BIOSECURITY AND BIOTERRORISM: BIODEFENSE STRATEGY, PRACTICE, AND SCIENCE Volume 4, Number 3, 2006 © Mary Ann Liebert, Inc.
Reducing Mortality from Anthrax Bioterrorism: Strategies for Stockpiling and Dispensing Medical and Pharmaceutical Supplies DENA M. BRAVATA, GREGORY S. ZARIC, JON-ERIK C. HOLTY, MARGARET L. BRANDEAU, EMILEE R. WILHELM, KATHRYN M. McDONALD, and DOUGLAS K. OWENS
A critical question in planning a response to bioterrorism is how antibiotics and medical supplies should be stockpiled and dispensed. The objective of this work was to evaluate the costs and benefits of alternative strategies for maintaining and dispensing local and regional inventories of antibiotics and medical supplies for responses to anthrax bioterrorism. We modeled the regional and local supply chain for antibiotics and medical supplies as well as local dispensing capacity. We found that mortality was highly dependent on the local dispensing capacity, the number of individuals requiring prophylaxis, adherence to prophylactic antibiotics, and delays in attack detection. For an attack exposing 250,000 people and requiring the prophylaxis of 5 million people, expected mortality fell from 243,000 to 145,000 as the dispensing capacity increased from 14,000 to 420,000 individuals per day. At low dispensing capacities (14,000 individuals per day), nearly all exposed individuals died, regardless of the rate of adherence to prophylaxis, delays in attack detection, or availability of local inventories. No benefit was achieved by doubling local inventories at low dispensing capacities; however, at higher dispensing capacities, the cost-effectiveness of doubling local inventories fell from $100,000 to $20,000/life year gained as the annual probability of an attack increased from 0.0002 to 0.001. We conclude that because of the reportedly rapid availability of regional inventories, the critical determinant of mortality following anthrax bioterrorism is local dispensing capacity. Bioterrorism preparedness efforts directed at improving local dispensing capacity are required before benefits can be reaped from enhancing local inventories.
T
2001 demonstrated the vulnerability of the U.S. to bioterrorism. Reviews of biodefense planning and tabletop preparedness exercises have highlighted the lack of consensus regarding two key aspects of bioterrorism preparedness planning: the level of inventory of medical and pharmaceutical supplies that should be held locally versus regionally and the necessary capacity for rapidly dispensing these supplies to an exposed population.1,2 HE ANTHRAX ATTACKS OF
During a response to bioterrorism, two primary inventories of medical and pharmaceutical supplies could be used: (1) local inventories and (2) the Strategic National Stockpile (SNS), a national inventory of antibiotics, chemical antidotes, and ventilators and other medical and surgical supplies developed to supplement local inventories.3,4 No national consensus exists regarding appropriate levels of local inventories.1,5 The 2002 U.S. Medicine Institute for Health Studies Forum on Surge Capacity
Dena M. Bravata, MD, MS, Jon-Erik C. Holty, MD, Emilee R. Wilhelm, Kathryn M. McDonald, MM, and Douglas K. Owens, MD, MS, are with the Center for Primary Care and Outcomes Research, Stanford University School of Medicine, and the StanfordUCSF Evidence-based Practice Center, Stanford, California. Jon-Erik C. Holty and Douglas K. Owens are also with the VA Palo Alto Health Care System, Palo Alto, CA. Gregory S. Zaric, PhD, is with the Ivey School of Business, University of Western Ontario; and Margaret L. Brandeau, PhD, is with the Department of Management Science and Engineering, Stanford University. 244
STOCKPILING AND DISPENSING SUPPLIES FOR BIOTERRORISM RESPONSES
recommended that hospitals have a 48-hour supply of antibiotics to treat or prophylax staff and other first responders, but it did not recommend that individual communities stockpile pharmaceuticals.5 In contrast, a 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 recommended “if financially feasible, some local stockpiling of certain antidotes, like antimicrobials.”1 These conflicting recommendations reflect the great variability across the country in the types and quantities of medical and pharmaceutical supplies held in local inventories.6,7 In addition to this lack of consensus regarding local inventories, considerable uncertainty exists concerning the capacity that communities should have for inventory distribution (i.e., management of the inventory up to the point of dispensing) and dispensing (i.e., provision of the supplies to end users such as patients or medical personnel).8 To receive bioterrorism preparedness and response funding, the Department of Health and Human Services requires that states prepare detailed plans for establishing dispensing centers. However, little published evidence is available to guide these planning efforts.9,10 Similarly, since 2004, the Centers for Disease Control and Prevention have provided funding to an expanding number of U.S. municipalities through the Cities Readiness Initiative Pilot Program.11 This funding is intended to facilitate the development of plans and infrastructure required for participating cities to be able to provide oral antibiotics to their entire populations with 48 hours. However, many of these plans remain in the early development phase and have yet to be evaluated. We developed a simulation model to evaluate the costs and benefits of various strategies for detecting and responding to a bioterrorism attack involving aerosolized anthrax. Specifically, we used the model to evaluate the costs, health benefits, and cost-effectiveness of four types of strategies to decrease anthrax-associated morbidity and mortality: strategies that enhance bioterrorism event detection, increase local dispensing capacity, increase local inventories of antibiotics, and increase the national inventories deployed to the site of an attack.
METHODS We modeled the local and regional supply chain for medical and pharmaceutical supplies required for a response to anthrax bioterrorism and the expected morbidity and mortality for a population exposed to anthrax. The model, implemented in Excel, is a dynamic compartmental model that distinguishes individuals by exposure status (not exposed, potentially exposed, and
245
exposed), awareness of an attack (unaware of attack, aware of attack), anthrax disease stage (asymptomatic incubation, prodromal, and fulminant), and treatment status (receiving no prophylaxis or treatment, receiving prophylaxis, and receiving in-hospital treatment). Individuals transition between compartments according to their attack awareness, disease progression (including, in some cases, death), and receipt of prophylaxis or treatment. Full details of the model are provided elsewhere.12 Figure 1 shows the four detection, distribution, and dispensing strategies that we analyzed. Table 1 summarizes the base case parameter values we used in our model.
Attack Scenario Our simulation evaluated the effects of a hypothetical aerosol release of Bacillus anthracis spores in a metropolitan area with a population of 5 million people. In the U.S., eight cities have metropolitan areas of 5 million or more people (e.g., New York, Chicago) and four cities have metropolitan areas of 4 million to 5 million people (e.g., Houston, Atlanta).13 We varied the number of people who were exposed from 50,000 to 250,000. The smaller attack scenario represents what might occur if an aerosol release occurred at a large sports stadium or entertainment venue. Because the number of people who are exposed during an aerosol release of anthrax spores depends on several factors—including the number of spores released, the wind speed and direction, and population density—we assumed that a larger attack would result in 250,000 people becoming infected. This value is within the range of attack scenarios used by others who have modeled emergency responses to anthrax bioterrorism.14,15 Under some attack scenarios (e.g., an aerosol release over a poorly defined geographic area), residents of the affected area may not be able to readily determine whether they have been exposed. Thus, we varied the number of individuals seeking prophylactic antibiotics from 100,000 to 5 million (i.e., 2% to 95% of the unexposed population).
