Lessons Learned from the Supply Chain Literature

From our review of the supply chain literature, we synthesized five key lessons learned.

1. Strategies to improve supply chain network designs, including regionalization of some elements of the supply chain, can reduce inventories, improve service, and save money. A number of traditional (manufacturing) supply chains have benefited from network redesign that balance the needs of local customers with the objectives of the overall supply chain (which typically include reducing inventories, maintaining high levels of customer service, and reducing overall costs). Appropriate design of the bioterrorism response supply chain network that facilitate effective and timely response to a bioterrorism event needs to balance considerations of adequate redundancy (to maintain adequate capacity in the event of a large-scale attack) while reducing excessive costs. The key elements of the bioterrorism response supply chain currently receiving considerable attention from policymakers are the distribution networks that dispense the materials from the Strategic National Stockpile.
2. For any given network design, effective inventory management, such as pooling of inventories at regional warehouses, can reduce levels of held stock, improve service, and save money. Regionally held inventories of supplies needed for a bioterrorism response may allow for less inventory to be held overall while still being available to local regions in a timely manner. This is the philosophy behind the Strategic National Stockpile. Because, some bioterrorism response supplies have limited shelf lives, minimizing (and/or rotating) inventory may be important from economic and logistics perspectives. This is a key consideration for locally held inventories.
3. Postponement of product customization and reliance on component modularization can reduce supply chain costs while improving manufacturing and delivery times. Several supply chain case studies demonstrated that redesigning final products to be assembled from a limited number of common component parts, even if those component parts are costly, can result in overall cost savings for the supply chain. For bioterrorism, “components” include antibiotics and medical supplies packaged in local and regional inventories, teams of trained personnel available for deployment to the affected locality, and protocols such as clinical practice guidelines for the diagnosis and management of populations exposed to a biothreat agent.
4. Coordination of activities of supply chain members, along with specific consideration of the incentives of all stakeholders in a supply chain, can improve service, increase efficiency, and reduce costs. The bioterrorism response supply chain is highly complex with numerous stakeholders. Specific delineation and alignment of their (sometimes competing) incentives may benefit programs to coordinate response efforts.
5. Timely and accurate information about customer demands facilitates inventory management and rapid delivery of products. For the bioterrorism response supply chain, information systems that can accurately characterize the available supply of goods and personnel and the ongoing needs of the community affected may benefit the response.

Summary of the Evidence of Existing Infrastructure for Local and Regional Responses to Bioterrorism

Numerous existing systems could play a key role during a bioterrorism response. These systems can be broadly categorized in two ways: by the level at which they primarily operate (e.g., local as opposed to regional) and by the type of response tasks they perform (e.g., systems to protect the public health, to increase hospital capacity, to increase the number of trained and equipped response teams, and to provide inventories of pharmaceutical and medical supplies).

The term “regionalization” has been used describe two phenomena—to describe the process of breaking down a centralized system into component regions and to describe the inverse process of collecting local systems into an overarching region. For the purposes of this Report, we define a local response to bioterrorism as occurring under the jurisdiction of the local public health officer. However, we recognize that many relevant local responders and response organizations (such as clinicians, first responders, hospitals, and emergency management professionals) may not organize themselves in relationship to the local public health jurisdiction. Thus, if an included article defined a local response in another way, we supply the authors' definition in our description of the evidence.

Regional responses fall into three broad categories: sub-state, multi-state, and federal. Most states utilize sub-state regions (often drawn according to county lines) for disaster planning. For example, in California these regions are an integral organizational element of the state's mutual aid agreements. (Mutual aid agreements are described in further detail in the section on “Synthesis of Evidence About Regionalization of Emergency Trauma Care”.) Multi-state regions are typically defined by one or more federal agencies. For example, there are ten Federal Emergency Management Agency (FEMA) regions, each made up of between four and eight states (described in greater detail below). These same regional designations are also used by other federal agencies, such as the Health Resources and Services Administration (HRSA)^* and the Department of Health and Human Services (HHS). Other federal agencies divide multi-state regions differently: the Department of Veterans Affairs (VA) and the Indian Health Service distribute medical care according to unique geographical designations.

Infrastructure of U.S. Public Health System

The mission of the U.S. Public Health system is “to promote physical and mental health and prevent disease, injury, and disability.”¹²⁵ During a bioterrorism event, public health officials have numerous responsibilities, including outbreak investigation; laboratory testing; provision of treatment and infection control guidelines to clinicians; development and implementation of plans for mass prophylaxis, treatment, isolation and quarantine; assessment of hospital/clinical resources; and communication with the public via the media. The organization and legal mandates of public health departments vary significantly across the United States. We briefly review the effects of these differences on chain of command, outbreak investigation, and the institution of epidemiologic control measures such as quarantine during a bioterrorism response. In subsequent sections, we review additional specific CDC programs such as the Laboratory Response Network, Strategic National Stockpile, and Epidemic Intelligence Service. We direct interested readers elsewhere for comprehensive reviews of the public health system.¹⁴²

Regionalization of Public Health Departments. Generally, eastern states such as New York, Connecticut, and Massachusetts with their large urban areas, define the local public health official jurisdiction by municipal boundaries. In contrast, western states typically define the local public health jurisdiction at the county level.¹⁴³ Overall, 60% of the nation's local public health departments are defined by county boundaries,^* 25% are defined by municipal boundaries (e.g., New York City), 7% are defined by overlapping city-county boundaries (e.g., San Francisco city/county), and 8% serve several counties/districts (e.g., Texas).¹⁴³

Local public health departments also differ according to whether public health services are provided within the context of a larger health department that provides other health services (e.g., direct patient services such as operating hospitals) and by the degree of autonomy given to the local departments. Centralized systems, such as those in Arkansas and Florida, retain a large degree of control at the state health department.¹⁰⁹ In contrast, decentralized systems such as those in Arizona and Colorado, direct authority to the city or county health department.¹⁰⁹ Some state health departments offer a “hybrid” approach—centrally coordinating some services while allowing local departments broad latitude in the provision of other services.¹⁰⁹ In several states, local health departments are grouped by the state departments of public health into regions (often with the borders of these regions drawn along county lines) with leadership provided by regional health directors and/or regional headquarters. Texas and Alabama are examples of this kind of regionalization (Figure 2).

Figure

Figure 2. Regionalization of Public Health Departments in Texas and Alabama. These maps are publicly available.

These differences in organizational infrastructure are likely to affect bioterrorism preparedness and response planning in three ways. First, to the extent that neighboring states have public health departments with different organizational structures, and thus different chains of command, identification of key public health decisionmakers with whom to coordinate response planning may be complex. Second, numerous Federal Response planning efforts (such as the Strategic National Stockpile) are managed through state health departments that are given responsibility for planning the distribution and dispensing of the Stockpile (and, often, associated funds and other resources), which they then pass on to local health officials. The local implementation of these federal programs will differ depending on the organizational relationships between state and local health departments. Third, because a bioterrorism response requires cooperation among public health officials and other first responders, coordination of these services may be complex if one group of first responders (e.g., fire/police) is organized at the municipal level whereas public health personnel are organized at the county level.^†

Public Health Legislation Affecting Bioterrorism Planning and Responses. A comprehensive review of the heterogeneous municipal, state, and federal legislation that directly affects the decision making responsibilities of public health officials, clinicians, and emergency management officials during a bioterrorism event is outside the scope of this Report. Moreover, many relevant laws are currently under revision following the recent national attention to the tradeoffs between personal privacy, civil liberties, and national security. For reviews of the salient legal issues, we refer interested readers elsewhere.^{89, 90}

In our brief review of the laws affecting key public health and clinical decisionmakers, we found two issues that may affect regionalization during a bioterrorism response. First, for some highly controversial policies such as quarantine, the legal authority to implement the policy may be inadequate for its practical implementation and enforcement.^{* 91, 92} For example, in many states the legal authority to institute a quarantine lies with the local health officer (and if local and/or state laws are inadequate for implementation of necessary quarantine or isolation, local public health officials may apply relevant federal laws).^{93, 110} In the event that local control is deemed inadequate, CDC's domestic quarantine regulations provide that CDC may institute appropriate quarantine or isolation.¹¹⁰ However, implementation of such authority during a bioterrorism event requires enforcement cooperation from police, elected officials, and other responders. In some jurisdictions, the relevant statutes do not obligate these responders to enforce quarantine decisions. Thus, the implementation of public health laws requires coordination with the laws governing related response organizations and personnel.

The second legal issue is that clinicians who are employed by a hospital are required to report to that hospital during a disaster or bioterrorism event. (However, even if clinicians do not report for work—as some have refused to do during responses to naturally occurring infectious disease outbreaks—it is not clear what repercussions they should expect.) For clinicians employed by more than one hospital, or who are members of more than one response unit (e.g., clinicians who are members of both their local Disaster Medical Assistance Team and the Army National Guard ), it may be unclear where their primary obligation for service resides. This raises the related conflict of double counting—when, for purposes of preparedness planning, each of these organizations counts the same individual as part of its team during a bioterrorism response.

The National Disaster Medical System

Background. The National Disaster Medical System is a federally coordinated program whose primary purpose is to provide medical and mental health assistance in the event of mass casualties, when local capabilities are overwhelmed.^{94, 95, 128} The National Disaster Medical System program falls under the auspices of the Department of Homeland Security and is co-managed by four federal agencies: HHS, VA, DoD, and FEMA. The three primary roles of the National Disaster Medical System are: to provide a rapid medical response through Disaster Management Assistance Teams, to evacuate patients through the military Aeromedical Evacuation System, and to provide hospital care to evacuated patients.^{95, 96} We describe Disaster Medical Assistance Teams in more detail in the next section. The Aeromedical Evacuation System, operated by the DoD, has a “specialized airlift mission supporting patient movement on any mobility airlift platform” to safely transport patients to a hospital for definitive care.¹¹¹ Hospital participation in the National Disaster Medical System is voluntary; however, hospitals that participate during a mass casualty event are reimbursed by the federal government. The resources available to the National Disaster Medical System are coordinated by 64 centers around the country.⁹⁴ The National Disaster Medical System program also supports Disaster Mortuary Teams to assist with mortuary needs, Veterinary Medical Assistance Teams to administer care to animal victims of disasters, and National Medical Response Teams who receive special training and equipment to respond to weapons of mass destruction terrorist attacks—although this training has traditionally emphasized hazardous materials/chemical weapons preparedness.^*

Regionalization of the National Disaster Medical System. The National Disaster Medical System is managed federally with coordination of available resources provided by 64 centers around the country. The ability of the National Disaster Medical System to respond to local events is based in large part on two factors: a robust communication system, and voluntary participation of hospitals and individual members of the various response teams. The National Disaster Medical System relies on a communication system that tracks available resources, facilitates the timely deployment of National Disaster Medical System teams, and coordinates ongoing resource demands and availability. The National Disaster Medical System has typically responded to isolated mass casualty incidents (although in 1999 it did respond to four different disasters concurrently).⁹⁴ Thus, the ability of the National Disaster Medical System to coordinate a response to a bioterrorism event that spans several geographic regions remains largely untested.

Disaster Medical Assistance Teams

Background. Disaster Medical Assistance Teams are voluntary specialty teams designed to provide medical care to victims during a mass casualty incident when local medical resources have been overwhelmed.^* Teams are organized and sponsored locally, and team members operate on a volunteer basis during training exercises and planning meetings. Local sponsors may provide considerable financial support to their Disaster Medical Assistance Teams, since the federal stipend for team organization is minimal.⁹⁹ Other team members provide their own funds for training and equipment (often these teams are incorporated as non-profit organizations to aid in fundraising).⁹⁹ Teams typically have about 35 members and include physicians, nurses, emergency medical personnel, and logisticians, among others.¹⁰⁰ Disaster Medical Assistance Teams are designated Level I, II or III depending on their capabilities and experience. To receive a Level I designation, a team must be self-sufficient for up to three days (provide food, water, energy, and other basic services to team members and patients in the event that local services are temporarily disabled by the disaster) and have between three and five people that can serve each Disaster Medical Assistance Team role (e.g., surgeon, logistician, etc.). Level II teams are not required to be self-sufficient, and are required to have two people for each team role. Level III teams are still in the developmental phase and may be deployed to assist a Level I or II team.⁹⁹ Teams may also have a specialty designation: currently, the program supports trauma, mental health, burn, pediatric, and weapons of mass destruction specialty teams. Disaster Medical Assistance Teams can be mobilized to provide support anywhere in the United States. When teams are mobilized, their members are paid as Federal employees, and their licenses and certifications are recognized by all the states during the activation period.

Regionalization of Disaster Medical Assistance Teams. Disaster Medical Assistance Teams are staffed, organized, and managed at a local level and deployed regionally. Disaster Medical Assistance Teams differ widely in terms of number of volunteers, types of expertise of the volunteers, level of experience of the volunteers under emergency circumstances, and equipment and supplies. Because members are volunteers, they can require as little as 24 to 72 hours to be ready for deployment, although times were much longer for some responses when the program was new.⁹⁵ This lag in time to deployment could significantly affect the utility of a Disaster Medical Assistance Teams during a bioterrorism response. (In some cases, Disaster Medical Assistance Teams have arrived at the scene of a disaster as many as five days after the disaster struck and for this reason have not been considered to be a completely reliable resource by some disaster planners.)¹⁰⁴

The Metropolitan Medical Response System

Background. In the event of a large-scale bioterrorism event, it may take considerable time for regional responders to arrive. During this initial response period, local responders will be primarily responsible for the management of the response. Recognizing the need to enhance the capacity of local responders, the Metropolitan Medical Response System^* was designed in 1996 as part of a coordinated national response to bioterrorism, providing preparedness plans and funds to purchase the equipment and supplies specifically required for bioterrorism preparedness and response.¹³³ Public Law 104–201 provided funding through HHS for 125 local emergency medical systems of the Metropolitan Medical Response System in selected cities in the United States (Figure 3).⁹⁶ The Metropolitan Medical Response System, now under the auspices of the Department of Homeland Security, provides funding to U.S. cities to create and enhance local capacity to respond to a covert release of chemical, biological, radiological, nuclear, and explosive weapons. In exchange for initial funds^† to purchase medical equipment, pharmaceuticals, and supplies, the Metropolitan Medical Response System program requires detailed plans from participating cities regarding preparedness for chemical and biological attacks.^{107, 108, 112} For a more detailed review of the Metropolitan Medical Response System, which is presently under expansion, and for an evaluation of the system, we direct interested readers elsewhere.^{9, 133}

Figure

Figure 3. The Metropolitan Medical Response System. This map is publicly available.

Regionalization of the Metropolitan Medical Response System. The Metropolitan Medical Response System is a local program designed to enhance the preparedness of municipalities; however, a large-scale bioterrorism event may necessitate regional coordination of efforts among municipalities. A lack of regional planning by Metropolitan Medical Response System members may complicate a bioterrorism response. For example, individual cities each write their own preparedness plans and there is no formal mechanism for exchange of information between Metropolitan Medical Response System cities during the preparation phase or during a response. Thus, cities may not have the capacity to readily determine the needs or the capacity of their neighbors. Also, most Metropolitan Medical Response System plans involve participation from numerous organizations, such as local public health departments, hospitals, and emergency personnel, which tend to have different organizational infrastructures across cities. A coordinated response among cities may be complicated by these organizational differences. Cities responding to bioterrorism also must work with the National Disaster Medical System to organize the evacuation of patients and the deployment of Disaster Medical Assistance Teams (and other response teams). Thus, an efficient bioterrorism response will require coordination of services between affected Metropolitan Medical Response System cities and non-Metropolitan Medical Response System cities, other federal programs (such as National Disaster Medical System, Medical Reserve Corps, Public Health Service's Commissioned Corps Readiness Force, the National Guard's weapons of mass destruction Civil Support Teams), and other regional response providers.^*

The Department of Homeland Security

Background. The Department of Homeland Security was established by Congress under the Department of Homeland Security Act of 2002,¹³⁴ to integrate and coordinate the leadership, assets, personnel, and functions of the numerous federal agencies involved in all aspects of homeland security, including bioterrorism preparedness and response.^113–116 The establishment of Department of Homeland Security required an enormous reorganization of the federal government, including the transfer of several programs in their entirety (e.g., U.S. Coast Guard, FEMA, U.S. Secret Service, and Metropolitan Medical Response System) and the selective transfer of key defense programs from other departments (e.g., from HHS the Office of Emergency Preparedness and the Strategic National Stockpile).¹³⁴ Ongoing efforts by these agencies (such as the use of surveillance technologies to monitor cargoes coming into the United States for weapons of mass destruction) represent promising initiatives in preventions efforts against bioterrorism. The Department of Homeland Security is organized into four main directorates: Border and Transportation Security, Emergency Preparedness and Response, Science and Technology, and Information Analysis and Infrastructure Protection. The role of the Emergency Preparedness and Response directorate is to “oversee domestic disaster preparedness training and coordinate government disaster response.”¹³⁵ This directorate coordinates efforts of FEMA, the Strategic National Stockpile, National Disaster Medical System, the Nuclear Incident Response Team (DOE), and the Domestic Emergency Support Teams (DOJ).

