Background: Mortality amongst nursing home (NH) residents increased by 43% during the first wave of coronavirus disease 2019 (COVID-19). We estimated the ‘contextual effect’ on mortality, tried to explain it by NH characteristics and identified resident- and NH-level risk factors for mortality. Methods: The contextual effect was measured for two cohorts of NH residents managed by the general scheme in metropolitan France (RESIDESMS data from 03/01/2020 to 05/31/2020 and 03/01/2019 to 05/31/2019) by the intraclass correlation coefficient (ICC) estimated from mixed-effects logistic regression. Results: Amongst 385,300 residents (5,339 NHs) included in 2020 (median age 89 years, 25% men), 9.1% died, versus 6.7% of 379,926 residents (5,270 NHs) in 2019. In the empty model, the ICC was 9.3% in 2020 and 1.5% in 2019. Only the geographic location partially explained the heterogeneity observed in 2020 (ICC: 6.5% after adjustment). Associations with mortality were stronger in 2020 than in 2019 for male sex and diabetes and weaker for heart disease, chronic respiratory disease and residence <6 months. Mortality was higher in 2020 (15.1%) than 2019 (6.3%) in NHs with at least one death with a mention of COVID-19 and more heterogeneous (ICC: 8.0%) than in the others (mortality: 6.7% in both years; ICC: 1.1%). Conclusion: Our results suggest that the COVID-19 crisis had a heterogeneous impact on mortality in NH residents and that geographic location explain a part of the contextual effect, which appears to have had little influence on mortality in NHs not being affected by the virus.