Objective: To evaluate the costs and health outcomes associated with primary care use by Indigenous people with diabetes in remote communities in the Northern Territory. Design, setting and participants: A population-based retrospective cohort study from 1 January 2002 to 31 December 2011 among Indigenous NT residents ≥15 years of age with diabetes who attended one of five hospitals or 54 remote clinics in the NT. Main outcome measures: Hospitalisations, potentially avoidable hospitalisations (PAH), mortality and years of life lost (YLL). Variables included disease stage (new, established or complicated cases) and primary care use (low, medium or high). Results: 14 184 patients were eligible for inclusion in the study. Compared with the low primary care use group, the medium-use group (patients who used primary care 2-11 times annually) had lower rates of hospitalisation, lower PAH, lower death rates and fewer YLL. Among complicated cases, this group showed a significantly lower mean annual hospitalisation rate (1.2 v 6.7 per person [P < 0.001]) and PAH rate (0.72 v 3.64 per person [P < 0.001]). Death rate and YLL were also significantly lower (1.25 v 3.77 per 100 population [P < 0.001] and 0.29 v 1.14 per person-year [P < 0.001], respectively). The cost of preventing one hospitalisation for diabetes was $248 for those in the medium-use group and $739 for those in the high-use group. This compares to $2915, the average cost of one hospitalisation. Conclusion: Improving access to primary care in remote communities for the management of diabetes results in net health benefits to patients and cost savings to government.