The project aims to develop more flexible and powerful statistical methods for infectious disease epidemiology based on survival analysis and contact intervals. The contact interval from person A to person B is the time between the onset of infectiousness in A and infectious contact from A to B, where infectious contact is defined to be a contact sufficient to infect a susceptible person. It can be right-censored if B is infected by someone other than A or if A recovers from infectiousness prior to infectious contact with B. Because of this censoring, survival analysis is the natural approach to estimating the contact interval distribution. During the K99 phase, this project will develop nonparametric estimates of the contact interval distribution and semiparametric regression models allowing the estimation of log hazard ratios representing the effects of covariates on infectiousness and susceptibility. Since many of these estimates can be represented as sums or averages over possible transmission trees, they provide a natural bridge between traditional epidemic data and phylogenetic data. During the R00 phase, generalizations of the K99 methods will be developed and placed into a Bayesian framework for the analysis of partially-observed epidemics. Survival analysis is one of the pillars of modern biostatistics, so it will be a rich source of novel study designs and analytical methods in infectious disease epidemiology.
The project aims to extend statistical methods from survival analysis, which are used throughout chronic-disease epidemiology, to the analysis of infectious disease data. The resulting methods will be more flexible and powerful than current methods in infectious disease epidemiology, allowing determinants of susceptibility and infectiousness to be identified with greater accuracy.
|Yang, Yang; Halloran, M Elizabeth; Chen, Yanjun et al. (2014) A pathway EM-algorithm for estimating vaccine efficacy with a non-monotone validation set. Biometrics 70:568-78|