The effect of choosing different units of analysis when estimating risk of presence of dental caries in the primary dentition

Abstract

Background: Data on caries are usually collected with the tooth surface or the tooth as the unit, but subsequently analysed by aggregating the data at the level of the individual.

Aims: To evaluate how different units of analysis may affect the result of the statistical analyses in a study of the association between deciduous dental caries and a set of risk factors.

Method: 293 children (mean age: 7.5 yrs) from two primary schools in Dar es Salaam, Tanzania were examined. Thus the study design was cross-sectional. Logistic regression analysis was used to determine the association between caries experience and nutritional status, socioeconomic background, and microbiological and salivary characteristics using the subject, the tooth, and the tooth surface as unit of analysis, allowing for the fact that teeth and surfaces within the same individual must be considered non-independent.

Results: When aggregated data were used, point estimates were larger in some instances. The precision of the estimates increased considerably when the tooth as compared to the individual was used as the unit of analysis. No or limited gain in precision was obtained when the tooth surface as compared to the tooth was used as unit of analysis.

Conclusions: The choice of unit of analysis may strongly influence the result of the statistical analyses and thus the conclusion of an investigation.