As the steady march toward population health management (PHM) in medicine continues, the critical role of radiology should not be overlooked. Radiologists touch a range of patients and, as such, can occupy a central role in patient care management. The key role that radiology plays in acute settings is well known; however, the untapped potential of radiologists in facilitating healthcare from a PHM perspective is arguably even greater. The increasing sophistication of AI algorithms, and their potential integration into the radiologist’s workflow, further foregrounds the specialty’s importance.
One key space in which radiology can lead is in the management of incidental findings. Heterogeneous strategies around incidental findings management have led to uneven results. Part of the difficulty in creating a robust solution is the lack of control that radiologists exercise in ordering follow-up studies due to existing legislation. As Gregory N. Nicola, MD, FACR, chair of the ACR Commission on Economics, notes, as diagnostic radiologists, we are not considered treating physicians and as such not allowed to manage patients, which includes ordering follow-up studies.1 In addition, the manual nature of how incidental findings are traditionally managed is not conducive to a systematic workflow. Leveraging the central role of radiologists in using AI to manage incidental findings, in collaboration with our clinical colleagues, has the potential to transform healthcare delivery —both from a diagnostic and a workflow perspective.
As a diagnostic aid to radiologists, machine learning algorithms can help in the assessment of coronary artery calcifications, bone mineral density, abdominal aortic aneurysms, and numerous other applications.2 In other words, these algorithms are increasingly becoming part of the radiologist’s arsenal. In addition, these algorithms can be configured to alert the radiologist as well as populate findings into the imaging report directly.
Imagine integrating these diagnostic algorithms with natural language processing (NLP) and the EHR. Such a scenario could facilitate consultations with an appropriate specialist in a timelier manner. For example, consider the assessment of coronary artery calcification by an algorithm. In a typical workflow, this incidental finding would not be acted upon until the primary care provider requests a consultation. However, what if the coronary artery calcification score identified by the AI algorithm and included in the radiologist report could be automatically routed to a cardiovascular team for management? There would be more timely proactive health management with fewer patients lost to follow-up. Using these types of algorithms, scenarios such as aortic aneurysms or lung nodules could also be similarly managed. Indeed, similar workflows could be implemented for many incidentalomas.
Successfully managing incidental findings is critically dependent on collaborations with our referring providers. As experts in the assessment of radiological findings, it is our responsibility to direct patients on the next best steps in their care pathway. This mindset can include identifying and even scheduling the specialist visit that is most appropriate in managing the issue. Developing these workflows in conjunction with our clinical colleagues will help us move away from the current siloed nature of healthcare delivery. Radiologists are poised to lead in this transformation. In addition to making diagnoses and communicating findings, radiologists should be active participants in directing the most appropriate next steps.
In a PHM setting in which accountable care organizations are responsible for overall patient health, the emphasis is on proactive health maintenance. Directing at-risk patients to appropriate care so they are less likely to present in the ED is the type of forward-thinking healthcare for which we should be striving. With the aid of machine learning and by tightly coordinating care with our provider colleagues, radiologists can help establish better healthcare for our patients.