Disease Progression Our disease progression model is based on a comprehensive review of the literature of 82 historical cases of inhalational anthrax, including 11 cases resulting from the 2001 U.S. anthrax attacks.16 We assumed that exposed individuals could move progressively through three anthrax disease stages: asymptomatic incubation stage, prodromal stage, and fulminant stage. Prodromal stage anthrax (sometimes referred to as latent stage) is characterized by a nonspecific flulike syndrome consisting of fever, nonproductive cough, shortness of breath,
Average time, prodromal to fulminant Average time, fulminant to death, prophylaxis started in prodromal staged Average time, fulminant to death, prophylaxis started in fulminant staged Individuals in Treatment Probability of progression, prodromal to fulminant Mortality rate, fulminant Average time, prodromal to fulminant Average time, fulminant to death, treatment started in prodromald
Average time, prodromal to fulminant Average time, fulminant to death Individuals Receiving Prophylaxis Adherence rate for prophylaxis Probability of developing symptoms if adherent when prophylaxis is given during asymptomatic incubation stage Probability of developing symptoms if nonadherent when prophylaxis is given during asymptomatic incubation stage Probability of progression, prodromal to fulminant, for prophylaxis given during prodromal stage Mortality rate, fulminant
Anthrax Disease Progression Individuals Receiving No Prophylaxis or Treatment Median incubation periodb Probability of progression, prodromal to fulminant Mortality rate, fulminant
Parameter
FOR THE
SIMULATION MODEL
B(0.75, 0.13)
B(0.05, 0.03)
B(0.88, 0.06)
B(0.80, 0.08) Not varied in probabilistic analysis U(4.1, 6.1) U(0.8, 1.2) U(1.3, 1.9) B(0.14, 0.01) B(0.98, 0.01) U(4.08, 6.12) U(0.8, 1.2)
0%
100%
80% 100% 5.1 days 1.0 day 1.6 days 14% 100% 5.1 days 1.0 day
16
16 16, 49 16
15, 16
15, 16
15, 16
16 16
Estimated
17, 48
18
15, 16 15, 16
c
64%
47 16 16
Source
U(8.8, 13.2) B(0.95, 0.03) Not varied in probabilistic analysis
Model for Sensitivity Analysisa
THEIR SOURCES
U(0.88, 1.32)
AND
3.8 days 1.1 days
11.0 days 100% 100%
Base Case Value
TABLE 1. BASE CASE PARAMETER VALUES
Treatment costsf Daily cost, prodromal stage
Inventories of Equipment and Personnel Intensive Care Unit Beds Ventilators Respiratory Technicians Ventilators per Push Pack Time Parameters Time lag until attack detected Time lag until local inventory available for dispensing Time lag between attack detection and Push Packs arrival Time lag between attack detection and arrival of regional VMI Costs Prophylaxis costs (any stage of disease)
Push Pack (doses)
Days of Therapy Local Inventory (doses)e
Push Pack (doses)
Average time, fulminant to death, treatment started in fulminantd Antibiotic Inventories Days of Prophylaxis Local Inventory (doses)e
SE(4, 1) SE(2.4, 9.6) SE(24, 12) SE(0.4, 1.6) SE(0.02, 0.1)
5.0 hours 12.0 hours 36 hours Ciprofloxacin: $2/day Doxycycline: $0.12/day
SE(186, 744) Fixed at 99% of the cost of the Cipro-based program
SE(48, 0)
48.0 hours
Ciprofloxacin: $930 Doxycycline: $921
SE(336, 0) SE(273, 0) SE(131, 0) U(80, 120)
970 total, 336 available 766 total, 273 available 188 total, 131 available 100
Not varied in probabilistic analysis U(10780.8, 16171.2) U(6412.8, 9619.2)
Not varied in probabilistic analysis U(172800, 259200) U(2001600, 3002400)
Ciprofloxacin: 40,190 Doxycycline: 24,338 Ciprofloxacin: 216,000 Doxycycline: 2,502,000 Ciprofloxacin: 703 Doxycycline: 78 Ciprofloxacin: 13,476 Doxycycline: 8,016
U(1.28, 1.92)
1.6 days
(continued)
56, 57
54, 55 19–21
53
53
Estimated
Estimated
50 50, 51 50 52
31
Estimated
31
6, 7, 29
16
SIMULATION MODEL
U(0.256, 0.384) U(0.032, 0.048) Not variedi
Ciprofloxacin: $0.32 Doxycycline: $0.04 Ciprofloxacin: $21.07 Doxycycline: $19.70
SE(532.8, 2131.2) Fixed at 99% of the cost of the Cipro-based program
Model for Sensitivity Analysisa
THEIR SOURCES (CONT’D)
U(22596, 33894)
AND
$28,245
Ciprofloxacin: $2,664 Doxycycline: $2,655
Base Case Value
FOR THE
60
60
58, 59
56, 57
Source
We express the sensitivity analysis as probabilistic models rather than ranges. We use the following notation: U(a,b) means that we assumed a uniformly distributed random variable over the interval (a,b); B(a,b) means that we assumed a beta-distributed random variable with mean a and standard deviation b; SE(a,b) means that we assumed a shifted exponential distributed random variable with lower limit b and mean a b. b The incubation period is defined as the asymptomatic stage between exposure and beginning of prodromal stage. c Based on a literature review of historical cases, many of whom did not receive antibiotics, we assumed that disease progression from prodromal to fulminant for individuals not receiving prophylaxis would occur at a constant rate of 6.2% per day for the first 3 days and 43.3% per day thereafter.16 This resulted in a very close approximation between the model’s disease progression rate and the anthrax survival curve derived from the literature of untreated inhalational anthrax.16 These two values were simultaneously varied by 20% (i.e., multiplied by a uniform random variable U(0.8, 1.2)). d From the review of cases of inhalational anthrax, it was found that the average time from onset of fulminant stage to death is shorter among those for whom prophylaxis (or treatment) was started in the prodromal stage rather than the fulminant stage. This is likely because antibiotics started in the fulminant phase, when the patient is naïve to antibiotics, may have a somewhat greater ability to delay death than in the patient who has been receiving antibiotics for some time.
