Roche and IBM have teamed up to parse half a million electronic health records (EHRs) to create a predictive algorithm for patients with early risk of chronic kidney disease (CKD) related to diabetes. The industry duo developed the algorithm based on real-world data.
Age, body mass index, and glomerular filtration rate as well as concentrations of creatinine, albumin, glucose, and hemoglobin (HbA1c) were selected as important predictors on the basis of “a data-driven and medical selection for the study,” according to the paper, published in Nature Medicine (1). CKD was defined as a microvascular long-term complication of diabetes.
The Indiana Bioscience Research Institute (IBRI), Eli Lilly, and Indiana University School of Medicine provided Roche and IBM with a real-world data set originating from almost 100,000 patients with diabetes obtained from the Indiana Network of Patient Care database.
Dan Robertson, PhD, director of IBRI’s Applied Data Sciences Center, stated, “We are continuing our work with our industrial partners to explore disease progression, patient stratification, digital diagnostics, and eventually moving toward identifying new therapeutic targets to improve patient health.”
In a comparison of clinical trials and databases with [real world data], the study authors note, “predictive analytics algorithms using [real world data] could achieve equivalent or even enhanced accuracy compared with those using clinical trial data.”
The data algorithm “outperformed published algorithms derived from clinical trial data in a one-to-one comparison, as well as in cohort studies,” Health Data Management reported.
Reference
Ravizza S, et al. Predicting the early risk of chronic kidney disease in patients with diabetes using real-world data. Nature Medicine 2019; 25:57–59.
