1

Traci Hale, Battelle Memorial Institute

Committee on Methodological Improvement to the Department of Homeland Security’s 2006 Bioterrorism Risk Assessment

10 February 2007

Traci Hale, Battelle Memorial Institute

Committee on Methodological Improvement to the Department of Homeland Security’s 2006 Bioterrorism Risk Assessment

10 February 2007

2008 DHS Bioterrorism Risk Assessment:

Planned Improvements

2008 DHS Bioterrorism Risk Assessment:

Planned Improvements

2

General overview of plans

The 2008 Risk Assessment will include: An expanded list of agents to be assessed (to include anti-agricultural,

engineered, and emerging agents) An expansion of scenarios for each target-type and associated revisions to

the Event Tree Review and improvement of all consequence models Improved data regarding mitigation strategies, and improved medical

mitigation models Improved calculation engine to decrease run times and simplify

configuration files Implementation of formalized elicitation process to obtain SME judgments

in specific subject areas Expansion of economic modeling to include indirect costs as well as

additional direct costs Expansion of tailored risk assessments and sensitivity studies

3

Agent Production

ScenarioConsequences

Mitigation

Selection Probability

Agent Mass

Bioagent

DisseminationEfficiency

Event Detection

InitiationFrequency

RISK

Selection Probability

Selection Probability

ScenarioProbability

Agent ReleaseModeling

AgentRisk

Ranking

EventTree

Quantification

AGENTRELEASE

TargetThreat Group

MitigationResponse

Dispersion

SUSCEPTIBLEINFECTED

ILL &

INFECTIOUS

RECUPERATING

UNTREATED

prob of infection

avg incubation time

No Treatment

DYING

Become

Infected

initial

infection

Are Dying

Susceptible

Vaccinated

prob contact

is susc

Spontaneous

Recovery

Are Treated

RECOVERED\IMMUNE

SUSC REC'ing

PROPHYL

Become Ill

RECEIV'NG

TREATEMENT Recuperation

Inf Receive

Prophyl

Susc Receive

Prophyl

Infected

Vaccinated

INF REC'ing

PROPHYL

DEAD

Die

End Susc

Prophyl

Ineff Treat

End Inf

Prophyl

Recovercontacts

per day

+contagious

e

K

K

1 2 3

4

5

6

7

8

9

10

10

A

C

C

D

D

E

E

F

G

H

HJ

J

B

*

a

b

c

d

DiseaseSpread

Scenario Analysis and Consequence Modeling

4

Branch Probabilities and Uncertainty Management

5

Consequence Uncertainty

For the 2006 Risk Assessment epistemic uncertainty in the consequence results was not considered Aleatory uncertainty, reflecting variation in results arising from

unknowable details of bioterrorism attack scenarios, was embedded in the consequence models

Aleatory uncertainty is reflected in the scenario specific consequence distributions

Consequence uncertainty was omitted due to the overwhelming processing requirements

6

For the 2008 Risk Assessment epistemic uncertainty in consequence results will be implemented

Improvements in the risk assessment software allowing specification of user defined functions that accept uncertainty parameters for components of the consequence calculation are the required improvement. For example:

Number of illnesses conditional on threat organization, target, surrogate, and mode of dissemination, [RI|MRE,TO,Target,Surrogate,ModeD,<uncertainty parameters>]

Percentage of fatalities mitigated by public health response conditional on number of illnesses, target, and event detection, [MFI|CI, Target, EventDetect, <uncertainty parameters>]

Moving Risk Assessment computing to a Linux cluster platform was required to make the computations required for consequence uncertainty feasible

Consequence Uncertainty

7

Consequence Calculation Equations

AREACFCIECE

RFATTACKSIINFMFICICF

RFATTACKSIINFMEIPSRVRMREATTACKSPSRVRMREIINFCI

MMADPSRQFRQFAMRMRE

PSPSTPSR

VFVFVTVR

QFQFQFQFQFMTMR

,,|

,,|*

,,|,,|,,|

,|6

62

54321

Symbol Description Symbol Description

MT Target mass AREA Area requiring decontamination

QF1-QF6 Factors to explain production/processing/storage/etc. losses ATTACKS Number of simultaneous attacks