a
Treatment
Dispensing Center Labor (per center/day)g Maintaining Local Inventory (per daily dose/yr)h Prophylaxis
Daily cost, fulminant stage
Parameter
TABLE 1. BASE CASE PARAMETER VALUES
Given the paucity of published data on local inventories of antibiotics, we estimated the base case local inventory from a variety of sources. A survey of 10 hospitals in New Jersey, serving a total population of 1.6 million, found that the median number of on-hand doses per hospital was 289 for ciprofloxacin and 175 for doxycycline.7 These numbers were similar to other reports.6,29,30 To estimate the total doses of on-hand prophylactic antibiotics for a hypothetical community of 5 million people, we first calculated the number of hospitals and pharmacies from data for the city of San Jose, California. The greater San Jose metropolitan area has a population of 1.6 million people,61 which is served by 10 hospitals and 79 outpatient pharmacies (not including military or VA facilities).62,63 For the base case, we assumed that local inventories of prophylactic antibiotics (including those from hospitals, pharmacies, and other local inventories) contained approximately 64,500 days of prophylaxis, consisting of ciprofloxacin for 40,189 days of treatment, doxycycline for 24,336 days of treatment, and sufficient intravenous antibiotics to provide 781 days of treatment.6,7,29 Levels of intravenous antibiotic inventories were estimated by a local hospital pharmacist. f The Department of Veterans Affairs (VA) currently purchases and manages the VMI. CDC has recommended that local inventories be purchased according to the VA model. Thus, we based most of the drug costs used in our model on the VA product pricing and fee schedules (VA 340B program).64–66 However, because it may be difficult for local officials to purchase medical and pharmaceutical supplies at the same price as the VA, we varied this price widely in sensitivity analyses. Because the VA purchasing prices for ciprofloxacin do not reflect the price contracted by HHS, the cost of oral ciprofloxacin was based on the agreement between Bayer A.G. and HHS.56,57 Drug costs include product cost, a wholesale markup of 0.02% (for storage and distribution),56,57 and a pharmacy dispensing fee ($3.00 per 100 tablets or 1 vial).56 Treatment costs include the costs of a triple antibiotic regimen, intravenous supplies, hospital or ICU beds, physician charges, pleural drainage, and respiratory or cardiac support as needed. Hospital and ICU bed costs were derived from the literature.19,20 Physician charges and pleural effusion drainage costs were based on the 2003 Medicare reimbursement schedule for the corresponding CPT codes.21 g The types and numbers of individuals needed to operate and staff a mass prophylaxis distribution center were derived from the Weill/Cornell Mass Prophylaxis/Vaccination model.58 Labor costs for the dispensing centers include salaries for staff working at the dispensing sites, 1 day of paid training per year for each staff member of the dispensing center, and administrative costs to maintain the centers, including maintaining a current roster of staff members and ensuring that training and contact information are current. Labor costs were taken from pay scales for federal employees.67 h We assumed that the annual costs to maintain a local inventory of antibiotics including annual rotation costs plus storage fees. The rotation fee was estimated to be 30% of the purchase price of the drugs.60 We assumed that all local stocks must be replaced if an attack occurs, since the local stockpiles would be depleted prior to the arrival of regional supplies, and that annual maintenance costs are incurred in perpetuity. i Not varied since we did not consider strategies in which this number was varied.
e
250
BRAVATA ET AL.
FIGURE 1. ALTERNATIVE STRATEGIES ANALYZED. This figure presents the key elements of the antibiotic distribution and dispensing supply chain for a response to anthrax bioterrorism. The four detection, stockpiling, distribution, and dispensing strategies evaluated in our analysis are in the gray boxes.
and nausea. Fulminant stage anthrax (also called acute stage) is severe symptomatic disease characterized by abrupt respiratory distress and shock that typically progresses rapidly to death. We assumed that only patients in the fulminant stage would die. Based on our review of historical cases, many of whom did not receive antibiotics, we assumed that disease progression from prodromal to fulminant infection for individuals not receiving prophylaxis would occur at a constant rate of 6.2% per day for the first 3 days and 43.3% per day thereafter.16 This resulted in a very close approximation between the model’s disease progression rate and the anthrax survival curve derived from the literature of untreated inhalational anthrax.16 All other transition rates through each of the three disease stage were based on average times for disease progression (Table 1).
Prophylaxis and Treatment We assumed that asymptomatic exposed individuals became aware of the anthrax attack through alerts from the media, public health officials, and emergency response officials. Once asymptomatic individuals became aware of their potential exposure, we assumed they entered a queue for prophylactic antibiotics (either ciprofloxacin or doxycycline).17 We assumed that pro-
phylaxis with either antibiotic, given full patient adherence, would be 100% effective in preventing disease progression from the incubation to the prodromal stage. Based on the 2001 U.S. experience, we assumed that 64% of patients would adhere to prophylactic antibiotic regimens but varied this up to 90% in sensitivity analyses.18 We assumed that nonadherent individuals would receive only partial benefit from prophylaxis, depending on how long they took antibiotics. Once exposed individuals developed symptoms of either prodromal or fulminant stage anthrax, we assumed that they entered a queue for treatment consisting of three antibiotics (ciprofloxacin or doxycycline with rifampin and clindamycin) administered intravenously in an intensive care setting with supportive care as necessary (e.g., pleural fluid drainage, and respiratory or cardiac support).17,19–22 Treatment was limited by the availability of intravenous antibiotics, intensive care unit beds, ventilators, and respiratory technicians. We assumed that if inventories of antibiotics were depleted, disease progression depended on how long individuals had received appropriate antibiotics. We assumed that antibiotics for both prophylaxis and treatment were dispensed on a first-come, first-served basis. We performed a probabilistic sensitivity analysis using Monte Carlo simulation to generate confidence intervals for the uncertain variables used in the model.
STOCKPILING AND DISPENSING SUPPLIES FOR BIOTERRORISM RESPONSES
Strategies to Reduce the Mortality from Anthrax Bioterrorism We evaluated four types of strategies to decrease the morbidity and mortality from an anthrax attack: strategies that (1) enhance bioterrorism event detection, (2) increase local dispensing capacity, (3) increase local inventories of antibiotics, and (4) increase the amount of inventory deployed from the SNS to the site of an attack (Figure 1). We calculated the costs, cost-effectiveness, and life years saved for each of these strategies up to 120 days after the attack. Strategies that Enhance Bioterrorism Event Detection Prophylaxis and treatment of exposed individuals cannot begin until the bioterrorism event has been recognized. Thus, we evaluated the effects on expected mortality of changing the time until event detection. Syndromic surveillance systems have been widely deployed for the early detection of illness resulting from bioterrorism; however, no published reports specifically describe the expected time to detection of a bioterrorism event.23 In the base case, we assumed that an attack would be detected 48 hours after release of anthrax spores. In sensitivity analyses, we varied the detection time from 0 hours after the release of the anthrax spores (as might happen during an event announced by the bioterrorist) to 144 hours (6 days) after the attack (assuming that at 6 days, the relatively large number of patients presenting with symptoms would alert public health officials). Strategies that Increase Local Dispensing Capacity Two factors determine the hourly local dispensing capacity: the number of dispensing sites and the number of individuals who can be served at each site per hour. Published reports of mass vaccination and antibiotic dispensing campaigns for naturally occurring infectious diseases have documented widely varying dispensing capacities (from 162 to 1,700 patients per hour), with some campaigns using one or two large dispensing sites and others relying on more than 10 sites.24–27 No published studies have evaluated the optimum number of dispensing sites for a given population size/density. In the base case, we assumed that the local community would have 10 dispensing centers, each of which would be able to dispense prophylactic antibiotics to 1,000 people per hour. We assumed that each dispensing center would require 48 hours to ramp up to full capacity and that, at full capacity, it would operate 14 hours per day (i.e., base case dispensing 10 centers 1,000 people per center 14 hours per day 140,000 people/day).27 In sensitivity analyses, we varied local dispensing capacity from 14,000 to 420,000 people per day.