FEMA, the agency responsible for coordinating federal responses to disasters, created the Federal Response Plan,^* which identifies the essential emergency response functions of the 27 participating federal departments and agencies, and the American Red Cross.¹¹⁷ The 12 essential emergency response functions include communications, medical services, and search and rescue.¹¹⁷ FEMA does not have its own response materials or equipment but does have emergency response professionals who advise and coordinate the requesters and providers of disaster relief. FEMA has coordinated successful regional relief efforts to hurricanes, earthquakes, tornadoes and volcanic eruptions; specific responses are described in the section on “Synthesis of Evidence About Regionalization of Emergency Trauma Care”. FEMA has four main responsibilities for disaster preparedness: preparedness, response, recovery, and mitigation.¹¹⁷ Since the September 11^th attack, FEMA has been designated as the lead agency for incident management.¹¹⁸

Regionalization of the Department of Homeland Security. At the time of writing this Report, the organization of Department of Homeland Security is underway. Once this reorganization is completed, it is expected that Department of Homeland Security will have a significant role to play in a regionalized response to bioterrorism. Currently, Department of Homeland Security has adopted a regional structure previously used by FEMA, HHS, and HRSA (Figure 4 depicts the ten organizational regions). For bioterrorism preparedness, Department of Homeland Security responsibilities include working with local preparedness planners to develop strategies to receive and distribute the Strategic National Stockpile (this effort is primarily directed through state coordinators rather than the ten regions). Because the reorganization of programs into Department of Homeland Security is ongoing, there is no published evidence regarding the effects of the developing regional infrastructure on a regional bioterrorism response.

Figure

Figure 4. Regions as Definined by FEMA, HHS, HRSA, and Others. This map is publicly available.

Department of Veterans Affairs

Background. The Department of Veterans Affairs (VA) has more than 1100 health care facilities across the country and maintains four mobile clinics that have been deployed in natural disasters. Under the Federal Response Plan, the VA is held responsible for “backup medical support for military personnel in wartime and the general public during natural, manmade, or technological emergencies.”¹¹⁷ In addition, the VA Emergency Management Strategic Healthcare Group plans and coordinates five other emergency management functions: 1) development of VA hospital emergency plans for disasters and coordinated mutual aid agreements for patient transfers, 2) backup of the military health care systems, principally through the 69 VA medical centers designated as primary receiving centers for DoD casualties, 3) joint administration duties of the National Disaster Medical System, 4) participation in the Federal Radiological Emergency Response Plan to respond to accidents at nuclear power plants and acts of terrorism, including training and support of the Medical Emergency Radiological Response Team (which consists of physicians and nuclear physicists), and 5) activities to support the continuity of federal government functions during a national emergency.⁹⁶ As Kenneth Kizer described in his report on the VA's role in federal emergency management, “A number of VA facilities are well equipped to handle decontamination and patient staging, and these capabilities could be expanded to additional VA medical centers. Enhancing and maintaining this capability at VA hospitals may well be easier, in many cases, than it would be in the private sector where managed care, market forces, and the effects of the Balanced Budget Agreement may make it impossible to sustain support for disaster contingency efforts over time.”⁹⁶

Regionalization of the VA Healthcare system. The VA is divided into 23 Veterans Integrated Service Networks across the county (Figure 5). Each Veterans Integrated Service Network includes certain states in its network and provides care to veterans in the regions it covers. Since each Network covers distinct geographic areas, the relationships of Veterans Integrated Service Networks with the local and state health departments in their areas may differ. The effectiveness of the VA in providing medical support during a bioterrorism event will likely depend on these types of relationships.

Figure

Figure 5. The 23 VA Veterans Integrated Service Networks. This map is publicly available.

Bioterrorism Preparedness Programs Sponsored by the Department of Defense

The DoD has sponsored several programs that translate its military expertise in preparedness and response to bioterrorism for civilian use. We direct interested readers to a recent evaluation by the U.S. General Office of Accounting of federally sponsored information technologies that could be used during a bioterrorism response (including several sponsored at least in part by the DoD).¹²⁰ This section briefly describes two military programs that would likely play a key role during a regionalized civilian response.

National Guard. During a bioterrorism event or other disaster, governors can mobilize the Army and Air National Guards of their state. Often these guardsmen and women provide security, medical treatment, debris clearing, transportation, shelter construction, and other services.^146–148 Under mutual aid agreements (described in greater detail in the “Synthesis of Evidence About Regionalization of Emergency Trauma Care” section), governors can also direct the use of their state National Guard units in a neighboring state. If the magnitude of the disaster has exceeded the capacity of local responders and state's National Guard, the governor can request federal assistance from the President of the United States.¹⁴⁶ Depending on the type of support required, FEMA, as coordinator of all federal assistance through the Federal Response Plan, may send active U.S. military forces to provide disaster relief.

Emergency response decisionmakers (such as the head of each state's National Guard and governor) differ with respect to the role that they envision for the National Guard during a bioterrorism response. For example, the DoD has developed National Guard Weapons of Mass Destruction Civil Support Teams^* to “assist local and state authorities in assessing the situation surrounding a weapons of mass destruction emergency; advise these authorities regarding appropriate actions; and facilitate requests for assistance to expedite the arrival of additional state and federal military assets.”¹²¹ The 1999 GAO Report on the use of National Guard response teams noted that whereas some states have integrated Weapons of Mass Destruction Civil Support Teams into their statewide preparedness plans, “the Federal Bureau of Investigation and the Federal Emergency Management Agency do not see a role for the [Weapons of Mass Destruction Civil Support Teams] in the federal response.”¹²¹ There have also been concerns that reductions in the size of the U.S. military and engagement of U.S. forces in operations abroad have restricted the availability of U.S. military personnel to respond to a domestic bioterrorism event or other disasters.¹⁴⁶

Biological Weapons Improved Response Program. The Biological Weapons Improved Response Program was developed by the DoD to assist with mass casualty care at the local level.^{122, 123} A key component of this program is the development of the Modular Emergency Medical System, which is intended to expand local hospital capacity, and integrate with other disaster medical resources.¹²⁴ The Modular Emergency Medical System has two components: Acute Care Centers that serve as mass inpatient care facilities for incident victims, and Neighborhood Emergency Help Centers that are designed for high-volume triage and dispensing of prophylaxis and information.¹²⁴ The Modular Emergency Medical System is designed to use the Incident Management System and communication links between local hospitals, Acute Care Centers, and Neighborhood Emergency Help Centers to coordinate patient care. For example, as an incident unfolds, the Modular Emergency Medical System might be used as follows: the initial victims are triaged in the usual manner to available hospitals until local hospitals have no additional capacity, at which point the city could mobilize the Modular Emergency Medical System and establish Neighborhood Emergency Help Centers and Acute Care Centers (at predetermined locations such as schools for the Neighborhood Emergency Help Centers and outpatient clinics for the Acute Care Centers), each of which is linked to a local hospital. The Acute Care Centers would accept non-critical patients from the local hospital. The Modular Emergency Medical System has been tested in full-scale operational evaluations but has not been deployed for an actual disaster response.¹²⁴

Lessons Learned from the Literature on the Existing Infrastructure for Local and Regional Responses to Bioterrorism

We derived three key lessons from our synthesis of the literature on the existing infrastructure for local and regional responses to bioterrorism.

1. Numerous response agencies with regional organizations could contribute to a bioterrorism response. Many of the agencies described in this section have successfully contributed to regionalized responses to bioterrorism and bioterrorism-related emergencies (e.g., infectious disease outbreaks and disasters). The literature suggests that numerous response agencies with regional organizations could contribute to a bioterrorism response including public health departments at local, state, national, and international levels; the National Disaster Medical System (a federally coordinated program that provides medical and mental health assistance including the evaluation of patients and provision of hospital care when local capabilities are overwhelmed); Disaster Medical Assistance Teams (voluntary specialty medical teams that can be deployed to provide a wide range of disaster medical services and resources); the Metropolitan Medical Response System (which expands municipal bioterrorism preparedness through grants that provide pharmaceuticals and other supplies and requires detailed preparedness planning by recipient cities); and the Department of Homeland Security (which has oversight responsibilities for many of the key bioterrorism response agencies and programs such as the Federal Emergency Management Agency and the Strategic National Stockpile).
2. Most of these agencies were designed independently. Efforts to coordinate them are ongoing or under development. Most of the agencies described in this section were developed to provide specific resources (e.g., trained personnel) or facilitate particular tasks during a response to different situations. The Department of Homeland Security has ongoing efforts to coordinate many of these response agencies.
3. Given the demonstrated ability of personnel to participate in more than one of the key response agencies, decisionmakers must be aware of the potential for double-counting. Double-counting can become an issue for responders who are members of more than one response organization (e.g., one person who is simultaneously a member of the Army National Guard, their local Disaster Medical Assistance Team, and a key part of the local response ). A single database or coordinated information system to record all response personnel and resources could address this issue.

Summary of the Evidence of Regionalization of the Response to the 2001 Anthrax Bioterrorism

Our search identified 373 citations regarding the 2001 anthrax attack and subsequent evaluation and response. Of these, we included 60 citations that described regionalized aspects to the response to bioterrorism: 30 articles, 14 government reports and 16 Web sites.^† None of the included articles provided a comprehensive evaluation of the response to the 2001 anthrax bioterrorism. Rather, most focused on a particular aspect of the response, such as the laboratory response.

Application of Evaluation Criteria to the 2001 Anthrax Bioterrorism Literature

In this section we discuss our application of the four evaluation criteria relevant to the 2001 anthrax bioterrorism literature: network design, inventory management, management of information, and incident command. We found no articles that specifically described efforts to employ postponement or modularization, coordinate the bioterrorism response supply chain, or manage the incentives of the numerous relevant responding agencies. Nor did we find articles analyzing volume-outcomes associations.

Network design. A loosely connected network of organizations, each member of which is responsible for a portion of a bioterrorism response in the United States includes national, state and local agencies. Regional coordination and mutual aid agreements between these agencies are designed to facilitate the rapid and efficient deployment of equipment and expert personnel to a local area that may lack resources or be overwhelmed by surge-capacity requirements.¹⁸³

National Response Organizations. Although the GAO has identified over 20 federal departments and agencies that have responsibilities during a bioterrorism event,¹⁶⁸ only four agencies, the FBI, CDC, HHS' Office of Emergency Preparedness, and FEMA, have primary responsibility to provide regionalized assistance to state and local agencies during a bioterrorism event.¹⁶⁹ The following paragraphs summarize the evidence about the role of three of these key federal agencies in a regionalized bioterrorist response and review their actions during the 2001 anthrax attack.

Centers for Disease Control. CDC has the responsibility “to provide national leadership in the public health and medical communities in a concerted effort to detect, diagnose, respond to, and prevent illnesses, including those that could occur as a result of bioterrorism.”¹⁷⁰ Prior to the 2001 attacks, CDC defined several regional priorities in preparing for response to a bioterrorism event, including disease surveillance, public health networks, medical consequence management, and the Strategic National Stockpile (then the National Pharmaceutical Stockpile).¹⁷¹ Thus, CDC developed the Laboratory Response Network, awarded grants to all 50 states and four major metropolitan health departments, and funded several communication programs such as Health Alert Network and the Epidemic Information Exchange System.¹⁷²

The Laboratory Response Network was developed in 1998 as a tiered system of laboratories across the United States with a hierarchy of capacity to handle specimens during a bioterrorism response.¹⁷² The Laboratory Response Network includes CDC's Rapid Response and Advanced Technology Laboratories, charged by Congress as having responsibility for rapidly identifying an infectious agent during a bioterrorism event. During the 2001 outbreak, the Laboratory Response Network's laboratories processed more than 121,000 samples for B. anthracis providing surge capacity for local laboratories.^{156, 173}

Communication was a key role of CDC during the anthrax attack.¹⁷⁷ Unfortunately, no single integrated information system connects all health departments with one another, nor health departments with local hospitals and first responder communities.¹⁷⁴ However, the included articles described two systems that facilitated some information sharing among members of the public health community during the 2001 anthrax bioterrorism: the Health Alert Network and the Epidemic Information Exchange System. The Health Alert Network is an electronic communications system that enables sharing of information by public health professionals nationwide during an infectious disease outbreak.¹⁷² During the 2001 bioterrorism events, the Health Alert Network allowed local public health agencies to request information from CDC, and provided general information such as how to handle suspicious letters and packages.¹⁷⁵ Because the Health Alert Network is limited to public health personnel, hospitals, infection control professionals, and first responders, physicians had no access to these essential reports.^{162, 165 *} The Epidemic Information Exchange System, is a Web-based communications network that provides a format for state, local and national health officials to share preliminary information regarding potential disease outbreaks.¹⁷² During the 2001 attack, the Epidemic Information Exchange System provided 90 reports to public health decisionmakers including state epidemiologists, local CDC investigative teams, and other public health officials.¹⁷⁶ However, one survey of the health response to the 2001 attack noted that no public health official cited the Epidemic Information Exchange System as a source of information.¹⁶² Thus, the likely utility of these systems to clinicians, public health officials, and the public during a large-scale bioterrorism response remains uncertain.

Office of Emergency Preparedness. At the time of the 2001 Anthrax attacks, the National Disaster Medical System was organized under the auspices of the Department of Health and Human Services' Office of Emergency Preparedness into 44 Disaster Medical Assistance Teams and four National Medical Response Teams. These included 7,000 volunteer health and support personnel to assist local response organizations.¹⁷² On October 21, 2001, units from the U.S. Public Health Service, including a Disaster Medical Assistance Team, established a clinic in the Washington, DC area to educate and provide prophylactic antibiotics to persons exposed at the Brentwood U.S. Postal Facility.¹⁸¹ Over a 14-day period, this clinic operated 14 hours per day, seven days per week with a total staff of 136 persons, and dispensed medications to over 18,000 persons. The productivity of this clinic (in terms of numbers of staff relative to numbers of patients served) has now become a reference standard for planning efforts for dispensing medications and providing immunizations during a future bioterrorism event.

