The True Efficacy of a Water Quality Intervention: Accounting for Systematic Bias
Joseph N.S. Eisenberg1, Kyle Enger2, Kara Nelson3
1Epidemiology, University of Michigan, Ann Arbor, MI, United States; 2Environmental Sciences, Michigan State, East Lansing, MI, United States; 3Environmental Engineering, University of California, Berkeley, CA, United States

Diarrheal disease is the third leading cause of infectious disease-specific mortality and the sixth leading cause of mortality worldwide. In developing countries, diarrhea causes 2.5 million deaths each year. Recently, in home point-of-use (POU) water treatment devices, designed to improve water quality by removing waterborne infectious agents, have been widely studied as a means to reduce the burden of such infections. The study design most often used to test these devices in the field is the intervention trial. Randomized control intervention trials (RCT) are thought of as the gold standard study design in epidemiology and other medical research; i.e., it is the study design that produces the least amount of systematic bias. Most of the published water and hygiene intervention trials conducted in the developing world, however, did not randomize or blind the participants. In addition to the biases associated with lack of randomization and blinding there are other biases such as recall bias and compliance to the intervention or non-intervention group that affect the internal validity of the estimate derived from the trial. In this talk we present results from an analysis of these biases in intervention trials. Specifically, we present a modeling framework to evaluate the impact of these biases on estimates from the intervention. This modeling framework uses a quantitative microbial risk assessment (QMRA) approach to simulate the intervention. We apply this modeling framework to an RCT conducted in the Democratic Republic of the Congo. The RCT was conducted to evaluate a POU device designed to filter out enteric pathogens from water. The QMRA model provides adjusted estimates for the effects of an intervention when biases are taken into account. In this study we present distributions for the adjusted estimates to reflect the uncertainty and variability of the model.

Keywords: Microbial risk assessment; Water quality; Pathogens; Developing countries

Biography: Dr. Eisenberg is an associate professor of epidemiology in the school of public health at the University of Michigan. He is an expert in water- and vector-borne transmission modeling, infectious disease epidemiology, and microbial risk assessment. His broad research interests integrate theoretical work in developing disease transmission models and empirical work in designing and conducting epidemiology studies. Specifically he has been interested in integrating environmental infection transmission systems into microbial risk assessment. His domestic work focuses on microbial risk assessment, such as examining risks of norovirus in drinking water, developing a dynamic population level microbial risk assessment framework, and examining optimal intervention strategies for influenza.