Grant: Project Grant

Background: Chronic kidney disease (CKD) affects ~5 million Canadians. Not everyone with CKD progresses to kidney failure, but those who do require dialysis or a kidney transplant to survive. Kidney transplant is often preferred, but many patients are not eligible for a transplant. Those that are not need dialysis to survive. For these patients, the ideal transition to dialysis is pre-planned and done as an outpatient before dialysis becomes an emergency. Unfortunately, half of all patients who start dialysis do so in an unplanned fashion as inpatients in the hospital. As a result, these patients require urgent placement of a dialysis catheter which makes infections and other complications more likely. Unplanned dialysis is strongly linked to poor health outcomes including death, anxiety, and diminished quality of life. It is also costly to the healthcare system. Purpose: To develop a computer-based model that will help doctors identify patients with CKD likely to require dialysis in the near future (6-12 months). Procedure: We will use artificial intelligence which is a state-of-the-art technology where computers are designed to 'learn' from trends in healthcare data to help doctors improve their decision-making. First, we will use routinely collected data from four CKD centers across Ontario to develop the model. Second, we will test the model in CKD centers in Alberta, Manitoba, Nova Scotia, and Québec. Third, we will explore the potential impact the model will have on both the patient (reductions in death, hospitalizations, and anxiety) and the healthcare system (reduced costs) through its prevention of unplanned dialysis. Outcome: We anticipate that this model will accurately identify patients likely to require dialysis within the next 6-12 months. Relevance to Patients: By helping to identify patients likely to require dialysis in the near future, this model will better allow for doctors to intervene and prepare patients for dialysis in a planned fashion.