Federal Emergency Management Agency. FEMA is the lead agency for consequence management following a bioterrorism event.¹⁸⁶ Due to the September 11, 2001 terrorist attacks, FEMA had fully activated the Federal Response Plan and all ten of FEMA's regions prior to the anthrax incidents.¹⁸⁷ Our review found no assessment of FEMA's actions during the 2001 anthrax attack. This is most likely because FEMA had a limited role in responding to the anthrax attack (they were involved in the World Trade Center disaster management at the time) and because no Emergency or Major Disaster Declarations were made by FEMA for the sites involved in the anthrax attacks.¹⁸⁸ Since the 2001 attacks, FEMA has been instrumental in providing assessments of state and local preparedness for future bioterrorism events.¹⁶⁷

State and Local Response Organizations. In the years prior to the 2001 anthrax attacks, several initiatives appropriated resources for state and local preparedness programs. The National Defense Authorization Act of 1997 appropriated $97 million for “domestic emergency assistance programs, including the implementation of programs providing advice, training, and the loan of equipment to state and local emergency response agencies and assistance to major cities in establishing medical strike teams.”¹⁶⁹ In 1999, CDC received Congressional funding “to enter into multi-year cooperative agreements aimed at upgrading state and local health department preparedness and response capabilities relative to bioterrorism.”¹⁹⁰ However, despite this funding, at the time of the 2001 anthrax attack, many of the state and local programs were operated independently, without significant regional coordination.^{166, 192}

The United States has more than 2900 local public health agencies.^{* 143} Prior to the 2001 anthrax attack, there were few efforts to regionalize local public health agencies through cooperative agreements within and among states.^{161, 192} Since the 2001 anthrax incident, some regionalized systems have evolved. For example, in New Jersey, a cooperative agreement to share resources during a bioterrorism event has been organized into three regions, each representing approximately ten counties, linked together via the Local Information Network and Communication System.¹⁹³ However, to date there have been no evaluations to assess the readiness of such regionalized plans and no evidence to suggest that particular methods of regionalization benefit the ability to respond to naturally occurring or bioterrorism-related illness.¹⁷⁴

Coordination Among National, State and Local Response Organizations. During the initial stage of the 2001 anthrax attack, local agencies worked closely with CDC and other federal agencies to rapidly confirm the diagnosis of anthrax and to initiate an epidemiological investigation.¹⁶² As the scope of the attack increased, so did the challenges to the available systems for coordinating local, regional, and federal response efforts.^{162, 165, 194, 195} For example, during the 2001 anthrax attack, the three separate health departments in the Washington, DC region (Maryland, Virginia, and the District of Columbia) had no established mechanism for coordination of information regarding epidemiologic assessments or medical recommendations.¹⁹⁴ Since the 2001 attack, these three local health departments have agreed to “coordinate disease surveillance, alerts, evacuation, and other emergency preparedness efforts” for future bioterrorist incidents.¹⁶²

Inventory Management. During the 1990s, Congress authorized establishment of the National Pharmaceutical Stockpile, which was renamed the Strategic National Stockpile after it came under the broader umbrella of the Department of Homeland Security. The Strategic National Stockpile provides a national repository of medical resources, including antibiotics that can be deployed at the local level during a bioterrorism event.¹⁹⁶ During the initial stages of an attack, 12-hour Push Packs from the Stockpile could be deployed. These consist of antibiotics, antidotes, and other medical supplies necessary to treat a wide range of possible biological or chemical agents over a short period of time.¹⁹⁷ If the attack requires “a larger or multi-phased response,” the second component, the Vendor Managed Inventory, could be shipped within 36 hours.¹⁹⁸ The Vendor Managed Inventory is a “tailored” supply of pharmaceuticals and medical supplies intended to treat a specific biological agent in a larger population over a prolonged period of time. After the 2001 World Trade Center attack, CDC was able to deliver a Push Pack to New York City within seven hours after it was requested.¹⁷⁸ During the 2001 anthrax attacks, the pharmaceutical stockpile team made 143 sorties to nine states and delivered 3.75 million antibiotic tablets between October 8, 2001 and January 11, 2002.¹⁵⁶

Since the anthrax attacks, some state and local agencies have obtained their own inventories of medical supplies. In 2002, the U.S. Medicine Institute for Health Studies forum on surge capacity during a bioterrorism event recommended that hospitals have 48 hours worth of antibiotics to treat or prophylax staff and other first responders, but did not recommend that individual communities stockpile pharmaceuticals.¹⁹⁹ However, the 2000 weapons of mass destruction tabletop exercise in Spokane, Washington suggested that local communities need to be self sufficient for at least 24 hours and, “if financially feasible, [to have] some local stockpiling of certain antidotes, like antimicrobials.”²⁰⁰ Currently, most local hospital and pharmacy inventories of antimicrobials are inadequate for a large-scale bioterrorism response. A survey in New Mexico found that, on average, local communities had four doses of ciprofloxacin on hand per 1000 persons.²⁰¹ A study in New Jersey noted that hospitals had an average of 289 doses of ciprofloxacin, which roughly equaled three doses per 1000 persons.¹³ No published evidence specifically evaluates what should be included in a local inventory or the costs and benefits of procuring and maintaining such inventories.

Information Management. Regional management of information needs during the anthrax attacks of 2001 fell broadly into three categories: management of information relating to epidemiological investigations, communication among responders, and communication with the public.²⁰⁸ Detection of the index case of inhalational anthrax occurred at the local level in a timely manner.¹⁵⁶ Six days later, when a coworker of the initial case became ill and was ultimately diagnosed with inhalation anthrax, the investigation then focused on the work environment of these two patients and administration of prophylaxis to potentially exposed coworkers. Subsequent to the Florida case, response times were markedly shorter: On October 15, 2001 a letter containing a powdery substance was opened in Senator Daschle's office with preliminary tests suggesting anthrax within 15 minutes.²⁰⁹ Approximately 45 minutes after the letter was opened, the ventilation system was shut down, potentially exposed persons had nasal swabs performed, and an initial 3-day supply of antimicrobial prophylaxis was distributed.

The included articles did not describe information systems that facilitated communication among responders or with the public.^{162, 210} Several untested but promising information systems have recently been deployed to facilitate local responder communication (e.g., San Diego County's Emergency Medical Alert Network, the Rochester Area Community Healthcare Information System, and the New Jersey Local Information Network and Communication System).²¹¹

Incident Command. Effective command and control is a required component of a successful response to a bioterrorism event. The included articles report that during the 2001 anthrax attack, incident command was complicated by three factors: 1) no single organization led the response across the nation, 2) no hierarchy of authority became clearly established or accepted among the numerous response organizations, and 3) there was no single system for effective information processing to facilitate incident command.^{165, 177}

Lessons Learned from the Regional Response to the 2001 Anthrax Cases

Five key lessons regarding a regionalized bioterrorism response system can be learned from the evidence about the 2001 anthrax attacks.

1. Cooperative agreements and regionalized response plans are needed. Pre-event regionalized planning and asset sharing agreements among local public health agencies and hospitals may facilitate enhanced surge-capacity and coordinate responses during a bioterrorism event.
2. Incident command must be well defined and familiar to responders. The chain of command during a bioterrorism event may benefit from pre-event planning, practice, and widespread acceptance by relevant response agencies.
3. Information systems for communication among responders and with the public must be implemented and tested prior to an event. A bioterrorism response is likely to benefit from a coordinated effort to provide and share accurate information at all levels: local, state, and federal. There is little evidence regarding current efforts to facilitate sharing of information across types of response professionals (i.e., among clinicians, public health officials, and criminal investigators) or with the public—critical gaps in the literature.
4. The costs and benefits of acquiring, storing, and maintaining local inventories of medical supplies have not been established. The federal government has spent millions of dollars to acquire and maintain the Strategic National Stockpile and to develop tested strategies for deploying the Stockpile to a local population within a few hours of an event. However, local and state organizations are creating their own inventories. No published evidence describes the costs and benefits of establishing these local inventories.
5. The Laboratory Response Network significantly enhanced regional laboratory surge capacity. Although most local and state public health laboratories were highly taxed, the Laboratory Response Network was able to offset many surge capacity requirements at the local level during the anthrax attacks. Whether the logistical framework exists to support transportation of samples from local clinical laboratories to regional Laboratory Response Network laboratories for processing during a bioterrorism event of larger magnitude remains untested.

Summary of the Evidence of Regionalization of Responses to Naturally Occurring Outbreaks

We reviewed 296 articles about the response to ten naturally occurring outbreaks involving SARS, Hantavirus Pulmonary Syndrome, pandemic influenza, meningococcal meningitis, smallpox (two outbreaks and the 2003 pre-event vaccination campaign), cryptosporidiosis, West Nile virus, and monkey pox. Of these, we included 181 articles, 10 government reports, and 33 Web sites that described one or more outbreaks and the regional responses to them.^* We found no specific evaluations of regionalization of responses to outbreaks. Table 10 presents the lessons learned from articles describing regionalized responses to the nine outbreaks. From each of the included articles, we abstracted information about the tasks that responders performed during outbreak responses, and the resources required to perform these tasks. We present those data in our answer to Key Questions 1 and 2 in the “Summary Synthesis of Evidence About Regionalization for Bioterrorism Preparedness and Response” section and Table 14.

Table

Table 10. Regionalization of Outbreak Response.

Table 14

Table 14. Examples of Tasks, Subtasks, and Resources Required for a Bioterrorism Response.

Application of Evaluation Criteria to the Responses to Naturally Occurring Outbreak Literature

In this section we discuss our application of the six relevant evaluation criteria to literature describing responses to naturally occurring outbreaks: network design, inventory management, postponement and modularization, coordination of outbreak response and management of incentives, management of information, and incident command. We found no article that specifically described volume-outcome associations. The following sections describe our application of the evaluation criteria to the literature describing regionalized outbreak responses.

Network Design. The effectiveness and timeliness of an outbreak response system is influenced by how key components of the system respond to the outbreak. Key components in the system (i.e., network) include field personnel, public health agencies, healthcare providers, laboratory professionals, and other governmental agencies. The included articles describe five key elements of regionalized outbreak responses performed by these response groups including: mobilization of trained personnel, the epidemiological response (including outbreak investigation and implementation of policies to reduce the spread of disease), the hospital responses, laboratory responses, and coordination of the response. This section synthesizes the available literature on how these five outbreak response tasks have been regionalized during responses to SARS and other outbreaks.

Mobilization of Trained Personnel. Response to an outbreak often requires pooled regional resources (throughout this section we regional responders include those at the state and federal levels). When local communities possess insufficient trained personnel to respond to a disease outbreak, trained personnel such as clinicians and epidemiologists can be mobilized from other communities or from national or international resources. In the United States, infrastructure exists to provide local communities with regional and national resources when necessary. For example, many leading outbreak investigation professionals have been trained through two-year assignments with the Epidemic Intelligence Service, which was established in 1951 by CDC as an early warning system against biological warfare and man-made epidemics. Each year, the 60 to 80 Epidemic Intelligence Service officers have played a key role in numerous outbreak investigations, eradication programs, and development and deployment of surveillance efforts.^*

Trained personnel mobilized during outbreaks may be supplied by national or regional organizations that recruit both paid and volunteer professionals. For example, during the 1918 swine/Spanish flu pandemic, nurses were in drastically short supply because more than one third of U.S. nurses were supporting the war effort overseas. The American Red Cross responded to this shortage by developing a network of professional and volunteer nurses (including retired nurses, student nurses, and eventually, women with even minimal nursing experience). Although these nurses were recruited from a national pool, they were often deployed to specific regions by the U.S. Public Health Service or state public health officials.²¹⁸

Deployment of regional resources can also occur on an international scale. During the SARS outbreak, the WHO, CDC, Institut Pasteur, Medecins Sans Frontieres, and National Health Service, among others, sent supplies and teams of experts to China, Vietnam, Singapore and Hong Kong.^{216, 219} Once in the field, WHO and related international teams performed a variety of tasks, including the review of medical records. These teams played a critical role in determining that the cases of atypical pneumonia in Guangdong Province, China, were “clinically and epidemiologically” related to later outbreaks in Vietnam and Hong Kong.²²⁰ During a bioterrorist attack, the ability to deploy highly trained personnel to the affected area may contribute significantly to the local response capacity.

Epidemiological Response. Epidemiological response to an outbreak by local, state, and regional public health agencies includes surveillance, case investigation, contact tracing and follow-up, issuing guidance to healthcare facilities, implementation of infection control measures a well as instituting public health policy such as issuance of travel advisories, isolation and quarantine. Epidemiological policy and legislation are often framed at the state, national, or international levels; however, epidemiological action often occurs at the local level and is performed by local and/or state responders. Thus, coordination of efforts among federal, state, and local officials is key to effective epidemiological responses. Conversely, differences in epidemiological approaches (and political will) between federal and state policymakers and local public health officials can complicate epidemiologic responses to outbreaks. For example, China's local responses to SARS may have been slowed by delays resulting from Beijing-based officials who discouraged clinicians and hospital administrators from fully disclosing the extent of the outbreak.^{221, 223, 434} Specifically, early in the outbreak, most hospitals were not reporting cases on a daily basis, and even as late as April 2003, two months after the outbreak was reported to have begun, contact tracing was not performed reliably.²²⁴ As the outbreak continued, China demonstrated dramatic improvement in its responses.²²⁵ For example, in Hong Kong, officials instituted a policy of heightened travel surveillance for one year, to immediately detect any resurgence in SARS cases.²²⁶ Hong Kong border screeners used infrared temperature scanners and required health declarations from all travelers.²²⁶ Similarly, local Hong Kong health officials quarantined all known contacts of SARS patients at home. These local responders, working in collaboration with their colleagues in the Department of Immigration, prevented quarantined patients from leaving the country and spreading the disease. China's improved epidemiological response was largely credited to changes in the attitude of high-ranking officials, and in two prominent situations, to the change of the officials themselves.⁴³⁴

Hospital Responses. Hospital responses to infectious disease outbreaks include provisions for definitive medical care, infection control, emergency management, and communication. Hospital responses may be organized locally at a single hospital facility or regionally within the context of a hospital network. During the SARS outbreak, several infection control practices were effectively instituted on a multi-hospital basis. For example, a team from the Taiwanese Center for Disease Control visited each of the 15 hospitals in Taiwan treating suspected or confirmed SARS cases.²²⁷ This team audited charts and implemented strict infection control measures, including education of health care workers and distribution of personal protective gear. Additionally, the infection control nurses at these 15 hospitals undertook surveillance of staff absenteeism.²²⁷

There are ongoing efforts in U.S. hospitals to pool their resources to prepare more efficiently for a bioterrorism event.^* For example, the MediSys hospital network includes three large hospitals in New York City (in Flushing, Jamaica, and Brookdale).²²⁸ These hospitals share a network-wide Nuclear, Biological, and Chemical Incident Task Force charged with providing assessments of and recommendations for pharmaceutical preparedness at both the individual hospital and network levels. This Task Force recommends that each hospital pharmacy retain a specified inventory, with a mutual aid understanding that needed supplies can be borrowed within the network, as available. Each pharmacy stocks supplies to treat staff and patients for three days, should drugs from the Strategic National Stockpile be delayed.²²⁸

Laboratory Responses. Laboratory responses to outbreaks have been regionalized for two primary purposes: to provide surge capacity for diagnostic testing of specimens during a large outbreak that overwhelms local laboratories, and to serve as the reference laboratory capable of testing for rare pathogens. The Laboratory Response Network, which serves both of these functions, was described in the sub-section “Application of Evaluation Criteria to the 2001 Anthrax Bioterrorism Literature”. During the 1999 West Nile virus outbreak in New York City, the primary laboratory with capacity for diagnosing West Nile virus was CDC's Division of Vector-Borne Infectious Diseases, in Fort Collins, Colorado. While the outbreak was developing in the New York City area, laboratory samples had to be mailed to Colorado using commercial mail carriers and express mail service for testing. Significant delays in diagnoses resulted from laboratory services not being available on the weekend²²⁸ and the lack of surge capacity to handle the large number of specimens requiring analysis (more than 2000 in three months).²¹⁷ As a result of this outbreak, the capacity to perform diagnostic analysis for West Nile virus has been expanded to other laboratories around the country—a critical step given the concerns about this agent becoming endemic in more geographic regions of the United States.²²⁸

Coordination of the Outbreak Response. The included articles describe several regional and international networks developed for coordination of disease surveillance, laboratory support, and outbreak response efforts.^*For example, a number of affected Asian nations came together during the SARS outbreak to sign an agreement that they would screen all incoming and outgoing international travelers, create an international emergency SARS hotline, freely exchange epidemiological information, cooperate on research and training, plan and attend meetings to devise other countermeasures to combat the outbreak, and commit to complete transparency where the outbreak was concerned.²³⁰

One of the most recognized efforts was the WHO-sponsored network of research laboratories established during the SARS outbreak.²³¹ Eleven laboratories in ten countries were invited to participate in the effort to identify the causative agent and to develop a diagnostic test for SARS. All accepted the offer, choosing to cooperate rather than compete.^{232, 233} Networked participants communicated via daily teleconferences and used a secure WHO Web site to post results and questions and to access research tools, such as polymerase chain reaction primers. Samples were shipped around the world within hours of a request.²³³ Such cooperation contributed greatly to the timely worldwide response to SARS. The laboratories announced the conclusive identification of the novel coronavirus responsible for SARS just one month after the creation of the network.²³¹ This rapid success can be attributed in large part to the design of the international network that enabled research to progress 24 hours per day with laboratories in every time zone. Also, laboratories had access not only to local strains of SARS, but also to samples from around the world.

During the SARS outbreak, WHO also erected parallel networks of epidemiologists and clinicians.²³⁴ The epidemiologists worked in all local areas with documented transmission of SARS and collaborated to resolve several key issues of epidemiological concern, including the transmissibility of SARS among asymptomatic individuals and during the incubation and acute illness phases of the disease.²³⁴ The clinicians' network included 80 clinicians from all 13 countries with SARS cases. This network compared the efficacy of different treatment regimes and reported on the results of clinical trials with ribavarin.²³⁴ This type of collaboration would be particularly relevant for a bioterrorism event occurring with a novel infectious agent, or a previously recognized biothreat agent with antibiotic/antiviral resistance.

