Machine Learning for Chronic Kidney Disease Detection & Risk Stratification

Cover image

Executive Summary

The purpose of this whitepaper is to describe the methodology and application of Cricket Health’s predictive analytics capabilities for early detection and risk stratification of patients with Chronic Kidney Disease (CKD). Though CKD is highly prevalent and represents a large economic burden in the United States, efforts to manage the condition at earlier stages, delay disease progression, and reduce complications have been fairly limited. Furthermore, a large proportion of persons with CKD remain undiagnosed. Because of this, Cricket Health saw a strong need to invest in developing machine learning (ML) models to predict estimated glomerular filtration rate (eGFR), which is a common proxy for kidney health. These models use claims data only to predict eGFR with a high degree of accuracy — no EHR or lab data is required to make a prediction. Utilizing in-house ML models in kidney care management ensures that Cricket Health is well-positioned to identify and risk-stratify undiagnosed or misclassified CKD patients for risk-bearing entities with access to claims data.