It took Gaurav Jain, MD, FASN, a professor in the Division of Nephrology at The University of Alabama at Birmingham, and his colleagues about 1½ years to get up and running with the CKD [Chronic Kidney Disease] Insights program. However, once they did, data from their patients’ electronic health records (EHRs) started automatically flowing into a dashboard that stratified their patients by disease stage and risk profile, helping them proactively identify patients in need of interventions, such as patient education, mental health care, or transitional care.
The team implemented the CKD Insights program to help them meet their metrics in the Centers for Medicare & Medicaid's Kidney Care First program. “[We were] finding ways to make the data easily accessible and then using that data and technology to try to make our lives easier, hopefully improve the quality of care, and keep the cost down as much as possible,” Jain explained. He pulled off the implementation with a small team that included himself and a nurse coordinator, who led the program's patient education, mental health components, and patient survey administration. They also worked to leverage existing resources to avoid adding unnecessary costs. But despite the ease of using the dashboard, many team members did not want to log into another program, and the program could not send information or action alerts back to the EHR system. “The only people using it were myself and our care coordinator,” he explained. “We had a really tough time trying to engage other physicians to use it.”
There were also challenges with determining how to pay for the program, and Jain's institution ultimately decided to sunset the program several months ago. He shared his perspective at ASN's Nephro-Economics in 2024: From the Big Picture to Advances in Transplant Policy symposium hosted by ASN and the Division of Nephrology, Columbia University Irving School of Medicine, on September 6, 2024 (1). Although Jain was disappointed in the project's outcome, he is still optimistic about the potential of such technology to help support value-based care initiatives and help ease the burden on an overstretched nephrology workforce. His optimism was shared by fellow speaker Navdeep Tangri, MD, PhD, professor in the Department of Community Health Sciences at the University of Manitoba, Canada. Tangri highlighted the potential of technologies like artificial intelligence (AI) to support nephrology care.
Jain and Tangri shared perspectives on the appropriate use of different technologies. They highlighted cases of both successful and unsuccessful use and offered tips on navigating these emerging tools and avoiding pitfalls. “The truth about these things is that we need to find sustainable models that work,” Jain said.
Finding partners
Despite the setback with CKD Insights, Jain and his colleagues persisted in their efforts to find a way to identify patients who are high-risk for kidney diseases. They found an approach by partnering with the dialysis company DaVita, which serves many of these patients. A consultant working with DaVita was able to share insights with Jain and his team from DaVita's Shared Patient Care Coordination program on 100 patients living with kidney failure who produce the highest costs.
DaVita's nurses and social workers risk-stratified shared patients and regularly met with Jain and his colleagues. They worked as a team to identify which patients may be at very high risk of hospitalization or an emergency department visit, may have run out of insulin, or need a wheelchair or other interventions to address their growing fall risk, and they developed plans to address the patients’ needs in collaboration. “The big win of this program is that we were all sitting together frequently talking about these patients and trying to solve problems,” Jain said. In fact, he noted that the annual cost of care for patients in the cohort decreased from $83,000 in 2021 to $73,000 in 2024.
Jain and colleagues are also collaborating with a program called REACH Kidney Care and Blue Cross Blue Shield to improve patient management. REACH's educators and navigators work with patients to check their blood pressure, manage diabetes, and share what they learn with the nephrologists. Although the program is small, it has helped ensure that 83% of participating patients who develop kidney failure receive a preemptive transplant or home dialysis. Jain's team also uses a program called IllumiCare, allowing them to create alerts in their health record system for patients whose proteinuria or potassium levels are concerning. For patients identified this way, the team has a contract with Blue Cross Blue Shield's Medicare Advantage plan that will pay for an annual nephrology wellness visit.
As a mandatory part of the Centers for Medicare & Medicaid's Kidney Care First program, Jain and his team also regularly survey roughly 180 patients to assess their engagement with their care. The 13-item questionnaire includes whether the patients know what each medication does and whether they are confident about contacting a physician or handling health problems independently. The survey helps Jain and colleagues identify patients who may need extra support or education. It also helps them connect patients who have mental health needs with a psychologist, who runs their clinic specializing in neurologic and psychiatric care for patients with kidney diseases, and patients who need medication support with a pharmacist. Over the past 2 years, they have improved their patient scores on the surveys with the help of multidisciplinary partnerships. “If you have the data, and you have targeted interventions, without a lot of resources, you are able to make a difference in a patient population of this size,” Jain said.
