Statistical models of hospital patient fatality rates after accidental falls in children

Abstract

Pediatric traumatic brain injuries represent a public health concern globally. Epidemiological studies are essential to understand their extent, uncover risk factors and outcome predictors, and establish efficient prevention policies. Etiological investigations also benefit from epidemiological data. In young, pre-linguistic children, distinguishing between accidental and non-accidental trauma is often challenging for frontline clinicians. Classification errors can have severe repercussions for children and their families. Past influential epidemiological studies have drawn general conclusions about the etiology of pediatric traumatic brain injuries based on specific statistical information, such as hospital patient fatality rates after reported accidental falls. In this article, we use simple mathematical models to revisit one of these studies and discuss the reliability of its conclusion. We show that apparent paradoxical discrepancies, such as hospital patient fatality rates that are higher after short versus long falls, may have mathematical justification and do not necessarily justify seeking alternative etiological explanations such as non-accidental trauma.