Functional data analysis to guide a conditional likelihood regression in a case-crossover study investigating whether social characteristics modify the health effects of air pollution
Abstract
In this study we are focused on exploring whether social characteristics modify the relationship between air pollution and hospitalizations due to asthma or chronic pulmonary obstructive desease (COPD) in El Paso, Tx. The case-crossover design with conditional regression analysis was used, here the controls and the case are the same subject at different times and has the advantage of removing confounding by permanently confounding factors. Social charateristics are included in the models as interactions with the pollutants, variables included are age, sex, ethnicity and insurance status as indicator for the socio-economic status. The pollutant's lags were chosen using the historical functional linear model to estimate the association between the response and pollutant at all lags simultaneously. The regression coefficient function was calculated by P-splines with the smoothing parameter chosen with a modified ridge trace method. We included single pollutant analyses for NO 2 and PM2.5 for both asthma and COPD diseases, adjusting for apparent temperature (combination of temperature and dew point) and wind speed. The lags for low and high wind speed were chosen, in the case of asthma, based on previous literature and in the case of COPD based on odds ratios. Subgroup analyses by ethnicity are presented, in order to compare Hispanics and Non-Hispanics without the assumption that the weather variables have the same effect for all subgroups. We found that when PM2.5 is equal to the 98% percentile of the daily values, Hispanics are more likely to be hospitalized due to asthma or COPD than Non-Hispanics, but when NO2 is equal to the 98% percentile for the daily values, contrary to PM2.5 , Non-Hispanics are more likely to be hospitalized due to asthma or COPD than Hispanics. For children with Medicare the probability of being hospitalized for asthma increases significantly when NO2 increases by one interquartile range. This exploratory study is looking for patterns that can be later compared with findings in other cities as part of a comprehensive review.
Subject Area
Applied Mathematics|Public health|Public policy
Recommended Citation
Herrera Hernandez, Juana Maribel, "Functional data analysis to guide a conditional likelihood regression in a case-crossover study investigating whether social characteristics modify the health effects of air pollution" (2013). ETD Collection for University of Texas, El Paso. AAI1545167.
https://scholarworks.utep.edu/dissertations/AAI1545167