Routing exams for optimal performance



Increase the number of sub-specialized interpretations,reduce turnaround time, and improve the work balance of radiologists by intelligently routing studies based on optimized staffing levels as well as radiologist training and expertise. The algorithm may also suggest when to request back-up assistance based on predicted workload and staffing.





Define-AI ID



Rich Heller, Woojin Kim, Worth Saunders

Panel Chair

Rich Heller

Non-Interpretative Chairs

Alexander J. Towbin, Adam Prater

Panel Reviewers

Business Facing Panel


Creative Commons 4.0 


Public Comment

Clinical Implementation

Value Proposition

In many places, there is an emphasis on sub-specialized interpretations and optimizing service levels for professional radiology services. Even in radiology practices that do not have the required metrics around sub-specialty interpretations, there is often a desire to have certain types of exams interpreted by radiologists with specific expertise. A system that can intelligently route exams to the most appropriate radiologist could reduce the turnaround time, meet specified requirements, and potentially improve interpretive quality and satisfaction of radiologists.


A 17 radiologist practice provides around-the-clock services to a large community hospital and its off-campus outpatient imaging center. The new head of neurosurgery has stated that at least 80% of all advanced neuroimaging (CT and MRI) should be interpreted by a neuroradiologist. The hospital has also begun offering transcatheter aortic valve replacement (TAVR) and the CT interpretations of these patients are done by select radiologists who work with the cardiovascular surgeons. There are also strict turnaround time requirements. The practice would like a system that can automatically route studies to the most appropriate radiologist, balancing the need for sub-specialty interpretations with turnaround times. The ideal system also improves quality by better balancing workload and increasing subspecialty reads, which ideally lowers the number of cases referred to the department for review due to an interpretive quality issue. In addition, since many radiologists prefer to read within their sub-specialty, the system may improve radiologist job satisfaction. 

Workflow Description

There are several groups of incoming data. Information on the available radiologists comes from the e-schedule. This can include information on who is scheduled to be working currently, who is logged on and actively working now, and who is supposed to be coming online later. The worklist data, containing information on the specific exams to be read, is also inputted and constantly updated. A separate prediction tool inputs data on expected incoming volume and case mix. Pre-set information on required turnaround times, percent sub-specialty interpretation, and specific exam requirements (i.e., TAVR only by Drs. X, Y, or Z) are also inputted. Additionally, historical data on radiologists is inputted. This can include information on radiologist expertise, including data on sub-specialty training and preference (i.e., they may not be fellowship trained, but they read mostly within a certain sub-specialty and are considered experts in that area) and productivity. 

Technical Specifications





The list of radiologists scheduled to be working, actually online and expected to be online at a certain time.

Potential Feature

Names of specific radiologists



Data from current worklist



Features of an exam such as modality, location of the exam, body part, wRVU, patient demographics, referring provider, and indication, as well as related exams obtained at the same time, all of which could be extracted from the RIS and PACS worklist manager.

Potential Feature(s)

Modality, Body part, patient demographics (age&sex), indication, referring provider, similar exams ordered at same time, wRVU (per CMS), location of the exam (point of service and location in hospital)


Data on predicted volume and case-mix


From a separate prediction tool, input the expected volume of exams as well as the predicted case-mix (ie types of exams)

Potential Feature(s)

Volume of exams, modality of exams, body part, indication, referring provider, exam location


Pre-set Information on required sub-speciality interpretations and certain exam types



The specific requirements to interpret a certain sub-category of the exam (like advanced neuroimaging) from a certain cohort of radiologists (i.e. neuroradiologists). For example, 80% of advanced neuroimaging to be read by neuroradiology. Another example, all TAVR exams to be ready by Rs. X, Y, or Z only.

Potential Feature(s)

Specific requirement for exam types


Radiologist Turnaround Times



Delivery of timely information to the healthcare team as needed; can be defined as the time from report available on the worklist for interpretation to the final report signed

Potential Feature(s)

Turnaround time


Data on radiologist expertise, preference & performance



Information on each radiologist in the practice based on the selection criteria of the facility. This may include radiologist’s preference, the rate of dictation, and subspecialty expertise, as inputted by training or imputed by reading preferences.

Potential Feature(s)

Radiologist productivity (i.e. exam types read per unit time), radiologist expertise


Primary Outputs


 Route to appropriate radiologist



Direct exams to the radiologist that will satisfy the preset goals




Study-Radiologist Pairing





Suggest when and what type of additional resources would be useful to satisfy preset goals