VT Target volume RF Percentage of index infected who become untreated fatalities

VF1,VF6 Factors to explain concentration/formulation volume changes MEI Epidemiological spread and prophylaxis factor

PST Target percent solids MFI Public health system mortality prevention efficacy

PS6 Factor to explain formulation percent solids change CI Illness consequences

QFR Respirable fraction CF Fatalitiy consequences

QFA Active fraction E Economic cost

IINF Number of index infected CE Economic consequences

8

Subject Matter Expert and Stakeholder Interactions

9

Subject Matter Experts and Stakeholder Interactions

Subject Matter Expert / Stakeholder interactions are taking several forms: Formal elicitations

START (bioagent selection probabilities) Selected psychologists from the IC (bioagent selection

probabilities) BTISWG (interdiction, frequency of initiation, bioagent selection,

multiple attack probabilities) Informal elicitations

BTISWG panel discussion of target, dissemination, and production probabilities

Stakeholder Working Group meetings IBRAWG (review and vet attack scenarios, production data,

medical mitigation data)

10

Subject Matter Experts and Stakeholder Interactions

IBRAWG Includes CDC, NIH, FDA, USDA, EPA, and the intelligence community, created to “provide interagency input guidance and to the DHS BTRAP. This

Working Group will assist DHS in identifying agents and scenarios for the 2008 Bioterrorism Risk Assessment and will provide technical review of risk assessment input and assumptions, establishing subgroups for this purpose if necessary. The IBRAWG will be a source of technical advice and expertise, and will serve as an interagency forum for sharing, reviewing, and vetting risk assessment data and results as they are generated.”

To date, this group has been responsible for the selection of the 2008 biological agent list, has provided input for attack scenarios, and has provided significant contributions which will play a role in consequence and mitigation modeling.

BTISWG Includes members of the intelligence community created for the express purpose of providing classified intelligence/threat information

and data to the Risk Assessment. This group will be responsible for the assignments of probability concerning terrorist decisions through both formal Subject Matter Expert elicitations as well as informal discussions.

11

Subject Matter Experts and Stakeholder Interactions

These Working Groups provide the stakeholders with the opportunity to review and discuss attack and mitigation input scenarios

and input data reach consensus regarding broad spectrum issues (such

as selection of the 2008 bioagent list) voice any issues or concerns regarding the assessment

while still in progress

12

Indoor Aerosol Dispersion Modeling

13

Indoor Aerosol Dispersion Modeling

In October 2006, the release of a biological agent in a subway system was specifically identified by DHS as a high priority attack scenario to address In the 2006 assessment, the subway attack scenario was

considered, but only as one of several small enclosed building surrogates

For the 2008 assessment, the subway (and other surrogates) will be individually assessable on the Event Tree

The model for the subway attack scenario will include the transfer of biological agent from the point of dissemination to additional downstream subway stations via movement of air in the subway cars and through the tunnels

14

Foodborne and Waterborne Contamination Modeling

15

Foodborne and Waterborne Contamination Modeling

Food Contamination Assessment Joint efforts are being initiated with BTSafety

BT Safety is participating in a collaborative effort with several federal agencies, including FDA, USDA, CDC and DHS NCFPD to develop a simulation model to estimate the impact of food supply contamination

This model is planned to be modified and incorporated into the DHS Risk Assessment calculation engine

16

Foodborne and Waterborne Contamination Modeling

Water Contamination Assessment Joint efforts are being initiated with EPA

To generate a more realistic decay model To calculate a more accurate mixing parameter based on

comparisons of the Risk Assessment mathematical model with EPA-held hydraulic models of real public water systems

Continuing discussions with EPA staff regarding other aspects of the Water Contamination scenarios, including building system contamination and post-attack decontamination

