“…Use of multivariate logistic regression in EHR datasets is frequently affected by the facts that (a) a significant number of patient records may not include essential covariates necessary to execute the model, and (b) large‐scale imputation may yield biased results (Aagaard et al, ; Pedersen et al, ; White, Daniel, & Royston, ). To address this limitation, a case–control design has been successfully used for population‐based risk factor estimation in studies utilizing EHR datasets, where cases with complete or almost complete sets of confounding variables are matched randomly with controls in a stratified manner (Castro et al, ; Messmer, Williams, & Williams, ). Thus, for the second aim of this study, that is, the identification of risk factors for peri‐implantitis, a nested case–control design was used, as described above.…”