State and territorial epidemiologists from 56 US jurisdictions submitted COVID-19 case reports to the Centers for Disease Control and Prevention (CDC) using a standard COVID-19 case report form; submissions were through direct data entry or comma separated value (CSV) files upload into CDC’s Data Collation and Integration for Public Health Event Response HHS Protect platform or through the National Notifiable Diseases Surveillance System (14–16). Jurisdictions included all 50 states, New York City (reporting separately from New York state), the District of Columbia (DC), and 4 territories (Guam, Northern Mariana Islands, Puerto Rico, and the US Virgin Islands). Jurisdictions voluntarily reported confirmed and probable COVID-19 cases based on the case definition promulgated by the Council of State and Territorial Epidemiologists (CSTE) at the time of reporting (15,16) (Appendix). Health officials in each jurisdiction determine the manner in which they obtain race and ethnicity information; common methods include interviews with patients and their families and reviews of medical records.

To create the hospitalization dataset, we identified jurisdictions in which >80% of case reports had the CDC hospitalization query not missing (i.e., answered yes, no, or unknown) for illnesses during May 1–December 1, 2020. If the date of illness onset was not provided or the person had no symptoms, the earliest date provided was used; this date was typically either the date the positive specimen was collected or the date of the case report. We used the same restriction to create the deaths dataset, requiring >80% of case reports with the CDC death query not missing for illnesses during May 1–December 1, 2020. This approach resulted in some variation in jurisdictions included in the hospitalization and death datasets. We then excluded from the hospitalization database any case record with missing information on hospitalization and excluded from the death database any case record with missing information on death.

For the hospitalization rate calculations, all numerators included confirmed and probable COVID-19 cases according to the CSTE definition at the time of reporting (Appendix). We calculated a lower bound for the rate by including in the denominator as not hospitalized those cases with unknown checked for hospitalization on the reporting form (N1 = 2,479,423). We calculated an upper bound for the rate by excluding from the denominator cases with unknown checked for hospitalization (n1 = 687,527 [27.7% of N1]). For the death rate calculations, all numerators included confirmed and probable COVID-19 cases according to the CSTE definition at the time of reporting (Appendix). We calculated a lower bound for the rate by including in the denominator as alive those cases with unknown checked for death on the reporting form (N2 = 4,708,444). We calculated an upper bound for the rate by excluding from the denominator cases with unknown checked for death (n2 = 756,133 [16.1% of N2]).

We chose a study period that reduced possible biases from variability in testing availability and data completeness. For each jurisdiction in the hospitalization analysis, we examined 2 metrics by week: the proportion of CDC’s COVID Electronic Lab Reporting (CELR) weekly testing volume reported as positive, and the proportion of cases reported to CDC that had symptoms using COVID-19 case report data. Those metrics had stabilized by May 1 and remained stable until December 1, 2020, which we chose as the last date included to eliminate any effects from vaccination. A brief survey was sent to health officials in jurisdictions included in the analysis to assess how data were being collected and reported.

We calculated case-hospitalization and case-fatality rates by sex (male, female), age group (<1, 1–4, 5–14, 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, and >85 years), race and ethnicity (American Indian or Alaska Native, not Hispanic or Latino; Asian or Pacific Islander, not Hispanic or Latino; African American or Black, not Hispanic or Latino; Hispanic; other or multiple races, not Hispanic or Latino; White, not Hispanic or Latino; and unknown), hospitalization and intensive care unit (ICU) admission (admitted to ICU; hospitalized but not admitted to ICU; and not hospitalized), and presence of any symptoms. We show race and ethnicity data in a combined variable; the other or multiple races category includes not Hispanic or Latino persons whose race was reported as other or for whom more than one race was reported, and persons whose record had both a racial designation and racial information unknown selected. The unknown category includes cases who have a known ethnicity of not Hispanic or Latino but either unknown or missing race; have a known race but either unknown or missing ethnicity; or have both unknown or missing race and unknown or missing ethnicity. Although included in the unknown category in the main analysis, we excluded persons from the race and ethnicity subanalysis if ethnicity was unknown, regardless of race.

To account for differences in age distribution by race and ethnicity, so rates among the groups can be compared, we adjusted hospitalization and fatality rates for race and ethnicity to the age distribution of the largest group in the 2019 national census of the US population: White, not Hispanic or Latino (17). We compared the census population distributions by sex, age group, and race and ethnicity for our subset populations to the full 2019 US census to determine whether our subsets were representative of the United States (Appendix Table 1) (17). We also describe demographic characteristics of all COVID-19 cases reported to CDC during May 1–December 1, 2020, by sex, age group, race and ethnicity, hospitalization and ICU admission status, presence of any symptoms, and deaths (Appendix Table 2). We used the 2019 US Census midyear estimates because they were the most recent data available at time of analysis.

We describe hospitalization and fatality rates of persons with COVID-19 reported through CDC surveillance, by age group, sex, race and ethnicity, and hospital and intensive care, during May 1–December 1, 2020, before the availability of COVID-19 vaccines and widespread commercial at-home testing. This study estimated rates in large US populations that included 2.4 million COVID-19 cases for measuring hospitalization rates and 4.7 million cases for fatality rates.