Inventory Management. Inventory management for outbreaks requires the acquisition of material resources such as antibiotics and vaccines, as well as their distribution to responders or members of the public. These tasks can be logistically challenging when resources are in short supply or when large numbers of people must receive prophylaxis or treatment. For example, during the 1947 smallpox outbreak in New York City, city leaders decided to vaccinate all six million residents but initially had only 250,000 individual doses of smallpox vaccine on hand. City workers quickly repackaged 400,000 bulk doses, and additional doses were made available from military stockpiles. However, these were still insufficient to meet the demand. The mayor responded to the shortage by calling an emergency meeting between vaccine manufacturers and the New York City health department. In response, manufacturers agreed to undertake a 24-hour per day schedule for packaging existing bulk vaccine and to divert all available vaccine to New York City.²³⁵

Postponement and Modularization. The concepts of postponement and modularization apply to two key aspects of regionalized planning for outbreaks. First, the distribution of pre-packaged antibiotics in unit doses rather than bulk packaging, which facilitates their dispensing to a wider variety of target populations, is an important example of modularization. If antibiotics are packaged in short course dose packs (e.g., seven-day supplies), public health officials uncertain about whether their current inventory of antibiotics will meet the demands of patients seeking prophylaxis, may choose to dispense an abbreviated course of antibiotics rather than a full course, allowing additional time for regional inventories to arrive. Second, public health recommendations and protocols for outbreak management can be generated at the state or national level and communicated to local users who may further customize these protocols for their use. This approach reduces the need for local officials to prepare their own materials and increases the likelihood that protocols within a region are concordant. CDC is working with professional organizations to educate responders about smallpox: CDC provides educational materials to organizations such as the Infectious Disease Society, the Dermatology Society, the Emergency Medicine Society, the Nursing Association, and the American Medical Association. These organizations repackage the information and distribute it to their members.²³⁷ Although there have not been evaluations of the costs and benefits of postponement or modularization techniques for bioterrorism preparedness, the available evidence suggests that the use of these methods may facilitate the timely communication of essential outbreak response information to relevant organizations for further local customization.

Outbreak Response Coordination and Management of Incentives. For an infectious disease outbreak to be contained as quickly and efficiently as possible, it is necessary that responders at local and regional levels, including clinicians, government agencies, and members of the public, have incentives to work together to contain the outbreak.

For communicable diseases, the incentives of clinicians must be considered if a stable, well-trained work force is to be maintained. One devastating feature of the SARS outbreak was its effect on health care workers: as of April 14, 2003, 46% of SARS cases in Hong Kong and 63% of cases in Hanoi were among health care workers.²²⁷ Because SARS infected such a high number of health care workers, many quit or did not come to work, resulting in reduced capacity just as health care systems were struggling to serve infected populations.^{2, 238, 239} In Toronto, officials reported difficulty finding physicians to staff the SARS clinics. To combat the problem, the Toronto health ministry issued a call for volunteer infectious disease specialists from the United States to serve temporarily in Toronto. The Infectious Diseases Society of America e-mailed its members advertising the need for U.S. physicians to work in Canada. Although compensation was $2000/day plus expenses, and Canada provided malpractice and temporary licenses and work permits, many American clinicians “were afraid to go.”²³⁹ Thus, without carefully addressing the incentives for responders, regionalized systems for enhancing local responder capacity, particularly during communicable infectious disease outbreaks, may be compromised.

Management of Information. During a regionalized outbreak response, three flows of information are critical: flows between responders and the public, horizontal flows among responders (between regional responders which include information systems for logistics management), and vertical flows among responders (from local responders to national responders or vice versa).

To prevent the spread of disease, outbreak-related information from responders to the public must be accurate, timely, and complete. The included articles describe several simple yet effective strategies to communicate to the public during the SARS outbreak. WHO used international press briefings and global videoconferences for healthcare workers. CDC sponsored regular conference calls with medical professional associations and state and local health departments and laboratories.² During a two-week period early in the outbreak, CDC hosted nine telephone press conferences to keep the media informed and to ensure that the information presented on television and in print was accurate and complete. Travel health alerts were made available in eight languages for more than 60,000 travelers returning from East Asian countries.^{2, 240}

Horizontal information flow among responders is also important. The day the WHO issued a global health alert for countries with documented local transmission of SARS, the director of the Los Angeles county health department issued a notice to all 81 hospitals in the county, informing them of the symptoms and requesting to be notified immediately of suspected cases. Similarly, during the 1999 outbreak of West Nile virus in New York City, the local health department used media releases, a public hotline, and printed materials in eight languages to teach an anxious public how to protect against the virus with considerable success: In two months of operation, the hotline received more than 150,000 calls.²¹⁷ Whether these methods of communication will adequately serve the needs of responders or the public during a large-scale bioterrorism response has not been evaluated.

Finally, outbreak responses require vertical information flow among regional and local responders. Technology can play an important role in this vertical information flow.³ For example, Web-based communication technologies served a key role during the SARS outbreak. Internet technology was used by the international network of laboratories to facilitate the isolation of the causative agent for SARS. Also, California's Health Alert Network was used to send SARS-related e-mails to local health departments and to many hospitals.^{2 *}Electronic versions of SARS case reports and related information were available to worldwide subscribers to the New England Journal of Medicine one month before they were available in print.²⁴¹

Other outbreaks have demonstrated the limitations of available communication systems. The 1999 outbreak of West Nile virus in New York City was multijurisdictional: patients lived and were treated in several cities and states. No agreements were in place regarding information sharing for outbreak tracking and surveillance before the outbreak occurred. Interim arrangements were unable to produce effective data management strategies.²¹⁷ Additionally, the health departments needed better integration with animal health agencies. In a GAO evaluation of the response to the 1999 West Nile virus outbreak, it was noted that, “the length of time it took to connect the bird and human outbreaks of West Nile virus signals a need for better coordination among public and animal health agencies.”²⁴² This outbreak stimulated ongoing efforts to coordinate collection and reporting of surveillance data for zoonotic and other naturally occurring and bioterrorism-related outbreaks.^{69, 435}

Incident Command. The included articles suggest that clear chain of command facilitates effective, expeditious containment of infectious disease outbreaks and communication of the ongoing epidemiological efforts to the public. The included articles described the emergency operations centers that were established worldwide during the SARS response to bring together local EMS, police, and community planners; however, most of these centers were not coordinated regionally or nationally.² CDC activated its Emergency Operations Center for which 800 “medical experts and support personnel” were available around the clock to coordinate a U.S. SARS response.² The Emergency Operations Center has only been activated twice before: for the 1999 West Nile virus outbreak and for the 2001 anthrax attack.²⁴³ The WHO's contribution to the SARS response was facilitated by its event management information system which “generates a dynamic picture of operations, aids organization of logistics, and provides a systematic way to…manage resources.”²⁴⁴

Lessons Learned from the Responses to Naturally Occurring Outbreaks

From our review of the outbreak response literature, we drew three main lessons regarding a regional response to outbreaks resulting from bioterrorism.

1. Communication and cooperation between health authorities of neighboring regions are needed. Infectious diseases can spread quickly and communication and cooperation among neighboring communities can facilitate a response. Rapid communication can be difficult to achieve through interim agreements. Thus, cooperation during a bioterrorism response may benefit from pre-event development and routine use of shared communication systems.
2. International surveillance and reporting may be required to combat infectious disease outbreak. The SARS response benefited from international efforts of disease surveillance and reporting of suspicious cases to the WHO. In the event of a bioterrorism attack, international cooperation to detect, report, and respond may reduce associated morbidity or mortality.
3. During a response to an outbreak resulting from an emerging or communicable biothreat agent, it is essential to carefully consider and address responder incentives. During bioterrorism event, strategies to protect responders and their families is an essential component of maintaining a robust work force.

Summary of the Evidence of Regionalization of Disaster Responses

We found 155 articles that described responses to natural disasters, including hurricanes (i.e., Andrew, Iniki, and Hugo), earthquakes (e.g., Loma Prieta and Northridge), tornadoes, and floods, and to manmade disasters (e.g., the Tokyo sarin attack, the Keystone chairlift collapse, the destruction of the Alfred P. Murrah Federal Building in Oklahoma City, airline crashes, and the terrorist attacks of September 11, 2001). Of these, we included 37 articles^* and 17 government reports and 12 Web sites.^† None of the included articles specifically evaluated the effects of regionalization on disaster management services. Table 11 presents the evidence of lessons learned from articles on natural and manmade disasters that can be applied to bioterrorism preparedness. For reviews of disaster epidemiology and response, we refer interested readers elsewhere.^{42, 441, 442}

Table

Table 11. Lessons Learned From Regionalized Responses to Disasters.

Most included articles describe response organizations or responses to particular disasters. Although our review found no specific evaluation studies, it highlighted issues regarding regionalization of disaster preparedness and response, which we synthesize in the next section.

Application of Evaluation Criteria to the Disaster Response Literature

This section discusses our application of the evaluation criteria to the disaster response literature: network design, inventory management, supply chain coordination and management of incentives, management of information, volume-outcome associations, and incident command. Use of these evaluation criteria facilitates an assessment of the structure, timeliness, and capacity of regionalized disaster response systems. None of the included articles presented information relevant to postponement and modularization.

Network Design. The key elements of regionalized disaster responses include mutual aid agreements, available hospital capacity, available trained personnel, and available morgue capacity. Procurement and placement of relevant material is also a key component of a well-designed network for disaster response; material and supplies will be discussed under Inventory Management and Volume-Outcome Associations.

Mutual Aid Agreements. Mutual aid agreements are the primary method of regionalizing disaster responses.^{438, 443–446} These agreements provide surge capacity to local jurisdictions when basic services such as law enforcement, firefighting, and health services are overwhelmed. Mutual aid agreements enable localities to expand surge capacity at minimal cost, as each individual locale does not have to maintain inventories of supplies and personnel at maximal-preparedness levels. In addition, such agreements facilitate multi-region involvement if a disaster becomes too large for a single area to manage alone or if a disaster encompasses multiple regions.

Having started as a statewide agreement among California fire and law enforcement agencies in the 1950s (Figure 6), statewide mutual aid systems now exist in many U.S. states.⁴³⁸ Under mutual aid, local jurisdictions who expect to exhaust their own resources call on neighboring jurisdictions for assistance, according to mutually agreed upon rules for deployment and reimbursement. If the resources of a mutual aid region are expected to be overwhelmed, the region may also request state assistance. Several states have developed information systems that link members of the mutual aid regions. or example, California's Response Information Management System (RIMS) is a secure Web-based system (with access limited by user type) designed to coordinate and manage the State's response to disasters and emergencies.^{502 *}

Figure

Figure 6. Office of Emergency Services Mutual Aid and Administrative Regions. Map publicly available.

Whereas mutual aid agreements are typically within states, neighboring states have developed regional agreements for disaster management (particularly important for states like Tennessee with multiple neighboring states). In 1992, the 19 members of the Southern Governors' Association, responding to the massive destruction caused by Hurricane Andrew in Florida and Louisiana, developed the Emergency Management Assistance Compact.^{447, 448} The design of the Emergency Management Assistance Compact was complicated by the heterogeneity in participating states' constitutional and statutory laws. Party states independently resolved legal questions associated with the compact such as “whether a governor, absent specific statutory authorization, had the power to grant good faith immunity to relief workers from another state or to spend money out of state on interstate assistance.”⁴⁴⁷ The Emergency Management Assistance Compact has been highly effective, and since being approved by Congress in 1996, has been adopted by all states and U.S. territories except for California and Hawaii.^{447, 449}

The Emergency Management Assistance Compact is currently working to develop a 50-state database of available assets sorted according to standardized categories for each type of disaster (standardized aid packages describe the number and types of personnel and equipment needed for a given disaster response).⁴⁴⁸ A barrier to the comprehensive utility of the Emergency Management Assistance Compact during a bioterrorism response (i.e., its ability to serve as the nation's mutual aid agreement) is that it is not in effect nationally. California, with its own long-standing agreements, has legislation and mutual aid contracts in place that conflict with the Emergency Management Assistance Compact.⁴⁴⁸ Despite this, those states that have ratified the Emergency Management Assistance Compact are using it.

There are numerous examples of the Emergency Management Assistance Compact partnerships resulting in expedited regionalized disaster responses. In September 1998, as Hurricane Georges, having devastated Puerto Rico and the Caribbean, headed north to the Florida Keys, officials knew from prior hurricane experience that they would not be able to evacuate patients with special medical needs from the Keys to the mainland on the single highway connecting the two. Florida officials requested use of Emergency Management Assistance Compact partner North Carolina's medical evacuation aircraft, which were immediately dispatched and effected a rapid rescue of this vulnerable population. The Emergency Management Assistance Compact also provides interstate aid for disasters other than hurricanes. In response to ten Emergency Management Assistance Compact states sending help to Florida during its 1998 wildfires, Florida Governor Lawton Chiles stated, “Thanks to our compacts with other states through our Emergency Management Assistance Compact program, we got the help we needed. The coordinated effort during the fires is proof of the value of these compacts.”^{447, 449}

Regionalization of response for disasters often requires a coordinated international response with Canada or Mexico, particularly in cities close to the U.S. border. Extensions of mutual aid agreements have also developed between U.S. cities and states and their neighboring communities in Canada and Mexico. The La Paz Agreement, signed in 1983 by Presidents Ronald Reagan and Miguel de la Madrid Hurtado, identified 14 pairs of cities along the United States-Mexico border in need of joint response plans for disasters.⁴⁵¹ An analysis of the disaster response plans, personnel and equipment covered by a mutual aid agreement between El Paso, Texas and Juarez, Chihuahua, Mexico highlighted key problems in developing a joint plan: the development of written plans was complicated by conflicting international laws, worker's compensation, insurance, and cross-border communication frequencies; and lack of international incident command structures, bilingual databases and community notification systems.⁴⁵¹

The mutual aid agreements that have been developed for disasters serve as a framework for regional agreements for bioterrorism preparedness and response. The literature on mutual aid for disasters highlights key considerations relevant to regional mutual aid for bioterrorism. First, the establishment of these agreements requires careful pre-event consideration of liability issues, remuneration, and licensing. Second, design of international mutual aid agreements with Canadian and Mexican regions require consideration of international law, multi-lingual information technologies, and heterogeneity of disaster resources.

Hospital Capacity. Following a disaster, hospitals must provide immediate care to injured victims, continue providing primary care to the populations they serve, and may themselves be the victims of the event (either directly through plant and equipment damage or indirectly through loss of staff who are victims themselves or who are caring for family members). Similar to a bioterrorism event, both natural and manmade disasters can rapidly generate large numbers of casualties requiring urgent medical attention.^479–481 Hospital response systems have been designed to assist victims of disasters in the hours and days following an event. For example, the Medical Disaster Response Model is designed to mobilize local clinicians to provide immediate medical care in a carefully planned, regionalized system where clinicians initially provide care to injured victims in the field then move to pre-designated disaster medical aid centers, finally converging on pre-designated regional casualty collection points.⁴⁸²

A critical component in effective triage during a disaster is having real-time data on regional hospital capacity and bed availability. For example, immediately after the Loma Prieta earthquake, California's Emergency Medical Services Administration began a statewide bed and resources inventory. Before it could be completed, the bed count was discontinued, as it was clear by then that hospital capacity in the Bay Area would be sufficient to meet the needs of victims injured in the quake. However, 2000 available beds had been identified in unaffected regions of the state.⁴⁸³ Similar activities occurred following other major disasters: after the September 11 attacks on the World Trade Center, the Greater New York Hospital Association worked to amass bed availability information and tracked patients so they could be located by families.⁴⁸⁴

These systems that track hospital capacity, numbers and locations of patients, and estimate the need for additional beds could serve a key role during a local or regional response to bioterrorism. However, none of the included articles specifically evaluated various strategies for estimating available surge capacity or for methods of enhancing local hospital capacity for bioterrorism-related events.

