It is inspiring to see innovative uses of technology come into play to meet the challenges faced by radiology practices. In this case, artificial intelligence (AI) is helping solve a thorny barrier to the effective implementation of Medicare's appropriate use criteria (AUC) program.
In an original article published in the September 30 issue of the Journal of the American College of Radiology, researchers from the Department of Radiology and Medical Imaging at the University of Virginia Health System (UVA) address a challenge to the effective implementation of AUC: the proliferation of unstructured "free text" entries from the ordering physician if matching predefined indications are not readily available using the traditional search-and-select approach. When predefined indications for advanced imaging exams are not a good match, ordering physicians may resort to using free-text orders.
The AUC program resulted from the Protecting Access to Medicare Act (PAMA), which was enacted in 2014. The program requires the physician who places orders for outpatient advanced diagnostic imaging (MRI, CT, PET, and nuclear medicine) studies for Medicare patients to review AUC by making use of a clinical decision support (CDS) system approved by the Centers for Medicare and Medicaid Services. Although the payment penalty phase of the AUC program was set to begin next year, CMS has now proposed in the 2022 Medicare Physician Fee Schedule rule to begin the payment penalty phase on January 1, 2023, or the first year following the end of the COVID-19 public health emergency, whichever comes first.
When consulting the CDS, the physician can review an array of indications that map to AUC to determine if the exam is "appropriate" or not. Here an approach using AI comes into play. The UVA researchers describe an AI tool capable of considering additional patient-specific data from the EHR when determining the appropriateness of advanced diagnostic imaging studies. An example cited in the JACR article is that a patient with an MRI-incompatible pacemaker could automatically receive a low appropriateness score or an alert advising against MRI for that particular patient. When given the opportunity, the authors note that physicians and all provider groups appear to prefer the free-text–AI approach over the traditional search-and-select approach in determining the appropriateness of imaging orders.
Bottom line, the innovation of the free-text–AI approach to ordering may ease or enhance PAMA compliance for determining appropriateness for advanced imaging order entry. Clearly, hospitals and imaging centers can benefit from such AI tools to assure compliance before the penalty phase of the AUC program goes live.