The ability to measure insulin sensitivity across the phenotypic spectrum of diabetes may contribute to a more accurate characterisation of diabetes type. Our goal was to develop and validate an insulin sensitivity (IS) score using the euglycaemic–hyperinsulinaemic clamp in a subset (n = 85) of 12- to 19-year-old youths with diabetes participating in the SEARCH study in Colorado, USA. Youths with a diagnosis of type 1 (n = 60) or type 2 diabetes (n = 25) underwent a 3 h clamp to measure glucose disposal rate (GDR, mg kg−1 min−1). Demographic (age, sex, race), clinical (BMI, waist, Tanner stage) and metabolic characteristics (HbA1c, lipids, blood pressure, urine albumin:creatinine) were used to estimate logeIS score via stepwise linear regression on a model-development set (n = 53). Estimated IS score was evaluated for reproducibility on two validation sets: youths with diabetes (n = 33) and healthy control youths (n = 22). The best model included waist, triacylglycerol (TG) and HbA1c levels (R 2 = 0.74). Diabetes type did not enter the model and there were no significant interactions between diabetes type and other predictors. Estimated IS score correlated well (r = 0.65, p