BACKGROUND: Among adults with incident diabetes, data are lacking about first antihyperglycemic initiation and whether medication choice aligns with recommendations. OBJECTIVE: To identify predictors of initiating any antihyperglycemic, and specifically sulfonylurea versus metformin. METHODS: This retrospective cohort study included 241 327 patients from 11 US health systems, 2005 through 2010. Assessments included antihyperglycemic initiation within 6 months of diabetes identification, first medication initiated, and initiation predictors. RESULTS: Only 40.3% (n = 97 350) started any antihyperglycemic; 75.2% (n = 73 221) started metformin. Glycosylated hemoglobin (HbA1c) predicted initiating any antihyperglycemic (HbA1c >9%, relative risk [RR] = 3.94, 95% CI = 3.82, 4.07, vs HbA1c >6.5%-7%). Age modified the HbA1c effect: at higher HbA1c, likelihood of starting antihyperglycemics differed little across ages; at lower HbA1c, older patients were less likely to start antihyperglycemics (P < .001). Individuals with elevated serum creatinine (SCr) were more likely to started on sulfonylurea (SCr = 1.4-2, RR = 2.21 [2.05, 2.39]; SCr >2, RR = 2.75 [2.30, 3.29] vs normal SCr), particularly as HbA1c increased: patients with HbA1c 8%-9% and SCr >2 were 5.59 times (2.94, 10.65) more likely to start sulfonylurea versus those with HbA1c >6.5%-7% and normal SCr. Age predicted sulfonylurea initiation (20-39 years, RR = 0.87 [0.79, 0.95]; >/= 80 years, RR = 2.41 [2.20, 2.65] vs 50-59 years). CONCLUSIONS: Among adults with incident diabetes, metformin was generally the first antihyperglycemic initiated. However, 59.7% did not start any antihyperglycemic at diabetes identification. HbA1c and age predict antihyperglycemic initiation; SCr and age predict sulfonylurea initiation.