Sociodemographic and clinical correlates of HbA1c testing longitudinally at the first and second SEARCH follow-up visits for type 1 diabetes (top) and type 2 diabetes (bottom). Data are OR (95% CI) for reporting ≥3 tests/year at both visits relative to those reporting <3 tests/year at either visit. *Core variables of age at diagnosis, diabetes duration, sex, race/ethnicity, and clinical site were not subject to stepwise selection. Candidate variables included parental education at baseline, household income at baseline, diabetes treatment, diabetes care provider, having a personal doctor, having problems with the cost of care in the prior 12 months, type of health insurance, and having continuous health insurance in the prior year. Nonsignificant core variables and nonselected candidate variables are not shown. †All variables were allowed to compete in the type 2 model because of the small sample size; only diabetes provider was retained.