[Observer error in interpretation of chest-radiophotographs (author's transl)]

A series of 434 radiophotographs (70 X 70 mm) containing 40 patients with known abnormal symptoms and 15 resp. 21 pictures with suspected abnormalities were interpreted by 85 chest physicians. 84% (average of 77 readers) of the known symptoms were found in the right localisation. The mean value of 8 diagnosticians with only little routine in reading radiophotographs, was 65%. The number of correct findings with respect of the 40 plus 15 resp. 21 abnormalities lies near 80%. Nearly 10% of the true negative cases were described as positive by the diagnosticians. The number of false positives varies considerably (between 0% and 50%) for different interpreters. The number of true positive findings when the films are read by two observers independently increases only when we combine two readers with poor results in finding true positive cases. But dual reading means nearly twice as much false positive findings. The repetition of reading the films by the same interpreter gives nearly the same results as the first reading.