Response Personnel. Disaster responses often require large numbers of personnel. Response personnel typically consist of local responders, who are usually first to the disaster scene, and regional response teams, which may be activated once news of the disaster spreads. Local response personnel include trained professionals such as fire, police, and emergency medical personnel and trained volunteers, as well as untrained volunteers. Trained local responders are often sufficient for key response tasks, even following a major disaster. For example, following the 1994 Northridge earthquake in Southern California, the Los Angeles County public health department created 12 interdisciplinary assessment teams that were sent primarily to “high-risk” areas to provide public health education, environmental health assessment, mental health counseling, and first aid to victims of the earthquake.⁴⁸⁵ Trained volunteers can augment the professional response. For example, prior to the 1989 Loma Prieta earthquake, the local American Red Cross chapter in one devastated Bay Area community had established close links to a local amateur (Ham) radio club. Immediately following the earthquake, when this community sustained extensive damage, the Red Cross chapter was able to leverage the pre-existing personal and professional relationships to provide emergency communication among its responders, even when telephones were not operating.⁴⁸⁶

Community members without specialty training can also contribute significantly to a local response. For example, taxis and taxi drivers played an important role following the 1995 Tokyo sarin attack because, unlike the EMS system, their communication and dispatch system was not overloaded. As a result, 25% of sarin victims were transported to the hospital by taxi. However, because the taxi dispatch system had no way to direct drivers to those hospitals able to accept patients, many taxi-transported patients arrived at hospitals lacking capacity. For future Tokyo disasters, an informal taxi surveillance system has been implemented in which taxi drivers now report disasters to their dispatch center as soon as they are observed. However, this system offers no protection to taxi drivers responding to chemical or biological attacks, as drivers have not been given protective equipment or training.³⁹

Other programs have invested more significantly in training and equipping community members for disaster response. For example, the Community Emergency Response Team Program originated in Los Angeles following the 1987 Whittier Narrows earthquake, and has since been promoted and supported by FEMA. Community Emergency Response Team trains community members to offer “immediate help to victims until professional services arrive.”⁴⁵² Residents are trained to extinguish small fires, provide basic medical aid, rescue victims safely, and collect disaster intelligence.⁴⁰ Community Emergency Response Team members are often organized into neighborhood response teams.⁴⁰ Forty-eight states and six foreign countries currently have Community Emergency Response Team programs.⁴⁰

Programs like Community Emergency Response Team underscore an important strength of local responders: they generally can arrive before regional responders. This timeliness of response may be critical during the response to a bioterrorist event. For example, following the Singapore Airline disaster in 2000, a contingency plan indicated that a Site Medical Team would be called to treat victims at the scene. However, the team did not arrive until 30 minutes following the crash—a delay that was concordant with planning and practice.⁴⁸⁷ When the team arrived, members discovered that all critically injured patients had already been triaged and transported by local emergency transport vehicles. These vehicles had arrived within 15 minutes of the crash, and their quick response was credited with saving numerous lives.⁴⁸⁷

In addition to local responders, regional response teams with expertise in emergency management are often deployed to disaster sites. Numerous organizations, including the federal government, military, and professional volunteer organizations sponsor and support such teams. For example, the International Federation of Red Cross and Red Crescent Societies have participated in many of the responses described in Table 11. Recent regional restructuring has allowed the Red Cross's Field Assessment and Coordination Teams to save money and better coordinate response efforts among local, regional, and international responding organizations.⁴⁸⁹ An international pool of approximately 200 trained Field Assessment and Coordination Team members with expertise in disaster assessment and response are on call at any given time. From this group, teams of six or seven members are typically deployed from the international Red Cross headquarters in Geneva within 12 to 24 hours following a disaster.⁴⁸⁸ This team makes a report that can be used to coordinate the response efforts of each local Red Cross organization (this is compared with earlier practices of numerous response organizations each performing their own needs assessment prior to mobilizing response personnel and equipment).⁴⁸⁹ Field Assessment and Coordination Teams have been augmented by Regional Intervention Teams that provide integration with regional responders who are often more familiar with the host country's language and culture.⁴⁸⁸

The federal government also deploys trained response teams to U.S. disaster scenes. For example, following severe hurricanes, Disaster Medical Assistance Teams are typically deployed and provide primary care services or augment overloaded local medical service providers.¹⁰⁴ Ten Disaster Medical Assistance Teams were placed on “standby alert” following a hurricane watch for Hurricane Andrew in 1992. A total of 16 Disaster Medical Assistance Teams were deployed to assist in the response and recovery following Andrew, which caused unprecedented destruction in southern Florida and Louisiana. Teams deployed later in the response also served as replacements for exhausted teams that had been deployed earlier.⁴⁹⁰

In addition to civilian government responders, various branches of the U.S. military also train and maintain regional response teams for disaster preparedness. Such teams may be deployed to assist civilian communities following a major disaster. Some teams receive specific training to respond to bioterrorist events. For example, the Army National Guard maintains a Weapons of Mass Destruction Civil Support Team in each federal region.⁴⁵³ These teams (also described in the “Bioterrorism Preparedness Programs Sponsored by the Department of Defense” section) consist of 22 Guard members, trained and ready for domestic deployment to either a chemical or biological incident.⁴⁵³ Following deployment, teams may fall under either state or federal control.⁴⁵⁴

There are two key advantages to regional response teams for disasters and bioterrorism. First, regional placement allows for fewer responders to be trained, since one team can serve a broad area. Additionally, teams may benefit from more consistent deployment, adding to responders' experience and reducing the risk that training will decay if not reinforced. The section below on volume-outcome associations discusses this concept in greater detail. However, we found no specific evidence that describes optimal strategies for regionalization of highly trained personnel for bioterrorism or disaster preparedness (e.g., evaluation of the number of highly trained response teams required for a region), alternative methods for enhancing the cross-training of local personnel (e.g., local pharmacists used for triage functions during a bioterrorism response), or for the best methods of coordinating the activities of local and regional response personnel.

Morgue Capacity. Due to their ability to cause a large numbers of deaths in a very short time, disasters can seriously strain morgue capacity. For example, in the case of the Loma Prieta earthquake, 57 of 63 deaths occurred within two minutes of the earthquake, and 42 of these deaths were in one small geographic area.⁴⁷⁹ Bioterrorism events have a similar potential: recent estimates suggest that a smallpox attack on a major U.S. city could cause as many as 100,000 deaths.⁴⁵⁵ However, we found no information regarding emergency preparedness or disaster response at morgues and other facilities for fatality management.

Inventory Management. Disaster responses can require enormous quantities of supplies, which are often needed immediately. Inventory management plays a critical role in ensuring that such needed supplies arrive in a timely fashion. While we found little discussion of inventory management in connection with disaster response in our literature review,^* three key inventory management issues are relevant to regionalized approaches to disaster response. First, a needs assessment must be performed for each locale to determine the amount and type of resources required for a disaster response.^{488, 489} Most often, this occurs after an event; however, for areas that are subject to annual or seasonal events (e.g., hurricanes), pre-event planning for expected disasters can take place. Second, determinations as to which organizations are responsible for purchasing, inventorying, and delivering different resources must be transparent and shared with all responders.^{488, 489, 500, 501, 503} Finally, inventory distribution plans must be developed. Inventories can be pre-placed so that they are readily available during a disaster or can be transported to affected regions after an event occurs.^{488, 489, 500, 501, 503} Alternatively, some humanitarian aid agencies have developed software systems, such as the Pan American Health Organization's Humanitarian Supply Management System (SUMA), to track the arrival and distribution of humanitarian supplies as they arrive following a disaster.⁵⁰⁴

Following its 1998 response to Hurricane Mitch, the International Federation of Red Cross and Red Crescent Societies determined that pre-positioned supplies were necessary to facilitate a timely response to victims' needs. Thus, the Red Cross established agreements with local vendors in hurricane prone areas.⁴⁸⁸ These agreements specified the price to be paid for relief supplies and stipulated that the supplies would be stored in the vendors' own warehouses. Such warehouses are consistently stocked at a certain inventory level according to Red Cross needs.⁴⁸⁸ According to these agreements, vendors assume the cost of warehousing and of stocking additional supplies, not the Red Cross.⁴⁸⁹ The Red Cross also maintains its own regional warehouses and logistics units for further pre-positioning of supplies.⁴⁸⁸ We found no specific evidence that evaluates key questions with respect to inventory management for disasters including assessments of methods to determine the size, location, or contents of pre-placed inventories.

Disaster Response Coordination and Management of Incentives. The included articles addressed the incentives of mutual partners, but did not specifically address the incentives of other responders. Since the establishment of California's comprehensive mutual aid agreements, numerous conflicts have occurred over payment for assistance provided, and liability insurance covering responders.⁴⁵⁰ The California master mutual aid agreement specifies that responding jurisdictions retain control of their own personnel and facilities and that no party is required to deplete its resources to a point that endangers its ability to manage a local response. However, mutual aid is mandatory during “conditions of extreme peril or declared disaster.” In some cases, mutual aid partners have resisted providing mutual aid, principally for economic reasons, despite being called on to do so.⁴⁵⁰ In contrast to California's intrastate system of mutual aid, states requesting help under the Emergency Management Assistance Compact are responsible for “reimbursement of all out-of-state costs and accept tort liability for out-of-state personnel.”⁴⁴⁹ “States ask for out-of-state assistance only when they need it and can cover the costs. States know sending aid will not be a financial or legal burden.”⁴⁴⁹

Management of Information. Management of information can affect every aspect of a disaster response. Some information about a disaster may be available before it occurs, allowing planners and the public extra time for coordination and planning. For example, because hurricanes are somewhat predictable, information from preceding years can inform future disaster preparations. The most critical information management and communication challenges often occur during the immediate post-event and disaster response phases, when infrastructure is likely to be overloaded and may be damaged or even destroyed. During these phases, effective communication among decisionmakers, responders, and the public is crucial.

Communication Among Decisionmakers and Responders. During a disaster, decisionmakers and responders have three main information requirements. First, decisionmakers require continuously updated information about local needs and available local and regional resources. Second, the information required to maintain a clear chain of command must be readily available to decisionmakers and to the responders under their command. Finally, information about response task performance and response outcomes is essential for iterative improvements of disaster responses.⁴⁴

Numerous commercial systems are available to facilitate disaster communications (e.g., California's Response Information Management System (RIMS). For example, most counties in the San Francisco Bay Area use the very high frequency Hospital Emergency Administrative Radio system in a disaster.⁴⁵⁶ This system links all hospitals in a single county to a dispatch system. The system is organized at the county level, which also controls the EMS, and allows overloaded hospitals to divert ambulances to other hospitals with excess capacity.⁴⁵⁶ However, such systems do not always function in a large-scale disaster. For example, following the 1995 Tokyo sarin attack, EMS communication systems lost contact with physicians at the Tokyo Metropolitan Ambulance Control Center. Because paramedics in Japan are prohibited from performing certain advanced life support activities without case-by-case permission from a physician, this loss of communication resulted in delayed treatment for victims.³⁹

Communication with the Public. Systems for communication with the public can be grouped broadly in two categories: alert/warning systems intended to provide generalized alerts and to notify the public of an increased risk of a bioterrorism event, as well as news and information outreach systems that provide detailed information about particular events.

Among the included articles, we found one evaluation of a public alert/warning system.⁴⁵⁷ This study evaluated the use of sirens as tornado warning systems. The study found that although people in communities served by a tornado siren are more likely to get their initial warning of an imminent tornado from a siren than from television or radio news, these residents are no more likely to seek shelter than those not served by tornado sirens (only about 30% of people in either group reported seeking shelter after receiving a tornado warning).⁴⁵⁷

A warning system of particular relevance to bioterrorism is the Homeland Security Advisory System. Created by presidential directive in March of 2002, this system uses five colors to represent perceived levels of threat to national security ranging from Red (severe threat) to Green (low threat) (Figure 7).⁴⁵⁸ The perceived threat levels are assigned by the Attorney General in consultation with the Secretary of the Department of Homeland Security and others based on intelligence information. Since its inception, this system has remained primarily at Yellow Alert (elevated). As of August 10, 2003, the system has gone to Orange Alert four times.⁴⁵⁸ The Advisory System requires the executive branch of the federal government to comply with a system of “Protective Measures” corresponding to each threat level. Additionally, the presidential directive that created the system encourages other departments and branches of government at every regional level to create their own list of “Protective Measures” to be instituted should the nation be placed under increased alert.^{459, 460} The Advisory System is also intended to integrate with other alert systems used by decisionmakers in other federal agencies (e.g., the Department of Transportation, the Department of Agriculture) as well as officials at the state and local levels. The Advisory System aims to provide a “common language” for each of the nation's alert systems, allowing responders and the public to more easily interpret risks and threats.^* However, according to the recent State of America's Cities survey conducted by the League of Cities, only 20% of American cities currently use the National Homeland Security Advisory System (although 71% of large cities either use the system or are working to implement it).^{461, 462} The system continues to be developed, and may benefit from providing more specific alerts to localities rather than issuing a warning for the nation as a whole (thus avoiding unnecessary spending by localities unaffected by the perceived threat), and from providing responders with more information about appropriate measures to take in response to a given threat.⁴⁵⁸

Figure

Figure 7. Homeland Security Advisory System. Figure publicly available.

During an emergency, the public can also receive information via disaster-specific media channels. After the 1994 Northridge earthquake, victims were kept updated by an extensive FEMA media effort that included “reports, memos, summaries and updates.”⁴⁶³ Six days after the earthquake, FEMA established the Recovery Channel, which was shown by local television stations and in disaster response centers such as shelters and disaster relief application centers. The following week, FEMA began publishing the Recovery Times, a newsletter with specific information for victims, updating them about the status of the recovery and providing information about where and how to apply for assistance.⁴⁶³

Volume-Outcome Associations. The included articles support the concept that increased disaster response experience is associated with better response outcomes.⁴⁸⁶ After the 1989 Loma Prieta earthquake, residents of one community who had responded to many disasters, including mudslides and wildfires, contributed significantly to a high-quality response.⁴⁸⁶ The ongoing risk of natural disasters increased the experience of local responders and compelled the community to plan and participate in disaster training exercises.⁴⁸⁶

Conversely, lack of experience has negatively impacted disaster responses. For example, during the 1990 Texas floods, one community's Red Cross chapter had only a single paid employee who was relatively new to the job and had no experience planning for or responding to disasters.⁴⁸⁶ As a result, the quality of the local response was inadequate, prompting deployment of national Red Cross volunteers who stayed for three weeks, providing the “backbone” of the local chapter's response.⁴⁸⁶

Such findings have important implications for regionalization of disaster and bioterrorism preparedness. First, personnel with experience in disaster response are often few in number. Staffing them at a regional level ensures wider availability of this scarce resource and allows local areas to spend their money elsewhere. (However, redundancy in the system through the allocation of some resources locally can enhance capacity when a large region is affected.) Second, because regional staffing models cover a wider geographic territory, they are therefore more likely to encounter disasters, thus improving their experience.

Incident Command. The 1900 Great Hurricane in Galveston, Texas resulted in more than 1000 casualties and led to the development of the modern approach to city management.^{* 491} The broader lesson of the Great Hurricane remains: the more complicated a response becomes, the more necessary it is to have a strong chain of command. This is particularly true when the response involves a large region and many diverse responders (especially regional responders who may never have previously worked together). This chain of command must be established and agreed upon before the event occurs. The chain of command must be integrated to include all responders and must allow resources to be rapidly deployable.^{14, 454}

During a regionalized disaster response, it is essential to establish whether a local, regional, or federal agency is in command. Typically, incident command begins with local responders, as they are usually the first to the scene. Understanding when and how command changes as additional resources are called in is important to maintain an effective response to a disaster. Protocol should be defined prior to an event, and should clearly lay out changes in command. Lack of such protocols might lead to difficulties such as those experienced during the response to the 1989 massive Exxon Valdez oil spill off the coast of Alaska.⁴⁹² Following the spill, some time-sensitive cleanup actions, such as the application of dispersing agents and the burning of oil, were delayed due to competing contingency plans and a confused chain of command.⁴⁹² There was no established authority under which Exxon could assume control of the cleanup from local port authorities. After Exxon assumed control, it became clear that the corporation had no standing contingency plan for the geographic area where the spill had occurred, and that all response logistics would have to be created de novo. Ultimately, the response and cleanup were controlled by a triumvirate consisting of the president of Exxon Shipping Company, a Rear Admiral of the U.S. Coast Guard, and the Commissioner of the Alaska Department of Environmental Conservation. However, this command structure was not in place until more than two days after the tanker ran aground.⁴⁹²

For efficient disaster response, chain of command must be established as quickly as possible following the occurrence of the disaster. The Incident Command Center of the New York City Department of Health was activated only 32 minutes after the first plane crash on September 11, 2001.⁴⁸⁴ The New York City Department of Health was aided by previous disaster planning and drills. The disaster literature provides compelling evidence that emergency management, including that which would be required for a bioterrorism response, is facilitated by a clear chain of command.^{505, 506}

Lessons Learned from the Disaster Literature for the Regionalization of Services for Bioterrorism Preparedness

From our review of the disaster response literature, we synthesized three key lessons learned.

