USING INDIVIDUAL-LEVEL RACE AND ETHNICITY TO IMPROVE REPRESENTATIVENESS IN CLAIMS-BASED SURVEY SAMPLING FOR A SURVEY OF WOMEN WITH VASOMOTOR SYMPTOMS OF MENOPAUSE
Author(s)
Judith J. Stephenson, SM, Brian J Harty, MA, Mia Si, PhD, Rebecca Conto, MA, Biruk Sisay, MS, Amanda Marshall, MBA, Sarah E. Daugherty, PhD;
Carelon Research, Wilmington, DE, USA
Carelon Research, Wilmington, DE, USA
OBJECTIVES: Identifying survey samples representing the racial and ethnic distribution of the population of interest can be challenging using administrative claims data as the sampling frame. This study aimed to develop and implement a sampling strategy incorporating individual-level race and ethnicity to generate a more demographically aligned survey sample of women for a vasomotor symptom (VMS) study.
METHODS: The Healthcare Integrated Research Database (HIRD®), a large administrative claims database, served as the sampling frame. Historically, race and ethnicity were unavailable, contributing to survey respondent populations that were disproportionately White, non-Hispanic. An individual-level race and ethnicity variable was integrated into the HIRD in 2023. We developed a stratified random sampling plan using the 2023 US Census distribution for women 40-65 years as a reference population and applied overadjustments based on historic race- and ethnicity-specific rates derived from six previously conducted HIRD-based surveys. Our aim was to mitigate systematic race and ethnicity underrepresentation in a target sample of 800 completed surveys from commercially insured women, aged 40-65 years. The sample list used for fielding the survey contained 74,775 eligible women.
RESULTS: Compared with prior Carelon Research survey respondent populations (82%-97% White, non-Hispanic), the VMS survey achieved a markedly more diverse respondent population. Of 828 completed surveys, 64.7% of women were classified as White, non-Hispanic, and 35.3% were classified as other racial and ethnic groups. This sampling strategy, incorporating historic differential rates, substantially improved alignment with the national reference population (59.5% White, Not Hispanic or Latino, and 40.5% other racial and ethnic groups).
CONCLUSIONS: Integration of individual-level race and ethnicity data with administrative claims data, and the application of historical survey metrics resulted in a more representative and inclusive survey sample. This approach yielded a substantially more diverse respondent population than prior claims-only methods.
METHODS: The Healthcare Integrated Research Database (HIRD®), a large administrative claims database, served as the sampling frame. Historically, race and ethnicity were unavailable, contributing to survey respondent populations that were disproportionately White, non-Hispanic. An individual-level race and ethnicity variable was integrated into the HIRD in 2023. We developed a stratified random sampling plan using the 2023 US Census distribution for women 40-65 years as a reference population and applied overadjustments based on historic race- and ethnicity-specific rates derived from six previously conducted HIRD-based surveys. Our aim was to mitigate systematic race and ethnicity underrepresentation in a target sample of 800 completed surveys from commercially insured women, aged 40-65 years. The sample list used for fielding the survey contained 74,775 eligible women.
RESULTS: Compared with prior Carelon Research survey respondent populations (82%-97% White, non-Hispanic), the VMS survey achieved a markedly more diverse respondent population. Of 828 completed surveys, 64.7% of women were classified as White, non-Hispanic, and 35.3% were classified as other racial and ethnic groups. This sampling strategy, incorporating historic differential rates, substantially improved alignment with the national reference population (59.5% White, Not Hispanic or Latino, and 40.5% other racial and ethnic groups).
CONCLUSIONS: Integration of individual-level race and ethnicity data with administrative claims data, and the application of historical survey metrics resulted in a more representative and inclusive survey sample. This approach yielded a substantially more diverse respondent population than prior claims-only methods.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
PCR74
Topic
Patient-Centered Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas