“Excess Mortality” During COVID-19 Varied by Race, Ethnicity, Geography

Mortality data used in this research come from an anonymized version of the Social Security Administration’s Numerical Identification (Numident) database, which contains date of birth, date of death and sex information for all individuals ever assigned a Social Security Number (SSN) (Figure 1).

Self-reported race/ethnicity and last known state of residence are found by linking individuals in the Numident to their responses to the 2000 and 2010 decennial censuses, 2001 to 2019 responses to the American Community Survey (ACS) and other administrative records.

Those with or without an SSN who could not be linked to a decennial census or ACS response were excluded from the analysis.

Excess mortality is calculated as the difference between the actual mortality rate and the mortality rate we would expect to see in the hypothetical absence of, in this case, the COVID-19 pandemic.

We drew mortality data for the April 1, 2020 to March 31, 2021 period from Numident death information. We estimated expected mortality by modeling the actual rates from April 1, 2010 to March 31, 2020 and extrapolating this decade-long trend into the pandemic era.

If the COVID-19 pandemic had not occurred, we’d expect a national mortality rate of approximately 9.3 persons per 10,000 per month (Figure 2). The rate of persons per 10,000 per month is referred to as “person-months” or “PMs”.

But the mortality rate actually observed in the first year of the pandemic was 11.1 per 10,000 PMs, resulting in a national excess mortality rate of 1.8 per 10,000 PMs or 216 per 100,000 persons per year. (The full paper provides details on how and why these estimates differ from National Center for Health Statistics (NCHS) estimates.)

By linking various population data sources, we can estimate excess mortality for detailed population subgroups.

Census Bureau data, which gathers self-reported race and ethnicity responses, is an improvement over the less reliable race and ethnicity assignments on death certificates made by coroners and next-of-kin. We use these responses to categorize individuals into one of seven race and ethnicity groups “alone” (not in combination with another race or ethnicity) or into a multiple race group.

Excess mortality was higher for men of all ages and race/ethnicities and for all those age 65 and older (Figure 3).

But differences by race/ethnicity were particularly striking:

• Excess mortality among non-Hispanic AIAN, Hispanic, and non-Hispanic Black individuals 65 and over far surpassed rates of their non-Hispanic White counterparts.
• Similarly, while excess mortality among non-Hispanic White individuals was almost entirely confined to those 65 and older, we found evidence of substantial excess mortality among younger working-age (ages 25-64) Hispanic, non-Hispanic Black and non-Hispanic AIAN populations. For example, while the excess mortality rate for non-Hispanic White men ages 65-74 was 3.2 per 10,000 PMs, their younger non-Hispanic AIAN counterparts ages 45-54 experienced a comparable rate of 3.7 per 10,000.

With or without the pandemic, we would expect to see differences in mortality among racial and ethnic groups, which complicates comparisons. For example, the average non-Hispanic White person in the United States is older than the average Hispanic individual and, because mortality increases with age, mortality rates tend to be higher on average for non-Hispanic White populations than for Hispanic populations.

The methodology we used to control for these differences applies group-specific mortality rates to the national age and sex distribution. The resulting adjusted mortality rates tell us how excess mortality would differ across racial/ethnic groups if all groups had the same age and sex structures.

Age- and sex-adjusted mortality rates (Figure 4) show that overall, the non-Hispanic AIAN population experienced the highest rates (4.7 per 10,000 PMs) of excess all-cause mortality during the first year of the pandemic, followed by the Hispanic (3.6 per 10,000 PM) and non-Hispanic Black (3.5 per 10,000 PM) populations.

In contrast, excess mortality was lowest among the non-Hispanic White and non-Hispanic Asian populations (1.4 and 1.5 per 10,000 PM, respectively).

However, the Hispanic population experienced the biggest increase in mortality during the pandemic — 49.1% higher than expected based on pre-pandemic mortality trends.

Excess mortality also rose among the non-Hispanic Some Other Race (35.1%); non-Hispanic AIAN (34.6%); and non-Hispanic Black (31.4%) populations.

The non-Hispanic White population saw the smallest increases (14.6%) relative to expected mortality.

Periodic spikes in excess mortality corresponded with shifts in the geographic dispersal of COVID-19.

During the pandemic’s first entire year, New York, New Jersey, Louisiana, Mississippi, and Washington, D.C. experienced the highest excess mortality (Figure 5).

Much of the excess mortality in New York and New Jersey occurred during the pandemic’s first wave (April 1-June 30, 2020). In the second wave, excess mortality dispersed to the deep South, particularly Mississippi and Louisiana. By wave 3, however, excess mortality was common across the South, had expanded to the Southwest and was also elevated in the Midwest.

In the decade leading up to the pandemic, the national mortality gap between the Black and White populations had narrowed substantially but still favored non-Hispanic White individuals.

The Hispanic population, however, held a long-standing mortality advantage over the non-Hispanic White population.

But the pandemic reversed improvements in the Black-White mortality gap and completely wiped out the Hispanic mortality advantage.

The Black-White and Hispanic-White mortality gaps shifted by an average of 2 deaths per 10,000 PMs in favor of the non-Hispanic White population in virtually all states during the pandemic’s entire first year (Figures 6 and 7).

These findings point to the social and economic determinants of health, which also contributed to racial and ethnic mortality gaps for years before the pandemic began and drove disproportionate increases in excess mortality among racial and ethnic minorities in the pandemic’s first year.