1. Information management, using common technology platforms, is essential for assessing the needs of the local community and the resources available to them, and for coordinating responses from regional agencies. The disaster literature emphasizes the need for updated information so that response resources can be dispatched to those in need in a timely manner. Bioterrorism responses may benefit from a common technology platform. These may be as simple as a common radio frequency or as complicated as multiple computer networks from different regions communicating with each other to exchange information on resource placement and availability.
2. A key component of effective regional responses to disasters includes mutually agreed upon pre-event protocols that establish chain of command structures. The evidence suggests that these protocols are most effective when they delineate the chain of command for an event at several levels: a chain of command utilizing only local responders should be established, as should plans for how the chain of command changes when regional and federal response agencies become involved.
3. Mutual aid agreements are key components for providing surge capacity for regional response to disasters. These agreements ensure that every locale does not have to be staffed and prepared for a maximal intensity event. They enable risk to be spread among several locales, and provide cost-sharing of disaster preparedness. Mutual aid agreements for bioterrorism are likely to benefit from careful pre-event consideration of liability issues, remuneration, and licensing.

Regionalization of Trauma Care

A regionalized trauma system includes classification of hospitals within the region according to the services that they can provide, protocols for pre-hospital trauma care and transportation by emergency medical personnel, and coordination of these pre-hospital and hospital-based services.⁵⁰⁹ Regionalization of trauma care efforts in the 1980s and 1990s was spurred by a report from West et al. documenting significant differences in the preventable death rates between San Francisco county, which had a regionalized system of care, and Orange County, which did not.⁵¹⁰ Regionalizing U.S. trauma care was further enabled by legislation such as the Emergency Medical Services Act in 1973 and the Federal Trauma Care Systems Planning and Development Act in 1990, which provided assistance to states to “plan, implement, and monitor organized trauma care systems.”⁵¹¹ Currently in the United States, five states have systems that meet all the criteria for a regional trauma care system. Only 24 states have trauma care systems that function across the entire state.⁵¹²

Depending on the state, trauma care hospitals in the United States are designated by one of four or five levels (Level I through IV or V). The criteria for designation of hospitals, established by the American College of Surgeons,⁵⁰⁸ are based on the type and amount of equipment, resources, and trained personnel available to the hospital.^{508, 511} Level I hospitals (also referred to as tertiary care centers) provide the most extensive trauma care. A region within the trauma care system is typically defined as a geographic area that is served by at least one Level I center. Rural areas are more likely to be served by a Level III or IV center that collaborates with the nearest Level I center located in a more urban area. Urban areas are more likely to have Level I and II centers. Regionalization of trauma care is based on the principle that injured patients will be appropriately transported to the nearest hospital that can provide the best possible care to the patient. The destination hospital is determined by emergency medical personnel according to the severity of injury. The most severely injured patients are transported directly to the nearest Level I center, bypassing the closest hospital if it is not a Level I center. Less severely injured patients are transported to the closest hospital capable of meeting the needs of the patient.⁵⁰⁸

Summary of the Evidence of Regionalization of Trauma Care

We reviewed 98 articles and reports on trauma care. Of these, 74 articles and two government reports described regionalization of trauma care.^* Thirty-nine of these articles presented evaluations of regionalization of U.S. and Canadian trauma care systems.^† Table 12 presents the evidence from the 16 evaluations of trauma care regionalization with the greatest relevance to bioterrorism preparedness and response.

Table

Table 12. Selected Evaluations of Regionalized Trauma Care Systems.

Most evaluations of regionalization of trauma care either compare patient outcomes before and after implementation of a regionalized trauma care system,^‡ or compare outcomes between trauma care centers and non-trauma care centers.^§ Mann and colleagues published a systematic review of the published evidence regarding trauma system effectiveness and found that most of the studies had systematic biases depending on the methodology used to assess trauma care effectiveness.⁵²³ For example, panel studies could lead to biased results depending on the quality of data available or whether assessments of preventability of death were made by individual physicians or in a group setting.^{523, 551} Other studies relied on data registries of questionable reliability due to incomplete data, data collected for another purpose, or outdated data used for comparison purposes.^{37, 521, 523}

Despite these potential sources of methodological bias, 29 evaluation studies reported improved outcomes associated with regionalization, including reduced mortality,^* pre-hospital time,^† costs of treatment, ^‡and improved appropriateness of patient care.^§ The reviewed evidence is comprised mostly of retrospective cohort studies or observational studies where the comparison group differed by location or time period. Additionally, the published literature mostly relied on hospital survival as an outcome measure, and restricted patient inclusion criteria by injury severity or geographic location.⁵²³ Despite the limitations of the published data, the results of evaluations of trauma care consistently favor regionalization. For example, a systematic review of nine regionalized trauma systems³⁷ reported a 15%-20% improvement in trauma-associated mortality after the implementation of regionalized trauma care.^{37, 523} Miller and Levy reported 15.5% lower costs per injury episode and $1,025 average per case savings in states with regionalized trauma care.³⁷

Four key elements of trauma care regionalization are primarily credited with these improved outcomes: 1) formal designation and accreditation of hospitals at the appropriate level,^{508, 532} 2) implementation of protocols specifying patient triage and transportation to appropriate trauma centers, 3) use of communication networks between pre-hospital care givers and hospital personnel, and 4) availability and ongoing training of medical personnel to assess, triage, and treat patients. In a prospective evaluation of mortality and pre-hospital time during the regionalization of a trauma care system, Sampalis and colleagues reported a decrease in mortality from 52% to 18%, with a decrease in mortality for each year of the study since the initiation of the regionalization process.⁵⁰⁹ The regionalization stages were classified as: initial designation of hospitals, establishment of triage and transfer protocols, and integration of hospital and emergency services.⁵⁰⁹ A regression model of the stage of regionalization and the risk of death showed significant decreases in the adjusted odds ratio for the risk of dying at each advanced stage of regionalization (i.e., each stage was significantly associated with decreases in mortality). The study also reported a 30% reduction in pre-hospital time, a critical factor in reducing morbidity and mortality.^{509, 574} In a systematic review of effectiveness of regionalized trauma systems, the elements contributing to improved mortality were the designation of trauma centers,³⁷ the development and implementation of appropriate patient triage and transfer protocols,^34–37 and improved pre-hospital care.³⁷

Two other systematic reviews of effectiveness of regionalized trauma care systems provide some evidence as to the importance of designation of hospitals in improving survival.^{521, 551} Simons and colleagues compared mortality rates between trauma centers that had been designated and accredited to those that had only been designated without formal accreditation, and found significantly improved outcomes in the center that met all the criteria for accreditation.⁵⁵⁶ Voeller and colleagues reported a relative risk of dying before establishment of a regional trauma center to its establishment of a regional trauma center of 2.7 (p < 0.03). The researchers suggested that the improved outcomes they observed were most likely due to the designation of a tertiary care center, the establishment of triage protocols, improved pre-hospital care through a communications network, and the availability of dedicated and specialized personnel.⁵³³

The evidence suggests that geographic variations do not affect the importance of the key components of regionalized trauma care systems. Communication, designation, triage, and transportation protocols still play an important role in establishing regionalized trauma care systems in rural areas.⁵³² However, it might be important to tailor particular components according to the location of the system.⁵¹⁶ For example, for regional systems in a primarily rural or remote area of the country, designation of hospitals may concentrate on Level IV or V hospitals rather than Level I hospitals, or protocols for patient triage and transfer might focus on patient treatment and stabilization before transport to a tertiary care center rather than reduction in pre-hospital time.^{516, 532}

Application of Evaluation Criteria to the Trauma Literature

The following section presents results of applying the relevant evaluation criteria to the trauma literature: network design, coordination of trauma services and management of incentives, management of information, volume-outcome associations, and incident command. None of the included articles presented information relevant to inventory management or postponement and modularization.

Network Design. The design of a regionalized trauma system provides for the designation of centers with highly specialized trauma care capabilities, including highly specialized personnel, surgical capacity, and specialized resources and equipment in urban areas, with less advanced centers in rural areas. This designation of hospitals, and the number and levels of centers, varies for different geographic regions. To achieve improved quality of trauma care at decreased cost, a trauma system requires a carefully determined number of hospitals at each level in the system. The number and level of centers established within a given area depends on the number of trauma patients expected in that area, the size of the population, and the population density. One report estimated that “75 to 90 macro-trauma/EMS regions would be appropriate for the nation.”⁵⁷⁴ Eastman and colleagues estimated that one trauma center could serve 500,000 to two million people, based on an estimate of 0.5 to 1 major trauma patients per 1,000 people per year.⁵¹⁸ They further calculated that in an urban area, with a population density of 10,000 people per square mile, one center could serve people in a 5.6 mile radius, whereas in a suburban area, with a population density of 1,000 people per square mile, the center could serve people in a 17.8 mile radius.⁵¹⁸ Some authors have suggested that the number of Level I centers serving a particular geographic region be limited to enable Level I centers to achieve a certain level of experience. Implementation of this policy led to a decrease in Level I centers in Oregon's statewide trauma system from five to two.^{36, 37} However, we found no evidence that specifically evaluates variations of the number and level of trauma centers within a geographic region. Efforts to regionalize bioterrorism preparedness will need to consider the character of the area served and the population density.

Coordination of Trauma Services and Management of Incentives. The included articles highlight two key motivators of regionalization of trauma care: legislative mandates and funding and reimbursement issues.

Regionalization of trauma systems in the United States occurred primarily due to legislative mandates (such as the Emergency Medical Services Act in 1973 and the Federal Trauma Care Systems Planning and Development Act in 1990).⁵¹⁸ Despite these legislative mandates, the government has provided limited financial support for the development and support of trauma systems,³³ an oversight that has been strongly criticized.³³ Eastman and colleagues estimated that the reimbursement for trauma patients on average results in a loss to the hospital of 19% of costs per patient.⁵¹¹ Economic disincentives, including inadequate reimbursement for medical care, lack of governmental assistance, high operating costs, and high levels of uncompensated care, have led to closures of trauma centers.³³ One study reported that 95 hospitals had dropped their trauma designations.⁵⁸¹ Furthermore, the specialized personnel providing care at trauma centers are inadequately reimbursed.⁵¹¹

Funding and reimbursement issues will play an important role in decisions regarding regionalization of bioterrorism preparedness. At this time it is not clear how hospitals participating in these efforts will be reimbursed. The primary payers of trauma care reimbursement, including private insurance companies, Medicare, and Medicaid, will have to engage the federal government and hospitals in dialogue, so that the incentives for hospitals to provide specialized bioterrorism care are clearly understood.

Management of Information. A key feature of effective regionalized trauma care is the implementation and maintenance of trauma care data registries that facilitate the ongoing evaluation of regionalized performance, and the identification of areas for performance improvement.^{518, 531, 574} Hospitals participating in statewide trauma registries can regularly audit their performance and compare it with that of similar hospitals in the system. This quality improvement process enables hospitals to identify gaps in care and target areas for focus. Iterative quality improvement processes are also relevant to bioterrorism preparedness because bioterrorism response systems will need to regularly monitor and update their plans and protocols.

Volume-Outcome Associations. Regionalized trauma care systems achieve cost savings and improved outcomes by concentrating the treatment of severely ill patients at trauma centers.^33–38 The premise underlying this volume-outcome association is that limiting the number of tertiary care centers, according to community need, leads to increased experience in treatment of severely injured patients, and thus improved outcomes. Cost savings are achieved by limiting the provision of expensive medical services to selected centers. The Oregon Trauma System, considered a model for a regionalized statewide system, specifically limits the number of tertiary care centers to achieve this level of experience.³⁷ Nathens and Maier concluded that the “limited direct evidence coupled with the extensive indirect evidence…[showed that] experience improves outcomes and that volume plays a critical role in the accrual of experience” for trauma care centers.³⁸ In a review of the Portland Trauma System, Mullins and colleagues reported that patient survival increased when the number of severely injured patients treated at each tertiary care center exceeded 350 patients annually.³⁴ Guidelines from the American College of Surgeons provide information as to the number of patients in different injury categories that should be treated at Level I centers for these centers to be cost-effective and maintain proficiency.⁵⁰⁸

Efforts to regionalize bioterrorism preparedness should take into consideration the volume-outcome association. It may be that restricting the hospitals providing care to patients during a bioterrorism event will result in those hospitals having better protocols and more experience in providing such care.

Incident Command. An established and centralized system for command and control is a key factor of trauma care regionalization. Because responders are often in the field or other disparate locations, communication systems form the foundation for command and control in trauma care.

An emergency communication system, like an emergency 911 system,⁵¹⁵ enables emergency medical personnel to reach the site of the accident in a timely manner. Furthermore, a centralized communications network permits the coordination of care in the field. It also serves as a link between hospital personnel and field personnel, thereby enabling hospital personnel to be informed of the status of injured patients, and to prepare for their arrival.⁵³³ It can also be used in areas with more than one appropriate destination hospital to check for availability of beds and to triage patients to that hospital. This centralized communication system could play a similar key role in a bioterrorism response.

Lessons Learned From Regionalization of Trauma Care

The evidence we reviewed suggests that regionalization of trauma care has led to decreased morbidity and mortality among severely injured patients. The literature further implies that these improved outcomes may be due to the large volume of patients seen at specified trauma centers as a result of the regionalization process. This association is highly relevant for bioterrorism response planning since the availability of skilled, specialized care could impact the outcome of exposed patients. Five issues from the process of trauma care system regionalization might be considered during the planning of a regionalized bioterrorism response system.

1. Pre-event hospital designation contributes to lower costs and improved patient outcomes. The evidence from trauma care regionalization suggests that a key component of high quality, cost-effective care is limiting high-cost specialty care to specifically designated hospitals with increased experience in treating severely injured patients. A bioterrorism response system may benefit from the pre-event designation of hospitals. Such designation may allow facilities to ensure that they have all the resources and personnel they need for a bioterrorism response, ensure that emergency medical personnel know where to take exposed patients, and may allow the public to be directed to the appropriate hospitals in case of a bioterrorism event. Furthermore, a hospital designated prior to a bioterrorism event could have the various infection control equipment in place (e.g., negative pressure control rooms) which could be used during a bioterrorism response but which could also be used routinely for treatment of other communicable diseases. The designation of hospitals allows a clear understanding among all medical services personnel about the role of each hospital and the services it will provide. Such designation also allows planners to determine the number of hospitals at each level that will be required to serve a specific geographic area. Hospital designation is likely to be relatively easy for non-contagious diseases; however, hospitals may be hesitant to be designated as specialty care hospitals for more contagious diseases.
2. Formalized protocols for pre-hospital and hospital care contribute to improved patient outcomes. Formalized protocols in the regionalized trauma care system for patient triage and transportation of patients to appropriate trauma care centers are directly relevant to the field assessment and triage that will need to be conducted during a bioterrorism response. A regionalized bioterrorism response may benefit from similar protocols so that first responders know where and how to transport exposed patients in a timely manner so as to contain exposure and ensure rapid delivery of definitive care to the patients.
3. An established communication network is essential to the coordinated regionalized provision of trauma care. It is likely that an established communication network such as that used by trauma care systems could also be used during a bioterrorism event. The existing traumacommunications network enables emergency medical personnel to reach the site of an accident in a timely manner and permits ongoing communication between the field and the destination hospital. Such a system could play an equally important role during a bioterrorism response. It could serve as a centralized command and control system to assist in managing the flow of patients to regional network hospitals that have the capacity to care for them.
4. Correctly aligned incentives, particularly sufficient funding, are critical to retaining the participation of designated trauma hospitals. The included articles suggest that regionalized trauma care systems do not provide sufficient incentives for some hospitals to remain in the system. A bioterrorism response system is likely to benefit from clearly outlined mechanisms for reimbursement to hospitals designated to care for patients within a given regional network so that hospitals are less likely to refuse to participate or to drop their designations.
5. Data registries contribute to continuous evaluation and improvement of trauma systems. Ongoing evaluation of trauma care systems is important to identify areas for performance improvement. One source of data to conduct such evaluations is a standardized data registry. Such a registry allows comparison of one system's performance with similar systems, and facilitates identification of areas for improvement. A bioterrorism response system may benefit from periodic, region-wide evaluations. Such evaluations could be facilitated by the use of data registries.

