Private data sharing between organizations can facilitate rapid innovation in healthcare, especially by enabling AI development using previously inaccessible data. However, there are multiple barriers healthcare institutions currently face when it comes to unlocking data.
By collaborating and sharing data without compromising privacy, speed and integrity of the data, healthcare institutions can predict future diagnosis and reduce complexities associated with internal and external data sharing, which often includes sensitive personal identifying information.
“The most significant barrier with data sharing in healthcare is the high cost and high effort level required to maintain compliance,” said Riddhiman Das, co-founder and CEO of Tripleblind.
Das, who will be sharing his insights on the benefits of private healthcare data next month at HIMSS22, said, pointing out more than 130 jurisdictions have data privacy regulations in place today – these include HIPAA in the U.S., GDPR in the EU and PDPA in Singapore.
“Even within the U.S. there are multiple regulations, such as California’s CCPA,” he added. “Another barrier is some healthcare systems consider themselves in competition with other systems, which makes them less likely to share data.”
TripleBlind’s Blind Learning technology enables private datasets to be used for model training without ever being moved or re-identified. The data remains usable by data scientists, while enforcing all privacy regulations.
From his perspective, there are two principal reasons why gaining access to private data accelerates healthcare innovation.
First, healthcare-focused algorithms, such as diagnostics and decision support, must be trained – and the larger and more diverse data with which that training is conducted, the more accurate the algorithm will be.
“Healthcare systems today train their algorithms on their data from local patients,” Das explained. “One system’s data may skew toward older, white, sicker patients. If that system can gain access to data from people of color, younger and generally more healthy patients, the algorithms they develop will be more accurate.”
In addition, Das said, some types of data, such as genomic information, has previously been difficult to share while maintaining patient privacy.
“Effectively sharing this type of data opens new opportunities for new diagnoses and treatments,” he said.
