Correcting for misclassification for a monotone disease process with an application in dental research

Area of Science:

• Biostatistics
• Epidemiology
• Oral Health Research

Background:

• Longitudinal studies in oral health often involve complex data with misclassification.
• Accurate estimation of prevalence and incidence is crucial for public health interventions.
• Existing models may struggle with unconstrained misclassification and covariate relationships.

Purpose of the Study:

• To evaluate binary Markov models for longitudinal oral health data with misclassification.
• To develop an extended model incorporating covariates and multiple classifiers.
• To assess the models' ability to estimate key epidemiological parameters without external data.

Main Methods:

• Utilized binary Markov models with unconstrained misclassification processes.
• Developed and implemented a Bayesian extension of the binary Markov model.
• Applied the models to a longitudinal oral health study dataset.

Main Results:

• The simple binary Markov model accurately estimates prevalence, incidence, and misclassification parameters without external data.
• Incidence estimators from the model outperformed previously proposed approaches.
• The extended Bayesian model successfully estimated parameters, including covariate relationships, without external information.

Conclusions:

• Binary Markov models are effective for analyzing longitudinal oral health data with misclassification.
• The proposed extended Bayesian model offers a robust framework for incorporating covariates.
• The methods provide reliable estimates of prevalence and incidence, crucial for oral health surveillance and intervention.