While kidney disease is the ninth leading cause of death in the United States, a staggering 90 percent of people affected do not even know they have it. Individuals are typically asymptomatic until the disease develops into later stages, and at that point, kidney failure has begun.
Although common chronic conditions, including diabetes and high blood pressure, increase the chances of a person developing kidney disease, vague symptoms do not scream for the attention that would normally merit swift intervention. This makes catching chronic kidney disease (CKD) early and delaying its progression challenging. Education around risk factors is one way to improve CKD progression, and with advancements in technology, these efforts can now be supported by predictive analytics that are powered by artificial intelligence (AI) and machine learning (ML) algorithms.
AI models offer an early look into who might develop kidney disease, who would benefit from early intervention, and the risk of further complications without a change in course. Individual health conditions are considered, along with social determinants of health (SDoH), which helps clinicians make great strides in kidney disease management. AI applications can fuel alerting systems, offer diagnostic assistance, and guide treatment decisions.
Fueling predictive analytics
Understanding probability and targeted forecasting can be more impactful than trying to address diagnosis after the fact. AI receives credit for lending insight into several novel and complex medical areas, including Covid-19, immuno-oncology, and kidney disease.
AI and ML enable predictive modeling. These algorithms are iterative, improving their accuracy each time they are run and fed additional data sets. By increasing the number of data feeds and the size of the patient population feeding into the system, more beneficial insights are derived. Results about one healthcare organization can be compelling – and more value can be derived through compiling organizational-agnostic data from a larger data pool while still maintaining the ability to drill down into specific locations and providers.
Combining data from several sources, including claims, clinical data, live feeds from health exchanges, dialysis machines and demographic information for social determinants of health, algorithms can predict adverse events, including kidney failure during a given time frame, a cardiology event, and even mortality. Additional determinants to be included involve social vulnerability index, including average income and education level. Adding pre-defined, rules-based alerts makes this information more actionable.
Predictive analytics make sense of massive amounts of data, cut through the noise, and allow clinicians to focus on efficiently and effectively delivering high-touch patient care. This results in drastic reductions in unplanned dialysis starts or “crashes” and hospitalizations, along with higher adoption of home dialysis.
Informing a value-based world
Per-person, per-year spending on ESRD (end-stage renal disease) patients averages nearly $80,000. Moving to more preventive, value-based care shifts the focus to patient outcomes and lowers costs long-term. As more organizations are moving to value-based kidney care models, predictive analytics can determine a general risk score when enrolling patients that includes estimated per patient spend and an analysis of comorbidities.
Helping organizations identify patients who should be included in a particular model will place more people on the right care path sooner, improve outcomes and their quality of life. Analytics also inform where to target efforts for mitigating high-cost areas and the programs in which providers should participate.
Dashboards offer visual representations of insights at the macro, population health level, showing clear metrics on how interventions improve both clinical and financial outcomes. Filters are set for patients by market and stage of CKD, details of which can be shared to help inform plan decisions. As population health management is increasingly applied to care for patients with CKD, these insights will play a broader role with multi-disciplinary teams creating and implementing strategies to improve patient care.
Using analytics and prediction models help inform a patient’s care journey, identify which patients are at risk for kidney failure and high-risk patients, prompting early interventions. Applying AI and machine learning to inform predictive analytics enables clinicians to get patients started with treatment regimens that may prevent declining kidney function.
Once in the care program, technology is imperative to running the required reporting for value-based programs as well as for continued patient care. Providers need to understand what is working and where adjustments should be made.
Most predictive analytics are run on cloud-based platforms to offer predictions using two or three approaches without slowing down the internal systems for a healthcare organization as they continue seeing patients. Over time, tracking predictions against actual future outcomes will prove the model’s accuracy.
One point that should be noted is AI’s potential role in perpetuating bias. As with any form of technology, AI is only as good as the data it is fed. The opportunity is there to either reinforce bias or act without it, depending on the data it is fed. It is worth taking a close look to ensure any bias or gaps are addressed.
While AI can make great strides in aiding treatment plans, kidney disease remains complex and requires high-touch, personal care. Technology informs care decisions, better enabling clinicians to deliver compassionate kidney care the way it should be done. It involves delivering the right insights to the right hands, at the right time.
Photo: MF3d, Getty Images