Turning Health Data into an Asset for Artificial Intelligence

Healthcare organizations are accumulating unprecedented volumes of data as the digital ecosystem continues to expand. 

From EHR data and imaging studies to lab results, medical device data, and socioeconomic information, there is a truly massive amount of data theoretically available to fuel analytics projects and help generate actionable insights.

Many organizations are currently exploring how to integrate artificial intelligence (AI) models into their decision-making processes and consumer experiences, and they are well aware that they need access to the right data at the right time to take full advantage of what AI has to offer.

In order to stay on the cutting edge of this transformational approach, organizations will need to start investing in the technologies, strategies, and talent to bring artificial intelligence into the clinical and operational environments.

“There is more proof every day that AI is ready to change healthcare,” said Esteban Rubens, Global Enterprise Imaging Principal at Pure Storage.  “The regulators are on board, the algorithms are becoming very mature, and the market is primed to embrace it.”

“There have been some false starts in the past, but the game changer now is that the technology is finally up to snuff.  All the components are in place for AI, which is extremely resource-intensive.  What we need now is the data.”

However, most providers have encountered challenges when trying to forge their loose collection of data sources into an effective tool for supporting their goals. 

Many organizations lack an enterprise-wide plan for infusing artificial intelligence into their operations, Rubens observed, which can create issues when setting goals, deciding on budgets, developing data governance, or choosing partners to work with.

“Artificial intelligence is moving very quickly, and organizations that do not establish a roadmap for leveraging their data assets are going to fall behind,” he cautioned.

“This isn’t something they can relegate to a committee that will take five years to develop a strategy.  It has to be done in the next few months, otherwise the industry will simply outpace them.”

The first step is to create a data governance framework that encourages data interoperability and appropriate access while safeguarding privacy and security.

“Start with answering the major questions about data use and access,” Rubens advised.  “Where will data be stored?  Who can access it for what purpose? When does data leave the organization, and what third parties should be involved?  Who gets compensated when an organizational data asset is used for some sort of monetary gain?”

Organizations should develop steering committees with representatives from across the enterprise, including the compliance department, Rubens said.

“Healthcare has strict regulations around data usage, for very good reason.  If you work in a vacuum, and then try to show it to your legal team, chances are good that the lawyers will poke a lot of holes in your plan and you’ll have to start from scratch,” he said.   

“The same thing will happen if you don’t consult everyone who touches the data or hopes to get a result from your AI work.  Involving all stakeholders from the beginning will save a lot of time course-correcting after the fact and allow you to get to your value proposition much sooner.”

Healthcare providers must also consider their infrastructure needs, which are likely to change as small-scale pilots and test projects develop into full AI deployments.

“Planning ahead is very important when it comes to storage and computing power,” Rubens said.  “It can be an expensive mistake not to invest in flexible, scalable, and future-proof architecture that can support long-term goals.”

Choosing the right storage solution will create a firm foundation for creating the partnerships and collaborations required to make artificial intelligence a success.  

“Not everyone has all of the AI experience they want within their organization,” Rubens said.  “But not everyone needs a full team of experts on staff.  There are many, many companies out there that will provide their expertise as a service to help organizations extract value from their data.”

“There are data cleanup companies out there that will help get the data into shape for training AI models.  There are companies that will help you integrate supplementary data sets, like socioeconomic data, into your predictive algorithms to enrich the results.  There are companies that can help set up the right infrastructure, the right security – everything you need to get started and succeed.”

For health systems with academic affiliations, help might be even closer at hand.

“If you are associated with a school of engineering or data science, there will be students and professors willing to help make sense of your data and solve the interesting problems that healthcare presents,” Rubens said. 

“There is a huge appetite for cross-collaboration and interdisciplinary work, so I would encourage organizations to take advantage of those opportunities.”

The right partners, strong infrastructure, and a well-defined governance model will help organizations choose initial use cases that have the potential to measurably improve care and reduce inefficiencies.

“My advice would be to start small,” Rubens said.  “Choose a pilot that has well-defined parameters and clear metrics for success.  It doesn’t have to fix everything right off the bat.  It can be something as simple as supplementing a workflow to reduce burnout or creating a risk score for a certain condition.”

“Those small victories will snowball into the ability to solve bigger problems, and will build confidence and momentum.  I believe that with the right approach, most organizations will be pleasantly surprised about how far they can get towards their goals with the data that they have right now.”