Age was a primary driver of SARS-CoV-2 hospitalization and death; rates had a U-shaped curve, being higher in infants, lowest in children 5 to 14 years of age, and highest among persons >65 years of age, confirming previous reports (3,17,18). Other studies have identified older age and underlying medical conditions as risk factors for severe COVID-19 outcomes, including hospitalization, admission to an ICU, requiring mechanical ventilation, and death (3,18). Those findings highlight the importance of vaccination and other mitigation methods to prevent SARS-CoV-2 infection in high-risk age groups.

Fatality rates were lowest for children 1–14 years of age (0.01%), in keeping with previous reports (19). We found fatality rates among infants to be slightly higher (0.05%). We have not found comparable data for infants in other studies. Higher fatality and hospitalization rates in infants may be attributable to other factors, including maternal COVID-19 infection or medical problems that may arise during the first year of life. Hospitalization rates and mortality rates are generally higher among infants than older children, primarily from complications of prematurity and birth defects (20,21). Our findings among infants should be interpreted with caution because only 10 deaths were reported among the 21,331 infant cases.

We found that male persons had higher hospitalization rates than female persons except in the 15–24- and 25–34-year age groups, and that male persons, except infants, had higher fatality rates than female persons. The reason for the disparity by sex is unknown; biologic, behavioral, and psychosocial factors could be involved (22–26). The 2 female age groups with higher hospitalization rates than the male group include female persons of childbearing age. Pregnancy and recent pregnancy may be associated with severe COVID-19 outcomes, including hospitalization and admission to ICUs (27,28).

Both unadjusted and age-adjusted hospitalization and fatality rates showed differences by race and ethnicity; the highest hospitalization rate was for the African American or Black, not Hispanic category, and the highest fatality rates were for the Asian or Pacific Islander and African American or Black, not Hispanic categories. Similar disparities in SARS-CoV-2 infections have been reported in other studies. Data in our analysis did not include information (e.g., socioeconomic status and occupation) to enable further examination of the reasons for these disparities (7). One study found no association between race and ethnicity category and severe COVID-19 outcomes after primary COVID-19 vaccination series (receipt of 2 doses), which suggests that access to COVID-19 vaccines can help mitigate racial and ethnic disparities (18).

We found that almost half of persons admitted to an ICU died, similar to studies of <400 persons conducted in Seattle (50%), Washington (52%), and New York City (78%) (22,29,30). This finding emphasizes the importance of early diagnosis and treatment of SARS-CoV-2 infection. Mechanical ventilation is often associated with high fatality rates (22,26,31).

One limitation of this analysis is the use of data from only 21 jurisdictions for hospitalization rate calculations and 22 for fatality rate calculations of the 56 jurisdictions that reported case information to CDC. However, the inclusion of only those jurisdictions is also a strength of the analysis because a relatively high percentage of case reports had valid, nonmissing hospitalization or death data. A high percentage of case reports were for persons with unknown race and ethnicity, in large part because reports without ethnicity were classified as unknown even if race was recorded (32); a smaller percentage were categorized as having other or multiple races. Had ethnicity been known, the hospitalization rate for racial and ethnic groups with added cases would be lower, and the effect on death rate would vary by group. Public health authorities have made recommendations to improve race and ethnicity reporting in surveillance data (33,34). Our comparisons of age-adjusted rates among racial and ethnic groups do not adjust for differences in prevalence of underlying diseases. Tests were performed only on persons who sought testing or medical care; therefore, cases in persons with no or mild symptoms were less likely to be identified and reported. Differences in the application of death definitions across jurisdictions highlight the need for standardized definitions during a pandemic; in December 2021, CSTE released a definition of COVID-19–associated deaths, developed with input from jurisdictional health departments (35). Limitations of using surveillance data to estimate fatality rates include preferential ascertainment of severe infections that may lead to overestimation of fatality rates, and the effect of specific interventions (e.g., hospitalization or hospitalization at a particular hospital) on survival (36). Although the National Center for Health Statistics has provisional data on COVID-19–related deaths, it does not have information on ill persons who did not die, so rates cannot be calculated from these data. We used surveillance data, which commonly, and in this study, have quality issues, including missing key variables such as race, ethnicity, and symptoms. Because of those limitations, we bounded our primary results by different assumptions on which outcome those with missing data had and we avoided statistical approaches, which, because of our large sample size, would have led to misleadingly narrow confidence bounds. Furthermore, the large sample size ensures that many differences would be statistically significant even if those differences may be of little practical significance.

Caution is needed before drawing conclusions from our study population. One might expect concerns with the quality of data collected during an emergency response above those in other surveillance data. Yet, such surveillance data can provide information that would not otherwise be available (37). Despite the limitations of our study dataset, the inclusion of data from most of the country, including many persons of all ages, races, and ethnicities, and our method of accounting for missing information in the analyses make these estimates valuable to the public health community.

This analysis presents case-hospitalization and case-fatality rates by age group, sex, and racial and ethnic groups before the introduction of vaccinations targeting SARS-CoV-2. The introduction of vaccinations and the presence of new strains of SARS-CoV-2 altered those rates. Moreover, the wide availability of at-home tests for detecting infection and the lack of national reporting for home testing results make it more likely that only severe infections will be reported. Our results document the severity of SARS-CoV-2 infections early in the pandemic, provide a baseline for future comparisons, and highlight the importance of preventing severe illness in high-risk populations (e.g., through vaccination, early identification of symptoms, testing, and isolation to prevent transmission). Devising appropriate control measures, with community input, for persons in historically underserved racial and ethnic groups and among adults >65 years of age is especially important (18).