Radiology artificial intelligence ROI calculator demonstrates ‘substantial’ benefits at 5-year mark
Experts have developed a comprehensive calculator for assessing radiology artificial intelligence, which has demonstrated “substantial” return on investment at the five-year mark following implementation.
Researchers built the tool to help imaging leaders quantify comparative costs, estimated revenues and the value of using an AI platform at a U.S. hospital. They built parameters for the calculator through expert interviews and a review of literature related to imaging AI, according to research published March 16 in JACR [1].
Introducing AI into the hospital radiology workflow resulted in labor time reductions and delivery of ROI at 451% over a five-year period. Returns soared to 791% when also considering radiologist time savings.
“We demonstrate a substantial five-year ROI of implementing an AI platform in a stroke management-accredited hospital,” Lauren Nicola, MD, CEO of Triad Radiology Associates in Winston Salem, North Carolina, and co-authors wrote Saturday. “The ROI calculator may be useful for decision-makers evaluating AI-powered radiology platforms,” they added.
The calculator was initially created in Microsoft Excel to assess the ROI for Calantic, a centralized architecture for AI applications offered by Bayer (company representatives also served as study co-authors). The platform was designed to automate time-consuming tasks, optimize workflows and support radiological detection from imaging data.
Nicola—who also is chair of the American College of Radiology’s Reimbursement Committee—conducted comprehensive workshops aimed at defining a framework for the calculator. The final tool accounted for the perspective of all stakeholders impacted by AI implementation, including the differing impact for hospitals versus diagnostic imaging centers. Fourteen different AI powered applications, hosted within the Calantic platform, were assessed in the study, all relating to thoracic and neurological indications.
According to the calculator, time savings for radiologists included over 15 eight-hour working days of waiting time, 78 days in triage, 10 days in reading and 41 days of reporting time. The AI platform also provided revenue returns for the hospital by bringing in patients for clinically beneficial follow-up scans, hospitalizations and treatment procedures, the authors noted.
“Results were sensitive to the time horizon, health center setting, and number of scans performed,” Nicola et al. wrote. “Among those, the most influential outcome was the number of additional necessary treatments performed due to AI identification of patients.”
The calculator utilized estimated annual scan volumes of X-ray, CT and MRI, derived from data provided by a representative U.S. group practice with an assumed annual increase in visits of 10%. Using that data and expert input, an algorithm categorizing scans by body region estimated the number of relevant exams sent for further analysis by the AI application. Over the five-year time horizon, the estimated revenues generated from platform applications were nearly $3.6 million while the estimated total costs were about $1.8 million. This amounted to a return of about $4.51 for each dollar invested.
“The driving forces behind the predicted positive ROI are manyfold,” the authors noted. “Firstly, substantial reduction in waiting, triage, reading and reporting times enhances radiologist productivity and expedites patient care, leads to clinical and operational benefits. Secondly, increased diagnostic accuracy and early disease detection leads to augmented clinical benefits. Lastly, as a platform the value of individual AI applications is enhanced by integrating processes that cover procurement, installation, and maintenance. The ROI calculator offers itself as a comprehensive and evidence-based calculator when assessment of financial viability and value proposition of integrating AI technology into radiology practices is needed.”
Read much more about the results, including potential study limitations, at the link below.