251
Strategies that Increase Local Inventories Little published data exists regarding the quantities and types of medical and pharmaceutical supplies in local inventories.1,28 Thus, we estimated the amount of antibiotics in local inventories from a variety of sources.6,7,29,30 In sensitivity analyses, we varied the amount of local inventory from 0 to 5 times the base case inventory. Strategies that Increase the Inventory Deployed from the SNS The SNS has two components, Push Packs and Vendor Managed Inventories (VMI). Push Packs contain antibiotics, antidotes, and other medical supplies necessary to treat a wide range of possible biological or chemical agents and reportedly would be available for local distribution within 12 hours after being requested.4,31 The VMI contain additional supplies of antibiotics and medical equipment tailored to the specific needs of responders and reportedly can arrive at local distribution and/or dispensing sites within 36 hours following the detection of an attack.4,32 In sensitivity analyses, we evaluated the effects on expected mortality of changing the time to arrival of national inventories. We also varied the amount of prophylactic antibiotics dispensed to each individual as a function of whether the VMI was available (from the base case of a 14-day supply to a complete 60-day supply).
Costs Base case costs, presented in Table 1, include the costs of prophylaxis and treatment, the costs of operating the dispensing center, and the costs of maintaining local inventories of antibiotics. Our model adopted the societal perspective with costs representing those of the ideal insurer.33 All costs were discounted at 3% per year. We assumed that increasing local inventories would require a one-time cost to purchase the antibiotics and other supplies, plus the annual cost of maintaining the inventory (e.g., stocks must be stored in a dry, temperature-regulated environment and must be rotated and replaced as they expire).
Role of the Funding Agencies This work was supported in part by the Agency for Healthcare Research and Quality (Contract Number 29002-0017) and by the Department of Veterans Affairs. The funding agencies had no influence over any portion of this working, including the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
252
BRAVATA ET AL.
RESULTS We first present the results of the base case analyses, then the specific results of evaluating each of the five strategies described in Figure 1.
Base Case Analyses Expected mortality was highly dependent on the local dispensing capacity, the number of individuals requiring prophylaxis, and the adherence to prophylactic antibiotics. For an attack that exposes 50,000 individuals, at a low dispensing capacity (e.g., 14,000 individuals per day), approximately 28,000 deaths are expected (even if only 2% of the unexposed population requires prophylaxis). If the local dispensing capacity is 140,000 individuals per day (base case), we would expect 21,000 deaths (Figure 2A). For larger attack sizes, the reductions in expected mortality from increasing dispensing capacity are more significant (Figure 2B). As the number of individuals requiring prophylaxis increases from 2% to 95% of the unexposed population, the expected mortality doubles (from 21,000 to 40,000 for an attack exposing 50,000 individuals and from 110,000 to 201,000 for an attack exposing 250,000 individuals). This suggests that technologies that can distinguish between individuals who have been exposed and those who have not are worthy of investigation. At low dispensing capacities, nearly all exposed individuals died, regardless of the rate of adherence with prophylaxis (at the lowest dispensing capacity the expected mortality is nearly identical for adherence rates of 64% and 90%). At dispensing capacities of 140,000 individuals per day or greater, expected mortality depends on the rate of adherence with prophylactic antibiotics. If adherence could be improved from 64% to 90% by a counseling session costing $10 per person, then the cost-effectiveness of improved adherence is $2,300 per life year gained.
Effects of Strategies that Enhance Bioterrorism Event Detection Enhanced bioterrorism event detection reduces mortality, particularly for large attacks and when the number of unexposed individuals requiring prophylaxis is low (Figure 3A). For example, an attack resulting in 250,000 exposed individuals and requiring the prophylaxis of only 5% of the unexposed population detected at 48 hours resulted in 138,000 deaths. This same attack detected at 144 hours results in an additional 35,000 deaths—about 6 deaths per minute of delay in detection. When dispensing capacity is low, most exposed individuals cannot receive prophylaxis in a sufficiently timely manner to prevent disease progression—thus, no significant advantage accrues from rapid attack detection (Figure 3B). In contrast, when dispensing capacity is high, if the
attack is detected quickly, it is possible to prevent considerable mortality. Even if local inventories and dispensing capacities were infinite, we would expect 107,000 deaths if the event were detected at 48 hours (and 92,000 deaths if the event were detected at 0 hours) assuming a large attack. This is primarily because of the 64% adherence rate with prophylaxis used in the base case.
Effects of Strategies that Increase Local Dispensing Capacity Strategies that increase local dispensing capacity are highly cost-effective up to about 280,000 individuals per day (Table 2). The cost-effectiveness of increasing the dispensing capacity depends on the annual probability of an attack. This is because the costs of training and maintenance for each dispensing site are the same regardless of the annual probability of an attack, whereas the operating costs (the cost of operating the sites and the costs of all antibiotics) as well as the health benefits of increased dispensing capacity are accrued only if an attack occurs. However, even at very small annual probabilities of attack (less than 1 106), increasing local dispensing capacity to approximately 100,000 individuals per day falls well below the usual cost-effectiveness threshold: Medical interventions that cost less than $50,000 per life year gained are generally considered acceptable.34 To put this in context, emergency preparedness experts estimate the probability of a major earthquake (i.e., magnitude 6.7 with expected human and economic losses similar to those of the 1995 Kobe earthquake) affecting the San Francisco Bay region in the next 30 years is 0.62 (95%CI: 0.37–0.87),35 whereas the probability of a 1-kmsized near earth object strike (which would deposit approximately 80,000 megatons of total energy and result in major regional destruction and global climate disruption) during the 21st century is 2 104.36
Effects of Strategies that Increase Local Inventories Development of a local stockpile is cost-effective only if the size of an attack is large, the probability of attack is sufficiently high, and dispensing capacity is adequate (Figure 4). For a large attack, local stockpiling reaches conventional levels of cost-effectiveness only if the annual probability of an attack is greater than 0.0004 (Table 3). If local dispensing capacity is sufficient, initiation and maintenance of a local inventory reaches conventional levels of cost-effectiveness if a large attack occurs, but not if a small attack occurs (data not shown).
Effects of Strategies that Vary the Availability of National Inventories The strategy of sending additional Push Packs to the attack site until the VMI becomes available results in
STOCKPILING AND DISPENSING SUPPLIES FOR BIOTERRORISM RESPONSES
253
A
B
FIGURE 2. EFFECTS OF DISPENSING CAPACITY, ADHERENCE WITH PROPHYLAXIS, AND NUMBER OF INDIVIDUALS REQUIRING PROPHYLAXIS ON EXPECTED MORTALITY. Figure 2A presents the expected mortality for an attack resulting in the exposure of 50,000 individuals, and Figure 2B presents the results for an attack resulting in the exposure of 250,000 individuals. This analysis assumes 90% adherence to prophylactic antibiotics. The denotes a dispensing capacity of 14,000 individuals per hour, the denotes a dispensing capacity of 140,000 individuals per hour, and the denotes a dispensing capacity of 420,000 individuals per hour.
only small reductions in mortality (and only in the case of a large attack or an attack in which large numbers of individuals require prophylaxis). This is because at the base case level of dispensing capacity, the supplies in the first Push Pack are likely to hold out until just before the VMI arrives at 36 hours.
If the Push Packs were delayed in arriving at the attack site from the base case of 12 hours to 36 hours after being requested, expected mortality would increase from 137,500 to 143,000. If the arrival of the VMI were delayed from the base case of 36 hours to 144 hours after attack detection, there would be no effect on expected
254
BRAVATA ET AL.
A
B
FIGURE 3. EFFECTS OF DELAYS IN THE DETECTION OF AN ATTACK. Figure 3A presents the effect of delays in detection of an attack given the number of people who require prophylaxis. The solid dark line represents the analysis when 90% of the unexposed population needs prophylaxis; the solid light line represents the analysis when 25% of the unexposed population needs prophylaxis; and the broken line represents the analysis when 5% of the unexposed population needs prophylaxis. Figure 3B presents the effects of delays in the detection of an attack given local dispensing capacity. The denotes a dispensing capacity of 14,000 individuals per hour, the denotes a dispensing capacity of 140,000 individuals per hour, and the denotes a dispensing capacity of 420,000 individuals per hour. The results presented in this figure are for an attack exposing 250,000 people assuming base case estimates of local inventories and antibiotic adherence. The base case assumptions are noted with an arrow. mortality at dispensing capacities of less than 28,000 individuals per day (because at this level, dispensing capacity is rate-limiting). At higher dispensing capacities, delays in VMI resulted in increased mortality, particu-
larly for a large attack with large numbers of individuals seeking prophylaxis. If a 60-day supply of antibiotics is dispensed to all individuals (without regard to the available inventory),
5,735 5,765 2,505 887 421 218
46,069,485 67,976,441 60,442,905 56,711,188 55,635,104 55,170,775
253,311
132,269
63,912
24,131
11,791
8,033
6,740
Incremental Cost per LY Gained ($/LY)
4.2 1005
2.2 1005
1.0 1005
3.6 1006
1.6 1006
9.5 1007
7.3 1007
Annual Probability of Attack Required for Each Increase in Dispensing Capacity to Be Cost-Effective ($50,000/LY)b
LOCAL DISPENSING CAPACITYa
b
This analysis assumed an attack resulting in 250,000 individuals exposed and 25% of the unexposed population requiring prophylaxis. Medical interventions that cost less than $50,000 per life year gained are generally considered acceptable.34 Note: Uncertainty exists regarding the annual cost to maintain each dispensing site and its employees, so we varied both of these factors in sensitivity analyses. Even if the annual probability of attack is small (0.0001) and the annual costs to maintain the dispensing site double, the incremental cost-effectiveness of increasing dispensing capacity remains below $10,000 per life year gained.
a
2,476
16,687,233
14,000 to 28,000 (1 to 2 times the base case) 28,000 to 70,000 (2 to 5 times the base case) 70,000 to 140,000 (5 to 10 times the base case) 140,000 to 210,000 (10 to 15 times the base case) 210,000 to 280,000 (15 to 20 times the base case) 280,000 to 350,000 (20 to 25 times the base case) 350,000 to 420,000 (25 to 30 times the base case)
OF INCREASING
Incremental Life Years Gained (LY)
Incremental Cost ($)
Increasing Local Dispensing Capacity (individuals per day)
TABLE 2. INCREMENTAL COST-EFFECTIVENESS
FIGURE 4. INCREMENTAL COSTS AND EFFECTIVENESS OF STRATEGIES FOR INCREASING LOCAL INVENTORIES AND DISPENSING CAPACITIES. This figure presents the costs and benefits associated with combinations of levels of local inventory (e.g., 10 local inventory 10 times the base-case level of local inventory) and local dispensing capacity (e.g., 2 dispensing capacity twice the base-case dispensing capacity).
52
53
61
2,052,898
2,056,509
5,940,836
Incremental Life Years Gained (LY)
100,013
96,437
94,897
0.0012
0.0011
0.0011
$20,000/LY
0.00043
0.00041
0.00041
$50,000/LY
0.00021
0.00020
0.00020
$100,000/LY
Annual Probability of Attack Required for Each Increase in Dispensing Capacity to Be Cost-Effective Over a Range of Cost-Effectiveness Thresholds
LOCAL INVENTORIESa
Incremental Cost per LY Gained ($/LY)
OF INCREASING
This analysis assumed an attack resulting in 250,000 individuals exposed and 1,437,500 individuals requiring prophylaxis.
a
No local inventory to the base case level of inventory The base case level of inventory to 2 times the base case level of inventory 2 times the base case level of inventory to 5 times the base case level of inventory
Increasing Local Inventories
Incremental Cost ($)
TABLE 3. INCREMENTAL COST-EFFECTIVENESS
258
rather than the 14-day supply assumed in the base case, then mortality will increase only slightly (from 138,000 deaths to 140,000). This is because in the base case, given a dispensing capacity of 14,000 individuals per day, the intervals between the depletion of local inventories, depletion of the Push Pack inventory, and the arrival of the VMI are relatively brief.
Probabilistic Sensitivity Analysis Figure 5 shows the results of our Monte Carlo simulation as an acceptability curve, which allows decision makers to determine the probability that pre-attack increases in local inventory are cost-effective at various willingness-to-pay thresholds. Probabilistic sensitivity analysis indicated that, in the event that a large attack did occur, the strategy of doubling local inventory was the
BRAVATA ET AL.
preferred strategy in 6% of simulations at an incremental cost-effectiveness threshold of $50,000 per life year gained and in 85% of simulations at an incremental costeffectiveness threshold of $100,000 per life year gained.
