Seamless Data Architecture Supports Artificial Intelligence Success

Healthcare data is immensely complex and immensely voluminous, which offers both advantages and challenges for the development of artificial intelligence.

While large, rich, multifaceted data sets are vital for creating accurate machine learning models and validating the accuracy of AI results, storing all of this data in a secure, accessible, interoperable manner can be a major undertaking for healthcare organizations.

As algorithms develop the capacity to more and more actionable insights from previously untapped data assets, healthcare providers will need to keep pace by developing seamless storage and transport pipelines for larger volumes of this highly valuable commodity.

“Artificial intelligence in healthcare has absolutely exploded over the past year,” said Josh Gluck, vice president of global healthcare technology strategy at Pure Storage.  “There are people doing amazing things in proteomics, radiomics, genomics, and so many other areas that have the potential to drastically improve the way we deliver care.”

“We’re going to continue to see AI move into the patient environment, and healthcare providers will need to prepare for that in a number of ways.”

The first step for healthcare organizations is assessing their long-term goals for artificial intelligence.  Many organizations are beginning the AI adoption process by approving limited pilots and research projects in order to generate real-world evidence that machine learning can solve their business problems.

While this approach can help providers avoid wasted investment in tools or systems that don’t end up bringing a return on investment, it can also expose organizations to the challenges of “shadow IT.”

Shadow IT occurs when smaller teams within an organization start to develop infrastructure without sufficient oversight from the IT department or a C-suite IT executive. 

One-off artificial intelligence projects can result in pockets of data stored separately from the organization’s main data assets, and may even expose the organization to privacy and security issues if the data is being stored, shared, or used in a manner inconsistent with established compliance protocols.

Cybersecurity and HIPAA compliance are critical concerns for healthcare organizations, and healthcare organizations already find it extremely challenging to secure the access points they know about – they are powerless to protect data that isn’t on their radar to begin with.

In order to avoid adding unwanted complexity to data storage, sharing, and security, organizations should work to develop a single data hub that becomes a “one-stop shop” for AI researchers, says Esteban Rubens, global enterprise imaging principal at Pure Storage. 

“When you create a single, seamless hub for your data, you ensure that all of your applications are drawing on a unified source of truth.  There is no need to duplicate data or maintain multiple versions of the same infrastructure, which can get very costly very quickly.”

“Data is only useful if it is accessible, and research is only useful if it is transparent and reproducible.  You can solve both those problems by investing in a data hub, which allows everyone to know exactly what data is being used, where it is, how to keep it private and secure, and what it looked like originally before you ran it through any translations or any AI models.”

A unified approach to data management can also allow organizations to innovate where it matters.  Instead of getting creative with how to access the data itself, researchers can focus on developing high-value algorithms that offer clinical decision support and other key insights that will support the delivery of high-quality, cost-effective care.

“Artificial intelligence is not going to be optional for most healthcare organizations in the near future,” cautioned Gluck, “so organizations that start getting their infrastructure prepared now will have an advantage in the short-term and the long term.”

As AI matures even further, organizations will start developing the competencies to extract more information from existing data sets while layering on additional sources of information to create rich, comprehensive portraits of individuals and populations.

Data from novel sources, such as environmental and community data, wearable device data, and genomic information, hold the potential to revolutionize the way providers make clinical decisions, Gluck said.

“The fact that there’s such a focus on it and there have been so many revelations in what it can achieve, it’s driving the cost down and it’s becoming very pervasive,” he said.  “It’s going to be commonplace.  But we need to push through the challenges we’re facing now with our data so we can get to that next iteration.”

Developing a robust, future-proof, and seamless data architecture can help ensure that a healthcare organization can be disruptive instead of disrupted.

“Disruptive technology has sometimes been held back a little bit in healthcare, and there is still a lot of work that needs to be done in that area.  Patient safety is critical, of course, as is data security.  But so is innovation,” said Gluck.  “AI is going to bring enormous positive changes to the organizations that can take advantage of it.”

“I’m excited about how AI is going to evolve as providers get more comfortable with it and the infrastructure continues to evolve to enable better insights at the point of care.”