Purpose: Adequate glycemic control is a cornerstone of high quality diabetes care. Improvement in glycemic control is the net of two vectors: the number of patients who newly achieve glycemic control (improvement vector), minus the number of patients who relapse from adequate to inadequate glycemic control (relapse vector). The purpose of this analysis is to quantify the improvement and relapse vector in a single large care delivery system. Methods: We identified all T2D patients having at least one visit and A1c between 1/1/2017 and 5/31/2018. We selected as the index date the first visit with an A1c < 8% for the controlled cohort; and the first visit with an A1c ≥ 8% for the uncontrolled cohort. We followed patients for 12 months after the index date and selected the first A1c that shifted their glucose status. Results: Although the percent of patients with base A1c >=8% who improved (59%) far exceeded the percent of those with baseline A1c <8% who relapsed (20%), the absolute number of patients relapsing (5,109) exceeded the number improving (4,132), with a 30 per 1,000 episodes net worsening in quality of diabetes control (Table 1). Conclusions: In this high-performing care system the relapse vector now exceeds the improvement vector, accounting for recent overall deterioration in population levels of glucose control. Strategies to proactively identify and manage patients at high risk for glycemic relapse are urgently needed.