Top Banner
Center for Risk Science and Public Health Risk Assessment with “Actuarial Data” George Gray Center for Risk Science and Public Health Department of Environmental and Occupational Health Milken Institute School of Public Health
25

Risk Assessment with “Actuarial Data”, George Gray

Aug 21, 2014

Download

OECD Governance

Presentation by Prof. George Gray, Director of the Centre for Risk Science and Public Health, George Washington University, at the Workshop on Risk Assessment in Regulatory Policy Analysis (RIA), Session 7, Mexico, 9-11 June 2014. Further information is available at http://www.oecd.org/gov/regulatory-policy/
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Risk Assessment with “Actuarial Data”

George Gray

Center for Risk Science and Public Health Department of Environmental and Occupational Health

Milken Institute School of Public Health

Page 2: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Actuarial vs Modeled Risks

•  Actuarial Risks •  based on previous experience with the same risk •  predictions can be made with a great deal of

precision •  examples include diseases, auto accidents, etc.

•  Modeled Risks •  based on data and theory not direct observation of

the risk •  predictions subject to considerable uncertainty •  examples include cancer risk from chemicals , global

warming, etc.

Page 3: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Actuarial Data

•  Often associated with insurance

•  Estimating likelihood of adverse outcomes •  Timing (life tables) •  Mitigation •  Insurance

•  Predictions based on experience with the specific outcome – usually with a stochastic component

Page 4: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Actuarial Data

•  Key point is that data are measuring the outcome we want to evaluate

•  Often need to model to make predictions of the effect of risk management interventions

•  Main sources of uncertainty are sampling (how well are we counting the outcome) and the role of other variables in influencing the outcome

Page 5: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Example: Auto Fatalities

•  US Fatal Analysis Reporting System •  FARS has gathering data since1975 •  Data on approximately 1 million motor vehicle fatalities •  Over 100 data elements are gathered on the crash •  Data collected on a voluntary basis through cooperative

agreements between NHTSA and each of the 50 States, the District of Columbia, and Puerto Rico

•  To be included in FARS, a crash must involve a motor vehicle traveling on a trafficway customarily open to the public and result in the death of a person (occupant of a vehicle or a non-occupant) within 30 days of the crash

Page 6: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

FARS

Page 7: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Using Actuarial Data

•  Kids Transportation Safety Act of 2007 directed the Secretary of Transportation to initiate rulemaking to expand the required field of view to enable the driver of a motor vehicle to detect areas behind the motor vehicle to reduce death and injury resulting from backing incidents

•  “Such standard may be met by the provision of additional mirrors, sensors, cameras, or other technology to expand the driver's field of view…”

Page 8: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Critical Point to Remember

•  We will look at the analysis of alternative methods to prevent backover crashes in the next session

•  In this case FARS data are incomplete: •  Most accidents do not occur on a public roadway •  May not involve police

Page 9: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Example: Occupational Injuries and Fatalities

•  The U.S. Department of Labor, Bureau of Labor Statistics (BLS) collates data from U.S. industries regarding the annual rate and number of work-related fatal and nonfatal injuries and illnesses

•  Organized by U.S. Census Bureau North American Industrial Classification System (NAICS) code •  Industries are identified by a six-digit code •  e.g., motor vehicle tire manufacturing is classified in the

NAICS code 326199

Page 10: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Estimating Years of Life Lost

•  As part of project to examine occupational injuries and fatalities in Life Cycle Analysis (LCA) we wanted to estimate years of life lost from occupational fatalities

•  Used BLS data (which includes age of fatal accident ) and census data on years of life remaining at a given age

Page 11: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Using BLS Data

Scanlon, K.A., Gray, G.M., Francis, R.A., Lloyd, S.M., and LaPuma, P. (2013) The Work Environment Disability-Adjusted Life Year for Use With Life Cycle Assessments: A Methodological Approach. Environmental Health 6:12-21

Page 12: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Example: Logging

Page 13: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Critical to Keep in Mind

•  How complete is the reporting of occupational injuries and fatalities?

•  Studies by academics, the BLS and others suggest undercounting of injuries and fatalities

•  Estimates of the undercount of nonfatal injuries and illnesses ranges from 0% to 70%*

*Leigh J.P., Marcin, J.P., & Miller, T.R. (2004). An estimate of the U.S. government's undercount of nonfatal occupational injuries. Journal of Occupational & Environmental Medicine, 46, 10-18.

