One academic medical center is detailing how it increased the use of its alert system to notify providers about critical imaging findings—a crucial safety issue in the specialty.  

The University of Mississippi Medical Center had witnessed a troubling 27% drop in the number of such alerts during a three-year period. This despite the institution recording a 15% uptick in the number of imaging exams in the same timeframe, experts wrote March 22 in Current Problems in Diagnostic Radiology [1].

To begin addressing the concern, UMMC launched a program to educate its radiologists about the importance of critical alerts. It also began a campaign of “systematic reinforcement,” along with implementing a time-stamp macro in the dictation system, and engaging with other departments to bolster communication.

They’re finding promising results, with interventions producing an increase in the number of monthly critical alerts—particularly those that require imaging follow-up. UMMC has seen a “steady” improvement in documentation, with compliance at nearly 97% and a 0.5% per-month increase in the number of alerts to providers with up-to-date contact information.

“In conclusion, it is possible to increase the utilization of the critical alert system through education and reinforcement efforts,” lead author Robert W. Morris, MD, and colleagues wrote March 22. “Similar efforts can also result in improved documentation, all without sacrificing time compliance goals. All of these efforts serve to increase timely and effective communication of critical radiology results for improved patient safety.”

Almost 10% of radiology reports contain a critical result, according to previous research, and communication of them is key. Among radiology, communication errors are one of the top-six reasons for malpractice claims, the authors noted.

At 720-bed UMMC, radiologists use a commercial software to track and communicate any critical alerts, which is integrated into dictation software. They’re categorized into three separate tiers: red for an emergency needing a response within 60 minutes, orange for “requires attention” within 12 hours, or yellow for nonurgent necessitating a 24-hour timeframe for response. The onus is then on the referrer to close the loop via either the alert system or the EHR.

To begin improving adherence, the hospital updated its reporting protocol to add more specificity on what warrants an alert. They simplified things by including all needed info on a single webpage screen, prominently placed it on their departmental webpage, and emailed residents and faculty about it. Morris and colleagues also created a new dictation macro, with an auto-populated date and time. Its use was incorporated into the new critical results protocol with a time stamp that ID’d when a red-level alert arose.

“Our two critical results coordinators monitored red alerts for the presence of the macro critical time stamp, and utilized our internal safety reporting system to alert radiologists and departmental leadership for when the protocol was not followed,” the study authors noted.

Coordinators tracked the use of alerts for 15 months following the change. They observed an average increase of about 23 per month (up from 11), which did not reach statistical significance. Adoption of the new time stamp was at 83.3% when the study period started but leapt to 96.9% by its conclusion. UMMC was able to maintain target closed-loop communication time compliance of greater than 99% for all alert types during the intervention period, the authors noted.

By the end of the study, the department was sending critical alerts on just 3% of reports, fewer than the 10% estimate in a previous study on this topic. This possibly suggests that some findings warranted this tag but were not getting it.

“We have attempted to do an analysis using keyword searches in our electronic medical record, but have found this manual process to be so time intensive as to be impractical,” Morris et al. wrote. “Moving forward, our goal is to utilize commercially available software using natural language processing to monitor for reports that should contain alerts.”