Survey-Weighted Stratified Proportional Allocation for Multiracial Population Estimation: An Application to Hawaiʻi PRAMS

Abstract

Race classification poses a fundamental methodological challenge in population health research, particularly in settings with large multiracial populations such as Hawai'i. Conventional approaches—including standard single-race classification and priority-based assignment—impose arbitrary decisions that may distort population composition estimates and health outcome measures. This study proposes a survey-weighted stratified proportional allocation approach with progressive aggregation, which extends prior proportional weighting methods by incorporating local sociodemographic context and complex survey design features. Using Hawai'i Pregnancy Risk Assessment Monitoring System (PRAMS) data from 2016 to 2022, we compare four race classification methods and evaluate their impact on both population distribution and estimates of smoking during pregnancy. Results show that classification method substantially influences estimated population composition and smoking prevalence, particularly for Native Hawaiian and Other Pacific Islander and American Indian/Alaska Native groups, while adjusted odds ratios remained largely stable across methods. Among multiracial respondents, 99.7% were allocated without relying on global statewide distributions, demonstrating the practical utility of fine-grained stratification. These findings underscore that race classification is a substantive methodological choice with meaningful implications for population measurement and health disparity research.