Background: The COVID-19 pandemic is quickly spreading throughout Brazil, which is rapidly ascending the ranking of countries with the highest number of cases and deaths. A particularly unstable federal regime and fragile socioeconomic situation is likely to have contributed to the impact of the disease. Amid this crisis there is substantial concern in the possible socioeconomic, geopolitical and ethnic inequity of the impact of COVID-19 on the country's particularly diverse population. Methods: We performed a cross-sectional observational study of COVID-19 hospital mortality using observational data from the SIVEP-Gripe dataset. We present descriptive statistics to quantify the COVID-19 pandemic in Brazil. We assess the importance of regional factors such as education, income and health either on a state-by-state basis or by splitting Brazil into a North and a Central-South region. Mixed-effects survival analysis was used to estimate the effects of ethnicity and comorbidity at an individual level in the context of regional variation. Findings: Our results show that, compared to branco comparators, hospitalised pardo and preto Brazilians have significantly higher risk of mortality, with hazard ratios and 95% CI of 1.47 (1.33-1.58) and 1.32 (1.15-1.52), respectively. In particular, pardo ethnicity was the second most important risk factor (after age). We also found that hospitalised Brazilians in North regions tend to have more comorbidities than in the Central-South, with similar proportions between the various ethnic groups. Finally, we found that states in the North have a higher hazard ratio as compared to the Central-South, and that Rio de Janeiro obtained one of the highest hazard ratios, similar to the ones of the more underdeveloped Pernambuco and Amazonas. Interpretation: Our results can be interpreted according to the interplay of two independent, but correlated, effects: i) mortality by COVID-19 increases going North (vertical effect), ii) mortality increases for the pardo and preto population (horizontal effect). We speculate that the vertical effect is driven by increasing levels of comorbidity in Northern regions where levels of socioeconomic development are lower, whereas the horizontal effect may be related to lower levels of healthcare access or availability (including intensive care) for pardo and preto Brazilians. For most states the vertical and horizontal effects are correlated giving a larger cumulative mortality. However, Rio de Janeiro was found to be an outlier to this trend: It has an ethnicity composition (horizontal effect) similar to the states in the North region, despite high levels of socioeconomic development (vertical effect). Our analysis motivates an urgent effort on the part of Brazilian authorities to consider how the national response to COVID-19 can better protect pardo and preto Brazilians as well as the population of poorer states from their higher death risk from SARS-CoV-2 infection.