DISCUSSION We evaluated the costs and benefits of various strategies for stockpiling and dispensing medical and pharmaceutical supplies during a response to anthrax bioterrorism. Our analysis yielded four main results. First, we found that, following an outbreak of inhalational anthrax resulting from a bioterrorist attack, mortality is critically dependent on the local dispensing capacity. At low dispensing capacities, nearly all exposed
FIGURE 5. COST-EFFECTIVENESS ACCEPTABILITY CURVE OF DOUBLING LOCAL INVENTORY. This figure shows the results of our Monte Carlo simulation as an acceptability curve, which allows decision makers to determine the probability that pre-attack doubling of local inventory is cost-effective at various willingness-to-pay thresholds. For this analysis, the probability of attack was fixed at 0.0001.
STOCKPILING AND DISPENSING SUPPLIES FOR BIOTERRORISM RESPONSES
individuals died, regardless of the rate of adherence with prophylaxis. At high dispensing capacities, the expected mortality is highly dependent on the rate of adherence with prophylactic antibiotics. Second, our analysis demonstrated that the cost-effectiveness of some strategies to prepare for bioterrorism is sensitive to the probability that an attack occurs. If the probability of a bioterrorism attack is high, increasing local inventories of supplies and increasing the number of local dispensing centers may be cost-effective. If the probability of an attack is low, increasing local inventories is not likely to be cost-effective. Third, at local dispensing capacities of 14,000 individuals per hour or more (for a large metropolitan area), expected mortality depends on the rate of adherence with prophylactic antibiotics. Programs to increase adherence may be highly cost-effective. Finally, when dispensing capacity is low, surveillance strategies to enhance attack detection do not result in reduced mortality. However, in communities with high dispensing capacities, significant reductions in mortality can be achieved via early event detection (such as might be achieved through ongoing syndromic surveillance or environmental surveillance programs such as BioWatch37). Our results suggest that preparedness planning efforts at the local level should focus on three actions: development of plans to maximize local dispensing capacity, strategies to encourage adherence with prophylaxis regimens, and development of technologies to identify people at risk of exposure (so as to minimize the number of unexposed individuals who would be given prophylactic antibiotics). Other strategies to improve local preparedness, including increasing local antibiotic stockpiles and instituting surveillance systems to reduce the delay in attack detection, are likely to be cost-effective only if the community can achieve a high dispensing capacity, if the probability of an attack is greater than 0.0001 per year, and if the attack is large. The cost-effectiveness of several strategies was sensitive to the annual probability of an attack. This is because many preparedness strategies such as the stockpiling and maintenance of inventories of supplies require an initial investment and ongoing maintenance costs that are accrued regardless of whether an attack takes place. However, benefits of a preparedness strategy are accrued only if an attack takes place. Thus, the cost-effectiveness ratio becomes more favorable as the probability of an attack increases. We performed probabilistic sensitivity analysis to assess the likelihood that pre-attack increases in local inventory are cost-effective. The simulations assumed that the parameters are uncorrelated, as no evidence exists regarding the type or extent of possible correlations among these parameters. Parameter correlations could be con-
259
sidered in such a simulation. An alternative approach is scenario planning, a method of considering the consequences of current actions under a range of uncertain future scenarios.38 Such a method, commonly used in military systems planning,39–41 is well suited to bioterrorism preparedness planning. Our analyses addressed a wide range of issues related to enhanced local preparedness for responses to anthrax bioterrorism, but two key issues remain unaddressed. First, our results demonstrated the importance of local dispensing capacity. However, we did not explore which specific elements of dispensing capacity—such as triaging symptomatic patients to receive treatment, obtaining informed consent for antibiotic dispensing, actual dispensing of the antibiotics, etc.—were most rate-limiting. Our model could be expanded to evaluate these specific dispensing tasks. Second, our prophylaxis and treatment model did not include the use of anthrax vaccine. Although anthrax vaccine has recently been added to the CDC postexposure prophylaxis recommendation in a 3dose regimen, it is not likely to be available for inclusion in emergency response stockpiles until 2008, and it is licensed neither for postexposure prophylaxis for prevention of inhalation anthrax nor for use in a 3-dose regimen (thus, its use would have to be conducted under an Investigational New Drug application to the Food and Drug Administration).14,22,42–45 Some evidence suggests that the use of anthrax vaccine may reduce the need for longer courses of prophylaxis and treatment.46 Our model could be readily expanded to include the costs and benefits potentially associated with adding vaccination to prophylaxis regimens. As presented, our results could be used to inform the preparedness planning efforts for anthrax bioterrorism and for related biothreat agents that have similar disease progression and treatment strategies. Our analysis depended on the assumption that the Push Packs would arrive at the attack site 12 hours after being requested. This assumption was based on experience with Push Pack delivery during the September 11, 2001, response and references from the SNS.4,31 However, if an attack is made on multiple U.S. cities simultaneously, availability of Push Packs and VMI could be significantly delayed. Our results suggest that even if regional inventories are delayed significantly, expected mortality is largely dependent on dispensing capacity. Our model could be used to evaluate the expected costs and mortality of other clinical and planning questions such as the effects of an attack with antibiotic-resistant (as opposed to antibiotic-sensitive) anthrax, the effects of changing costs of particular antibiotics (e.g., the availability of low-cost antibiotics), or the use of other medical therapies (e.g., pleural fluid drainage). In the U.S., medical, public health, and emergency response experts are currently preparing for the enormous
260
BRAVATA ET AL.
logistical challenges of stockpiling, distributing, and dispensing prophylaxis and treatment to populations exposed to bioterrorism. Given the availability of the resources available through the SNS, primary planning efforts are likely to benefit most from increased local capacity to rapidly dispense these inventories, to reliably identify exposed individuals, and to encourage adherence with prophylaxis and treatment regimens.
11.
12.
ACKNOWLEDGMENTS We thank Drs. Nathaniel Hupert and Sara Cody for their comments on the manuscript.
13.
REFERENCES
14.
1. Terriff CM, Tee AM. Citywide pharmaceutical preparation for bioterrorism. Am J Health Syst Pharm 2001;58(3): 233–237. 2. Inglesby TV, Grossman R, O’Toole T. A plague on your city: observations from TOPOFF. Clin Infect Dis 2001; 32(3):436–445. 3. Koplan J. CDC’s strategic plan for bioterrorism preparedness and response. Public Health Rep 2001;116(Suppl 2): 9–16. 4. U.S. Centers for Disease Control and Prevention. Strategic National Stockpile. Atlanta: U.S. Centers for Disease Control and Prevention; April 14, 2005. Available at: http:// www.bt.cdc.gov/stockpile/index.asp. Accessed April 6, 2004. 5. U.S. Medicine Institute for Health Studies. Surge Capacity: Is It Time to Move Beyond ‘Just-in-Time’? Washington, DC: U.S. Medicine Institute for Health Studies; June 1, 2002. Available at: www.usminstitute.org/jun02surgeagenda.html. Accessed August 13, 2003. 6. Benson B, McKinney P. Selecting & stocking antidotes to biological/chemical agents. Paper presented to: Texas Society of Hospital Pharmacists Annual Meeting; 2003. 7. Case GG, West BM, McHugh CJ. Hospital preparedness for biological and chemical terrorism in central New Jersey. N J Med 2001;98(11):23–33. 8. Bravata DM, McDonald KM, Owens DK, et al. (Agency for Healthcare Research and Quality). Regionalization of Bioterrorism Preparedness and Response. 2004 April. Evidence Report/Technology Assessment Report No. 04E016-2. Prepared by Stanford-University of California San Francisco Evidence-based Practice Center under Contract No. 290-02-0017. 9. HHS announces $1.1 billion in funding to states for bioterrorism preparedness [press release]. Washington, DC: U.S. Department of Health and Human Services; January 31, 2002. Available at: http://www.hhs.gov/news/press/2002 pres/20020131b.html. Accessed February 16, 2004. 10. Hupert N, Cuomo J, Callahan MA, Mushlin AI, Morse S. (Agency for Healthcare Research and Quality). Community-Based Mass Prophylaxis: A Planning Guide for Public
15.
