16.78% | 4,501,861 | 17.78% 246,856 | 17.83% 45,479 | 21.32% (0.60-0.79) Very High ery e 8.65% | 2,096,425| 8.28%| 111,645| 8.07% 23,589 | 11.06% (0.80-1.00) Theme 2: Household Composition or Disability Very Low 44.83% | 11,508,946 | 45.45% | 615,049 | 44.43% 87,440 | 40.98% (0-0.19) po 23.59% | 6,108,288 | 24.12%| 334,683 | 24.18% 48,018 | 22.51% (0.20-0.39) M oderate 14.15% | 3,534,841| 13.96%| 202,963 | 14.66% 34,261 | 16.06% (0.40-0.59) High Ib 11.86% | 2,856,746 | 11.28% | 165,681 | 11.97% 28,476 | 13.35% (0.60-0.79) Very High 5.58% | 1,315,481 | 5.19% 65,884 | 4.76% 15,154| 7.10% (0.80-1.00) Theme 3: Minority Status or Language Vv ery Low 6.80% | 1,477,448} 5.83% 65,535 | 4.73% 15,266| 7.16% (0-0.19) Low : o , , . o , ; ) ; ; () eee 6.70% | 1,471,022} 5.81% 71,521 | 5.17% 14,804| 6.94% Moderate 10.22% | 2,348,469| 9.27% 102,231 | 7.39% 20,900| 9.80% (0.40-0.59) ee 20.19% | 5,044,663 | 19.92%| 274,195| 19.81% 42,198 | 19.78% (0.60-0.79) Very High 56.09% | 14,982,699 | 59.16% | 870,779 | 62.91% 120,181 | 56.33% (0.80-1.00) Theme 4: Housing Type or Transportation (019) 11.23% | 2,680,947| 10.59% | 107,398! 7.76% 20,667| 9.69% December 2021 RESEARCH REPORT 36 Table S-7. Estimated Associated Reductions in COVID-19 Related Outcomes by Social Vulnerability Index Social ieee popu ction Cases (%) Hospitalizations (%) Deaths (%) Category (0.20-0.39) 12.37% | 3,048,956 | 12.04% 128,622 9.29% 25,242 | 11.83% '0.40.0.55) 16.55% | 4,002,013 | 15.80% 196,960 | 14.23% 32,088 | 15.04% (0.20-0.79) 28.17% | 7,230,523 | 28.55% 455,584 | 32.91% 61,259 | 28.71% 080.1 00) 31.68% | 8,361,863 | 33.02% | 495,697 | 35.81% 74,094 | 34.73% Notes: CDC's Social Vulnerability Index (SVI) values range from 0 (least vulnerable) to 1 (most vulnerable). Population data from the 2019 American Community Survey. Source: ASPE analysis of multiple data sources for December 2020-July 2021 December 2021 RESEARCHREPORT 37 Table S-8. Valuing Morbidity and Mortality Risk Reductions Estimated Outcome Associated Reduction VSL or VSC Value of Risk Reduction Mortality 213,349 $11,501,365 $2,453,807,201,000 Morbidity Mild 23,940,041 $5,846 $139,963,056,240 Severe 982,825 $13,104 $12,878,938,613 Critical 188,086 $1,814,400 $341,263,844,956 Total Value of Mortalityand Morbidity Risk Reductions $2,947,913,040,809 Notes: VSL denotes value of statistical life; VSC denotes value of statistical case. Value of risk reduction is estimated by multiplying the number reduced by the VSL or VSC. All cases are considered mild and are adjusted for reductions in mortality and hospitalizations. Specifically Mild Morbidity is calculated by subtracting 1,384,260 hospitalizations from 25,324,302 cases to avoid double-counting. Critical Morbidity is assumed to be associated with ICU admissions and is calculated assuming that 29 percent of all hospitalizations are ICU admissions and by subtracting 213,349 deaths to avoid double -counting. December 2021 RESEARCH REPORT 38 Table S-9. Sensitivity Analysis of Valued Mortality and Morbidity Risk Reductions Low Valueof High Value of Low High Risk i ) VSL or VSC VSL or VSC Reduction Reduction RTT Renee, Renee) Estimated Outcome Associated Mortality 213,349 $5.36 million | $17.51 million | $1,145 $3,735 Morbidity Mild 23,940,041 | $2,728 $8,900 $65 $213 Severe 982,825 $6,115 $19,947 $6 $20 Critical 188,086 $846,720 $2.76 million | $159 $519 Total Value of $1,376 $4,487 Mortality and Morbidity Risk Reductions Notes: VSL denotes value of statistical life; VSC denotes value of statistical case. Estimates assume that 29 percent of hospitalizations are criticaland require ICU admission. Value of risk reduction is estimated by multiplying the number reduc ed by the VSL or VSC. All cases are considered mild and are adjusted for reductions in mortality and hospitalizations. Specifically, Mild Morbidity is calculated by subtracting 1,384,260 hospitalizations from 25,324,302 cases to avoid double-counting. Critical Morbidity is assumed to be associated with ICU admissions and is calculated assuming that 29 percent of all hospitalizations are ICU admissions and by subtracting 213,349 deaths to avoid double-counting. December 2021 RESEARCHREPORT 39 U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES Office of the Assistant Secretaryfor Planning and Evaluation 200 Independence Avenue SW, Mailstop 434E Washington, D.C. 20201 For more ASPE briefs and other publications, visit: aspe.hhs.gov/reports s'il el ABOUT THE AUTHORS Nicholas Holtkamp is an Economist in the Office of Science and Data Policy at ASPE. Allison Kolbeis a Health Science Policy Analyst in the Office of Science and Data Policyat ASPE. Trinidad Beleche is a Senior Economist in the Office of Science and Data Policy at ASPE. SUGGESTED CITATION Holtkamp, N., Kolbe, A., and Beleche, T. COVID-19 Vaccination Associated with Reductionsin COVID-19 Mortality and Morbidity inthe United States, and an Approach to Valuing these Benefits. Washington, DC: Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services. December 2021. COPYRIGHT INFORMATION All material appearing in this reportis in the publicdomainand may be reproducedor copied without permission; citation as to source, however, is appreciated. For general questions or general information about ASPE: aspe.hhs.gov/about December 2021 RESEARCHREPORT 40