The MEPS is a national probability survey of the noninstitutionalized civilian population of the United States, conducted by the Agency for Healthcare Research and Quality. The MEPS collects data on the financing and use of medical care and is designed to provide nationally representative estimates of health care use, expenditures, sources of payment, and insurance coverage. The 3 components of the MEPS include the Household Component (HC), the Medical Provider Component (MPC), and the Insurance Component (IC). The MEPS-HC survey uses the National Health Interview Survey (NHIS) as its sampling frame. The NHIS is a multipurpose, nationally representative survey of the noninstitutionalized civilian population of the United States administered by the National Center for Health Statistics and the Centers for Disease Control and Prevention. The MEPS-HC draws on the previous year’s NHIS sample through an overlapping panel design; therefore, 2 calendar years of information are collected from each household through in-person interviews (eg, the 1999 MEPS collects information from the 1998 and 1997 NHIS). The MEPS-HC collects detailed self-reported data in-person using a computer-assisted personal instrument. These data include demographic characteristics, household income, health and functional status, health insurance coverage, and access to care. The MEPS-MPC supplements and validates information on medical care events reported in the MEPS-HC by contacting medical providers identified by respondents. Weighted sequential hot-deck imputation is used to estimate missing data on the basis of responses from similar respondents. The MEPS includes detailed data on insurance coverage that allow for estimates of monthly insurance status and type of insurance coverage (eg, private, Medicaid, Medicare) for each respondent in a given survey year.
We analyzed data from the 24 149 children (younger than 18 years) who were sampled in the 1999–2001 MEPS linked to the 1997–2000 NHIS. Consistent with previous studies and given the limitations of the MEPS and the NHIS,25–28
chronic conditions included asthma, attention-deficit/hyperactivity disorder, arthritis, autism, cerebral palsy, congenital heart disease, cystic fibrosis, diabetes, mental retardation, muscular dystrophy, and sickle cell disease as determined by asking parents, “Has a doctor or other health professional told you that your child has. . .?” As in other studies,29
children were classified into the following racial/ethnic groups: non-Hispanic white, black, Hispanic, Asian, Alaskan/Pacific Islander, and Native American. Because of the small number of children in the Asian, Alaskan/Pacific Islander, and Native American groups, we had inadequate power to conduct analyses for each of these specific subgroups. On the basis of household income in relation to the federal poverty level, children were classified into the following categories: poor (< 100% of federal poverty level), near-poor (between 100% and 200% of the federal poverty level), middle income (200%–400% of the federal poverty level), and high income (< 400% of the federal poverty level). Within the NHIS, child participation in TANF is ascertained by asking adult respondents whether the child had received TANF or related public cash assistance payments during the year before the interview.30
TANF recipients included any child who was reported to have received TANF or related public cash assistance payments; all others were considered nonrecipients. Other covariates included age, gender, and maternal education. The outcome variable was insurance status classified as continuously insured for the entire year (12 months), uninsured for the entire year, or lapses in coverage. Children who were uninsured for at least 1 month in the survey year were considered to have lapses in insurance coverage. For insured children, each month of coverage was classified as private (employer-based insurance) versus public (Medicaid, State Child Health Insurance Program, Medicare, etc).
To obtain nationally representative estimates, we used MEPS person-level weights, provided by Agency for Healthcare Research and Quality, that reflect population distributions and account for the household probability of selection, ratio adjustment to national population estimates at the household level, and adjustment for nonresponse. Estimates of variability were obtained using a Taylor Series estimation approach. Variance estimation strata and primary sampling unit variables were provided with the MEPS-HC data.
We used χ2 tests to compare distributions of categorical covariates (eg, gender, ethnicity, poverty) between children with continuous insurance coverage and lapsed coverage and those who were uninsured. We used t tests to compare differences in mean age between these groups.
Multivariate ordinal logistic regression models were used to examine the independent association of race, poverty, and presence of chronic conditions with the likelihood of insurance coverage (continuous coverage, uninsured, or lapsed) after taking into account differences in age, gender, maternal educational, TANF status, and year of survey. Statistical tests were 2-tailed and were performed on the nationally weighted representative population using SAS-callable SUDAAN. This study was exempted from review by the Institutional Review Board of The Children’s Hospital of Philadelphia.