The diabetes population was estimated at 9.3 % (463 million people) globally in 2019 and was projected to rise to 10.2 % (578 million) by 2030 and 10.9 % (700 million) by 2045.1 A substantial proportion of people with diabetes developed chronic kidney disease (CKD) or had CKD progression, leading to end-stage kidney disease (ESKD) eventually. The clinical manifestation of CKD in the setting of diabetes included albuminuria, reduced glomerular filtration rate (GFR), or both. Guidelines have recommended that renal function should be regularly monitored in these patients, to ensure early CKD diagnosis and treatment.

Approximately 4.5 % of Medicare beneficiaries had CKD and type 2 diabetes (T2D) and accounted for >7 % of all Medicare costs in 2018, which was an estimation of $22,130 per person per year.2 Patients with CKD and T2D were 1.87 times more likely to have inpatient service and 1.92 times more likely to have emergency room visits, respectively, than those with T2D only.3 With the CKD severity increased, healthcare resource utilization and expenditures increased.4 Complications like heart failure, anemia, and hypertension are highly prevalent in CKD and T2D and led to worse health outcomes. People with CKD and T2D are three times more likely to die from a heart attack or stroke than patients with T2D alone.5 Early stages of kidney disease may be reversible. Interventions in earlier stages may slow or prevent the progression to later stages.

Many investigators made efforts to develop and validate risk prediction models for CKD progression, however, most of them focused on the general population or patients with established CKD. In addition, existing models were not very popular among physicians and patients due to the accessibility of the variables and the lack of generalizability. Therefore, our study aimed to develop a risk prediction model for CKD progression among patients with T2D and CKD using routine parameters and frequently used information, validate the model externally among patient-level data, and transform the complicated statistical model into an applicable risk score that will be easier to use for physicians and patients.