Summary of the Evidence of Regionalization of Surveillance Systems for Bioterrorism

Our search identified 117 articles on surveillance systems. Of these, 32 articles, two government reports, and one Web site described regionalization of surveillance systems.^* None of the included articles specifically evaluated the effects of regionalization on surveillance processes. CDC has proposed a draft framework for evaluating syndromic surveillance systems with some components that cover aspects of regionalization; however, an application of this framework to an existing surveillance system has yet to be published.⁵⁸⁴ Few surveillance systems have been evaluated for their detection capabilities (e.g., sensitivity, specificity, or timeliness).⁴³⁵ Thus, several ongoing evaluative efforts have focused on optimizing surveillance algorithms and comparing surveillance data with other routinely collected data (e.g., correlating surveillance data with seasonal influenza outbreak data).⁷⁴ Some systems have been compared to simulated outbreaks in weapons of mass destruction exercises, but not with respect to issues of regionalization.⁵⁸⁵

Table 13 presents the evidence from 14 articles about 13 regionalized surveillance systems. Typically, surveillance requires three key processes: data collection, data analysis, and decision making/response. In the following section, we present evidence concerning the regionalization of each of these surveillance processes.

Table

Table 13. Syndromic Surveillance Systems.

Data Collection. Ideally, surveillance data should be collected from the majority of exposed or ill persons (or representative samples of these) in a geographical region.⁵⁸⁶ Current bioterrorism surveillance systems collect a variety of non-traditional data including emergency room,^{74, 588, 604} health clinics,^{585, 590} calls to poison control,⁵⁹¹ and calls to 911,⁶⁹ school absenteeism,^{67, 592} and pharmacy sales⁵⁹³ surveillance data sources. Thus, the geographical areas and patient populations under surveillance will differ dramatically depending on the surveillance data collected. For example, the Electronic Surveillance System for the Early Notification of Community-Based Epidemics (ESSENCE system) developed by the Department of Defense-Global Emerging Infections Systems collects International Classification of Diseases, 9^th Revision (ICD-9) data from military installations worldwide.⁵⁹⁴ This system has been modified to collect data from civilian facilities in the geographic region around Washington, DC to increase the population under surveillance.⁶⁷ The geographic area under surveillance also depends upon which health systems, hospitals, and clinics participate in a given surveillance effort. For example, a surveillance system in the Boston area that utilizes ambulatory data from the Harvard Vanguard Medical Associates is limited to the 5–10% of the population of Eastern Massachusetts covered by this medical group.⁵⁹⁰ In contrast, the Syndromic Surveillance Information Collection system,⁷³ a collaborative project between the University of Washington and the public health department of Seattle/King County, includes four health care systems in King County, and collects data from three emergency departments/urgent cares and nine primary care clinics in the area. While these sites were chosen initially out of “convenience,” they were later noted by the system's designers to represent good geographic dispersion with wide patient catchment areas and diverse populations.⁷³

Data Analysis. The objective of analyzing surveillance data is to determine whether the observed number of cases exceeds the expected number for a given region or time period. Regionalized analysis of surveillance data can occur in three ways. First, the manner in which data are pooled for analysis can serve to change the geographic region under surveillance. Data collection occurs locally, but data analysis typically integrates data from several local sources (e.g., clinicians, hospitals, and county health departments) to assess the likelihood of an outbreak within a larger geographic region. Second, data analysis can attempt to integrate a variety of data sources in a given geographic region (e.g., West Nile virus in sentinel flocks, clinician reports of flu-like illness, and work absenteeism data). Third, regionalized data analyses collapse data into various time periods (e.g., weekly analyses of the incidence of influenza-like syndrome may be a more sensitive marker of influenza than analyses using daily incidence).⁷⁴

All bioterrorism surveillance systems centralize analysis functions to some degree, and many have automated analysis processes. For example, ESSENCE creates 2,700 syndrome- and location-specific graphs each day; these are automatically analyzed for patterns that require additional investigation.⁵⁹⁵ A regionalized approach to data analysis offers four potential benefits. First, it may increase the ability to detect an outbreak over a larger geographic region. Second, it may improve the ability to identify aberrant trends if data received from one region can be compared to other regions. Third, it may increase cost efficiencies related both to scale (i.e., reduction of redundant analytic activities) and standardization (i.e., single system rather than multiple systems). Fourth, many local health departments do not have personnel with sufficient statistical and epidemiological training to perform the necessary data analysis.⁶⁵ Reliance on state or other regional resources may be the only feasible means for performing adequate data analysis. These arguments might suggest that a completely centralized approach to data analysis (i.e., a single, national analysis center) would be the most effective and efficient. However, as Mostashari and Hartman describe, such an approach has several important limitations.⁴³⁵ Centralized data analysis from hundreds of localities could lead to many alarms triggered by chance alone.⁴³⁵ Furthermore, centralized data analysis still requires substantial infrastructure investments at the local level to respond successfully to outbreaks.⁴³⁵ Finally, analysis at a single rather than multiple levels creates the possibility of a single point of catastrophic system failure.⁴³⁵

Decision-Making and Response. A key principle of public health surveillance is the use of surveillance data to take appropriate action to minimize resulting morbidity and mortality.⁵⁸⁶ Traditionally, this decision-making process occurs primarily at the local level. Ideally, astute clinicians report unusual disease patterns to their local public health departments, which then have jurisdiction to investigate the potential outbreak and disseminate information to public health officials at the state and national levels as appropriate. Clinicians, who typically receive little, if any, training in public health reporting, are not always familiar with standard reporting procedures. They may bypass their local public health department and report directly to the state health department or to CDC. As automated syndromic surveillance systems are implemented at regional and national levels, this issue of where data reporting and decision-making occurs will become increasingly important. While a number of surveillance systems, such as the ones operating in Santa Clara County⁷⁴ and Seattle/King County⁷³ perform decision-making processes primarily at the local level, others, such as the systems operating in New Mexico⁷⁶ and Connecticut⁶⁴ perform these processes at the state level. Other larger regional systems allow for decision-making at multiple levels. For example, the National Bioterrorism Syndromic Surveillance Demonstration Project, a collaborative project between CDC, American Association of Health Plans, Harvard Medical School, and five health plans based in Massachusetts, Minnesota, Colorado, and Texas, will cover more than 20 million individuals from all 50 states.⁶⁸ Data will be analyzed at a centralized data center and decisions on new cluster identification will be performed centrally, but identification of an abnormal cluster of events will generate an alarm that will be automatically sent to the local health department in the affected area for further decision making and response.⁶⁸

Given the lack of experience interpreting syndromic surveillance data, response protocols have yet to be fully defined. In general, the public health officials responsible for analysis of surveillance data are also responsible for the initial outbreak investigation. However, questions remain as to who owns these surveillance systems. With ownership comes responsibility and accountability for decisions made based on surveillance data and their analyses. As some authors have previously noted, many organizations are key stakeholders of surveillance systems, including local, state and national public health departments; academic centers; and commercial entities.⁴³⁵ While local public health officials often have the legal mandate for decision making and response, expertise and resources are often concentrated at the state and national public health level or in the academic and commercial sectors.⁴³⁵ Successful surveillance will likely require a collaborative effort between key stakeholders, but ultimate ownership still remains to be determined.

Application of Evaluation Criteria to the Surveillance Literature

This section discusses the results of applying the relevant evaluation criteria to the surveillance literature: network design, coordination of surveillance and management of incentives, management of information, volume-outcome associations, and incident command. None of the included articles presented information relevant to inventory management or postponement and modularization.

Network Design. The concept of network design as applied to surveillance systems describes how the structure of a surveillance system affects its detection characteristics. An ideal surveillance network will not only detect aberrant disease patterns but will do so in an efficient and timely manner. Our review highlighted four key network design issues affecting implementation and evaluation of regionalized surveillance systems.

First, the geographic range of a surveillance system should be sufficiently far-reaching to enable detection of an outbreak that may be spread over a large geographic area. For example, detection of a bioterrorism event occurring in the New York City subway system requires a surveillance system with detection capability not only in Manhattan and other boroughs of New York City, but also in the outlying suburban regions of New York State, New Jersey, and Connecticut. Currently, many surveillance systems are designed along county or state lines. As the recent outbreaks of SARS and monkey pox have shown, disease is not limited to these politically-oriented geographic divisions. Accordingly, it is likely that current artificial geographic boundaries will have to be overcome if surveillance systems are to be maximally effective.

Second, a key objective in analyzing surveillance data is to maximize detection of the signal to noise ratio. Research is currently ongoing to assess how best to divide a surveillance region into subregions so as to maximize the ability to detect abnormal events. For example, ESSENCE uses SaTScan—free software developed with federal funds to detect geographic trends in cancer incidence—to facilitate detection of abnormal clusters within surveillance regions.⁵⁹² Other systems perform similar analyses.^{596, 597} Determining the optimum size and location of data collection regions is critical for effective surveillance. A region that is too small may result in missed cases and unacceptable data variability due to the small sample size. A region that is too large may result in reduced ability to detect small variations due to increased scale.⁵⁹² It is often not possible to perfectly subdivide geographic regions given pre-existing constraints on supplied data. This can be particularly problematic when combining surveillance data from civilian hospitals within a given geographic region with data from local military or VA facilities, which tend to have patients from significantly broader geographic areas. Interpretation of these surveillance data may be enhanced by temporospatial analyses.

Third, the literature provides no consensus about how much data must be collected to have a representative sample of the population under surveillance. Evidence suggests that a surveillance system need not cover the entire population to be able to detect an outbreak, and that coverage as low as 5–10% of the population surveyed may be clinically useful.⁵⁹⁰ Additionally, a system that collects surveillance data from only a representative segment of the population may be more cost-effective than collecting data on the entire population in a geographic region, as has been proposed.⁴³⁵ It is unknown what the ideal distribution of surveillance collections sites is for a cost-effective bioterrorism surveillance system.

Fourth, an ideal surveillance network incorporates redundancy so as to avoid system failure at any process step. Redundancy in surveillance systems can occur in a number of ways; two are highlighted here. First, while an effective surveillance system will have centralized analysis functions, some (limited) analysis could also be performed by other entities within the system, or even by entities outside of the surveillance system. Such overlap will help prevent detection failure due to incorrect analytic methodology or data types, as may have occurred during the drop-in surveillance in Japan during the 2000 G8 meeting when the system missed an outbreak of Vibrio parahemolyticus among policemen who ate prepared lunches.⁵⁹⁸ Second, an effective system will likely include a large number of potentially redundant data sources to increase detection capability. An evaluation of ESSENCE performance during a simulated bioterrorist outbreak using four data sets (ER visits, office visits, over-the-counter influenza medication sales, and school absenteeism) showed improved detection capability when all four data types were used for analysis rather than just ER visits or school absenteeism alone.⁶⁷

Coordination of Surveillance System and Management of Incentives. A clear understanding and alignment of incentives of key stakeholders involved in each of the surveillance processes may facilitate bioterrorism surveillance. Key stakeholders involved in surveillance systems include: data collectors, data analysts, and decisionmakers.

Data Collectors. Data collectors for surveillance systems include: those providing traditional clinical surveillance data such as outpatient clinicians, emergency department personnel, and clinical laboratories, as well as non-traditional data providers such as schools (providing absenteeism data), animal control personnel, and retail pharmacies. Providing incentives to data collectors is critical and difficult for three reasons. First, a surveillance system is only as good as its data. Second, many data collectors have few available resources to devote to surveillance, as they are already overburdened by their routine workloads. In an implementation of the Real-time Outbreak Detection System (RODS) system in the Pittsburgh area, local health departments requested that the RODS laboratory monitor output of the surveillance system because they did not have sufficient resources.⁶⁵ Third, data collectors may have the least to gain from implementing surveillance systems. Some may be fulfilling legally mandated requirements, while others satisfy a sense of civic responsibility, but data collectors may not perceive any other benefits. Indeed, many institutions may have disincentives to providing data. A hospital that provides data that sets off an alarm may be closed until further investigation; a pharmacy that provides over-the-counter sales data that becomes public might be at a competitive disadvantage; and a school that reports increased absenteeism may lose funding due to high absenteeism rates.

Chavin and Valleron studied the motivation of 560 French general practitioners for participating in a public health surveillance network.⁵⁹⁹ These clinicians reported an interest in contributing to the public health (39.5%), a scientific interest in epidemiology (24.8%) or in receiving epidemiologic feedback (17%), and receiving gratification for their supplementary role (10.5%) as their primary motivations for contributing to the surveillance network.⁵⁹⁹ Efforts to motivate data collectors, provide them with timely feedback, and emphasize “dual use” characteristics of the data or their analyses may lead to improved collection of surveillance data. In general, designers of surveillance systems have recognized the need to reduce the burden on data collectors. One goal of the National Bioterrorism Syndromic Surveillance Demonstration Project will be to utilize data that is already routinely collected as part of daily operations, so as to generate minimal incremental costs and resources.⁶⁸ Economic benefits, ranging from resources to help implement such surveillance systems, to tax benefits for those who participate, may provide additional incentives to data collectors.

Public Health Data Analysts and Decisionmakers. The incentives for public health officials to implement and maintain bioterrorism surveillance systems include: their legal mandate and responsibility for the protection of public health, resources made available for these activities, and the potential costs of inaction. However, given the lack of evidence demonstrating effectiveness of surveillance systems, some public health departments have chosen to allocate bioterrorism preparedness resources away from surveillance programs and toward clinician education, the purchase of decontamination equipment, and other preparedness programs. Thus, a priority in incentivizing public health decisionmakers to deploy and maintain surveillance systems is to demonstrate that surveillance systems can fulfill their stated purpose. While this may be difficult, particularly in the absence of a large-scale bioterrorism event, effectiveness in simulated scenarios and capability to detect other naturally occurring outbreaks may lend credence to such systems. In addition, while legal authority to investigate outbreaks is often given at the state and local level, funding for surveillance systems is increasingly provided at the federal level. Should a national surveillance system be implemented, a disincentive might arise, particularly at the local level, if such a system bypasses the traditional reporting and decision-making hierarchy and leaves local health officials effectively out of the loop. Because local public health officials often have limited resources and personnel to devote to surveillance, they might prefer that analysis be performed by other entities.⁶⁵ Therefore, a careful balance must be achieved between not over-burdening local health officials with surveillance activities while keeping them intimately involved in the surveillance process. Finally, to the extent that public health decisionmakers are motivated by the cost of inaction, it may be useful to convey the potentially high cost of inaction during a bioterrorist attack. Conservative estimates of the cost of a bioterrorist attack range from $478 million per 100,000 persons exposed to brucellosis to $26.2 billion per 100,000 persons exposed to anthrax.⁶⁰⁰

Volume-Outcome Associations. An increased volume of data collection and analysis should improve surveillance to the extent that it results in systems with more rigorous data sets, smaller confidence intervals, and analysts with improved capacity for recognizing abnormal disease clusters. This reasoning has contributed to the expansion and centralization of surveillance networks. For example, increasing data volume has allowed ESSENCE investigators to refine their surveillance algorithms and improve detection capability.⁶⁷ However, increased volume does not always ensure improved outcomes. Increased volume will necessitate increased resources for data collection and analysis. If routinely collected data are used, the incremental costs may be small, but substantial investments still must be made in computing infrastructure and personnel training as a system increases in size and complexity. Such investments may be barriers to implementing a high-volume system. In addition, high volume does not necessarily ensure good outcomes if there is not a concomitant emphasis on quality. Without quality controls in surveillance processes, an increase in volume will only result in propagation of errors. Therefore, evaluation of a surveillance system's volume and outcomes must also include an assessment of incremental costs and overall quality.