16.
17.
18.
19.
20.
21.
22.
23.
Health Preparedness. 2004 August. Prepared by Weill Medical College of Cornell University, Department of Public Health, under Contract No. 290-02-0013-3. AHRQ Pub No. 04-0044. U.S. Centers for Disease Control and Prevention. Cities Readiness Initiative (CRI) Guidance. Atlanta: U.S. Centers for Disease Control and Prevention; July 17, 2006. Available at: http://www.bt.cdc.gov/planning/coopagreement/ #fy06. Accessed July 26, 2006. Zaric GS, Brandeau ML, Bravata DM, Holty J-EC, McDonald KM, Owens DK. Modeling the logistics of response to anthrax bioterrorism. Palo Alto, Calif: Department of Management Science and Engineering, Stanford University; 2006. U.S. Census Bureau. Statistical Abstract of the United States. Report No. 26. Large Metropolitan Statistical Areas—Population: 1990 to 2002. Washington, DC: U.S. Census Bureau; 2003. Brookmeyer R, Johnson E, Bollinger R. Modeling the optimum duration of antibiotic prophylaxis in an anthrax outbreak. Proc Natl Acad Sci U S A 2003;100(17):10129– 10132. Wein LM, Craft DL, Kaplan EH. Emergency response to an anthrax attack. Proc Natl Acad Sci U S A 2003;100(7): 4346–4351. Holty JE, Bravata DM, Liu H, Olshen RA, McDonald KM, Owens DK. Systematic review: a century of inhalational anthrax cases from 1900 to 2005. Ann Intern Med 2006; 144:270–280. Bell DM, Kozarsky PE, Stephens DS. Conference summary: clinical issues in prophylaxis, diagnosis and treatment of anthrax. Emerg Infect Dis 2002;8(2):222–225. Shepard CW, Soriano-Gabarro M, Zell ER, et al. Antimicrobial postexposure prophylaxis for anthrax: adverse events and adherence. Emerg Infect Dis 2002;8(10):1124– 1132. Rapoport J, Teres D, Zhao Y, Lemeshow S. Length of stay data as a guide to hospital economic performance for ICU patients. Med Care 2003;41(3):386–397. Weber RJ, Kane SL, Oriolo VA, Saul M, Skledar SJ, Dasta JF. Impact of intensive care unit (ICU) drug use on hospital costs: a descriptive analysis, with recommendations for optimizing ICU pharmacotherapy. Crit Care Med 2003;31(1 Suppl):S17–S24. U.S. Department of Health and Human Services. Medicare and Medicaid 2003 Reimbursement Schedules. Washington, DC: Center for Medicare and Medicaid Services; U.S. Department of Health and Human Services; 2003. Available at: http://cms.hhs.gov/. Accessed September 5, 2003. Center for Infectious Disease Research and Policy (CIDRAP) and Infectious Diseases Society of America (IDSA). Clinical Pathway: Anthrax Inhalational Exposure. Minneapolis: Center for Infectious Disease Research and Policy and Infectious Diseases Society of America; 2002. Available at: http://www.cidrap.umn.edu/cidrap/files/17/ anthrax-clinical-pathway.pdf. Accessed September 9, 2003. Bravata D, McDonald K, Smith W, et al. Systematic review: surveillance systems for early detection of bioterrorism-related diseases. Ann Intern Med 2004;140(11):910– 922.
STOCKPILING AND DISPENSING SUPPLIES FOR BIOTERRORISM RESPONSES 24. Osterholm MT. How to vaccinate 30,000 people in three days: realities of outbreak management. Public Health Rep 2001;116(Suppl 2):74–78. 25. Krause G, Blackmore C, Wiersma S, Lesneski C, Gauch L, Hopkins RS. Mass vaccination campaign following community outbreak of meningococcal disease. Emerg Infect Dis 2002;8(12):1398–1403. 26. Mass Immunization Campaigns: A “How to” Guide Based on Experience During the February 2000 Capital Health Meningococcal Immunization Campaign. 2000. 27. Hupert N, Mushlin AI, Callahan MA. Modeling the public health response to bioterrorism: using discrete event simulation to design antibiotic distribution centers. Med Decis Making 2002;22(5 Suppl):S17–S25. 28. U.S. Government Accounting Office. GAO Congressional Report: Bioterrorism: preparedness varied across state and local jurisdictions. Washington, DC: U.S. Government Accounting Office; 2003. 29. Wetter DC, Daniel WE, Treser CD. Hospital preparedness for victims of chemical or biological terrorism. Am J Public Health 2001;91(5):710–716. 30. Parizek B. Department of Health Public Health Preparedness Update. Florida Department of Public Health; November 1, 2002. Available at: http://www.foma.org/ assets/DOHpreparedness11-1-02.pdf Accessed May 5, 2004. 31. U.S. Centers for Disease Control and Prevention. Receiving, Distributing, and Dispensing the National Pharmaceutical Stockpile: A Guide for Planners. Version 9, Draft. Atlanta: U.S. Centers for Disease Control and Prevention; April 2002. 32. U.S. Centers for Disease Control and Prevention. National Pharmaceutical Stockpile. Atlanta: U.S. Centers for Disease Control and Prevention. Available at: www.bt.cdc.gov/ DocumentsApp/NationalStockpile/national_pharmaceutical_ stockpile.pdf. Accessed August 27, 2003. 33. Gold MR, Siegel JE, Russell LB, Weinstein MC, eds. CostEffectiveness in Health and Medicine. New York: Oxford University Press; 1996. 34. Owens DK. Interpretation of cost-effectiveness analyses [editorial]. J Gen Intern Med 1998;13:716–717. 35. U.S. Department of the Interior; U.S. Geological Survey. Earthquake Probabilities in the San Francisco Bay Region: 2002–2031. Washington, DC: U.S. Department of the Interior; U.S. Geological Survey; 2003. Open-File Report 03-214. 36. Workshop on Near Earth Objects: Risks, Policies and Actions. Sixth Meeting of the Organisation for Economic Cooperation and Development (OECD) Global Science Forum; Frascati, Italy; 2003. 37. Center for Infectious Disease Research and Policy. BioWatch program aims for nationwide detection of airborne pathogens. CIDRAP News. Minneapolis: Center for Infectious Disease Research & Policy; Academic Health Center, University of Minnesota; March 10, 2003. Available at: http://www.cidrap.umn.edu/cidrap/content/bt/bioprep/news/biowatch.html. Accessed July 26, 2006. 38. Wilson I, Ralston WKJ. Scenario Planning Handbook: Developing Strategies in Uncertain Times. Boston, Mass: South-Western Educational Publishing; 2006.
