Decode Health Unlocks Better Patient Outcomes with AI

For us to be able to respond that quickly leveraging the AI platform was a clear signal that the framework Decode built could really speed time to innovation, not only for a COVID use case, but for other healthcare use cases as well.

Unlocking Better Healthcare with Data

Decode Health began with an ambitious vision: combine genomic and other health data with machine learning to help anticipate patient needs.

“Decode Health unlocks discovery with data,” said Chase Spurlock, PhD, CEO and Co-Founder. “Our goal is to help healthcare organizations be more proactive. Today, we’re the innovation partner to groups like large diagnostic companies, big pharma, as well as forward-thinking personalized healthcare companies.”

As a healthcare AI company, Decode Health partners with industry-leading organizations to deliver products, services and data assets that improve patient outcomes, lower costs, and advance precision health and health equity.

To realize its goals, Decode relies on powerful predictive analytics. In the company’s early days, modeling demanded exhaustive manual effort with considerable time preparing data, waiting on models, recalibrating, and waiting again.

“We didn’t have the tools to do this at scale,” said Julia Polk, Co-Founder and Executive Advisor. “Analytics were accurate, but they were slow. If we could accelerate that model building process, that would be a huge advantage to us.”

Models in Hours Not Weeks

Decode Health found a scalable solution in DataRobot AI Platform. With the platform, the company enhanced its framework with automated machine learning that streamlines predictive analytics end to end. In the past, a comprehensive analysis of a single dataset could take weeks, with multiple data team members working around the clock; now, they complete the same task in a matter of days. 

DataRobot accelerates the modeling process so that our team can focus on the data problem and the various data elements that help us target, predict, and understand a particular outcome.

As a trial, Decode Health ran a scenario in the AI platform where it already knew the answer. Within a few minutes, they were able to produce a model tuned to the expected accuracy, sensitivity, and specificity.

“In the space of a couple of days versus a couple of weeks, we produced a deployable model,” Spurlock said. “This proof-of-concept study showed our team how incorporating this tool into our framework could help us solve complex healthcare problems faster.”

Staying Ahead of Chronic Illnesses and COVID-19

The company uses AutoML to predict a variety of health outcomes for chronic illnesses such as autoimmune diseases and more. So far, the company has identified predictive patterns across multiple disease areas, enabling patients and providers to take proactive steps to improve patient outcomes.

When COVID-19 hit, government entities reached out to Decode Health asking for help in understanding the social determinants of health (SDoH) and risk personas in communities. They were tasked with turning around a report in a day.

“In the earliest days, this request would’ve been a week or two endeavor,” Spurlock said. “Fortunately, we were actively investing resources to enhance our platform to make it more scalable, with quicker turnaround times. These efforts paid off, enabling us to produce results to help stakeholders understand the risk persona of those most at risk for COVID hospitalization or death.”

Spurlock notes that, with a single model-by-model approach, they simply could not have delivered in the needed time frame.

“For us to be able to respond that quickly leveraging the AI platform was a clear signal that the framework Decode built could really speed time to innovation, not only for a COVID use case, but for other healthcare use cases as well,” he added.

In its home state of Tennessee, Decode Health predicted the top counties to be impacted with 95 percent accuracy. These results helped community leaders make informed public health decisions.

Decode Health has also applied predictive modeling approaches to identify rising COVID-19 hospitalizations and designed methods to uncover patients at risk for developing long COVID. 

Predictions at Speed and Scale

The ease of use of the AI platform has made it accessible to business and IT users alike, aided by DataRobot University.

“As a non-traditionally-trained data scientist, I find it’s all there and ready for me to investigate, explore, and test,” said Jamieson Gray, Chief of Staff. “Complementing the tools we’ve developed at Decode, we can launch hundreds of projects with different project design iterations.”

“We’re able to do the kind of work that we do for our partners without a team of 25 data scientists,” Polk added.

That’s because they automate data preparation, creating and comparing models, and managing them in production. By accelerating the process, the organization can focus more on the data elements needed for powerful predictions.

“It would go from a single model that would take days to build and days to get results, to hundreds of models built in hours,” Gray said. 

For the Decode team, unlocking speed and scalability to deliver AI-driven insights means more opportunities to improve patient outcomes, reduce costs, accelerate R&D for diagnostics and drug discovery, and enhance health equity.

Decode routinely employs fairness and bias mitigation strategies to ensure consistent data patterns across multiple models. These efforts safeguard against AI models anchored on just one or two elements with outsized influence leading to poor performance and varying results across patient populations.

Decode looks forward to continuing to fuel a future with better, less expensive, and more equitable healthcare.

“The future for diagnosing and managing autoimmune diseases, as well as other chronic conditions is very bright,” Spurlock said. “If you can intercept disease risk before it’s too late, you can make a significant difference in the patient’s life. Not only from an outcome standpoint, but also from a cost-savings standpoint, which benefits all of healthcare.”