Scaling structured radiology reporting template use across multiple health systems to build a data registry
Scaling a structured reporting template across multiple health systems to build a clinical data registry is possible with a few key pieces of advice, experts reported Friday in JACR.
Back in 2017, three Pennsylvania academic imaging departments agreed to start using such templates to categorize potentially malignant findings in the kidney. During the two years that followed, physicians transmitted all adult exam information across ultrasound, MRI and CT to a centralized data site.
Those involved have now gathered a set of 161,000 examinations from 103,000 unique patients. Average physician template use is also north of their goal of 90% at all sites included in the study.
“Ultimately, we scaled implementation of a mandatory structured reporting template coding the malignant likelihood of kidney findings in all abdominal imaging reports from a single academic health system to two other health systems,” Tessa Cook, MD, PhD, director of the Center for Practice Transformation in Radiology at the University of Pennsylvania, and colleagues wrote March 25. “Through the implementation of mandatory structured templates, we discovered multiple barriers and facilitators to radiologist compliance with the use of templates, trial team and site engagement, and site IT and enterprise IT coordination,” they added later.
Cook and colleagues used artificial intelligence to extract categories from the reports, while radiologists were asked to addend them with missing or incomplete templates after the first month. Each site also submitted sociodemographic and clinical data on subjects. Matching info from both of those categories was obtained at a 94%-97% rate across the three sites.
Meanwhile, use of the templates after addendum requests climbed to 95% or higher at the participating institutions. Cook et al. offered five take-home points for others looking to do the same: (1) engage and educate radiologists and nonclinical stakeholders; (2) continuously monitor registry data for accuracy; (3) address each site’s unique data-sharing and physician buy-in needs; (4) anticipate differences in each health system’s readiness for change; and (5) build trust among radiologists and imaging informatics professionals.
“The multisite database generated from this trial can provide insight into radiologist and patient factors driving variability in categorizing the malignant potential of imaging findings in the kidney as well as evidence-based patient outcomes,” Scott and co-authors concluded.