Reliability of grading retinal hemorrhages in abusive head trauma

Abstract

To study the intra- and inter-observer reliability of pediatric ophthalmologists in grading retinal hemorrhages in abusive head trauma. Pediatric ophthalmologists attending the 2009 annual meeting of the American Association for Pediatric Ophthalmology and Strabismus participated in an onsite survey. The participants were required to compare a collection of fundus photographs of patients with abusive head trauma to two standardized photographs, A and B. Participants graded photographs as normal (no retinal hemorrhages), better than A but not normal, same as A, between A and B, same as B, or worse than B. The survey was divided into four parts: part 1 consisted of 20 photographs to be graded; part 2 was a questionnaire about participant's professional experience; part 3 consisted of 20 photographs to be graded (including 15 graded in part 1 but reordered); and part 4 asked about the criteria used to grade the photographs. Of the 95 participants, 92 (97%) were licensed physicians with a mean of 14 years in practice. The intra-rater reliability was found to be high, with overall average of 85% pre- and post-agreement on the common pictures. Kendall's coefficient of concordance was used as the statistical measure of inter-rater reliability and had a high value (0.82). Pediatric ophthalmologists showed high intra- and inter-observer reliability in grading retinal hemorrhages in abusive head trauma.