Developing Weighting Methodology for an Epidemiological Study Among American Indians: The Strong Heart Liver Study

• Activity Number: 289 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jean Leidner* and Sixia Chen and Michael Middleton and Claude Sirlin and Walter Henderson and Justin Dvorak and Tauqeer Ali and Alvin C Silva and Jason G Umans and Shelley A Cole and Ying Zhang
• Companies: University of Oklahoma Health Sciences Center and University of Oklahoma Health Sciences Center and University of California, San Diego and University of California, San Diego and University of California, San Diego and University of Oklahoma Health Sciences Center and University of Oklahoma Health Sciences Center and Mayo Clinic-Phoenix and Georgetown University Medical Center and Texas Biomedical Research Institute and University of Oklahoma Health Sciences Center
• Keywords: Survey weighting; Stratified random sampling; Multiple imputation; Hepatic steatosis; Fibrosis; American Indians

Abstract

The Strong Heart Liver Study (SHLS, 2018-19) examined metabolic and environmental risks for development of hepatic stenosis and fibrosis in American Indians (AI). SHLS participants were recruited from those of the Strong Heart Study Phase V exam (SHS5), the largest prospective study of cardio-metabolic diseases in AI. Approximately 35% of SHS5 participants met the SHLS inclusion criteria, and 15% enrolled. Although stratified random sampling was used to select the recruitment pool, distributions of risk factors differed between the original population and enrolled sample. One or more stratification and auxiliary variables were missing for 6%-7% of SHS5 participants. Survey weighting and multiple imputation (MI) were used to address these issues. Separate weights were computed for participants in each combination of 3 stratification variables: binge drinking, diabetes, and obesity. Stepwise regression and multiple imputation were used to ensure that the final survey weights, ranging 2.3 to 7.9, accurately reflect the characteristics of the original study population. The developed methods provide a useful tool in epidemiological study of chronic diseases.