Background: The National Cancer Database (NCDB) introduced COVID-specific variables in 2020 to assess the impact of the pandemic on cancer care. Variables include (1) any SARS-CoV-2 test and (2) SARS-CoV-2 test result. We evaluated patterns of breast cancer reporting and mortality compared to pre-pandemic. Methods: From the 2020 NCDB breast cancer participant user file, we separately examined reporting and mortality in the years 2017-2020. Pre-pandemic years (2017-2019) were used as a baseline to evaluate secular trends not related to the pandemic. For patients diagnosed in the study period, patient/hospital factors associated with complete reporting (CR) or no reporting (NR) of COVID variables were compared. For patients who died during the study period, yearly mortality rate and overall survival (OS) were examined overall and by stage. Patient/hospital factors were examined for associations with year of death. Results: Fewer breast cancer cases were reported in 2020 (232,023) compared to 2019 (268,210) and 2018 (260,259). In 2020, there were 97,810 (42%) with NR and 116,455 (50%) with CR of COVID variables. On multivariable analysis, patients with NR were more likely to be > 70 years old, of Asian race, with Medicare coverage, lower education quartile, and treated at community hospitals and those in the Middle Atlantic region. There were 60,933 deaths in 2020, 64,502 in 2019, 63,865 in 2018, and 61,930 in 2017. There were significantly fewer stage 4 deaths in 2020 compared to prior (Table). Although there was a steady trend of increasing OS across the 4-year period, the incremental increase from 2019 to 2020 was higher than expected based on previous years. This finding was particularly true for stage 4 patients (median OS: 31.4 months in 2020, 20.2 months in 2019). On multivariable analysis, 2020 mortalities were less likely in patients > 80 years old (OR 0.88, p < .01) or with stage 3 (0.81, p < .01) or stage 4 (0.64, p < .01) disease, and more likely at academic centers (1.05, p < .01) than in 2017-2019. Conclusions: Non-reporting of COVID data is not at random, and is more common in populations at risk. In addition, fewer breast cancer mortalities were reported in 2020, with higher than expected OS. These findings suggest that OS was artificially increased in 2020 due to underreporting of patients who may be at higher risk of mortality. Accurate and complete data is crucial for evaluating the pandemic's impact on patient care. Underreporting of patients who may be most vulnerable limits our ability to comment on the true impact of the pandemic. These limitations must be considered to avoid blind spots in evaluation of pandemic data and risk deepening disparities.

N: number of deaths OS: OS in months.