Page 14: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Key Points: Actuarial Data

•  Actuarial data usually focus on the outcome of interest directly – no extrapolation necessary

•  Sources of uncertainty are related to the accuracy of the data and applicability of the data to the specific analysis

•  Uncertainty is usually smaller than that from risks requiring greater extrapolation and generalization (e.g., health risks from chemicals)

Page 15: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Ranking and Scoring Systems

•  Risk analysts sometimes like to use ranking systems when quantitative estimates of risk are difficult

•  Qualitative approach that is intended to give a semi-quantitative feel for the magnitude of risks

•  As analysts you should look very closely – systems have the potential to mislead

Page 16: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Example: USDA’s Proposed Qualitative Approach to PRA

•  Goal of Pest Risk Analysis (PRS) is characterization of pest risk potential for weeds, insects, or diseases that my be introduced

•  Requires •  Identification of pests •  Assessing the consequences of introduction •  Characterizing introduction potential

•  Proposed approach involved assigning scores (1, 2, or 3) depending on attributes of pest in each category - final characterization is sum of scores from each category

Page 17: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Strengths of the Qualitative Approach

•  Explicit identification of information to be considered

•  Well designed flow of information including identification of stopping points

•  Documentation of attributes of pests that contribute to greater or lesser pest potential

Page 18: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Concerns About the Qualitative Approach

•  Scores may not adequately reflect risk

•  Does not convey degree of uncertainty in estimates of pest potential

•  Does not address effectiveness of risk management schemes

Page 19: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Scoring in Qualitative Analysis

Risk Element #4: Economic Impact Introduced pests are capable of causing a variety of direct

and indirect economic impacts. These are divided into three primary categories (other types of impacts may occur):

• Lower yield of the host crop, e.g., by causing plant mortality or acting as a disease vector. • Lower value of the commodity, e.g., by increasing costs of

production, lowering market price, or a combination. • Loss of foreign or domestic markets due to the presence of

a new quarantine pest

Low (1): Pest causes any one or none of the above impacts Medium (2):Pest causes any two of the above impacts High (3): Pest causes all three of the above impacts

Page 20: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Do Scores Reflect the Risk? Hypothetical Example

% Yield Decline % Price Decline Market Loss? Pest 1 2 2 Yes ($1M) Pest 2 0 No Yes ($50M)

Economic Impact Score Pest 1 - High (3) causes all three impacts Pest 2 - Low (1) causes one impact

•  Which is the greater risk?

Page 21: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Scoring Host Range

Risk Element #2: Host Range The risk posed by a plant pest depends on both its ability to

establish a viable reproductive population and its potential for causing plant damage. For arthropods, risk is assumed to be correlated positively with host range. For pathogens, risk is more complex and is assumed to depend on host range, aggressiveness, virulence and pathogenicity; for simplicity, risk is rated as a function of host range.

Low (1): Pest attacks a single species or multiple species within a single genus

Medium (2): Pest attacks multiple species within a single plant family

High (3): Pest attacks multiple species among multiple plant families

Page 22: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

The Importance of Uncertainty: Hypothetical Example

•  For many pests complete data on host range, especially in a new region, is unlikely to be available

•  Assigning scores requires judgment but may be subject to uncertainty

P(low) P(medium) P(high) Pest 1 0.10 0.85 0.05 Pest 2 0.90 0.05 0.05

Host Range Score Pest 1 - Medium (2) Pest 2 - Low (1)

Page 23: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Why Uncertainty Matters

•  Risk managers may care very much about low probability but high consequence events

•  If Pest 2 will cause significant damage if it can attack multiple species among multiple plant families then even a small probability of this event can be significant

•  Characterization of uncertainty is critical to understanding the value of new information

Page 24: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Ranking and Scoring

•  Ranking and scoring systems often used as part of a tiered approach to risk assessment (e.g., Persistent, Bioaccumulative and Toxic (PBT) chemicals)

•  Have utility for priority setting

•  Be very careful about their use as risk metrics in RIA

Page 25: Risk Assessment with “Actuarial Data”, George Gray

Center for Risk Science and Public Health

Look Carefully at the Numbers