Introduction: COVID-19 vaccinations have reduced COVID-19 cases and mortality for nursing home (NH) residents. However, little is known about how the discriminability of COVID-19 death changed before and after vaccination. The objective of this study is to examine factors that predict COVID-19 death before and after vaccination. Methods: We conducted a retrospective cohort study on NH resident data collected using the Resident Assessment Instrument Minimum Data Set Version 2.0. The cohort included 14977 residents who tested positive for COVID-19 between March 7, 2020, and July 31, 2021. The cohort was split into two groups, COVID-19 deaths before and after January 1st, 2021. Logistic regression, LASSO regression, and random forests methods were used to evaluate the predictive ability of resident characteristics and COVID-19 mortality. Model performance was assessed using the area under the receiver operating characteristics curve (AUC). Variable importance was measured by the change in AUC. Results: Age, sex, diabetes, declining cognition, and deteriorating activities of daily living were the most informative predictors for COVID-19 mortality before and after COVID-19 vaccination. COPD, emphysema, asthma, and emphysema were informative of COVID-19 mortality after vaccination only. The logistic regression, the LASSO regression, and the random forest model display similar predictive ability for COVID-19 mortality in their respective cohorts. A similar discrimination was reached for COVID-19 mortality before and after vaccination (AUC = 0.67, AUC=0.68, and AUC=0.644 respectively). Conclusions: The factors associated with COVID-19 mortality are multifactorial and may be modifiable. Closer attention to these factors may help reduce COVID-19 mortality. Although the discriminability of the models was poor, advanced knowledge of NH resident characteristics can support upstream decision-making to prioritize care for NH residents who are at the greatest risk of COVID-19 death. Future studies are required to validate these Findings and demonstrate the utility of this model in pandemic preventability.