Data were obtained from the 2000–2003 Medical Expenditure Panel Survey (MEPS), a nationally representative subsample of the National Health Interview Survey. The MEPS is an annual survey of families and individuals, their medical providers, and employers across the U.S. (6
). In order to have an adequate number of persons in certain population subgroups, the MEPS oversampled blacks and Hispanics in all years and began oversampling of Asians in 2002 (7
). The sample for the current analysis is comprised of 40,215 adults, with 2,397 (~6%) reporting having only diabetes and 427 (~1%) reporting both diabetes and VI.
Participants were classified as having diabetes if they reported being told by a healthcare provider that they had diabetes. To determine VI status, participants were asked a set of visual impairment questions. Their answers were summarized into a five-category VI status variable: 1, no difficulty seeing; 2, some difficulty seeing, can read newsprint; 3, some difficulty seeing, cannot read newsprint, can recognize familiar people; 4, some difficulty seeing, cannot read newsprint, cannot recognize familiar people, but is not blind; and 5, blind. In order to preserve a large enough sample for comparing individuals with both diabetes and VI, participants falling into categories 2–5 were combined into one single category designated as “any VI.” Sampling weights were applied in order to ensure that the resulting sample was nationally representative of the U.S. population and include adjustment for oversampling of race/ethnic groups and survey nonresponse.
In 2000, MEPS added the EuroQol 5D (EQ-5D), a preference-based health status classification scale that asks participants to report the degree of problems (none, mild, and severe) encountered across five domains: mobility, self-care, daily activities, pain, and anxiety/depression. EQ-5D generates a health-related quality of life (HRQL) score between 0 (death) and 1 (perfect health), using U.S. health utility preference weights (8
). For the purposes of illustration, consider a person in otherwise perfect health who cannot walk because of diabetes-related complications and who has a HRQL score of 0.6. This indicates that for the average person who cannot walk but is in otherwise perfect health, each year of life lived is valued at 60% of that of a person who can fully walk (9
). This HRQL score can be used to multiply the average life expectancy for an individual (based on age, sex, race, and disease status) to determine the adjusted number of remaining years of perfect health he/she is expected to have. Continuing with the example above, if this individual has an HRQL score of 0.6 at age 60 and is expected to live for 10 more years, the remaining quality-adjusted life expectancy at age 60 is 0.6 × 10 = 6 QALYs.
For this study, individual-level EQ-5D scores from the MEPS were summarized by race (black or white) and diabetes/VI status (no diabetes/no VI, diabetes only, VI only, and diabetes plus VI). Multivariate analyses were conducted to estimate differences in EQ-5D scores across the sample adjusting for age, sex, and race. To calculate life expectancy, life tables based on the general U.S. population were used by which the risk of mortality was adjusted by age, sex, race, diabetes, and VI status. Following the method outlined in Muennig and Gold (10
), we multiplied the total number of person-years in each age category by the age-specific EQ-5D score to get quality-adjusted life expectancy. Differences in quality-adjusted life expectancy by race and diabetes/VI status represent the incremental QALYs (i.e., marginally greater years of optimal health) experienced by whites relative to blacks.
All statistical analyses were performed using SAS 9.2 (SAS Institute Inc., Cary, NC). To account for the complex survey design of the MEPS data, analyses were completed with adjustments for sample weights and design effects. The University of Miami institutional review board reviewed and approved this study.