Despite the development of effective pharmacological therapy to prevent both macrovascular and microvascular complications and adverse events, diabetes control remains sub-optimal. Poor adherence to recommended regimens is a causal factor in preventable morbidity and mortality in diabetic patients. Sub-groups of patients with differing longitudinal profiles of adherence may yield differing treatment outcomes. Identifying characteristics associated with longitudinal profiles can potentially alert clinicians to patients at risk for poor clinical outcomes allowing for early intervention and follow-up. Furthermore, few studies have examined the role of adherence improvements as a mediator of intervention effect on glycemic control. In this work we sought to identify patterns, correlates and outcomes of adherence to oral hypoglycemic agents. We also assessed whether adherence improvements mediated a brief interventions effect on glycemic control. Longitudinal analysis via growth curve mixture modeling was carried out to classify 180 patients who participated in an adherence intervention according to patterns of adherence to oral hypoglycemic agents across 12 weeks. Adherence was assessed using the Medication Event Monitoring System. Hemoglobin A1c assays were used to measure glycemic control as the clinical outcome. Individual patient residential data was geo-coded at the tract level. Three patterns of adherence to oral hypoglycemic agents were identified: adherent, increasing adherence, and nonadherent. Both individual and neighborhood level factors were identified that were associated with patterns of adherence. Patients with an increasing adherence pattern were more likely to have a Hemoglobin A1c (HbA1c) < 7% (adjusted odds ratio = 14.52, 95% CI [2.54, 82.99]) at 12 weeks in comparison with patients with the nonadherent pattern. Across the whole sample, longitudinal adherence profiles mediated 35.2 % (13.2, 81.0 %) of the effect of a brief adherence intervention on glycemic control [from odds ratio (OR) = 8.48, 95 % confidence interval (CI) (3.24, 22.2) to 4.00, 95 % CI (1.34, 11.93)]. These findings imply that the identification of patients with type 2 diabetes at risk of nonadherence is important for clinical prognosis and the development and delivery of interventions.