Test employs AI to assess kidney disease risk among adults with type 2 diabetes
Click Here to Manage Email Alerts
How physicians assess chronic kidney disease and its progression among adults with type 2 diabetes may be changing due to a new prognostic tool.
KidneyIntelX (RenalytixAI) employs machine learning and random forest models to predict kidney disease progression for those with type 2 diabetes and early stages of CKD. The system incorporates data from electronic health records as well as tumor necrosis factor receptors 1 and 2 and kidney injury molecule-1 — biomarkers found in the blood — to project the risk for kidney failure. Results then determine whether a patient is at high, intermediate or low risk, and this can then help drive treatment decisions.
Better results
“Our primary goal is to improve upon the rather simplistic model that is currently used in clinical care, which is largely dependent on cut-points of eGFR and the urine albumin to creatinine ratio,” Steven Coca, DO, MS, associate chair of clinical and translational research for the department of medicine at Mount Sinai and co-founder of RenalytixAI, told Healio. “Really, the strength of the algorithm derives from taking into account multiple other lab variables, and most of these are basic clinical chemistries that are ordered in most patients in primary care settings.”
Coca said work on this project began 2 years ago. He and colleagues have published data showing how effective this type of algorithm can be. In fact, the odds of kidney failure were calculated as 10 times greater with a high-risk reading from the system compared with a low-risk reading, according to a press release from RenalytixAI.
“Unlike typical cohort studies in which patients have routine visits at standard intervals and all the variables of interest are populated, in the real world, patients are seen in different intervals and different amounts of testing is done,” Coca said. “Missing variables wreak havoc on standard regression models, whereas these machine learning algorithms can handle the variations in frequency and density that are present when extracting data passively from electronic medical records.”
Coca said he hopes the system makes it easier to understand just how much risk patients have with respect to CKD and ultimately encourage them to seek proper medical help to receive the treatments they need, including referral to nephrologists and use of antagonists of the renin–angiotensin–aldosterone system and optimal blood pressure control.”
“Awareness is also very unsatisfactory at multiple stages of kidney disease,” Coca said. “We hope that the output of a validated kidney risk score from this test, as well as educational content that will accompany the test, will motivate physicians to prescribe the appropriate treatments and patients to adhere to these treatment regimens. Moreover, we hope to develop evidence to show how our test integrates with SGLT2 inhibitor prescription and application, and we think it can serve as another guiding force to decide who would be the best candidate for SGLT2 prescription. There are multiple ways the test can be used to change how we are currently managing patients.”
Next steps
Coca also sees a larger scope for the effect this tool can have for health care in general.
“It will also help the public health of our nation, and it will also help institutions manage these patients appropriately, decide their allocation of resources in terms of specialists, etc,” Coca said. “We anticipate that by getting the patients to the right place — ie, earlier referral to nephrologists — it will reduce the burden of progressive kidney disease.”
The system is a predictive tool only, and there will still be patients who require treatment no matter when their risk is assessed. Coca said he still sees an opportunity to assist these patients with this diagnostic system.
“Even if those patients are destined to progress and don’t respond to those therapies, the silver lining still is that instead of the status quo, which is that about 40% in the U.S. start dialysis without seeing a nephrologist in the prior year, these patients will be hooked in with a nephrologist earlier in the course of their CKD, and they can start discussions about home dialysis or kidney transplant much sooner,” Coca said.
The future for the KidneyIntelX is fast approaching. According to Coca, the test, which the FDA has given breakthrough designation, could be available as early as 2020 and was approved for reimbursement by CMS. With the test that much closer to use on a larger scale, Coca said, there is still work to do.
“[The] test alone may not drive improvements in care without a robust technology package or educational package, and we are developing all of that in parallel so that when physicians order the test, and when patients get the results, they will need to know how to interpret it and how to respond to it,” Coca said. “That will be in place to help guide both the patients and providers, especially in the early stages of usage.” – by Phil Neuffer
Reference:
Nadkarni GN, et al. bioRxivorg. 2019;doi:10.1101/587774.
For more information:
Steven Coca, DO, MS, can be reached at Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, Box 1243, New York, NY 10029; email: steven.coca@mssm.edu.
Disclosure: Coca is a co-founder of RenalytixAI and reports that he serves as a member of the scientific advisory board and receives consulting fees, equity and stock options. Coca also reports that he has received consulting fees from CHF Solutions, Takeda, Janssen, Bayer, Relypsa and pulseData, as well as stock options from pulseData.