OBJECTIVES: Medication Possession Ratio (MPR) has been extensively used to assess compliance issues. Adherence Index (AI) is a measure, which simultaneously evaluates participant level compliance and participant level persistency. This study investigates the drivers of AI for people with diabetes and the impact of an increasing AI on total health care costs compared to MPR. METHODS: Data were obtained from pharmacy and medical claims of the participants with diabetes who continuously enrolled during January 1, 2001 through December 31, 2002, yielding a total study population of 19,824. Participant level compliance and participant level persistency were calculated. Total health care costs included costs of drugs and all medical services. Participants classified into high and low compliance groups using mean AI or MPR as a cutoff point were compared. Univariate and multivariate analyses were applied. RESULTS: Mean age for participants with diabetes was 72 years and 48% were female. The participants in high AI group (AI = 0.91) had significantly lower annual total health care costs [$16,138; Standard Deviation (SD) = 37,502] compared to those in low AI group [AI <0.91; $17,365, SD = 44,280]. However, there was no significant difference in annual total health care costs between high MPR [MPR = 0.90; $16,268; SD = 38,640] and low MPR group [MPR <0.90; $16,554; SD = 42,218]. In addition, diabetic participants with older age [Parameter Estimate (PE) = 195; p <0.0001], male gender (PE = 2,498; p <0.0001), hypertension (PE = 3,618; p<0.0001), congestive heart failure (PE = 12,178; p <0.0001), other cardiovascular diseases (PE = 2,915; p <0.0001), chronic pain (PE = 2,470; p <0.0001), or behavioral health diseases (PE = 8,940; p <0.0001) had significantly lower AI. CONCLUSIONS: AI better differentiates those participants who have increased costs due to an increase in ER visits, hospitalizations, or office visits. Moreover, AI is better at assessing medication compliance issues than MPR since AI takes into account detailed participant levels of both compliance and persistency.