New clinical decision support software guidelines highlight keys to self-regulation

With the Food and Drug Administration adopting a limited role in regulating software that assists physicians with clinical decisions, a coalition of providers and developers released voluntary guidelines aimed at helping the industry self-regulate.

The 21st Century Cures Act excluded clinical decision support (CDS) software from FDA oversight as long as providers can review the information that is fed to them and make an independent decision. To that end, the Clinical Decision Support Coalition has released (PDF) draft voluntary guidelines to ensure CDS software remains as a support tool. The Coalition, which has previously called on the FDA to released case study guidance, is asking for public comments by July 1.

RELATED: 3 ways Trump’s FDA nominee could reshape digital health

Allowing software to take on a broader role in clinical assessments is “risky,” writes Bradley Merrill Thompson, a medical device attorney with Epstein Becker Green and the Coalition’s general counsel wrote in a LinkedIn post. Therefore, the guidelines focus on four key factors to ensuring software remains as a support tool for independent clinical diagnoses:

  • Transparency: In addition to providing treatment recommendations, CDS software should give clinicians the clinical evidence that supports any recommendation.  
  • Competent end users: It may seem obvious, but CDS software should be used by a trained professional with experience in the disease that software supports. As Thompson points out, a psychiatrist should not use CDS software to recommend cancer treatment.
  • Time: CDS software shouldn’t be used to make split-second diagnoses. Physicians need adequate time to reflect on any machine-driven recommendations in order to make an independent judgment.
  • Access to outside information: CDS software shouldn’t be a clinician’s primary resource. This recommendation blends in with transparency by ensuring the physician has access to all sources of information considered by the software. For software that reviews more targeted data points, developers should consider whether a user is likely to have other information available from a physical exam, lab results or images.

“The coalition’s goal is to develop a framework that will provide consistency in user empowerment across software developers,” Thompson wrote. “The key for this unregulated portion of CDS is ensuring that the healthcare professionals can gather and analyze the information needed to make the best recommendations to their patients.”

Despite the rise of machine learning, research shows human doctors are still more effective than symptom checker apps and websites. But hospitals are exploring the use of CDS software to improve care and reduce costs.