Abusive head trauma is a leading cause of traumatic death and disability during infancy and early childhood. Evidence-based screening tools for abusive head trauma do not exist. Our research objectives were 1) to measure the predictive relationships between abusive head trauma and isolated, discriminating, and reliable clinical variables and 2) to derive a reliable, sensitive, abusive head trauma clinical prediction rule that—if validated—can inform pediatric intensivists’ early decisions to launch (or forego) an evaluation for abuse.Design:
Prospective, multicenter, cross-sectional, observational.Setting:
Acutely head-injured children less than 3 years old admitted for intensive care.Interventions:
None.Measurements and Main Results:
Applying a priori definitional criteria for abusive head trauma, we identified clinical variables that were discriminating and reliable, calculated likelihood ratios and post-test probabilities of abuse, and applied recursive partitioning to derive an abusive head trauma clinical prediction rule with maximum sensitivity—to help rule out abusive head trauma, if negative. Pretest probability (prevalence) of abusive head trauma in our study population was 0.45 (95 of 209). Post-test probabilities of abusive head trauma for isolated, discriminating, and reliable clinical variables ranged from 0.1 to 0.86. Some of these variables, when positive, shifted probability of abuse upward greatly but changed it little when negative. Other variables, when negative, largely excluded abusive head trauma but increased probability of abuse only slightly when positive. Some discriminating variables demonstrated poor inter-rater reliability. A cluster of five discriminating and reliable variables available at or near the time of hospital admission identified 97% of study patients meeting a priori definitional criteria for abusive head trauma. Negative predictive value was 91%.Conclusions:
A more completeunderstanding of the specific predictive qualities of isolated, discriminating, and reliable variables could improve screening accuracy. If validated, a reliable, sensitive, abusive head trauma clinical prediction rule could be used by pediatric intensivists to calculate an evidence-based, patient-specific estimate of abuse probability that can inform—not dictate—their early decisions to launch (or forego) an evaluation for abuse.