<%BANNER%>

Cost-Effectiveness of Workplace Closure and Travel Restriction for Mitigating Influenza Outbreaks: A Networkbased Simulation

University of Florida Institutional Repository
  • STANDARD VIEW
  • MARC VIEW
MISSING IMAGE

Material Information

Title:
Cost-Effectiveness of Workplace Closure and Travel Restriction for Mitigating Influenza Outbreaks: A Networkbased Simulation
Physical Description:
Conference Papers
Language:
English
Creator:
21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems ( Conference )
Mao, Liang
Publisher:
The Association for Computing Machinery
Place of Publication:
Orlando, FL
Publication Date:
Copyright Date:
2013
Edition:
ISBN 978-1-4503-2529-5/13/11

Notes

Abstract:
Background: Social distancing strategies, such as workplace closure and travel restriction, have been widely considered as alternative measures to contain influenza viruses, particularly when vaccines and antiviral drugs are under development. However, their cost-effectiveness in large urbanized populations is poorly understood. Method: To fill this knowledge gap, this research builds a spatially-explicit network-based model to simulate influenza transmission, mitigation strategies, and their associated costs. This model represents a spatio-temporal network of individuals' daily contacts, which enables the simulation of local infection and long distance dispersion of influenza. The workplace closure and travel restriction strategies, as well as their combinations with antiviral prophylaxis, are incorporated into this model to estimate their cost-effectiveness in mitigating seasonal flu and pandemic flu. The metropolitan area of Buffalo, NY, USA, with a population about 1 million, is selected as the study area. Results: Without any intervention, the seasonal flu and pandemic flu cost $234.1 million and $331.1 million, respectively. The closure of 30% affected workplace is the most cost-effective single strategy with $12.9K per case averted for seasonal flu and $34.9K for pandemic flu. The travel restriction is not cost effective if applied alone, but a 50% travel restriction in combination with antiviral prophylaxis and workplace closure forms the best strategy, which only costs $10.6K per cases averted for seasonal flu and $13.5K for pandemic flu. Conclusions: For a large urbanized area, the closure of affected workplaces could be an effective and cost-saving strategy to mitigate influenza outbreaks, while the highest cost-effectiveness can be achieved by combining the travel restriction with antiviral prophylaxis and workplace closure.
Acquisition:
Collected for University of Florida's Institutional Repository by the UFIR Self-Submittal tool. Submitted by Liang Mao.
Publication Status:
Published

Record Information

Source Institution:
University of Florida Institutional Repository
Holding Location:
University of Florida
Rights Management:
All rights reserved by the submitter.
System ID:
IR00003547:00001

  • STANDARD VIEW
  • MARC VIEW
MISSING IMAGE

Material Information

Title:
Cost-Effectiveness of Workplace Closure and Travel Restriction for Mitigating Influenza Outbreaks: A Networkbased Simulation
Physical Description:
Conference Papers
Language:
English
Creator:
21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems ( Conference )
Mao, Liang
Publisher:
The Association for Computing Machinery
Place of Publication:
Orlando, FL
Publication Date:
Copyright Date:
2013
Edition:
ISBN 978-1-4503-2529-5/13/11

Notes

Abstract:
Background: Social distancing strategies, such as workplace closure and travel restriction, have been widely considered as alternative measures to contain influenza viruses, particularly when vaccines and antiviral drugs are under development. However, their cost-effectiveness in large urbanized populations is poorly understood. Method: To fill this knowledge gap, this research builds a spatially-explicit network-based model to simulate influenza transmission, mitigation strategies, and their associated costs. This model represents a spatio-temporal network of individuals' daily contacts, which enables the simulation of local infection and long distance dispersion of influenza. The workplace closure and travel restriction strategies, as well as their combinations with antiviral prophylaxis, are incorporated into this model to estimate their cost-effectiveness in mitigating seasonal flu and pandemic flu. The metropolitan area of Buffalo, NY, USA, with a population about 1 million, is selected as the study area. Results: Without any intervention, the seasonal flu and pandemic flu cost $234.1 million and $331.1 million, respectively. The closure of 30% affected workplace is the most cost-effective single strategy with $12.9K per case averted for seasonal flu and $34.9K for pandemic flu. The travel restriction is not cost effective if applied alone, but a 50% travel restriction in combination with antiviral prophylaxis and workplace closure forms the best strategy, which only costs $10.6K per cases averted for seasonal flu and $13.5K for pandemic flu. Conclusions: For a large urbanized area, the closure of affected workplaces could be an effective and cost-saving strategy to mitigate influenza outbreaks, while the highest cost-effectiveness can be achieved by combining the travel restriction with antiviral prophylaxis and workplace closure.
Acquisition:
Collected for University of Florida's Institutional Repository by the UFIR Self-Submittal tool. Submitted by Liang Mao.
Publication Status:
Published

