AIMS: Examine potential associations between inclusion of mid-level providers in United States Veterans Affairs (VA) primary care programs and diabetes control. METHODS: We established a cohort of diabetes patients (alive October 1, 1999) using the VA Diabetes Registry and VA corporate databases. 1999 VA Survey of Primary Care Practices data were combined with individual-patient information. We used a two-level hierarchical model to determine the relationship between staffing characteristics and hemoglobin A1c (HbA1c), among 88,682 patients from 198 clinics. RESULTS: Inclusion of nurse practitioners (NPs) at relatively limited levels (% of all providers who are NPs) in the primary care program was significantly associated with HbA1c lower by 0.31 percentage points (95% CI, -0.50% to -0.12%) compared to programs that did not include NPs. Having some level of NP staffing vs. no NP staffing was associated HbA1c lower by 0.25%. Inclusion of physician assistants (PAs) in primary care programs was generally not associated with a statistically significant difference in HbA1c. The exception is that moderate levels of PA staffing were associated with slightly higher HbA1c [0.18%, 95% CI, 0.02-0.34)]. CONCLUSIONS: Diabetes control among primary care patients appeared to benefit from inclusion of NPs, while an analogous association was not found for PAs.