Incident Command. Effective surveillance systems will almost certainly facilitate incident command during bioterrorism responses. Identifying which entity has ultimate decision-making capability and responsibility is often complicated by the many stakeholders involved in the surveillance process. Ideally, the structure of public health response should be determined before an outbreak and should depend upon the severity of outbreak. Drop-in surveillance efforts provide good examples of how incident command and response protocols can be established before an event occurs. For example, the surveillance program developed for the 2000 Olympic Games in Sydney, Australia established a Health Olympic Coordinating Centre which received input from the regional Department of Health, metropolitan Sydney public health units, area sentinel hospitals, and other health-related facilities.⁶⁰¹ This center assessed health surveillance information and coordinated data analysis.⁶⁰¹ Participation in this surveillance effort resulted in strong ties between the regional health department, local public health units, local governments, and other area health facilities and providers.⁶⁰² Perhaps more importantly, this program led to an improved understanding of “respective roles and functions” in surveillance preparedness, a key benefit that will help in responding to potential future outbreaks.⁶⁰²

Management of Information. Surveillance systems must transfer and manage information between participants within and outside of the system. As surveillance networks expand to include participants with different computing infrastructures, management of information will likely become increasingly complex. This may involve integrating different computing systems and data streams, while allowing information and analysis to flow upwards and downwards within the system to appropriate end-users. Several barriers impede efficient management of information, including lack of national reporting standards, lack of conversion and translation systems for communication between systems, and legal requirements regarding patient confidentiality. CDC has initiated the National Electronic Disease Surveillance System partly to address the lack of standardization among surveillance systems.⁵⁸⁷ This program focuses on developing national standards for computing infrastructure and reporting for surveillance systems. Other initiatives, like the Frontlines of Medicine Project, aim to develop non-proprietary, “open” standards for communication between emergency rooms.⁵⁸⁷ The development of multiple industry standards may be avoided by increased collaboration among major systems developers, as has occurred in the National Bioterrorism Syndromic Surveillance Demonstration Program.⁶⁸ Integration of new participants into surveillance programs is often hampered by differing computing platforms. In the Syndromic Surveillance Information Collection System⁷³ in Seattle, many of the information technology groups involved in the network developed their own reporting and transfer strategies.⁶⁵ In response, investigators had to develop central conversion capabilities. Such translation capabilities will become increasingly important as surveillance systems expand to include data collectors who may have legacy computing platforms. Finally, new privacy regulations from the Health Insurance Portability and Accountability Act (HIPAA) require additional measures to be undertaken to ensure patient confidentiality.⁶⁰³ Existing systems that lack standardized confidentiality and security procedures may violate HIPAA standards. One of the National Electronic Disease Surveillance System's goals is to establish standardized security and confidentiality protocols.⁶⁰³ Surveillance systems have evolved as well. Several state-of-the-art systems strip away individual-level information, utilizing these data only when an abnormal disease pattern has been identified.⁶⁸

Lessons Learned From Regionalization of Surveillance Systems

Regionalization of surveillance systems has important implications for key processes related to data collection, analysis, and decision making and response.

1. There is a need for evaluations of regionalization of surveillance data collection and analysis. Although numerous syndromic surveillance efforts in local areas are promising, there has not been an evaluation of the tradeoffs in terms of costs and benefits of regionalizing some components of bioterrorism surveillance (e.g., costs of data collection, analysis and reporting, false positive and false negative rates). Given this lack of evaluative data, we developed a simulation model to evaluate some of these costs and benefits. This model is presented in the next section.
2. Local data collectors may not be incentivized to collect surveillance data, particularly if they are analyzed regionally. If a regionalized surveillance system is based on a model of local data collection with regional analysis, considerations of means to reduce costs of data collection and to share relevant analyses with local data collectors may enhance local participation.
3. Decision-making and chain of command must be clear to all stakeholders. Decision-making and response could occur at the local, state, regional, or national level, and would depend on outbreak characteristics and the level of response needed.
4. A common technology platform may facilitate the collection and analysis of surveillance data. This common platform may be composed of new technologic infrastructure or translation tools that enable existing legacy systems to communicate with one another.
5. Privacy issues are a key concern as increasingly detailed surveillance data is collected and disseminated. Collection and analysis of surveillance data must be able to protect individual privacy while being sufficiently detailed to detect new outbreaks.

Changing Thresholds When Pooling Surveillance Data

To illustrate the general tradeoffs that occur when analysis of surveillance data is regionalized, we simulated a syndromic surveillance system that collects daily reports of a syndrome of interest from two regions (i.e., cases of flu-like illness). Analysis of these surveillance data could be unpooled (in which case reports from the two regions are analyzed separately) or pooled (in which case data from the two regions are combined before analysis). We assumed that if the number of cases of flu-like illness exceeds a threshold, then a warning will be issued. Since the number of patients with these symptoms varies from day to day, there is a chance that normal variation will cause the observed number of cases to exceed the warning threshold even when there is no outbreak resulting from bioterrorism. We refer to such a situation as a false positive (i.e., a false alarm). We assume that the thresholds are set so that the probability of a false positive does not exceed a level that represents a decisionmaker's preferences for false positives.

Figure 8 illustrates the detection thresholds for two regions. The number of cases of flu-like illness for Region 1 is plotted on the x-axis and the number of cases for Region 2 is plotted on the y-axis. For each of the two regions we illustrate the threshold above which a warning will be issued indicating an unexpected peak in cases of flu-like illness. The entire shaded area in Figure 8 represents the numbers of case reports from the regions that would cause a warning to be issued.

Figure

Figure 8. Thresholds for Detection of Outbreaks in Two Regions.

When data from two regions are pooled, the warning thresholds may need to be modified to prevent an increase in the probability of false positives. This situation is illustrated in Figure 9. A simple way to pool data is to calculate the sum of cases in both regions. Figure 9 shows a diagonal line that represents a threshold for the pooled data. The space between the diagonal line and the individual thresholds (indicated by an asterisk) represents case reports in the two regions for which a warning would not be issued if data from the two regions were analyzed individually (unpooled) but a warning would be issued it the data were pooled (using the sum). If the thresholds for the two regions are not adjusted, then the use of pooled data will increase not only the probability of detecting an outbreak associated with bioterrorism but also the probability of a false positive. The overall probability of a false positive can be held constant by increasing the values of the individual thresholds. In Figure 9, the thresholds for Regions 1 and 2 have been increased so that the probability of exceeding either threshold decreases by an amount that offsets the increase in probability associated with using pooled data.

Figure

Figure 9. Illustration of Pooled and Individual Thresholds for Detection of Outbreaks in Two Regions. Unpooled thresholds are denoted by the solid lines. Pooled thresholds are denoted by the broken lines. *The area indicated with an asterisk indicates (more...)

The Effect of Inter-Regional Correlation on Pooled Surveillance Data

To investigate tradeoffs that occur when pooling surveillance data, we simulated data representing a bioterrorism attack in two regions that collect syndromic surveillance data. We simulated the daily number of cases for a syndrome of interest (i.e., fever and rash) representing both naturally occurring cases and an increase in the number of cases associated with bioterrorism. The baseline number of cases for each region was normally distributed with a mean of 100 cases per day and standard deviation of 100 cases per day. We varied the attack size so that between 0 and 350 additional patients with fever and rash presented per region per day. We set the threshold for detecting an outbreak as the total number of cases more than two standard deviations above the daily average.

We simulated correlations that may exist in surveillance data collected from two sources eligible for pooling. The correlation between two regions describes the extent to which the number of cases in each region is similar on a day-to-day basis. The correlation varies between 1 and -1. A correlation of 1 indicates a perfect positive relationship between the number of cases in each region and a correlation of -1 indicates a perfect negative relationship between the number of cases in each region. If the inter-regional correlation was 1, an increase of one case in Region 1 would always correspond with an increase of one case in Region 2. A correlation of zero would indicate that there is no relationship between the numbers of cases in the two regions. The correlation between two sources of surveillance data may vary according to the geographic proximity of the two regions, the types of surveillance data collected (e.g., one region may report cases of patients presenting to triage nurses with fever and rash as opposed to cases of fever and rash collected from ICD-9 codes from outpatient clinics), and the patient populations under surveillance (e.g., pediatric as opposed to nursing home patients). We varied the correlation between the two regions between 0 and 0.5.

We compared the probability of detecting an outbreak of cases of fever and rash when using pooled data as opposed to using individual data. We found that pooling surveillance data may yield modest benefits in the probability of detecting an outbreak, particularly when there is no correlation between the regions. The benefits of pooling are smallest when attack sizes are very large or very small. We found that the rate of false alarms increases as the correlation between the pooled data sources increases. We evaluated a number of different strategies for pooling data and determining thresholds. For all strategies that we considered, all large attacks were readily detected. However, pooling did yield some benefit for smaller attacks.

Lessons Learned for Regionalization of Analysis of Surveillance Data

Our simulations demonstrate two primary effects of pooling surveillance data.

1. Large outbreaks are relatively easy to detect when using either pooled or unpooled data. However, smaller outbreaks can be difficult to detect under both strategies. In part, it is difficult to detect small outbreaks because the standard deviation about the mean is relatively large compared to small increases in numbers of cases. When the pooled threshold is increased to avoid increasing the false positive rate, small outbreaks can be missed.
2. Pooling strategies may improve detection capabilities but are highly situation-specific. Pooling of surveillance data enhanced detection capabilities when the correlation between the individual regions was very low. We plan addition analyses to determine the effects of pooling over time and extending our analyses to pool more than two regions.

Summary of Inventory Management Simulation Model

We simulated a bioterrorism attack with Bacillus anthracis on a metropolitan area with a population of 5 million. Selection of parameter values was informed by our review of the U.S. experience with inhalational anthrax.¹⁵² We assumed that exposed individuals enter the incubation phase of the disease and may then progress to first stage anthrax, second stage anthrax, or death.^{152, 605, 606} We defined first stage anthrax as symptomatic disease with a nonspecific flu-like syndrome consisting of low-grade fevers, nonproductive cough, myalgias, and malaise (sometimes referred to as the “latent” phase of the disease).^{152, 605, 606} We defined second stage anthrax is severe symptomatic disease characterized by abrupt respiratory distress, shock, and death within 24 hours (sometimes called the “acute” phase of the disease).^{152, 605, 606} We assumed no anthrax-related deaths would occur except among individuals in the second stage of infection.

We assumed that all exposed individuals were initially unaware of the attack and that exposed individuals might become aware of the attack either through development of symptoms or through the media or public health alerts. After becoming aware of the attack, individuals who are aware of their potential exposure may seek post-exposure prophylaxis and enter a queue for prophylactic antibiotics.⁶⁰⁷ For exposed individuals, we assumed that the rate of disease progression (from incubation to first stage) is reduced by prophylactic antibiotics and that prophylactic antibiotics are distributed on a first-come, first-served basis.^{607, 608} We assumed that the rate at which individuals receive prophylactic antibiotics is limited by the availability of prophylactic antibiotics and the distribution capacity for these antibiotics. We based estimates of adherence to prophylactic antibiotic regimens on the 2001 experience.^{152, 206, 605}

Individuals who develop symptoms associated with first or second stage anthrax may enter a queue for treatment. Treatment consists of a triple antibiotic regimen (ciprofloxacin or doxycycline with rifampin and clindamycin) administered intravenously in an intensive care setting and may also include supportive care (e.g., pleural fluid drainage, respiratory or cardiac support, etc.).^{* 607, 611} We assumed that treatment would be restricted to symptomatic individuals. We assumed that the ability to treat patients is limited by the inventory of intravenous antibiotics, the number of available intensive care unit beds, the number of available ventilators, and the number of available respiratory technicians. We assumed that when patients begin treatment a small supply of antibiotics is reserved for them so that there is little chance that they will need to leave treatment due to a lack of antibiotics.

We considered scenarios that differ in the number of people exposed to anthrax. In the low-exposure scenario (such as an aerosol release over a sports stadium), we assumed that 50,000 people were exposed.^{605, 612} In the high-exposure scenario (such as an aerosol release from an airplane over the downtown area of the city), we assumed that 250,000 people were exposed.^{605, 612} We also considered scenarios that differ in the number of people likely to require prophylactic antibiotics. If an attack occurs in such a way that responders can determine whether or not a person was likely to have been exposed, the number of people requiring prophylaxis will be lower than if an attack occurs such that responders cannot make this determination. In the latter scenario, we assumed that only exposed people would request prophylaxis; whereas, in the former, we assumed that both exposed and non-exposed people would request prophylaxis, resulting in inefficiently allocated resources, and slowed queues.

We assumed that responders would first use the local inventories of antibiotics (if available), then Push Packs from the Strategic National Stockpile, then regional Vendor Managed Inventories or additional Push Packs.^{613, 614} We based our estimates of the antibiotic inventories of local communities on a survey of ten hospitals in New Jersey¹³ and other reports.^{† 201, 616–619} We based our estimates of the types and numbers of individuals needed to operate and staff a mass prophylaxis distribution center from the Weill-Cornell Mass Prophylaxis/Vaccination model.⁶²⁰

Summary of Preliminary Results of an Inventory Management Simulation Model

We found that mortality associated with anthrax bioterrorism in our simulation is very sensitive to the number of people seeking prophylactic antibiotics. For both the low and high exposure attacks, we found that approximately 30% of the exposed population died given a low demand for prophylaxis (i.e., the situation in which exposure can be readily determined) compared to approximately 45% of the exposed population given high demand for prophylaxis (i.e., the situation in which exposure cannot be readily determined so many unexposed persons seek prophylaxis). This difference represents an increase in mortality of approximately 8000 exposed individuals for the low exposure attack and 35,000 exposed individuals for the high exposure attack attributed to high demand for prophylactic antibiotics.

We compared the mortality that resulted if only a single Push Pack was delivered to the site of attack (and then the local responders had to wait until the regional Vendor Managed Inventory arrived) compared with an alternative strategy in which a Push Pack was delivered every six hours following the arrival of the first Push Pack, until either all 12 Push Packs have been delivered or the regional Vendor Managed Inventory becomes available. We refer to these as Single Push Pack and Multiple Push Pack strategies. We found that the Multiple Push Pack strategy has little benefit for a low exposure attack but can have a significant benefit for the high exposure attack (i.e., 1–2% reduction in mortality compared with the Single Push Pack strategy).

We evaluated whether increasing or decreasing the number of Push Packs would change the delay in delivery of a Push Pack to an attack site. There are currently 12 Push Packs in the United States that are reportedly able to arrive at the site of an attack within 12 hours of a request.⁶²¹ The locations of these Push Packs are not publicly available. If we assume that the Push Packs are geographically distributed across the United States, then every location in the United States could be reached with a maximum of approximately one hour of flight time. We found that our model was much less sensitive to increasing or decreasing the number of Push Packs than it was to increasing or decreasing the time it takes to dispense the contents of the Push Pack. However, for a high exposure attack, an additional Push Pack may have some benefit when the Multiple Push Pack strategy is used. The risk of a bioterrorism event may be proportional to population density. Thus, we plan to evaluate the effects of changing the number of Push Packs, if the Push Packs are initially distributed according to population density.

We evaluated the cost effectiveness of changing the size of the local inventory of antibiotics. Increasing the local stockpile would require a one-time cost to purchase the supplies, plus annual costs associated with maintaining the inventory. Because the net present value of all costs and health benefits is a function of the probability of an attack, we estimated the cost effectiveness of increasing the size of the local inventory if the annual probability of an attack ranges between 0.01% to 1%. We found that the cost effectiveness of increasing the size of the local stockpile is highly dependent on the annual probability of an attack. However, the total costs are not very sensitive to the probability of an attack. This is because the annual maintenance costs of the additional stock accounts for 80–85% of the total costs, and these costs are completely insensitive to the probability of an attack, as they must be incurred every year regardless of whether an attack takes place.

Lessons Learned from the Inventory Management Simulation Model

We draw four lessons from the preliminary results of our inventory management simulation model.

1. The mortality associated with anthrax bioterrorism may be highly sensitive to the number of people seeking prophylactic antibiotics. Whereas responders and the public may be able to estimate the probability of exposure for an attack in a localized area (e.g., an attack in a particular building), establishing whether an individual has been exposed to a biothreat agent may be difficult for many types of bioterrorism events (e.g., an aerosolized attack over a metropolitan area).
2. For a large-scale bioterrorism event, delivering multiple Push Packs until delivery of the regional Vendor Managed Inventory may reduce mortality. In our preliminary analyses, we explored two strategies for delivering Push Packs to an affected area and found that a strategy of deploying a second Push Pack before the first one is depleted may save lives during a large-scale bioterrorism event. In our subsequent analyses, we plan to evaluate other strategies for dispensing antibiotics on the basis of the on-hand inventory (e.g., dispensing only short courses of prophylactic antibiotics if the on-hand inventory is low or if the demand for antibiotics is high).
3. Assuming uniform geographic distribution of the Push Packs, we found no significant change in the time required to deliver a Push Pack to an attack site with changes in total Push Pack numbers. However, the risk of a bioterrorism event may be higher in areas of the greatest population density. In future analysis, we plan to evaluate the impact of changes in the number of Push Packs if they are geographically distributed according to population density.
4. Increasing the size of the local inventories of antibiotics may be cost effective if the annual probability of an attack is high. The cost effectiveness of increasing the size of local inventories of antibiotics was sensitive to both the probability of an attack and the costs of the inventory.