Chest
Volume 68, Issue 3, September 1975, Pages 278-282
Journal home page for Chest

Clinical Investigations
Disagreements in Chest Roentgen Interpretation

https://doi.org/10.1378/chest.68.3.278Get rights and content

Each of 100 chest radiographs selected randomly from a hospital population were read by five experienced radiologists. The films were rich in abnormal findings. Disagreement analysis was performed considering the entire content of the interpretation. The disagreements were graded by a panel as to type (false negative, false positive, and indeterminate) and significance. The overall repeatability of the panel's performance was 75 percent. Forty-one percent of the reports contained potentially significant errors and 56 percent indeterminate disagreements. Among all errors 78 percent were false negatives and 22 percent were false positives. The error rate varied among readers. An average of 25 percent of important findings were omitted by an individual reader. An association between specific abnormalities and the types of disagreement was found.

Section snippets

Case and Reader Selection

The admission chest radiographs of 100 patients, representing every fifth patient examined at the Peter Bent Brigham Hospital as of March 1, 1968, were included.

Five staff radiologists served as readers for the study; they had similar training and experience (a minimum of three years beyond residency). The identity of the individual readers was not known to the panel. Each reader read five sets of 20 films so that at the end of the study the same 100 films had been read by all readers, thus

The Number and Nature of Significant Abnormalities in the One Hundred Study Films

Grade 2 and/or 3 abnormalities were found in 73 films (Table 1). The total number of these abnormalities was 173. The classification and incidence of abnormalities are given in Table 2.

Incidence of Errors Among the Five Readers

There was marked variation in the number of errors among the five readers (Table 3). Considering only more significant groups (grades 2 and 3), the number of false negatives ranged from 27 to 59 per reader. This represents 16 percent to 34 percent of the abnormalities present (Table 2). In grades 2 and 3, 73

Discussion

Because our aim was to consider all statements encountered in the reports it became necessary to develop a method for judging errors in this complex situation. We chose disagreement analysis because this approach eliminated the need for the difficult and many times impossible task of precise pathologic correlation of every radiologic finding. Our intention was to limit the role of the panel as much as possible. For these reasons, we narrowed the task of the panel to evaluation and

ACKNOWLEDGMENT

We would like to express our appreciation to our research assistant, Miss Marguerite Murphy, and to Miss Linda Tuttle for her aid in the preparation of the manuscript.

References (6)

There are more references available in the full text version of this article.

Cited by (0)

This work was supported in part by USPHS grants GM18674, GM01910, and GM14940.

Manuscript received December 8; revision accepted February 17.

View full text