BACKGROUND:: Lives saved predictions are used to quantify the impact of certain remedial measures in nurse staffing and patient safety research, giving an indication of the potential gain in patient safety. Data collected in nurse staffing and patient safety are often multilevel in structure, requiring statistical techniques to account for clustering in the data. OBJECTIVE:: The purpose of this study was to assess the impact of model specifications on lives saved estimates and inferences in a multilevel context. METHODS:: A simulation study was carried out to assess the impact of model assumptions on lives saved predictions. Scenarios considered were omitting an important covariate, taking different link functions, neglecting the correlations coming from the multilevel data structure, and neglecting a level in a multilevel model. Finally, using a cardiac surgery data set, predicted lives saved from the random intercept logistic model and the clustered discrete time logistic model were compared. RESULTS:: Omitting an important covariate, neglecting the association between patients within the same hospital, and the complexity of the model affect the prediction of lives saved estimates and the inferences thereafter. On the other hand, a change in the link function led to the same predicted lives saved estimates and standard deviations. Finally, the lives saved estimates from the two-level random intercept model were similar to those of the clustered discrete time logistic model, but the standard deviations differed greatly. CONCLUSIONS:: The results stress the importance of verifying model assumptions. It is recommended that researchers use sensitivity analyses to investigate the stability of lives saved results using different statistical models or different data sets.