Thousands of end stage renal disease experts turn to Renal Therapies Group for pharmaceuticals and the most prescribed technologies in the industry.
Browser Upgrade Recommended: Our website has detected that you are using a version of Internet Explorer that will prevent you from accessing certain features on FMCNA.com. We strongly recommend that you use a different browser to optimize your viewing experience. Supported browsers include Chrome, Edge, Firefox, and Safari.
We are actively monitoring the novel coronavirus (COVID-19) pandemic. Learn About the steps that we are taking to protect our patients, employees, physicians, and partners. If you are a patient with questions or concerns, please find additional information at FreseniusKidneyCare.com and AzuraVascularCare.com
Success! The link has been copied to your clipboard.
A new predictive tool designed specifically to identify dialysis patients most at risk for hospitalization has been embraced by nurses who have come to see it as essential.
The predictive tool is currently available for nurses in the Care Navigation Unit (CNU) at Fresenius Medical Care North America. The unit supports dialysis patients in a wide range of value based care programs, which work to lower costs while improving outcomes. Unnecessary hospitalizations are one of the biggest drivers of higher cost to the healthcare system.
“Without this new tool, I would likely spend three to four times longer to find the information available in one snapshot on this report,” said Bobbie Werth, a nurse coordinator. “It identifies the issues clearly, and it provides the nurse a starting point for the high-risk concerns.”
Senior data scientist Andy Long created this new tool, named the Imminent Hospitalization Predictive Model (IHPM), using more than a thousand variables. The predictive model uses machine learning to help identify the combination of factors that require intervention which may simply lie hidden beneath all the data. The IHPM returns a risk score as well as the top 10 reasons why that specific patient is at risk for hospitalization.
“The ultimate goal is to reduce the rate of hospitalization using all the data we have available,” said Long. “We incorporate the history of hospitalization, treatment vitals, laboratory measurements, and comprehensive assessments. We also include the notes entered from patient charts and use natural language processing to interpret them.”
Long makes clear that predictive models like the IHPM are not a panacea, nor replacement for a clinician’s expertise, but rather an extra tool available to nurses and doctors to help them make good decisions on behalf of patients.
“It’s not a silver bullet and the challenge is to integrate it into existing workflows,” said Long. “The best way we use it is to take an existing workflow and help make it better with a predictive model.”
That is why nurses say the new dashboard has been so easy to use and important for their daily monitoring of patients. Working with the nurses directly, Long created an easy-to-read dashboard highlighting the most important information up front. Over time, that dashboard has been refined based on input from the nursing team.
“It allows me to help more patients each day because of the format,” said Werth. “Time management is an important nurse skill, and this dashboard improves ours significantly, allowing us to be more effective and reach more people.”
Research is currently underway to determine the actual effectiveness of the predictive model in reducing hospitalizations, and the model itself may see further enhancements this year. Long sees this effort as just one of many initiatives to demonstrate how predictive models and machine learning can improve patient care. He says there are potentially strong benefits for patients on home modalities as well.
“We collect so much data every day there is no way for us to look at it all, but we can have a computer look at it for us,” said Long. “Especially for our home patients, if a computer can help us monitor those patients for us, we can hopefully create even better alerts for our home patients based on predictive models.”