Automatically predicting the likelihood of a screening mammogram recall using EMR data
Scientists have deduced how to predict a patient’s likelihood of experiencing a screening mammogram recall using data from the electronic medical record, according to a new study published Monday.
Women who are younger, have a surgical history of breast cancer, and have not experienced a prior breast imaging exam or recall are likelier to experience one, experts note. Researchers hope to build on the success of their model to conduct a prospective investigation, gauging the value of EMR-based tools to help predict recalls at the time of scheduling.
“Our study demonstrates the feasibility of using patient and breast care variables that can be automatically extracted from the EMR to predict the likelihood of a screening recall,” Jikai Zhang, with the Department of Electrical and Computer Engineering at Duke University, and co-authors wrote July 17 in the European Journal of Radiology [1]. “A major challenge of this link of investigation is the incompleteness of EMRs, but this study demonstrates that, as a proof of concept using a real-world setting, it is still feasible. Institution-wide efforts to improve the completeness of EMR data entry will likely improve the performance of models further,” he added later.
For the study, Zhang et al. identified all 21,543 women age 40 or older who underwent a digital breast tomosynthesis exam at a single, unnamed U.S. institution in 2018. Their primary outcome was those who experienced a screening recall recommendation of BI-RADS (Breast Imaging-Reporting and Data System) 0.
Women in the final sample were an average age of 59 and skewed mostly white (64%) and Black (27%). Of the total, 2,182 or 10% experienced a screening recall recommendation. Multiple variables were “significantly” associated with a future recall. For instance, a 50-year-old woman was 26% less likely to be recalled when compared to someone at age 40. Those with a history of breast cancer were 2.3 times likelier to experience a recall versus individuals without one. Women with a previous recall in the last five years were 23% less likely to experience another one compared to women without a recall history. And women with a prior screening mammogram during the past 18 months were 40% less likely to have a recall and 32% less likely for individuals with a screening mammogram in the last 18-30 months.
Zhang and co-authors see important implications from their work as practice leaders seek to improve efficiency and the patient experience.
“While previous studies have identified individual patient, imaging and radiologist characteristics predictive of screening recall, this work utilizes structured reporting elements within the EMR to facilitate the development of EMR tools that can influence mammogram scheduling,” the authors noted. “By triaging women with a higher likelihood of a screening recall recommendation, institutions can schedule women earlier in the day and at a location that offers same-day screening interpretations and subsequent diagnostic services. This will help to reduce women’s anxiety and thus minimize one of the principal harms of breast cancer screening. Furthermore, we calculated different recall thresholds based on a practice’s capacity and desire to accommodate potential same day screening recall examinations.”
Read much more in EJR at the link below.
Marty Stempniak has covered healthcare since 2012, with his byline appearing in the American Hospital Association's member magazine, Modern Healthcare and McKnight's. Prior to that, he wrote about village government and local business for his hometown newspaper in Oak Park, Illinois. He won a Peter Lisagor and Gold EXCEL awards in 2017 for his coverage of the opioid epidemic.