Abstract P109: Predictors of Hemoglobin A1c Trajectories in a Cohort of Patients with Diabetes

Abstract

Background: Predictors of different hemoglobin A1c (HbA1c) trajectories in diabetes have not been adequately studied. Methods: In a cohort of 8,669 individuals with incident type 2 diabetes from a managed care system from 2000 to 2009, we determined HbA1c trajectories over the 5 years following diagnosis using a procedure which isolates distinct trajectories and fits a mixture model to calculate the probability of membership in each latent class for each patient. Results: A five HbA1c trajectory model provided the best balance of fit and parsimony (Figure). The largest groups had HbA1c's below 7% (At Goal, 62%, 5472 of 8669) or near 8% (Near Goal, 23%, 1957 of 8669). Smaller groups had HbA1c's that rapidly improved and then remained near 7.5% (Improving, 6%, 502 of 8669), or were elevated and either gradually worsened (Worsening, 5%, 412 of 8669) or improved slightly (Poor Control, 4%, 326 of 8669). In multivariate models, compared to the At Goal group, individuals in the other groups were generally younger, more likely to be male, and more likely to have diagnosed hyperlipidemia. The Worsening and Poor Control groups had lower adherence to oral diabetic medications (OR (95% CI) for 10% increase in the proportion of days covered: Worsening vs. At Goal, 0.77 (0.73-0.81) and Poor Control vs. At Goal, 0.79 (0.75-0.84)), and were much more likely to be on insulin. The Improving and At Goal groups had indistinguishable adherence levels. Conclusions: Trajectory analysis holds promise as an approach for defining subgroups of diabetics who may require different care strategies, and may help focus clinical conversations on behaviors such as adherence that are necessary to achieve glycemic control.