261
39. Nickel RH, Mangel M. Weapon acquisition with target uncertainty. Naval Research Logistics 1985;32:567–588. 40. Froggett S, Liu S. System performance modeling in C41SR/ weapon system design and development. Naval Engineers J 2001;113:55–63. 41. Brown GG, Clemence RD, Teufert WR, Wood RK. An optimization model for modernizing the Army helicopter fleet. Interfaces 1991;21:39–52. 42. U.S. Centers for Disease Control and Prevention. Anthrax Q & A: Preventive Therapy. Atlanta: U.S. Centers for Disease Control and Prevention; March 25, 2005. Available at: http://www.bt.cdc.gov/agent/anthrax/faq/preventive.asp. Accessed July 26, 2006. 43. Brook I. The prophylaxis and treatment of anthrax. Int J Antimicrob Agents 2002;20(5):320–325. 44. From the Centers for Disease Control and Prevention. Additional options for preventive treatment for exposed persons to inhalational anthrax. JAMA 2002;287(5):579. 45. From the Centers for Disease Control and Prevention. Recommendations for antimicrobial prophylaxis for children and breastfeeding mothers and treatment of children with anthrax. JAMA 2001;286(21):2663–2664. 46. Fowler RA, Sanders GD, Bravata DM, et al. Cost-effectiveness of defending against bioterrorism: a comparison of vaccination and antibiotic prophylaxis against anthrax. Ann Intern Med 2005;142:601–610. 47. Brookmeyer R, Blades N, Hugh-Jones M, Henderson DA. The statistical analysis of truncated data: application to the Sverdlovsk anthrax outbreak. Biostatistics 2001;2(2):233– 247. 48. Jefferds MD, Laserson K, Fry AM, et al. Adherence to antimicrobial inhalational anthrax prophylaxis among postal workers, Washington, D.C., 2001. Emerg Infect Dis 2002;8(10):1138–1144. 49. Jernigan JA, Stephens DS, Ashford DA, et al. Bioterrorism-related inhalational anthrax: the first 10 cases reported in the United States. Emerg Infect Dis 2001;7(6):933–944. 50. Rinnert KJ, Parris T, Zibulewsky J. Local perspectives on bioterrorism: CME, Part 3 of 3. Baylor Univ Med Center Proc 2001;14(3):231–238. 51. Pegues D. UCLA Hospital in a Potential Emergency. Los Angeles: UCLA International Institute; 2002. Available at: http://www.isop.ucla.edu/article.asp?parentid1917. Accessed September 3, 2003. 52. Hejik K. Pharmacy responds to terrorist attacks. (www.pharmacist.com). Available at: www.pharmacist. com/articles/h_ts_0001.cfm. Accessed September 3, 2003. 53. Cohen ML. Bioterrorism: CDC’s Public Health Response. Testimony before the Committee on Governmental Affairs and Subcommittee on International Security, Proliferation and Federal Services, United States Senate. October 30. Available at: www.hhs.gov/asl/testify/t011030.html. Accessed September 3, 2003. 54. Bradsher K. Bayer agrees to charge government a lower price for anthrax medicine. New York Times 2001 October 25. 55. Department of Veterans Affairs. VA Federal Supply Schedule Service Pharmaceuticals: Current Pharmaceutical Products and Pricing. Washington, DC: Department of Veterans Affairs, Office of Acquisition and Materiel
262
56.
57. 58.
59.
60.
61.
62.
63. 64.
BRAVATA ET AL. Management’s National Acquisition Center. Available at: http://www1.va.gov/oamm/acquisitions/nac/fsss/pharmfss. htm. Accessed September 3, 2003. Office of Inspector General, U.S. Department of Health and Human Services. Cost Containment of Medicaid HIV/AIDS Drug Expenditures. Report No.: OEI-05-9900611. Washington, DC: U.S. Department of Health and Human Services; July 2001. MarkeTips. Veterans Administration Supply News 2000 January/February. Hupert N, Cuomo J. Weill/Cornell Mass Prophylaxis/Vaccination Model. Available at: http://www.ahcpr.gov/research/biomodel/index.asp. Accessed September 3, 2003. U.S. Office of Personnel Management. 2003 Federal Employee Pay Tables. Washington, DC: U.S. Office of Personnel Management; 2003. Available at: www.opm.gov. Accessed September 5, 2003. U.S. Department of State. International Information Programs. Washington, DC: U.S. Department of State. Available at: http://usinfo.state.gov/topical/pol/terror/01102416. htm. Accessed September 5, 2003. U.S. Census Bureau. Data Profiles. Washington, DC: U.S. Census Bureau. Available at: http://www.census.gov/acs/ www/Products/index.htm. Accessed September 3, 2003. Dartmouth Atlas of Health Care, the Pacific States. Dartmouth: Center for the Evaluative Clinical Sciences, Dartmouth Medical School. Available at: www.dartmouthatlas.org/. Accessed September 3, 2003. SMARTpages. Available at: www.smartpages.com/. Accessed September 3, 2003. Tommy G. Thompson, Secretary, Department of Health and Human Services; Statement on Federal Efforts to Co-
ordinate and Prepare for Bioterrorism: The HHS Role. Committee on Governmental Affairs and Subcommittee on International Security, Proliferation and Federal Services. Washington, DC; 2001. 65. Kevin Yeskey, MD; Director, Bioterrorism Preparedness and Response Program, National Center for Infectious Diseases, Centers for Disease Control and Prevention; Statement on Bioterrorism: CDC’s Public Health Response. House Committee on Veterans Affairs, Subcommittee on Health. Washington, DC; 2002. 66. Evans L. Southern California Plans Defenses against Bioterror Attack. Los Angeles: UCLA International Institute; June 5, 2002. Available at: http://www.international.ucla.edu/article.asp?parentid1917. Accessed September 3, 2003. 67. U.S. Office of Personnel Management. Salary Tables and Related Information. Washington, DC: U.S. Office of Personnel Management; 2004. Available at: http://www.opm. gov/oca/04tables/index.asp. Accessed February 2, 2004. Manuscript received April 6, 2006; accepted for publication August 2, 2006.
Address reprint requests to: Dena M. Bravata, MD, MS Center for Primary Care and Outcome Research 117 Encina Commons Stanford, CA 94305-6019 E-mail:
[email protected]
