Chronic kidney disease (CKD) is hard to detect due to its asymptomatic nature in the early stages and may get a diagnosis when the disease progresses to advanced, which may eventually lead to kidney failure. Even in patients with severe (stage 4) CKD, less than half were aware of their kidney damage.1 During the later stages, dialysis or a kidney transplant is the only way to maintain life.

Diabetes is the leading cause of end-stage renal disease (ESKD) for CKD patients in the US.2 It is the most common cause of kidney disease for approximately 45 % of patients who receive dialysis therapy. In addition, due to the interconnected chronic complications, adverse events get worse in diabetes than in patients with a single CKD condition, which increases the health cost. The estimation of annual cost per patient for those who developed ESKD among patients with diabetes (>$90,000) is almost doubled that of patients without diabetes (>$50,000) during the 1st year of treatment.3

Diabetes patients with CKD have an especially high risk of morbidity and mortality.4., 5. Many CKD patients die due to cardiovascular complications before they get into dialysis.6 For patients with CKD in the diabetes population, the prevalence of hypertension increases from 36 % in CKD stage 1 to 84 % in more advanced CKD stages 4 and 5.7 Hypertension is associated with a more rapid progression of CKD and is the second leading cause of ESKD in the US.8., 9., 10. Diabetes management affects the severity of a patient's kidney dysfunction and the development and progression of other complications among the diabetes population with CKD.11

During the last decade, a few prediction models for ESKD have been created from cohorts based on either general or kidney-specific populations, however, there is little evidence of their regular use in nephrology. This might be because clinicians are aware of the limitations inherent in the existing prediction tools. For example, some prediction models may not be applicable due to the monotony of the population for model development (e.g. Asian only, Caucasian only) and many models were not validated in the external data to demonstrate generalizability.12., 13., 14., 15. Thus, in our study, we aimed to develop and validate a predictive model on incident ESKD among people with T2D and CKD, and further transform the model into a 0–100 integer-point-based risk score. We followed the TRIPOD statement on prediction model development and validation, to analyze data and report results.