Machine-learning model predicts CKD progression with ‘readily obtainable’ lab data
Click Here to Manage Email Alerts
Key takeaways:
- The risk classifier model accurately predicted eGFR decline greater than 30%.
- The researchers did not evaluate the role of clinical variables such as blood pressure on the performance of the model.
A machine-learning model developed by researchers at Sonic Healthcare USA accurately predicted the progression of chronic kidney disease using readily available laboratory data.
“CKD is a major cause of morbidity and mortality,” Joseph Aoki, MD, senior vice president, population health, at the Austin, Texas-based company, and colleagues wrote in a study. While more research is needed, “our results support clinical utility of the model to improve timely recognition and optimal management for patients at risk for CKD progression.”
The investigators used a retrospective observational trial to analyze deidentified laboratory information services data from a large U.S. outpatient laboratory network. It involved 110,264 adultswith initial eGFR values between 15 mL/min/1.73 m2 and 89 mL/min/1.73 m2.
Researchers developed a seven-variable risk classifier model using random forest survival methods to predict eGFR decline of more than 30% within 5 years.
Results showed that the risk classifier model accurately predicted eGFR decline greater than 30% and achieved an area under the curve receiver-operator characteristic of 0.85.
“The most important predictor of progressive decline in kidney function was the eGFR slope,” the authors wrote, followed by the urine albumin-creatinine ratio and serum albumin slope. Other key contributors to the model included initial eGFR, age and sex.
“Our progressive CKD classifier accurately predicts significant eGFR decline in patients with early, mid and advanced disease using readily obtainable laboratory data.”
The authors wrote that the study did have limitations: It did not evaluate the role of clinical variables such as blood pressure on the performance of the model. Further prospective work is warranted to validate the findings and assess the clinical utility of the model, the researchers wrote.
Used as a complement to and in conjunction with other well-established predictive models, the progressive CKD risk classifier “has the potential to significantly improve timely recognition, risk stratification and optimal management for a heterogeneous population with CKD at a much earlier stage for intervention,” Aoki and colleagues wrote.
Collapse
Read more about
Click Here to Manage Email Alerts