Importance: It is important to understand differences in coronavirus disease 2019 (COVID-19) deaths by nursing home racial composition and the potential reasons for these differences so that limited resources can be distributed equitably.

Objective: To describe differences in the number of COVID-19 deaths by nursing home racial composition and examine the factors associated with these differences.

Design, Setting, and Participants: This cross-sectional study of 13 312 nursing homes in the US used the Nursing Home COVID-19 Public File from the Centers for Medicare and Medicaid Services, which contains COVID-19 cases and deaths among nursing home residents as self-reported by nursing homes beginning between January 1, 2020, and May 24, 2020, and ending on September 13, 2020. Data were analyzed from July 28 to December 18, 2020. Exposures: Confirmed or suspected COVID-19 infection. Confirmed cases were defined as COVID-19 infection confirmed by a diagnostic laboratory test. Suspected cases were defined as signs and/or symptoms of COVID-19 infection or patient-specific transmission-based precautions for COVID-19 infection.

Main Outcomes and Measures: Deaths associated with COVID-19 among nursing home residents. Death counts were compared by nursing home racial composition, which was measured as the proportion of White residents.

Results: Among 13312 nursing homes included in the study, the overall mean (SD) age of residents was 79.5 (6.7) years. A total of 51606 COVID-19-associated deaths among residents were reported, with a mean (SD) of 3.9 (8.0) deaths per facility. The mean (SD) number of deaths in nursing homes with the lowest proportion of White residents (quintile 1) vs nursing homes with the highest proportions of White residents (quintile 5) were 5.6 (9.2) and 1.7 (4.8), respectively. Facilities in quintile 1 experienced a mean (SE) of 3.9 (0.2) more deaths than those in quintile 5, representing a 3.3-fold higher number of deaths in quintile 1 compared with quintile 5. Adjustment for the number of certified beds reduced the mean (SE) difference between these 2 nursing home groups to 2.2 (0.2) deaths. Controlling for case mix measures and other nursing home characteristics did not modify this association. Adjustment for county-level COVID-19 prevalence further reduced the mean (SE) difference to 1.0 (0.2) death.

Conclusions and Relevance: In this study, nursing homes with the highest proportions of non-White residents experienced COVID-19 death counts that were 3.3-fold higher than those of facilities with the highest proportions of White residents. These differences were associated with factors such as larger nursing home size and higher infection burden in counties in which nursing homes with high proportions of non-White residents were located. Focusing limited available resources on facilities with high proportions of non-White residents is needed to support nursing homes during potential future outbreaks.