This study evaluates and compares between fixed and mixed effects logistic regression models to estimate the performance of health care providers in Dutch nursing homes, using secondary data from a 2002 cross-sectional survey of 42 nursing homes. The key factors examined include dietary intake, mobility, activity, sensory perception, moisture, ward specifications, and friction-and-shear. These factors are examined in relation to the prevalence of pressure ulcers among elderly residents. The study aimed to assess the statistical power and false positive control of fixed and mixed effect models using the Wald statistic, Likelihood Ratio Test (LRT), and Empirical Bayesian Estimates (EBEs). It also aimed at investigating the influence of shrinkage on inference accuracy. The study results revealed significant variability in coefficients between the two models, with the mixed effects model showing narrower confidence intervals and better control of false positives than the fixed effects model, particularly when using the EBE-based approach. The mixed effects model also identified three nursing homes with lower-quality care that were not detected by the fixed effects model. EBE effectively accounted for the number of patients in each institution thus ensuring accurate performance assessments by shrinking estimates for smaller settings toward the mean. Furthermore, the results suggest that the mixed effects model is a more reliable choice for statistical testing in covariate analysis. However, further research is recommended to refine EBE under heteroscedasticity assumptions and address its limitations. Besides, alternative methods such as Hierarchical Bayes Estimates (HBEs) should be explored for enhanced comparison and performance evaluation accuracy