Telehealth and remote monitoring
Telehealth and remote monitoring have also proven valuable for Jain's team. He noted that they increase access to patient education and mental health care and reduce the burden for patients on home dialysis who live far from the clinic. “It is clearly beneficial,” he said. “It improves compliance, increases access to nephrologists and transplant nephrologists, and increases access to education.”
However, he emphasized that it is essential to have systems and support in place to deliver telehealth care successfully. Jain said that his health system is working to integrate telemedicine into its EHR system to streamline the process further. He also noted a need to ensure that EHR systems at health systems, dialysis facilities, and other specialist offices can effectively communicate. Additionally, Jain's organization hired patient navigators to facilitate telehealth visits between nephrologists and patients during dialysis sessions. “It's been very beneficial in decreasing the burden of disease for patients,” he shared.
They also use remote monitoring integrated into their home dialysis machines to flag potential problems. Jain called it an “eye-opener,” enabling the team to learn that many patients who had good laboratory results and who were thought to be adherent were not undergoing even half of their recommended treatments. This indicated that many patients may have undergone treatments before the laboratory tests to produce acceptable results. He said that creating workflows and assigning responsibility for such monitoring tasks are critical. He also suggested being mindful to ensure that the use of technology does not disproportionately benefit those who are more technology literate or have more resources, which could reinforce socioeconomic disparities.
Jain recommended that nephrologists, particularly those participating in value-based care programs, look at all of the programs and resources available in their area. He said that it has helped his team be successful in the Kidney Care First program. “We’ve actually been able to pay for all the resources, and we are still left with some money to reinvest in the program to make care better,” he shared.
AI applications
Tangri highlighted several ways that AI may help improve kidney care. It can detect cases of acute kidney injury, analyze kidney biopsies, predict adverse events like intradialytic hypotension during dialysis, or help streamline physician note-taking. For example, Tangri said the Cleveland Clinic created an AI tool to predict acute kidney injury in patients undergoing cardiac surgery using six variables from the patient's first postoperative metabolic panel (2).
“You’ve got to have high accuracy, good calibration, and external validation,” Tangri said. He also emphasized the importance of optimizing AI-driven diagnostic or predictive tools for specificity to ensure that they do not generate too many false alarms for clinicians, causing alarm fatigue.
Deep learning models are also being used to analyze imaging data. Tangri noted that 87% of US Food and Drug Administration-approved AI medical devices are for imaging. For example, AI can be trained to detect changes in pathology slides or predict graft failure. It may also be used to help automate the process of biopsy analysis or help assess fibrosis via ultrasound or magnetic resonance imaging.
Tangri and his colleagues developed and externally validated a tool to predict CKD progression risk that generates a report directly in the patient's EHR. “To get people to take action on [patients who are high risk], you need to integrate it into the workflow,” he said.
The report includes a set of physician orders appropriate for a patient's risk level to help nudge their physicians to take appropriate action. Tangri noted that these recommendations can also be built into an EHR smart order set and best practices alerts.
Clinicians are also using generative AI tools to help alleviate administrative burdens. Many are using AI medical scribes to record the entire patient encounter and generate notes that the clinician reviews and edits as necessary. A study in two academic centers showed that AI scribes save, on average, between about 5 and 8 minutes of time (3). Tangri said that could free up additional time for more important patient care tasks or help reduce the after-hours work burden on physicians. “They decrease your note time and your pajama time,” he said. “That's quite meaningful.”
Tangri cautioned that not all studies of AI generative-based interventions have found benefits, and he said there is a need for more rigorous tool studies moving forward.
Jain noted that leveraging new technology to further value-based care will require local commitment and national policy changes. “It takes a lot of innovation, and it takes policy,” he said. “We are seeing results both at an organizational and national level.”
References
- 1.↑
American Society of Nephrology. Nephro-Economics in 2024: From the Big Picture to Advances in Transplant Policy. September 6, 2024. Accessed September 25, 2024. https://www.asn-online.org/NephroEconomics2024
- 2.↑
Demirjian S, et al. Predictive accuracy of a perioperative laboratory test-based prediction model for moderate to severe acute kidney injury after cardiac surgery. JAMA 2022; 327:956–964. doi: 10.1001/jama.2022.1751
- 3.↑
Rotenstein L, et al. Virtual scribes and physician time spent on electronic health records. JAMA Netw Open 2024; 7:e2413140. doi: 10.1001/jamanetworkopen.2024.13140