Under a proposed rule by the U.S. Department of Health and Human Services' (HHS) Office of the National Coordinator for Health Information Technology (ONC), EHR developers would have to disclose to providers how AI and algorithms within its software work and what data these components are trained on to make predictions.
By being more transparent, providers would have a better understanding of what solutions work best for them, based on their patient population needs, and the rule could also reduce algorithmic bias, according to ONC, which is calling it the Health Data, Technology, and Interoperability: Certification Program Updates, Algorithm Transparency, and Information Sharing (HTI-1) rule.
HTI-1 would be implemented as a new category of the ONC’s Health IT Certification Program,
according to FedScoop.
“The idea is that you should have a standardized nutrition label for an algorithm,” Micky Tripathi, leader of ONC’s health IT division, told FedScoop.
The comment period for the proposal ended on June 20, and acquired over 200 comments. Among the organizations that provided feedback were the American College of Radiology and the Healthcare Information and Management Systems Society. Here are their recommendations:
ACR – Collaborate with the FDA
In 2016, Congress passed the 21st Century Cures Act. Section 3060 of the law excludes certain medical software functions from what defines a medical device, including EHR technology. Because of this, most, if not all, EHR solutions submitted for ONC certification are not FDA-regulated devices. ONC certification is voluntary, but frequently sought by companies to ensure their solutions comply with HHS' CMS regulatory programs.
"Consistency between HHS oversight expectations for unregulated and regulated AI will help broadly advance transparency and promote innovation," said ACR.
It also recommends that ONC identify "qualified intended users" who can assess input and output data by AI and provide feedback for addressing performance issues and reducing risks.
HIMSS – Issue requirements to AI developers, not EHR manufacturers
Because decision support interventions (DSIs) and other AI features within EHR systems are often created by clients and third-party companies, HIMSS says that transparency requirements should be directed to them instead of EHR developers, who would find it difficult to search for and disclose source information, making engagement in "intervention risk management" challenging.
Recommendations from other organizations included sharing information in the public domain, making United States Core Data for Interoperability (USCDI) modules mandatory, ensuring clinician authority is not overridden, and creating narrower definitions for software used in EHR technologies.
The final rule is expected to be released later this year.