All Stanford University & Hospital ID holders are now welcome to visit Lane Library! Learn more
Today's Hours: 8:00am - 10:00pm

Search

Filter Applied Clear All

Did You Mean:

Search Results

  • Book
    Yihan Guan.
    This dissertation develops data-driven methods for healthcare risk modeling to facilitate risk identification and evaluation, focusing on post-marketing drug safety surveillance and breast cancer incidence prediction. For risk identification, we first propose a novel methodology for post-marketing drug safety surveillance: "Follow the money'', which monitors cost in health insurance claims data to detect increased spending related to adverse drug events (ADE). By using real claims data from 2.4 million insured individuals, we demonstrate that this method enables early detection of signals of ADE. Second, we recommend statistical study designs to detect rare and time-dependent ADE resulting from long-term drug use and highlight statistical considerations that drug surveillance designers should focus on when implementing a long-term surveillance. We characterize statistical properties and compare performances of a variety of surveillance designs by applying them to simulated event data as well as real health insurance claims data. For risk evaluation, we develop a stochastic simulation model of the Surveillance, Epidemiology and End Results database to quantify the impact of menopausal hormone therapy (MHT) on U.S. breast cancer incidence. Our model quantifies the promoting effect of MHT on breast tumor growth and the reduced mammographic detectability. By modeling unobservable tumor natural history dynamics and simulating individual life history at a population level, we evaluate the contribution of MHT use cessation to the observed decline in breast cancer incidence between 2002 and 2003. We also predict U.S. breast cancer incidence in the absence of MHT.
    Digital Access 2012
Live Chat