2 Methods
2.1 Data Source
This study used 2005–2009 MEPS data. The MEPS is a large, nationally representative survey which collects information on health care use and spending of the US civilian, non-institutionalized population. The MEPS has an overlapping panel design with five rounds of computerassisted personal interviews over a 2.5-year period. It contains both longitudinal panel and cross-sectional files enabling researchers to perform both longitudinal and cross-sectional analyses. The MEPS uses a stratified, multistage probability design that includes oversampling of minorities and individuals with disabilities. Clusters, strata, and person-level weights are used in MEPS to account for
an individual’s probability of selection in the sample and to generate a nationally representative sample of non-institutionalized persons.
This study used three files from the MEPS data: household component, medical conditions, and prescribed medicine files. The household component file is the core
survey and contains detailed data on demographic characteristics, health conditions, use of medical services, health care expenditures, health insurance coverage, and income
and employment among surveyed individuals and their households. The medical condition file provides information on each participant’s self-reported medical condition
which can be identified using either International Classification Disease-9-Clinical Modification (ICD-9-CM) or Clinical Classification Codes (CCC). Prescription medicines
files are the event-level files where each record represents a prescription medicine that has been purchased by the household respondent. Data in the prescription medication
file of MEPS is self-reported by survey participants, and the information is validated by the pharmacy where the prescription was purchased. All these files have a common
identifier (DUPERSID) which enables linking of these files across years. Additional information pertaining to MEPS can be found elsewhere [23, 24].