Stakeholder input (FDA, USDA, NCFPD, EPA) is considered critical to a successful assessment

17

Medical Mitigation and Epidemiological Modeling

18

Medical Mitigation and Epidemiological Modeling

Substantial changes are in progress for the modeling of the public health response and spread of contagious diseases following a bioterrorism event As described yesterday, SEIR modeling is being applied

Model and data review In late February, a review of the SEIR model and input data

will be performed by experts identified by DHS. In early March, IBRAWG members will review models and

input data as well Risk perception

The SEIR model incorporates effects of ‘worried well’ Impacts on available supplies time to treatment due to increased distribution times

19

Anti-agricultural Scenarios

20

Anti-agricultural Scenarios

Joint efforts with Lawrence Livermore National Laboratories and Texas A&M LLNL MESA model Texas A&M FAZD model

Incorporates stakeholder input from USDA (APHIS, ARS, CREES) DHS FAZD LLNL BKC

21

Economic Analysis

22

Economic Analysis

Economic consequences are significantly affected by the impact of risk perception on human behavior

CREATE (primarily Adam Rose, Peter Gordon, Jim Moore, and Bumsoo Lee) collaborating on development of I-O models to capture bioterrorism attack economic impacts

CREATE (primarily Adam Rose) developing CGE models for a small set of surrogates to compare with I-O models

23

Economic Analysis

Direct Costs: Economic models have four components:

Human health (Agent specific) Hospitalization, treatments, etc.

Fatalities One year of lost final demand

Decontamination (Agent and surrogate specific) Lost usage of buildings, clean-up, animal disposal

‘Conceptual model’ (Agent and surrogate specific) Risk perception based losses, e.g., reduced air travel

after airport or airplane attacks

24

Economic Analysis

Indirect Costs: Fatality and conceptual model costs implemented in I-O primarily

by reductions in final demand. Some specific surrogates (for example, Mall) transfer

demand from industry impacted (for example, Clothing and Entertainment) to another industry (for example, On-line shopping)

Human health and decontamination costs are assumed to be funded by the government Positive impacts to medical and decontamination industries, Non-defense government spending is reduced (budget cuts)

and household spending is reduced (taxes) to pay for public health response and decontamination.

25

Tailored Risk Assessments and Sensitivity Studies

26

Tailored Risk Assessments and Sensitivity Studies

Increased capacity for tailored assessments and sensitivity studies Faster computing through hardware upgrades and software

improvements

Example tailored assessments requested Use of high Ro agent, similar to measles

Injection of high expertise terrorists

Example sensitivity studies under consideration Impact of additional modeling detail

Water modeling using a hydraulic simulation of a public water system versus the analytical model

27

Calculation Engine Improvements to Enable Quick-turn around Tailored Assessments

Speed Improvements: Movement of code from .Net C# to ANSI standard C/C++

Gets rid of .Net overhead Porting of code from Windows to Linux

Gets rid of Windows overhead Parallelizing code

Employs multiple processors to perform calculations more quickly

Conceptual Improvements: Configuration files redesigned

Better mapping of component consequence distributions to all scenarios to which they apply

Development of user routines More convenient drawing of consequence components from distributions

conditional on other consequence components and uncertainty parameters Compartmentalizing intermediate consequence calculations

Speeds tailored risk assessments and sensitivity studies by allowing computations to start from point at which results change

28

2008 Bioterrorism Risk Assessment Planned Improvements Associated with NAS Issues

*Data updating approach Data replacement and Bayesian updating

Stakeholder/SME Interactions Bioagent selection, scenario definition, model and data review, direct

elicitation input *Standardize lexicon *Non-traditional agents *Data quality matrix Incorporation of risk perception

Indirect costs and worried well Sensitivity Studies

Performed based on specific DHS requests or to further investigate internally-identified areas of interest

* indicates activities which will be discussed in detail during the NAS response discussions