Enhancing Socioeconomic Status Prediction for Cavities: A Hybrid Method.

Area of Science:

• Public Health
• Biostatistics
• Social Epidemiology

Background:

• Socioeconomic status (SES) is crucial for understanding health disparities, traditionally measured using principal component analysis (PCA).
• PCA, while data-driven, may lack specificity for particular health outcomes.
• Decision tree analysis (DTA) offers an outcome-specific approach but may not fully capture SES breadth.

Purpose of the Study:

• To investigate if combining DTA and PCA enhances predictive accuracy for SES compared to PCA alone.
• To explore the utility of using significant loading indicators (SLIs) from PCA within a DTA-PCA framework to simplify SES predictors without sacrificing accuracy.
• To assess the predictive accuracy of these novel SES composites for dental caries.

Main Methods:

• Analyzed 12 SES indicators from 2,182 children in the SMILE birth cohort study.
• Developed five SES composites: DTA-only, PCA-only (full component and SLIs), and DTA-PCA (full component and SLIs).
• Evaluated predictive accuracy for dental caries using root mean squared error with 5-fold cross-validation.

Main Results:

• DTA-PCA combined SES composites showed superior predictive accuracy for dental caries over PCA-only methods.
• The DTA-PCA approach using SLIs outperformed PCA with the full component and was non-inferior to the DTA-only method.
• This hybrid method effectively simplified SES predictors while maintaining or improving predictive performance.

Conclusions:

• The DTA-PCA hybrid method provides a more accurate and potentially more precise framework for developing SES composites to predict dental caries.
• Incorporating SLIs in the DTA-PCA approach offers a practical way to create parsimonious yet powerful SES predictors.
• This methodology is adaptable for creating composite predictors from multi-item measurements across diverse health outcomes.