The Contribution of Exposure Measurement Uncertainty on the Total Uncertainty in Epidemiology Study ResultsEPA Grant Number: U916158
Title: The Contribution of Exposure Measurement Uncertainty on the Total Uncertainty in Epidemiology Study Results
Investigators: Jurek, Anne M.
Institution: University of Minnesota - Twin Cities
EPA Project Officer: Hahn, Intaek
Project Period: January 1, 2003 through January 1, 2006
Project Amount: $145,344
RFA: STAR Graduate Fellowships (2003) Recipients Lists
Research Category: Fellowship - Public Health Sciences , Academic Fellowships , Health Effects
The objective of this research project is to apply uncertainty analysis to epidemiologic studies to provide a more honest reporting of the uncertainty in epidemiologic study results. All epidemiologic study results have some amount of error, both random and systematic. One important source of systematic error in study results is faulty measurement of study exposures. Some exposures, such as exposure to environmental tobacco smoke, provide greater challenges in measuring than others such as gender. However, whatever the exposure, the ability or inability to quantify the error can have a large effect on a study's impact on public health decisions.
Routine epidemiologic practice involves formally quantifying only random error, thereby leaving other sources of error, such as exposure-measurement error (EME), to be dealt with informally, typically in the discussion section. Based on a survey I conducted of the epidemiologic literature, EME is sometimes completely ignored when interpreting study results; in other cases, EME is merely qualitatively interpreted. Given the current state of epidemiologic research, when addressing EME, many questions that must be answered to conduct accurate public health research cannot be answered. These questions include:
1. How much EME is present?
2. In what direction is this error in the study results?
3. What is my uncertainty about this error in the study results?
Because of these questions, my research uses uncertainty analysis, a probabilistic approach to sensitivity analysis, to clearly describe the effect of EME on the uncertainty in epidemiologic study results. Numerical simulation methods are used to generate the probability distributions for EME. Uncertainty analysis provides a more accurate representation of the uncertainty in epidemiologic study results, allowing for better-informed public health decisions.