Radiology residents are generally accurate in their interpretations of emergency department CT studies performed on overnight call, though they do miss a handful of important findings, concluded a study from University Hospitals Case Medical Center in Cleveland.
Among the more than 6,000 CT exam interpretations in the study, residents were most accurate in the evaluation of head CT scans, with errors in fewer than 2% of cases, followed by pediatric body CT at about 3%. Miss rates were less than 5% in adult abdomen and pelvis studies.
"They did very well: We had very low discrepancy rates, and when compared to other institutions we were similar," said Dr. Jennifer Montgomery, who originally presented the results at the 2012 RSNA meeting.
On-call resident interpretations
At Case, as in other academic institutions, on-call radiology residents initially interpret overnight emergency radiologic studies, she said. The studies are then reviewed the following day by subspecialty-trained attending radiologists.
"The residents primarily take care of the emergency department, but we are also available for any stat readings that would occur for the inpatients," Montgomery said. Specialty-specific attendings are also available on call at night, but residents primarily make the interpretation on their own.
The study aimed to determine the discrepancy rates of emergency department CT exams and rank them in order of importance compared to a retrospective review. The final attending dictation was also reviewed for any indication of a discrepancy, such as verbal notification of the ordering physician, or the current physician in cases where the patient had been admitted.
CT exams were performed on scanners from multiple vendors, all with 64 or more detector rows, including iCT 256 from Philips Healthcare, Aquilion One from Toshiba Medical Systems, and Somatom Sensation Cardiac 64 from Siemens Healthcare.
"We wanted to focus on, if there was a discrepant read, [how] that would have clinically impacted the patient," Montgomery said.
Clinical impact
The group classified discrepant reads into four categories based on the clinical impact in the emergency setting:
- Discrepancies without significant immediate clinical impact, and/or findings that required outpatient follow-up
- Discrepancies that could have affected patient care on that visit if not corrected; however, the patient received the appropriate care based on other correctly interpreted imaging findings or symptomatology
- Discrepancies that changed patient management in the emergency department or during the course of patient admission
- Sentinel events for discrepancies with potentially serious adverse outcomes if not corrected, such as an intracranial bleed, central pulmonary embolism, perforated appendicitis or abscess, or perforated viscous or organ trauma
The 6,050 CT exams included 2,736 head scans, 1,733 abdomen and pelvis scans, 553 chest exams using the pulmonary embolus protocol from the adult emergency department, 859 neurologic scans, and 169 abdomen and pelvis scans from the pediatric emergency department.
Overall, the residents delivered the lowest discrepancy rates in adult (1%) and pediatric (2%) head CT scans. Pediatric body CT interpretations came in at 3%, and the highest discrepancies were seen in adult pulmonary embolism cases, as well as abdomen and pelvis scans (4% to 5%). The results were in line with recently published literature showing discrepancy rates less than 5%, Montgomery said.
The majority of discrepant reads were classified as category 1 or 2, except for pediatric body CTs, where 57% of the discrepancies were category 3, though this represented only four patients over the course of the year. The single category 4 discrepancy was a missed perforated appendicitis with abscess.
Summary of significant discrepancy rates
|
Areas that were identified for focused educational improvement include otitis/mastoiditis in pediatric neurologic CT, and CT findings of pyelonephritis and thickening of the gastrointestinal tract, particularly the colon and antrum/duodenum, Montgomery said.
"With the neurologic CTs, we didn't really find too many things that were very significant ... but one area we did identify was the possibility of otitis/mastoiditis," she said. "It's not that people were misinterpreting the findings; we saw it as more of a blind spot for the residents, and so we wanted to identify those things and bring them to the attention of our clinical attendings so they could then bring it to the attention of the residents."
Subtle pyelonephritis was also being overlooked, so it was important to make sure that residents knew the signs and looked for them, she said. No discrepancies were seen in severe cases, however, and the patients were generally treated appropriately based on clinical findings.
"Findings can be fluid in the mastoid air cells, and I think the mastoid air cells can be quickly overlooked ... residents are focused on making sure there are not more serious issues like an intracranial bleed or cerebral edema," Montgomery said. The idea is to make sure the search pattern is thorough, she added.
The program director and the section heads have received the study results and will incorporate the findings into their lectures -- in particular, a spring lecture series that focuses on call issues for new and returning residents, she said.
Overall, the group performed a little better than some similar institutions, and "when compared to the 2.6% discrepancy rate in RadPeer data, where residents review each other, we were similar or lower than published data for discrepancies," she said.
The ACR's online RadPeer program enables peer review during routine interpretation of images by expert readers, and it provides summary statistics by participating physician and modality.
After the educational initiatives are fully implemented, the study will be worth repeating to see if the results in specific areas have improved, she said.
"We're not concerned about our discrepancy rates, but anything we can do to lower them we think is worthwhile," Montgomery said.
