News Release

Cricket Health Announces New Machine Learning Model for Detecting Chronic Kidney Disease

Share this story:

White paper details kidney care provider’s leading predictive analytics capabilities

SAN FRANCISCO (November 6, 2019) — Today, Cricket Health, a comprehensive kidney care provider, published a white paper, “Machine Learning for Chronic Kidney Disease Detection & Risk Stratification,” announcing its proprietary machine learning model to identify patients at-risk for or living with chronic kidney disease (CKD). Designed by Cricket Health’s interdisciplinary data team, the predictive model marks a significant advancement in the ability to accurately risk stratify a patient population for CKD using administrative data and without the need for electronic health record (EHR) clinical or lab test results data.

“Far too many patients are unknowingly living with CKD and won’t discover their condition until there is irreversible damage to their kidneys,” said Cricket Health CEO Arvind Rajan. “At Cricket, we set out to change this by empowering providers and payers with an accurate, convenient model to identify those at high risk for or living with undiagnosed CKD among their patient population at earlier stages in the disease’s progression.”

CKD and end-stage renal disease (ESRD) are incredibly costly and prevalent in the United States. Over 35 million American adults are estimated to have CKD, with millions more at risk of developing it. Over half a million are living with ESRD, meaning they have experienced kidney failure. The prevalence of both conditions is expected to grow in the coming years, with Medicare alone already spending more than $110 billion caring for beneficiaries with CKD and ESRD. Despite the prevalence and cost, nine out of 10 people with stages 1-3 CKD do not know they have it, and about half of those who have progressed to severe loss of kidney function, who are not on dialysis, remain unaware.

Lack of awareness is in part because CKD remains largely asymptomatic until it has caused severe kidney damage. Additionally, barriers to early detection exacerbate the problem, including: a lack of CKD knowledge among primary care providers and lack of support for CKD management; a lack of clear guidelines for CKD management; inconsistent CKD screening practices and low test compliance rates; and fragmented care.

Cricket Health’s machine learning model can predict estimated glomerular filtration rate (eGFR), a common proxy for kidney health, and the probability of a patient having each of the late-stages of CKD at a high degree of accuracy. Without the need for lab testing or patient involvement, Cricket Health can run its model frequently, and at little or no cost, across an entire patient population. This gives health plans, at-risk providers, and other payers insight into their entire patient population, not merely a subset of individuals who have been tested.

“Diagnosis is the first step to helping patients live their best possible lives with kidney disease. By identifying patients earlier, we can intervene to deliver the care they need and give them the tools they need to get the right medical management and avoid health complications, control transitions to ESRD, and, ultimately, to improve health outcomes,” said Dr. Carmen A. Peralta, chief medical officer at Cricket Health, a nephrologist who helped develop the machine learning model and co-authored the white paper.

Once the model identifies patients who are at risk for or likely to have CKD, Cricket Health can work with the payer to conduct outreach to each patient’s primary care provider or specialist, confirm their CKD status via lab test, and to deliver patient-centered, personalized kidney care to each patient.

The predictive model is powered by a variety of predictors, such as patient demographics, comorbidity data, and utilization patterns. The whitepaper explains the development, training, applicability, and performance metrics of the company’s machine learning algorithm and demonstrates its predictive value. The model was developed this year by a group combining clinical, epidemiological, statistical, and data science expertise.

For more on Cricket Health’s proprietary machine learning model and its technology platform, read the full white paper here:

What's New