Our long-term objective is to decrease the variability in evaluations and diagnoses of injured children by child abuse specialists. Variation in the evaluation and diagnosis of childhood injuries may be due to many sources including an incomplete understanding of the biomechanics of injuries and a lack of gold standard diagnostic tests. This project concentrates on whether physician bias is a source of variability in evaluations and diagnoses of childhood injuries for abuse. If the child's social history causes the physician to expect an injury was abusive or not abusive, this "expectation bias" may influence the child's subsequent evaluation or diagnosis.
The specific aims of this study are (1) to determine what components of the child's social history influence the examining physician's perception of risk for abuse;(2) to determine if the child's evaluation or diagnosis is changed by the physician's risk perception;and (3) to create three consensus evaluations to act as comparators to the physician's evaluations. Physicians will both submit and review cases of three injury types (traumatic brain injury, long-bone fracture, and skull fracture) using a web-based template. Physicians will rate their perceptions of social risk for each injury case. The texts of the cases submitted by the examining physician will be analyzed using qualitative analysis techniques to identify a priori defined social risk factors and de novo defined social cues. Social cues are physicians'impressions of the family and the child's social setting included in the text of the evaluation that give a more complete picture of the child's environment. Quantitative analysis, using linear mixed effects models, will be used to examine which elements encoded in the child's case evaluation contribute to the physician's perception of risk for an abusive injury.
For Specific Aim 2, physicians review each others'submitted cases. The first reviewer will read the full written report. The second reviewer will read a report altered by the investigators to exclude positive or negative prejudicial language, social history, and social cues. Evaluations will be scored by whether they met or did not meet the consensus evaluation developed in Specific Aim 3. Evaluations will be compared to determine whether physicians with different amounts of social information about the family evaluate children differently. Generalized estimating equations using the logit-link function will be used. Likewise, comparisons of diagnoses will be made by physician role (examiner versus reviewer) and by risk perception score. Finally, the consensus evaluation for each injury category will be developed by the participating physicians using an internet assisted Delphi process. The consensus evaluation is developed last in order to avoid influencing examiners'evaluations. These analyses are designed to understand what contributes to risk perception and to find if bias plays a role in physician behavior.
Each year 900,000 children are found to have suffered abuse or neglect. Correct diagnosis of childhood injuries as abusive or not abusive is important: incorrect diagnosis could lead to a child being improperly taken from his home or left to suffer further abuse. This project will address whether physician bias influences the evaluation or diagnosis of physical child abuse.