Sickle cell disease is the most common inherited blood disorder, characterized by misshapen red blood cells that can block blood flow and lead to infection, pain, and organ damage. Chronic kidney disease is a common long-term outcome of sickle cell disease, estimated to affect nearly 30% of adults with the condition, but identification and early intervention remain lacking.
In this study, the researchers used a chronic kidney disease phenotyping algorithm to predict whether an individual with sickle cell disease had progressed to chronic kidney disease. The algorithm was based on a set of 31 clinical features, and was applied to 13,284 individuals with physician-attested diagnosis of sickle cell disease.
The algorithm identified 2,767 cases of chronic kidney disease and 959 non-chronic kidney disease controls. Further analysis with a random forest classifier showed high accuracy, correctly predicting chronic kidney disease in 85% of patients with a known diagnosis. Using Shapley Additive exPlanations, a machine learning model, the researchers further identified a set of the strongest risk factors for developing chronic kidney disease, including increased age, female sex, and lower minimum levels of hemoglobin, the protein in the blood responsible for carrying oxygen throughout the body.
The researchers plan to expand their analysis to further improve this tool and, subsequently, identification, prevention, and intervention strategies.
