Describing Intersectional Health Outcomes: An Evaluation of Data Analysis Methods

Area of Science:

• Quantitative Social Science
• Epidemiology
• Health Disparities Research

Background:

• Intersectionality frameworks are increasingly used in quantitative research.
• Methods for analyzing disparities across numerous social identity intersections require evaluation.

Purpose of the Study:

• To evaluate the accuracy of different quantitative methods for intercategorical intersectional analysis.
• To compare estimation accuracy for continuous outcomes across various statistical approaches.
• To assess method performance in simulated epidemiologic data and a real-world health dataset.

Main Methods:

• Evaluated cross-classification, regression with interactions, multilevel analysis of individual heterogeneity (MAIHDA), and decision-tree methods (CART, CTree, random forest).
• Used simulated data to assess estimation accuracy of intersection-specific means.
• Applied methods to National Health and Nutrition Examination Study (NHANES) systolic blood pressure data.

Main Results:

• MAIHDA, CTree, and random forest yielded more accurate estimates for high-dimensional intersections, especially at smaller sample sizes.
• CART performed poorly for variable selection and estimation accuracy across sample sizes.
• Different methods produced varying systolic blood pressure estimates in the NHANES example, emphasizing method selection importance.

Conclusions:

• Identified more accurate methods for estimating intersectional health outcomes and disparities across different sample sizes.
• Highlighted that method choice significantly impacts results in intersectional analyses.
• Suggested pairing methods to address challenges in analyzing complex health disparities.