Back in 2016, Nathan Patrick Taylor, then a clinical informatics consultant for post-acute care provider Symphony Care Network, was skeptical of automated machine learning in healthcare.
The technology, still considered new today, was all but unknown four years ago. Automated machine learning had a “mystique” to it, Taylor said, with vendors claiming it could expedite the building and deploying machine learning models by automating time-consuming processes.
Some people assume that it can completely automate the creation of machine learning models, removing the need for a data scientist entirely, he said. But he thought that was unrealistic, so, when a colleague texted him to check out a DataRobot booth at a conference, Taylor wasn’t sure what to expect.
Limited expectations
“I rolled my eyes,” he said, skeptical that DataRobot, then a four-year-old startup vendor of automated machine learning, likely couldn’t accurately automate what a data scientist does.
The demo DataRobot showed at its booth showed that while the platform couldn’t build a model without any human input, it could build accurate models quickly, with limited data scientist work. Taylor was impressed but still wary.
Symphony Care Network shared a use case involving readmissions with DataRobot, and within three or four days, the provider had a model up and running. After two weeks of running it in a test environment, Symphony’s technical team pitched it to the company’s top executives. They liked it.
Thank you for reading this post, don't forget to subscribe to our AI NAVIGATOR!
After that, using DataRobot was “a no-brainer,” Taylor, now CIO at Symphony Care Network, said. “It just makes sense for us to go down this path.”
Machine learning in healthcare
Since Symphony Care Network’s introduction to DataRobot in 2016, the Illinois-based transitional care and assisted living provider has used the automated machine learning platform for a number of applications, including one that uses data on falls to predict readmission cases.
After seeing research indicating that if certain patients fall, they are likely to fall again, Taylor and his team built a model with DataRobot that can help predict if a patient is expected to fall.
The model incorporates data on the type of medication a patient is on and if the drug makes a patient more susceptible to falls, such as by making them dizzy. It also takes in data on patients’ conditions and whether a patient has fallen before, among other data.
The DataRobot-built model can mine patient data and identify patients who are likely to fall. That can help Symphony staff better accommodate the patient, while reducing fall rates and readmissions.
“It’s not so much about the prediction itself, but it’s about what it’s telling you will cause the outcome,” Taylor said.
Symphony Care Network also uses DataRobot for other applications, including to create models to predict the forecast demand for hip and knee replacements and how different treatments affect COVID-19 patients. The coronavirus hit some of Symphony Care Network’s facilities hard, with more than 150 reported deaths due to COVID-19 in its assisted living complexes. However, the healthcare provider also reported more than 200 coronavirus recoveries.
The DataRobot platform enables Taylor’s team to build and deploy models quickly. It’s easy to experiment with, he said, enabling Symphony Care Network to make and tweak models quickly.
DataRobot’s MLOps tool, designed to better help customers monitor and govern models, has enabled Symphony Care Network to keep its models up to date. The tool enables he healthcare provider to identify data drift in its models, which Taylor said he hadn’t known was a problem.[…]
An assisted living and transitional care provider uses DataRobot to automate the process of building and deploying machine learning models, enabling it to deploy models quickly.
copyright by searchenterpriseai.techtarget.com
Back in 2016, Nathan Patrick Taylor, then a clinical informatics consultant for post-acute care provider Symphony Care Network, was skeptical of automated machine learning in healthcare.
The technology, still considered new today, was all but unknown four years ago. Automated machine learning had a “mystique” to it, Taylor said, with vendors claiming it could expedite the building and deploying machine learning models by automating time-consuming processes.
Some people assume that it can completely automate the creation of machine learning models, removing the need for a data scientist entirely, he said. But he thought that was unrealistic, so, when a colleague texted him to check out a DataRobot booth at a conference, Taylor wasn’t sure what to expect.
Limited expectations
“I rolled my eyes,” he said, skeptical that DataRobot, then a four-year-old startup vendor of automated machine learning, likely couldn’t accurately automate what a data scientist does.
The demo DataRobot showed at its booth showed that while the platform couldn’t build a model without any human input, it could build accurate models quickly, with limited data scientist work. Taylor was impressed but still wary.
Symphony Care Network shared a use case involving readmissions with DataRobot, and within three or four days, the provider had a model up and running. After two weeks of running it in a test environment, Symphony’s technical team pitched it to the company’s top executives. They liked it.
Thank you for reading this post, don't forget to subscribe to our AI NAVIGATOR!
After that, using DataRobot was “a no-brainer,” Taylor, now CIO at Symphony Care Network, said. “It just makes sense for us to go down this path.”
Machine learning in healthcare
Since Symphony Care Network’s introduction to DataRobot in 2016, the Illinois-based transitional care and assisted living provider has used the automated machine learning platform for a number of applications, including one that uses data on falls to predict readmission cases.
After seeing research indicating that if certain patients fall, they are likely to fall again, Taylor and his team built a model with DataRobot that can help predict if a patient is expected to fall.
The model incorporates data on the type of medication a patient is on and if the drug makes a patient more susceptible to falls, such as by making them dizzy. It also takes in data on patients’ conditions and whether a patient has fallen before, among other data.
The DataRobot-built model can mine patient data and identify patients who are likely to fall. That can help Symphony staff better accommodate the patient, while reducing fall rates and readmissions.
“It’s not so much about the prediction itself, but it’s about what it’s telling you will cause the outcome,” Taylor said.
Symphony Care Network also uses DataRobot for other applications, including to create models to predict the forecast demand for hip and knee replacements and how different treatments affect COVID-19 patients. The coronavirus hit some of Symphony Care Network’s facilities hard, with more than 150 reported deaths due to COVID-19 in its assisted living complexes. However, the healthcare provider also reported more than 200 coronavirus recoveries.
The DataRobot platform enables Taylor’s team to build and deploy models quickly. It’s easy to experiment with, he said, enabling Symphony Care Network to make and tweak models quickly.
DataRobot’s MLOps tool, designed to better help customers monitor and govern models, has enabled Symphony Care Network to keep its models up to date. The tool enables he healthcare provider to identify data drift in its models, which Taylor said he hadn’t known was a problem.[…]
Read more: searchenterpriseai.techtarget.com
Share this: