Multiple imputation of dental caries data using a zero-inflated Poisson regression model

Area of Science:

• Dental Public Health
• Biostatistics
• Epidemiology

Background:

• Dental caries data often exhibit excess zeros, posing challenges for statistical analysis.
• Missing data in dental caries assessments require robust imputation methods.

Purpose of the Study:

• To demonstrate a straightforward technique for multiple imputation of missing dental caries data.
• To apply a zero-inflated Poisson (ZIP) regression model for handling excess zeros in caries data.

Main Methods:

• Utilized a dataset of 24,403 children from a North Carolina dental program.
• Estimated a ZIP regression model using non-missing caries data (n=17,766).
• Employed model coefficients to predict and impute missing caries counts.

Main Results:

• The imputation technique produced caries counts with distributions similar to the original data.
• The method effectively preserved the excess zeros characteristic of the non-missing caries data.
• Imputed data demonstrated comparable distributional properties to observed data.

Conclusions:

• The demonstrated zero-inflated Poisson regression technique offers a simple and effective solution for imputing missing dental caries data.
• This method is easily applicable in dental research and public health surveillance.
• Accurate imputation of caries data, especially with excess zeros, enhances the reliability of dental health assessments.