Comparative Performance of Four Prevalence Estimators for Untreated Dental Caries: Application to KERCADRS Phase III study
نویسندگان
1 Australian Women and Girls’ Health Research Centre, School of Public Health, Faculty of Medicine, The University of Queensland, Brisbane, QLD, Australia
2 Oral and Dental Diseases Research Center, Kerman University of Medical Sciences, Kerman, Iran
3 Physiology Research Center, Institute of Neuropharmacology, Kerman University of Medical Sciences, Kerman, Iran
4 Endodontology Research Center, Kerman University of Medical Sciences, Kerman, Iran.
5 HIV/STI Surveillance Research Center, and WHO Collaborating Center for HIV Surveillance, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran
6 HIV/STI Surveillance Research Center, and WHO Collaborating Center for HIV Surveillance, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran
doi
10.34172/.johoe.2509.1793چکیده
Background: The prevalence of untreated dental caries (UDC) is a critical indicator in dental public health. This study evaluated four techniques to estimate UDC prevalence: (1) overall prevalence, (2) average individual prevalence, (3) generalized estimating equations (GEE), and (4) random effects models (REM). Methods: A simulation study generated hypothetical populations under two scenarios, with intraclass correlation values of 0.05, 0.1, and 0.2: Scenario 1: UDC prevalence (5%, 10%, 20%) independent of missing teeth; Scenario 2: UDC prevalence dependent on the number of missing teeth. Four estimation methods were compared: 1. Overall Prevalence Estimator: calculated as the total number of UDC divided by total teeth; 2. Average Individual Prevalence Estimator: mean of individual prevalence values, 3. GEE: logistic regression with participant-level clustering effects, 4. REM: random effects logistic regression modeling prevalence at both the individual and tooth levels. Performance was assessed using mean squared error (MSE), bias, confidence interval (CI) coverage, and CI length. For practical implications, the simulation study results were applied to estimate the UDC in Phase III of the Kerman Coronary Artery Disease Risk Study (KERCADRS). Results: When UDC was independent of missing teeth, GEE and the average individual prevalence methods yielded the most reliable estimates (lower MSE and higher CI coverage). When UDC depended on missing teeth, no method performed optimally. However, GEE achieved comparatively better results. Analysis of the KERCADRS data showed a significant correlation between the number of UDC and the number of missing teeth (r = 0.15, P < 0.001). Accordingly, the UDC estimated using the GEE method was 26.3% (95% CI: 25.9%, 26.8%). Conclusion: In contexts where UDC and missing teeth are uncorrelated, GEE and average individual prevalence methods are recommended. When dependencies were present, the GEE method performed slightly better than the other methods. UDC prevalence in Kerman is high, and urgent action is needed to address it.
