The U.S. Food and Drug Administration's latest plan for overseeing clinical decision-support software leaves too much room for interpretation over what types of software the agency intends to regulate, according to health systems and software vendors.
While stakeholders largely applauded the FDA's vision of a risk-based regulatory approach in theory, a major point of contention involves confusion over the FDA's definition of what software would be included under the FDA's oversight.
FDA released an outline in September for how it plans to regulate clinical decision-support software, or software systems that analyze data to help clinicians and patients make treatment decisions. The draft guidance uses a risk-based approach, with the FDA focusing its oversight based on the risk to patients if software malfunctions.
For categories deemed at low-risk to patients, the FDA won't enforce applicable regulatory requirements. FDA's principal deputy commissioner Dr. Amy Abernethy said that would give developers more room to create and adapt their software.
Under the draft guidance, software meant to "drive" management of clinical care would be considered a type of medical device falling under the FDA's regulatory oversight, while software simply meant to "inform" clinical management would exempt. That distinction is meant to focus the FDA's oversight on tools that actively guide the next diagnostic or treatment a patient receives, and away from software that's helpful, but not necessary, for clinical decisionmaking.
"Drive functions provide enhanced support beyond simply supporting or providing a recommendation about prevention, diagnosis, or treatment of a disease or condition," according to the FDA's guidance.
But that line—dividing tools that "inform" from those that "drive"—isn't useful, according to some health systems and software vendors.
"We believe there is no actual distinction between software functions intended 'to guide' clinical decision-making … and those intended 'to provide recommendations' for clinical decision-making," Kaiser Permanente wrote in a letter to the agency. "We recommend FDA resolve this ambiguity in the final guidance by simply eliminating the distinction between these two concepts."
Epic Systems Corp., one of the largest providers of electronic health record software, agreed.
"Determining whether a CDS function 'aids' in treatment or diagnosis (drives) or simply 'supports (healthcare professionals') decision-making' (informs) … is an exceptionally difficult and subjective exercise, and will inevitably lead to inconsistent application across the industry and by FDA," the company wrote.
CommonSpirit Health, Mayo Clinic, Cerner Corp., Allscripts and the Advanced Medical Technology Association also expressed confusion over the terminology, though not all agreed with arguments to combine the two phrases into one category. Some asked for a clarification of the two terms or suggested different ways for the agency to think about clinical decision-support software.
CommonSpirit recommended the FDA instead draw a distinction between tools that provide recommendations to clinicians, and those that make decisions for clinicians.
"The FDA could draw a much clearer line," CommonSpirit wrote in a letter. "If an algorithm takes the agency and decision-making authority away from the clinician and dictates the next course of diagnostic testing or treatment without any reasonable opportunity for intervening clinical judgment, then such software functions should be regulated as devices."
If the FDA maintains its plan to regulate software that "drives" clinical management as a medical device, that could lead "many existing CDS algorithms being subject to the FDA approval process," according to a comment from the American Hospital Association—possibly hindering innovation down the line.
Software vendors also expressed skepticism at one of the criteria the FDA said it would use to determine what software merits regulatory oversight.
As part of the risk-based approach, the FDA said it plans to focus oversight on clinical decision-support software that meets two criteria: those that aim to help providers and patients manage serious or critical clinical conditions, and those that do not explain to users how the software reaches its recommendations.
Software vendors said they want more clarification around what type of information they need to provide to allow users to independently evaluate a recommendation, so that their tools can be exempt from the FDA's regulation.
"We believe that examples need to be developed that help illustrate to the manufacturer what it means to meet the test of independently evaluate in terms of timeliness, context, form, content and proximity to the clinical workflow," Cerner wrote in a letter to the agency.