Imaging Informatics: Tuning Up Physician Communications
By Katherine Gray, PhD
Radiology Today
Vol. 20 No. 9 P. 5

How to Minimize Unnecessary Scans — and Make the Most of Those That Happen

Communication between the ordering physician and imaging facilities is more important than ever. The health care system faces increasing demands for quality, while cost remains a factor. Changes facing providers today include the use of clinical decision support (CDS), the required use of EHRs, and new applications.

Radiology is a more complex environment than ever, increasingly encompassing exams such as CT scans, MRIs, diagnostic mammography, bone scans, thyroid scans, thallium cardiac stress tests, and PET imaging. Older technologies, such as ultrasound and basic X-rays, have undergone considerable refinements. This complexity can lead to medical errors, if the information shared between ordering doctor and facility is incomplete; Johns Hopkins recently published a report showing that medical errors, including diagnostic errors related to the use of imaging, represent the third leading cause of death in the United States.

One response from the federal government, intended to improve the initial step in communication between doctors and imaging facilities, is the use of CDS. This will be required on January 1, 2021 (with a test year in 2020), for all Medicare patients receiving outpatient advanced imaging. This means that a standard process will be in place for each ordering physician to document the specifics of the patient’s clinical symptoms with recommendations based on the medical literature as to the most appropriate procedure for evaluating the patient’s condition. These details are transferred electronically to the imaging facilities, along with the proper codes for ease of documenting the patient’s records.

Research has shown that when ordering doctors provide specificity on the patient’s symptoms and the reason for ordering the exam, radiologists can provide a better interpretation to both the physician and the patient. In addition, a study from the University of Chicago found that, when the imaging facility improved its coding, they had more complete billing and the payment lag was decreased by 28%. In other words, the whole system worked more efficiently and effectively.

Technological Refinements
Recently, there has been increased use of algorithms powered by AI to assist in the imaging process. Most of the focus has been on reviewing images with AI support to improve humans’ accuracy and speed. However, a few vendors have focused on the information in the interpretation, including tracking incidental findings to deliver suggestions for clinical correlations—even when the test comes back negative for the original diagnosis. In other words, despite a negative finding, valuable medical information can still be derived from the dictated results. Improved interpretation of this kind not only enhances patient outcomes but also can improve reimbursement for the imaging facility, as there can be more focused and productive follow-up once certain diagnoses are ruled out.

The efficacy of these new AI tools has been demonstrated by research conducted by the University of Minnesota. The study focused on testing for encephalopathies from 2014 to 2017, looking at data from almost 2,000 patients. The research showed that by a large majority—72.6% of the time—imaging results were negative. There was a clear need to predict more accurately when testing was necessary; furthermore, there was ample opportunity to harvest additional benefits from results that proved negative. The research concluded that there is a clear need for predictive algorithms to aid clinicians in making better assessments when ordering tests, and better categorization of data to aid radiologists in suggesting alternative testing and treatments when a test comes back negative.

AI algorithms can review a negative finding and search for similar phrasing in research data, extracting—based on potential correlation—and categorizing information for easier consumption with 95% accuracy. In this way, a clear “trail of breadcrumbs” can be provided to caregivers, enabling better care and, potentially, identifying the need for additional testing. Over time, this type of data collection as part of treatment will also improve the clinical processes and quality of care.

In a sense, these new tools are a logical extension to existing, qualified CDS mechanisms. The Centers for Medicare & Medicaid Services has recognized CDS solutions as a mechanism that is qualified to provide evidence-based imaging recommendations supported by appropriate use criteria. It’s past time to use advanced AI technology to get the maximum out of every test ordered, for provider, payer, and patient.

According to an old adage, first there are data, which can become information and then, eventually, they become knowledge. And over the long run, knowledge becomes wisdom. New diagnostic tools powered by AI can hasten that journey for improving the value from imaging results and, especially, improving the communications between ordering physicians and imaging facilities.

— Katherine Gray, PhD, is president of Sage HMS.