Black race was the main covariate of interest. Other covariates were as follows: diagnosis of glaucoma (suspected glaucoma—omitted reference group, primary open-angle glaucoma [POAG], narrow-angle glaucoma, other), ability to read newsprint with or without glasses, other ocular comorbidities—DR (background, proliferative, unspecified), cataract, ARMD (dry, wet, unspecified), death during the year, death during the next year, number of complications from diabetes (count of ICD-9-CM 250.xx codes), diagnosed dementia, DxCG score (measure of case mix severity based on all diagnoses/procedures recorded in claims in the year before the observational year),18 age, male sex, years of schooling, family income, number of limitations in activities of daily living, marital status, children living in the elderly person’s home, children living within an hour of the elderly person but not in the home, Medigap coverage, number of months during the year the person was enrolled in Medicaid, number of months during the year the person was enrolled in an HMO, number of ophthalmologists in the primary sampling unit per capita, standard metropolitan statistical area or county for persons not living in a standard metropolitan statistical area, number of optometrists per capita in the primary sampling unit, nonresidence in a standard metropolitan statistical area, living in nursing home, and binary variables for each year. We included interactions between glaucoma and diagnoses of other major eye diseases—cataract, ARMD, and diabetes mellitus. In some specifications, we also included interaction terms for black race and years since year of first diagnosis of glaucoma but excluded these covariates from the results presented because they did not alter the results. When information for a variable from NLTCS was not available, we set the variable to zero and included a binary variable to identify the value as missing. Standard errors were adjusted for repeated observations in the same individuals.