Machine Learning Analyzes EHR Data to Uncover Kidney Disease
A machine learning algorithm automatically scans EHR data to alert providers to patients in the earliest stages of chronic kidney disease.
Source: Getty Images
By Jessica Kent
- Machine learning can quickly analyze EHR data to identify chronic kidney disease, a condition which often goes undetected until it causes irreversible damage, a study published in npj Digital Medicine revealed.
The model can automatically scan a patient’s EHR for results of blood and urine tests and uses a combination of established equations and machine learning techniques to process the data. The algorithm may be able to alert clinicians to patients in the earliest stages of chronic kidney disease.
“Identifying kidney disease early is of paramount importance because we have treatments that can slow disease progression before the damage becomes irreversible,” said study leader Krzysztof Kiryluk, MD, associate professor of medicine at Columbia University Vagelos College of Physicians and Surgeons.
“Chronic kidney disease can cause multiple serious problems, including heart disease, anemia, or bone disease, and can lead to an early death, but its early stages are frequently under-recognized and undertreated.”
According to the researchers, approximately one in every eight American adults is believed to have chronic kidney disease. However, only ten percent of people in the disease’s early stages are aware of their condition.
Among those who already have severely reduced kidney function, just 40 percent are aware of their diagnosis.
There are several possible reasons for underdiagnosis, the team noted. Because people in the early stages of chronic kidney disease usually have no symptoms, primary care physicians may prioritize more immediate patient complaints.
Additionally, there are two tests needed to detect asymptomatic kidney disease: One that measures a kidney-filtered metabolite in blood, and another that measures leakage of protein in urine.
“The interpretation of these tests is not always straightforward,” Kiryluk said. “Many patient characteristics, including age, sex, body mass, or nutritional status, need to be considered, and this is frequently under-appreciated by primary care physicians.”
With this new algorithm, providers may be able to overcome these obstacles. The machine learning tool automatically scans EHRs for test results, performs the calculations that indicate kidney function and damage, stages the patient’s disease, and alerts physicians to the condition.
The team found that the algorithm performs nearly as well as experienced nephrologists. When testing the model using EHRs from 451 patients, researchers saw that the algorithm correctly diagnosed kidney disease in 95 percent of the kidney patients identified by two experienced nephrologists.
The tool also correctly ruled out kidney disease in 97 percent of the healthy controls.
The algorithm can be used in multiple types of EHR systems, including those with millions of patients. Organizations can also easily incorporate the tool into a clinical decision support system that suggests appropriate stage-specific medications.
Additionally, users can easily update the algorithm if standards of diagnosing kidney disease change in the future, and the model is freely available for use by other institutions.
Researchers noted that a potential limitation of the algorithm is that it depends on the availability of blood and urine tests in the EHR. While the blood test is fairly routine, the urine test is typically underutilized in clinical care.
Despite this drawback, the team believes that the algorithm could increase awareness of kidney disease, and potentially reduce the number of people who lose kidney function.
In addition to identifying chronic kidney disease, the model could also improve the power of other research studies.
The team has already applied the model to a database of millions of Columbia patients to find previously unrecognized associations between chronic kidney disease and other conditions. For example, depression, alcohol abuse, and other psychiatric conditions were considerably more common among patients with mild kidney disease compared to patients with normal kidney function.
In the future, researchers think the algorithm could be used to better understand the inherited risk of chronic kidney disease. The algorithm empowers genetic analysis of millions of people to discover new kidney genes.
“Our analysis also confirmed that a mild degree of kidney dysfunction is often present in blood relatives of patients with kidney disease,” said Ning Shang, PhD, associate research scientist in the Kiryluk lab and the lead author of the paper. “These findings support strong genetic determination of kidney disease, even in its mildest form.”
