The Utility of Quantile Regression in Disaster Research

Abstract

Following disasters, population-based screening programs are routinely established to assess psychological consequences of exposure. These data sets are highly skewed as only a small percentage of trauma-exposed individuals develop adverse mental health outcomes. Commonly used statistical methodology in disaster research generally involves population-averaged models, such as linear and logistic regressions. However, these models offer only a partial explanation of the complex relationships between the extent of disaster exposure, individual characteristics and adverse mental health outcomes. The aim of this report is to illustrate the benefits of using quantile regression in disaster research by analyzing the effects of a selected exposure variable, perceived threat to one’s life, and education level on post-traumatic stress symptomatology among n=2960 non-rescue disaster workers exposed to the World Trade Center (WTC) disaster in New York City on 9/11. The findings of the study are in line with previous WTC research that documented the link between perceived danger associated with disaster work, low education level and elevated post-traumatic stress symptomatology. However, the use of quantile regression demonstrates the robust and differential association between these variables throughout the entire distribution of post-traumatic stress symptomatology. Specifically, we show that the effect of high perceived danger and low education level were more strongly associated with post-traumatic stress symptoms in the upper tail of the distribution, after adjusting for covariates. Quantile regression methodology has the potential to enrich disaster research by tackling research questions that were previously unanswered. This method may be particularly useful in analyzing large population-based screening programs.