Ignoring the cluster effect is a common statistical oversight that is also observed in endodontic research. The aim of this study was to explore the use of multilevel modeling in investigating the effect of tooth-level and patient-level factors on apical periodontitis (AP).Methods
A random sample of digital panoramic radiographs from the database of a dental hospital was evaluated. Two calibrated examiners (κ ≥ 0.89) assessed the technical quality of the root fillings and the radiographic periapical health status by using the periapical index. Descriptive statistical analysis was carried out, followed by multilevel modeling by using tooth-level and patient-level predictors. Model fit information was obtained, and the findings of the best-fit model were reported.Results
A total of 6409 teeth were included in the analysis. The predicted probability of a tooth having AP was 0.42%. There was a statistically significant variability between patients (P < .05). Approximately 53.16% of the variability was accounted for by the patients, leaving 46.84% of the variability to teeth or other factors. Posterior tooth, inadequate root filling, and age were found to be significantly associated with AP (P < .05).Conclusions
Multilevel modeling is a valid and efficient statistical method in analyzing AP data. The predicted probability of a tooth having AP was generally small, but there was great variation between individuals. Posterior teeth and those with poor quality root filling were found to be more frequently associated with AP. On the patient level, advancing age was a factor significantly associated with AP.