C Siordia1. 1. C. Siordia, Department of Epidemiology, 130 North Bellefield Ave, University of Pittsburgh, Pittsburgh, PA 15213. Phone: +1 (412) 383-1708. E-mail: cas271@pitt.edu.
Abstract
BACKGROUND: Item allocation (the assignment of plausible values to missing or illogical responses in survey studies) is at times necessary in the production of complete data sets. In the American Community Survey (ACS), missing responses to health insurance coverage questions are allocated. OBJECTIVES: Because allocation rates may vary as a function of compositional characteristics, this project investigates how seven different health insurance coverage items vary in their degree of allocation along basic demographic variables. METHODS: Data from the ACS 2010 1-year Public Use Microdata Sample file are used in a logistic regression model and to calculate allocations rates. RESULTS: The findings reveal that: males; people aged 65 and older; those who speak English "very well" or "well"; US citizens; those out-of-poverty; and all racial/ethnic minority groups have higher odds of experiencing a health insurance item allocation relative to their counterparts. CONCLUSIONS: Since health insurance coverage allocations vary by demographic characteristics, further research is needed to investigate their mechanisms of missingness and how these may have implications for frailty related research.
BACKGROUND: Item allocation (the assignment of plausible values to missing or illogical responses in survey studies) is at times necessary in the production of complete data sets. In the American Community Survey (ACS), missing responses to health insurance coverage questions are allocated. OBJECTIVES: Because allocation rates may vary as a function of compositional characteristics, this project investigates how seven different health insurance coverage items vary in their degree of allocation along basic demographic variables. METHODS: Data from the ACS 2010 1-year Public Use Microdata Sample file are used in a logistic regression model and to calculate allocations rates. RESULTS: The findings reveal that: males; people aged 65 and older; those who speak English "very well" or "well"; US citizens; those out-of-poverty; and all racial/ethnic minority groups have higher odds of experiencing a health insurance item allocation relative to their counterparts. CONCLUSIONS: Since health insurance coverage allocations vary by demographic characteristics, further research is needed to investigate their mechanisms of missingness and how these may have implications for frailty related research.