A1489
Title: Modeling events with competing causes of known exposure: Risk of concussion due to sports in children
Authors: Florin Vaida - University of California San Diego (United States) [presenting]
Wenjing Meng - University of California San Diego (United States)
Abstract: In the ongoing ABCD study, over 10,000 children at the age of 9-10 are followed up for up to 10 years, with extensive information collected at baseline and yearly. We are concerned with the risk of mild traumatic brain injury (mTBI), which may be caused by playing sports, as reflected in the baseline study visit. About 4\% of children report mTBI at baseline, but the cause of concussion is not reported. Information about sports played in the past, and the exposure, in the form of average time per week, is recorded. We model the risk of mTBI as a function of individual sports played, as well as other demographics and socio-economic characteristics of the children and families, using an extension of Poisson regression, via a mixture of censored Poisson distributions, with censoring at 1 event. The exposure determines the mixing probabilities. The maximum likelihood estimator is computed using the EM algorithm, with standard errors determined via the Louis formula. We show that the hazard of mTBI is more than 10 times when playing a sport such as soccer (football), compared to time spent not playing sports.
