To err is human; analysis finds radiologists very human

December 1, 2010

Radiologist errors identified from an analysis of 656 imaging exams paints a unique picture describing why radiologists make errors. The analysis also points to ways such mistakes can be prevented.

Radiologist errors identified from an analysis of 656 imaging exams paints a unique picture describing why radiologists make errors. The analysis also points to ways such mistakes can be prevented. 

Dr. Young W. Kim, a radiologist who practices at Tripler Army Medical Center in Honolulu and Brooke Army Medical Center in San Antonio, reported Tuesday at the 2010 RSNA meeting that nearly one-third of the delayed diagnoses he studied at those two military hospitals from July 2002 to January 2010, were not only missed on the first exam, but were also not recognized on subsequent radiological examinations.

Moreover, Kim’s analysis provides insight into the very human world of radiological practice.

In Kim’s analysis, diagnostic errors are defined as missed diagnoses that are later corrected with a definitive review. Though Kim’s definition seemed abstract, the implications of botched interpretations are quite real. They are the leading cause of malpractice litigation, accounting for twice as many claims as medication errors, Kim said.

The legal liabilities associated with radiology are increasing in step with its growing influence on medical practice as a whole, Kim said. From the medical literature, the daily error rate from dictation is possibly 3% to 4% when considering all interpreted exams, but the rate is as high as 33% for abnormal studies.

Kim and colleagues identified 1279 errors examining 656 exams with delayed diagnoses collected over an eight-year period at the Tripler and Brooke Army Medical Centers. Cases were selected from difficult case conferences and the author’s files. Each case was reviewed by two radiologists. When appropriate, the case was assigned more than one type of error. Date of the original exam, imaging modality, and dates of incorrect diagnosis and corrected diagnosis were noted. Whether the diagnosis was missed on subsequent radiologic exams and the imaging technique from which the correct diagnosis was made were also noted.

Fifty-four percent of the errors arose from the interpretation of general radiography. CT and MR reading errors accounted for 23% and 9% of the errors respectively. Nuclear medicine and ultrasound contributed 7% and 1% respectively.

Under-reading was the main source of errors, Kim said; it accounted for 42% of the total. Errors associated with “satisfaction of search,” or the radiologist’s failure to thoroughly examine anatomy after seeing abnormalities relevant to a primary diagnosis, was the second most common error. These errors accounted for 22% of the total. Faulty interpretation was identified as the third leading source; it generated 9% of the total, Kim said.

Other sources of error included:

  • Failing to notice pathology in peripheral or unexpected locations. Tunnel vision and missing pathology in the first and last images in a series (7%)
  • Over-reliance on the prior radiology report (6%)
  • A lack of comparisons with prior studies (5%)

Kim recommended the adoption of mental or written checklists of essential things to cover to reduce under-reading. Individual effort is necessary to minimize satisfaction-of-search problems by staying alert to serendipitous findings. More training and experience is the tonic needed to address faulty interpretations, he said. 

The findings from the two military centers are similar to those of published reports on radiology error rates going back to the 1940s, Kim said in response to a question from Diagnostic Imaging. The main difference is the addition of cross-sectional MR and CT in his study.

Kim believes what an attending radiologist once told him: If you are not missing something, you are not working hard.