An Expert Judgment Assessment of the Concentration Response Relationship Between PM_2.5 Exposure and Mortality

Walker K.D., Roman HA, Conner L., Hubbell B., Richmond H.,

and

Allocating Efforts in Characterizing a Portfolio of Risks:
A Value of Information Approach

Jeff Keisler, PhD.

Wednesday, February 16, 2005
4:05-4:30 PM Social gathering, light snacks
4:30–6:30 PM Program

Conference Room, CDM
One Cambridge Place, 50 Hampshire Street,
Cambridge, MA

RSVP Required to Korin Scheible at CDM, ScheibleKA@cdm.com by noon the day of the meeting to facilitate security sign in.

An Expert Judgment Assessment of the Concentration Response Relationship Between PM_2.5 Exposure and Mortality

Walker K.D., Roman HA, Conner L., Hubbell B., Richmond H.

Summary: Uncertainty about the true public health benefits of reductions in PM2.5 levels in the U.S. rests, to some degree, on uncertainty in the relationship between PM2.5 exposure and premature mortality. This paper presents the results of a recently completed pilot study to characterize uncertainty in the concentration response (C-R) relationship between ambient PM2.5 concentrations and mortality using probabilistic distributions obtained through formal elicitation of expert judgments. Five experts, chosen through a peer nomination process, were interviewed using a structured elicitation protocol that included both qualitative and quantitative questions about the C-R relationship. The quantitative questions required the experts to estimate: 1) the percent change in annual non-accidental mortality in the adult population associated with a permanent 1 ug/m3 decrease in 24-hour average PM2.5. Each expert provided minimum, maximum and median estimates of the mortality effect, plus 5 th, 25 th, 75 th, and 95h percentile values for the distribution describing his uncertainty about the C-R relationship. The experts’ judgments were presented both individually and as a combined distribution reflecting equal weight given to each expert. The central tendency of the combined expert distribution for the quantitative questions reflects the influence of the major published epidemiological studies on this topic; however, the uncertainty in these distributions is greater than that reported in many of these studies, reflecting the incorporation of additional uncertainty factors into the expert’s subjective probability distributions. These findings suggest that the elicitation of expert judgments may be a useful tool for developing a more complete characterization of uncertainty in PM-related avoided mortality estimates for regulatory benefit analyses.

Biography:

Katherine Walker holds a doctorate in Environmental Health Sciences from Harvard School of Public Health. She has spent 20 years in the application of risk assessment techniques to environmental policy. Her current research interests are focused on the characterization of uncertainty and variability in environmental exposures and risks. She is involved in the development of a full-scale expert judgment elicitation project to characterize uncertainty in the concentration response relationship between PM2.5 and premature mortality and is also involved in the use of PBPK models to characterize inter-individual variability and developmental differences in human doses from exposure to acrylamide

Allocating Efforts in Characterizing a Portfolio of Risks: A Value of Information Approach

Jeff Keisler, PhD.
University of Massachusetts

Summary: When faced with a number of risks and limited resources to reduce risk, it is common to prioritize risks in some way and allocate resources accordingly. With a standard cost-benefit type approach, risk analytic methods may be used to quantify the potential impact of various risks, and a decision maker may then weigh these against the cost of mitigation. This is somewhat related to the project selection problem in Operations Research, and the decision analytic approach to portfolio management.

His recent research has used a value-of-information approach to compare different analytic tactics for this problem, and the results to apply to risk problems as well as follows: The risk portfolio manager may do any of the following before allocating funds to better characterizing different risks:

• Improve estimates of the expected loss due to each risk (and the cost of reducing each risk);
• Improve estimates of tradeoff weights when multiple criteria are involved,
• Evaluate not just the result of a maximum cost and a minimum cost plan for reducing each risk but also evaluate the in some way the impact for a continuous range of expenditures for each risk, and finally,
• Consider potential for certain interactions within certain clusters of risks, e.g., where the requirements for reduce one risk
  would also play a role in reducing a second risk, there are cost-efficiencies that might make it more desirable to allocate funds toward reducing such a pair of risks rather than some comparable risk that has no such potential efficiencies, or alternatively, there might be certain additional problems that are eliminated if two or more individual risks are also eliminated. 

This talk will describe results from simulation studies on these questions and synthesize them within the value-of-information framework, and discuss possible risk analytic applications.

Biography: Jeff Keisler received his MBA from the University of Chicago and PhD in Decision Sciences from Harvard University as a student of Howard Raiffa. He worked for ten years as a decision analyst at General Motors, Argonne National Laboratory, and Strategic Decisions Group, consulting to a number Fortune 100 companies and several government agencies, before joining the faculty of the University of Massachusetts at Boston, where he is currently Assistant Professor of Management Science and Information Systems. He is a member of the INFORMS Decision Analysis Society governing council and twice chaired the Decision Analysis Cluster at the INFORMS national conference, and also serves as secretary of the INFORMS Spreadsheet Research Productivity section. He has written numerous papers including articles for journals such as Risk Analysis, Interfaces and Decision Analysis. His research interests include portfolio resource allocation, value of information, decision process design, decision-analytic risk management, and multi-criteria decision-making .