Radiology can provide immense value to the healthcare system, not only through accurate diagnostic interpretations, but by helping direct the appropriate use of imaging.
But first, providers need to make sure they are leveraging their PACS and workflow technologies to the fullest.
Case in point: review of prior images. Many patients will have relevant prior scans that can serve as a helpful tool for making differential diagnoses. Often, recommendations for additional imaging or interventions (RAIs) can be avoided through a review of priors. The key is to make sure the technology available to radiologists is helping them manage what could be a large number of prior examinations, says Ankur M. Doshi, MD, radiologist at NYU Langone Medical Center in New York City.
“In patients with complex medical histories and numerous imaging studies, it can sometimes be difficult for the radiologist to identify key prior examinations if the PACS environment isn’t optimized to highlight relevant priors,” he says.
Doshi, who is also an assistant professor of radiology at NYU School of Medicine, and colleagues wanted to test just how many recommendations for follow-up imaging or procedures could be avoided if a comprehensive review of prior images was conducted, given that estimates indicated more than 10 percent of all radiologic exams and more than 30 percent of abdominal CT exams contain RAIs.
In a study published earlier this year in the Journal of the American College of Radiology, Doshi and colleagues retrospectively reviewed more than 1,000 abdominopelvic CT and MRI reports containing RAIs at their facility. Previous imaging for each patient was then assessed, including all relevant body parts and modalities, to see whether any of the RAIs could have been avoided.
The results were encouraging to Doshi and his team, as just 41 of the 1,006 RAIs (4 percent) could have been avoided through comparison to prior imaging examinations. This indicates the facility is doing a good job of comprehensively reviewing prior exams.
At the same time, it’s not clear hospitals across the country would have the same results as one of the top academic medical centers, and even if the results were generalizable nationwide, 4 percent of follow-up imaging requests would add up to make for significantly higher costs and the potential for many patients to face anxiety or complications needlessly.
“Radiologists add additional value in patient care by not only informing referring physicians of the best imaging test to order,” says Doshi, “but if pathology is discovered, they are really valuable in establishing long term stability to help determine benignity.”
Doshi says PACS vendors can help make the process of reviewing priors easier for radiologists by optimizing the viewing environment. They may offer the ability to automatically detect and display relevant prior imaging, but the key is striking a balance between providing sufficient information to help make a diagnosis and providing redundant or irrelevant information. This includes the display of imaging from different modalities or other body regions that happen to overlap the anatomy of interest. For example, when pathology related to the adrenal gland is discovered on an abdominal MRI, a prior chest CT can be helpful for comparison since this exam includes portions of the upper abdomen.
“PACS could have the ability to detect the study you’re reading—say CT of the abdomen—and highlight the most relevant priors, including prior CTs, MRs and ultrasounds of the abdomen, and then in some way call to attention other imaging that may cover the area of interest, such as chest CT or lumbar spine MR,” says Doshi.
Here, predictive algorithms could help enhance the radiologist’s experience and display only useful information. Doshi notes that some prior imaging may not be as relevant or informative based on the detected pathology, such as the case of a prior abdominal radiograph when encountering a liver lesion, for example.
On his wish list for an optimized PACS is something along the lines of a two-tiered display of relevant prior examinations, with one column featuring the most direct priors and an additional column with relevant prior examinations identified by an algorithm, possibly even a dynamic list generated based on pathology the radiologist is actively dictating.
Ultimately, Doshi looks to giants in other industries as the model. Amazon and Netflix, for instance, can quickly assess users’ interests