Affiliations: [a] Center for Financing, Access and Cost Trends, Agency
for Healthcare Research and Quality, 540 Gaither Road, John M. Eisenberg
Building, Rockville, MD 20850, USA. E-mail: wyu@ahrq.gov
Correspondence:
[*]
Corresponding author
Abstract: The Medical Expenditure Panel Survey Household Component (MEPS-HC) is designed to
provide nationally representative annual estimates of health care use,
expenditures, sources of payment, and insurance coverage for the US civilian
noninstitutionalized population. The expenditure data from MEPS have been shown
to exhibit a marked positive skewness, with a few high expenditure respondents
and many low or zero expenditure respondents. As a consequence of this departure
from the normal distribution, the frequency with which a conventional confidence
interval for a MEPS expenditure estimate will not capture the true population
parameter may be higher than the probability stated for the confidence interval.
Based on repeated sample simulations using data from the 1996 to 2001 MEPS-HC,
this paper evaluates and compares the actual probability achieved for confidence
intervals derived from expenditure data by types of expenditure and varying
sample sizes. The results are also compared to estimated confidence
probabilities obtained from repeated sample simulations for other types of
variables that do not exhibit as marked a positive skewness as health care
expenditures.
