Diabetes mellitus (DM) is among the most widespread non-communicable diseases and poses a substantial global health challenge. The aim of this study was to examine the incidence of DM and its nutritional, anthropometric, laboratory, demographic, and behavioral determinants, as well as comorbidities, within a Kurdish population residing in western Iran. This research was conducted in the Ravansar Non-Communicable Disease (RaNCD) cohort study, followed 9170 participants aged 35–65 years, for an average ± SD of 7.11 ± 1.26 years, from 2015 until 2023. A hierarchical Cox regression model was used to estimates the adjusted hazard ratios (HRs). The incidence of DM was 4.45 (95% CI 3.96, 4.99) per 1000 person-years. We found several significant predictors for DM incidence, including prediabetes, comorbidity, urban residence, total antioxidant capacity (TAC), and the interaction between gender and body mass index (BMI). Prediabetes emerged as the strongest predictor of DM incidence, with a hazard ratio of 10.13 (CI 7.84, 13.09). Additionally, having two diseases (HR = 2.18; 95% CI 1.44, 3.29) or three and more diseases (HR = 3.17; 95% CI 2.06, 4.90) increased the risk of developing DM. Also, the hazard ratios for the effects of gender on DM incidence in the normal, overweight, and obese BMI groups were 0.24, 0.81, and 1.01, respectively. The presence of prediabetes and obesity serve as the crucial indicators for the onset of DM, emphasizing the pressing need for interventions to prevent DM in these circumstances. Furthermore, there are notable disparities between urban and rural populations in this study, warranting further investigations to ascertain the underlying causes of such variations.