In this issue, we talk with Matthew S. Davenport, MD, Service Chief and Associate Chair for Operations in the Department of Radiology at Michigan Medicine, about the role of the ACR® Reporting and Data Systems (RADS)
in providing standardized terminology, assessment structure and classification for reporting and data collection in patient imaging.
Q. What are the ACR RADS?
The ACR RADS provide a standardized framework to report imaging findings and make recommendations. The goal of the RADS is to reduce variability and ambiguity in radiology reporting to promote effective communication between radiologists and referring providers, guide clinical management and enable data-driven performance improvement. Most RADS include image acquisition technical recommendations, reporting terminology and definitions, categories for assessing probability of disease, guidance for report organization and management recommendations. Generally, the RADS are modality dependent. The risk assessments are provided in terms such as normal or negative, benign, probably benign or intermediate risk, and likely malignant or highly likely malignant. Many RADS also provide tools such as a lexicon, risk stratification system, atlas, flash cards, report templates and white papers.
Q. How are the RADS developed?
. The RADS are developed by committees of volunteer member radiologists and relevant referring providers. In the current state, each RADS functions relatively autonomously and reports to the Chair of the ACR Commission on Quality and Safety. The spirit, knowledge, intelligence and enthusiasm of the volunteers serving the RADS program are the reason the RADS are as successful as they are.
Q. What is the history of the RADS and how many are currently available?
The RADS programs have been in existence for decades. The earliest RADS product was BI-RADS®
. Before BI-RADS, mammography reporting was heterogeneous and inconsistent — it was often hard for referring providers to interpret what radiologists were saying in their reports and, by extension, what to do next. The BI-RADS atlas provided standardized breast imaging terminology, report organization, assessment structure and a classification system that now includes mammography, breast ultrasound and breast MRI. BI-RADS created a consistent and coherent report, so that no matter which radiologist was providing the interpretation, the referring provider could understand it and take appropriate meaningful action.
The RADS products take analog information — words that a radiologist uses to convey findings — and converts them into a digital format — a code on an ordinal scale, for example from one to five — that expresses risk and informs management.
Because of the success of BI-RADS, there is substantial interest in developing similar products for other disease states like liver cancer, lung cancer, and head and neck cancer. Currently, there are 10 RADS products overseen by the ACR, with even more in the pipeline.
Q. Why not have a RADS program for every disease state?
The RADS programs are powerful because they take complex information and simplify and homogenize it. The data output of a RADS enables effective communication and data-driven performance improvement. However, the RADS programs also have challenges. They are complex and can be intimidating to learn. Also, they are highly focused, usually on a single disease (e.g., breast cancer or lung cancer). To accommodate all relevant human diseases, one could imagine hundreds or thousands of RADS products — an untenable proposition in the current state.
So, it is important to determine which disease states are common and meaningful enough where we need to have a RADS framework, and what disease states don't need that. It’s a tradeoff between the complexity and administrative oversight required and the potential value and impact on patient care.
Q. How do you manage the increasing oversight needed to bring more RADS to market?
The RADS programs have been effective, in part, because they have been developed by teams of focused, highly motivated volunteers functioning in a nimble committee structure. That framework works well when you have 10 RADS products, but as we scale up to 20, 30 or 40 RADS products, we start to deal with a massive amount of administrative oversight and complexity. This is similar to how the needs of a small business are not the same as the needs of medium or large business. New issues can emerge such as the need for reasonable harmonization across various RADS, the need to standardize the evidentiary basis required to approve new RADS or to make updates to existing RADS.
There has to be a shift in the way we think about administering the RADS programs as we grow. As we scale, there comes a need for better administrative oversight. That’s why we have started to explore forming a RADS Steering Committee to write guidelines and help inform the RADS products, including evaluating what should and shouldn’t be a RADS product, and how RADS products should be created, updated and governed.
At the same time, we don't want to create bureaucracy for bureaucracy’s sake. These RADS groups are extremely successful. So, the intent of the steering committee is to try to maintain the nimbleness, creativity and volunteer spirit that exists in RADS, while providing some administrative oversight that allows this valuable program to scale without collapsing under its own weight.
Q. You recently co-authored an article in the JACR® about PI-RADS. What are the key takeaways from that paper?
PI-RADS is a widely used RADS program that’s informed by people from across the world to improve early diagnosis and treatment of prostate cancer. The paper our group (senior author: Prasad Shankar) published in the JACR — Prostate Imaging and Data Reporting System Version 2 as a Radiology Performance Metric: An Analysis of 18 Abdominal Radiologists
— shows how you can use PI-RADS scores as individual radiologist quality assurance measures to ensure you're getting the results you expect. It is desirable to have a narrow band of expected positive predictive values for clinically important cancer at each PI-RADS score from three to five. Likewise, it is desirable to have a narrow band of expected negative predictive values for clinically important cancer at each PI-RADS score from one to two. We want those data to be in a consistent range so that the referring urologist is not receiving markedly different results depending on who reads the scan. If the data is out of expected boundaries, the radiologist can course correct on future scans.
This is one of the beauties of the RADS: In medicine, we often think about physicians wanting to be exceptional and go above and beyond. But, ironically, being exceptional in this case means that other radiologists are not performing to the same standard, so all patients are not getting the same outcome. Here, we are aiming for everyone to do a consistent, excellent job. In the paper, we try to establish benchmarks for what a radiologist should be seeing in terms of positive predictive values for PI-RADS 3, 4 and 5.
There's a Goldilocks zone — a sweet spot of exactly where you want to be. If the positive predictive value is too high, it suggests that the radiologist isn’t calling enough findings, and if the positive predictive value is too low, it suggests that the radiologist is either calling too many findings or is unclear about the rules for that category. Our paper tries to determine what the Goldilocks zones should be. Again, broadly speaking, RADS products take analog word data and convert them into digital data that can enable effective communication and inform performance improvement.
Q. What actionable steps should radiologists take now to start implementing RADS, if they haven't already?
If a group is interested in pursuing a RADS adoption in their practice, change management is key. There will be people in the group who want to do it, people who are on the fence, and people who don't want to do it. Consider discussing the implementation of RADS as a team. What are the pros and cons? Consider including in the discussion the people who are ordering the exams and receiving the reports. How will this affect their (and your) practice?
If a group decides to use one of the RADS, it is helpful to have education sessions (for example, take advantage of some of the new RADS educational modules). Sometimes a 100-page lexicon can be a bit intimidating. Another helpful tactic is to use peer learning sessions, where cases are discussed in a welcoming group environment, and the reason for assigning a particular RADS score can be discussed. Implementing a new RADS often requires a champion to socialize the idea inside the group and among the referring providers, and to deliberately systematize the RADS so its full potential can be utilized (e.g., automated data extraction to promote performance improvement).