ABSTRACT
Human-induced climate change has led to more frequent and severe flooding throughout the globe. We examined the association between flood risk and the prevalence of coronary heart disease, high blood pressure, asthma, and poor mental health in the UnitedStates, while taking into account different levels of social vulnerability. We aggregated flood risk variables from First Street Foundation by census tract and used principal component analysis to derive a set of five interpretable flood risk factors. The dependent variables were census-tract level disease prevalences generated by the Centers for Disease Control and Prevention. Bayesian spatial conditional autoregressive models were fit on this data to quantify the relationship between flood risk and health outcomes under different stratifications of social vulnerability. We showed that three flood risk principal components had small but significant associations with each of the health outcomes, across the different stratifications of social vulnerability. Our analysis gives the first United States-wide estimates of the associated effects of flood risk on specific health outcomes. We also show that social vulnerability is an important moderator of the relationship between flood risk and health outcomes. Our approach can be extended to other ecological studies that examine the health impacts of climate hazards.
ABSTRACT
The Cox proportional hazard (PH) model is widely used to determine the effects of risk factors and treatments (covariates) on survival time of subjects that might be right censored. The selection of covariates depends crucially on the specific form of the conditional hazard model, which is often assumed to be PH, Accelerated Failure time (AFT) or proportional odds (PO). However, we show that none of these semi-parametric models allow for the crossing of the survival functions and hence such strong assumptions may adversely affect the selection of variables. Moreover, the most commonly used PH assumption may also be violated when there is a delayed effect of the risk factors. Taking into account all of these modeling assumptions, this study examines the effect of the PH assumption on covariate selection when the data generating model may have non-PH. In particular, variable selection under two alternative models are explored: (i) the penalized PH model (using the elastic-net penalty) and (ii) the linear spline based hazard regression model. We apply the aforementioned models to the ACTG-175 data set and simulated data sets with survival times generated from the Weibull and log-normal distributions. We also examine the effect on covariate selection of stratifying the analysis on the off-treatment indicator.