Record Information

Source Institution:
University of Florida Institutional Repository
Holding Location:
University of Florida
Rights Management:
All rights reserved by the submitter.
System ID:
IR00003547:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Cost Effectiveness of Workplace Closure and T ravel Restriction for Mitigating Influenza Outbreaks: A Network based Simulation Liang Mao Department of Geography University of Florida Gainesville, FL, 32611 352 294 7516 liangmao@ufl.edu ABSTRACT Ba ckground: Social distancing strategies, such as workplace closure and travel restriction have been widely considered as alternative measures to contain influenza viruses, particularly when vaccines and antiviral drugs are under development. However, thei r cost effectiveness in large urbanized populations is poorly understood. Method: To fill this knowledge gap, this research builds a spatially explicit network based model to simulate influenza transmission, mitigation strategies, and their associated cost s. This model represents a spatio temporal network of individuals' daily contacts which enable s the simulation of local infection and long distance dispersion of influenza. The workplace closure and travel restriction strategies, as well as their combinat ions with antiviral prophylaxis, are incorporated into this model to estimate their cost effectiveness in mitigating seasonal flu and pandemic flu. The metropolitan area of Buffalo, NY, USA, with a population about 1 million, is selected as the study area. Results: Without any intervention, the seasonal flu and pandemic flu cost $234.1 million and $331.1 million, respectively. The closure of 30% affected workplace is the most cost effective single strategy with $12.9K per case averted for seasonal flu and $ 34.9K for pandemic flu. The travel restriction is not cost effective if applied alone, but a 50% travel restriction in combination with antiviral prophylaxis and workplace closure forms the best strategy, which only costs $10.6K per cases averted for seaso nal flu and $ 13.5K for pandemic flu. Conclusions: For a large urbanized area, the closure of affected workplaces could be an effective and cost saving strategy to mitigate influenza outbreaks, while the highest cost effectiveness can be achieved by combini ng the travel restriction with antiviral prophylaxis and workplace closure. Categories and Subject Descriptors I.6 [ Simulation a nd Modeling ]: I.6.5 Model Development I.6.6 Simulation Output Analysis. General Terms Algorithms, Design, Economics, Human Fact ors Keywords Influenza, Mitigation Strategies, Cost Effectiveness, Social Network GIS, Agent based Simulation 1. INTRODUCTION Recent outbreaks of influenza around the world, such as the H5N1 bird flu and new H1N1 flu, have triggered a new wave o f exploring effective mitigation strategies [ 1 ] A wide variety of strategies has been tested for their effectiveness in disease control, including pharmaceutical strategies, for example, the vaccination and antiviral prophylaxis, and non pharmaceutical social distancing strategies, such as the workplace/school closure and travel restriction [ 2 4 ] In addition to their control effectiveness, the economic costs of these strategies are also of importance for health policy makers to consider, for instance, monetary costs of vaccines and loss of productivity from work absenteeism. A strategy that brings extremely heavy burden to the socio economy is often not recommended, even though it is effective to control disease spread. A feasible mitigation strategy should not only be effective, but also cost effective. A prior knowledge on the cost effectiveness of different mitigation strategies is essent ial for influenza risk management. In the current literature, cost effectiveness has been primarily evaluated for pharmaceutical strategies through medical experiments [ 5 8 ] Due to resource limits, these studies were often focused on a small group of population (100~1000 people in size), for example, healthcare workers, school children, and elderly people. The costs and benefits of pharmaceutical strategies remai n unknown for large populations. Further, social distancing strategies, such as the city wide workplace/school closure and travel restriction, are difficult to experiment in real practice, and their cost effectiveness is poorly understood. The lack of such knowledge may cloud decision makers when facing influenza epidemics in a large city, or nationwide pandemics. With the recent rise of computer based epidemic simula tion models, there have been a small number of studies attempting to estimate the cost effe ctiveness of influenza mitigation s trategies for large populations. For instance, Scuffham and West had evaluated vaccination and antiviral prophylaxis for elderly Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the ful l citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. HEALTHGIS 13 November 5 8, 2013 Orlando Florida USA Copyright 201 3 ISBN 978 1 4503 2529 5/1 3/11

PAGE 2

populations in three European countries [ 9 ] This work employed a linear epidemic model, and thus could not account for the dynamic, nonlinear effec ts of interventions in infectiou s disease s likely underestimating the cost effectiveness [ 10 ] To improve, Sander et al. [ 11 ] and Perlroth et al. [ 12 ] had proposed stochastic agent based models to simulate the spread of influenza and mitigation strategies in the United States. However, both studies assumed a hypothetic population with an age and sex make up matched to the overall US average. Hence, the ir results may not be applicable to dense urban populations where demographic characteristics and contact patterns are distinct [12]. Further, many details of populations were highly simplified, such as the geographic location of workplaces and the daily t ravel behaviors of people. The latest work by Milne et al. [ 13 ] intended to gauge mitigation strategies in a realistic town population in Australia, but the details about workplaces and human mobility were not considered either. The lack of these details prevents these studies from evaluating the workplace closure and the travel restriction strategies, which are widely recognized as potential policies to c ontain influenza outbreaks [ 3 ] To fill the knowledge void, t his article intends to evaluate the cost effectiveness of workplace closur e and travel restriction in mitigating influenza outbreaks for a realistic urban area. A spatially explicit network based model is developed as a tool for evaluation. The section that follows introduces the structure of the model with three major component s, including a flu simulator, a cost calculator, and mitigation strategies. The third section presents the simulation results and identifies cost effective strategies. The last section concludes this article with discussion. 2. MATERIALS AND METHODS In hea lth economics, the cost effectiveness of a mitigation strategy is commonly measured as a ratio between incremental costs and averted consequence of this strategy, as given in Equation 1: where i denotes a specific mitigation strategy and CER(i) is its cost effectiveness ratio ($ per case averted) The incremental or net costs of a strategy includes the cost of the strategy Stra tegy Costs(i) minus the savings from health outcomes averted by the strategy The savings are the difference between the costs of influenza with out mitigation strategies FluCosts(0) as a base line scenario and those with the mitigation strategy FluCosts(i) The denominator is expressed as the number of influenza cases averted by the strategy i i.e., the differences in case numbers between the baseline scenario and the strategy i Following Equa tion 1, a computer model is designed to simulate the influenza spreading process, the mitigation strategy against influenza, and the resulting costs, respectively. Therefore, the model has three major components : a flu simulator a cost calculator, and a s et of mitigation strategies (Figure 1) 2.1 Component 1: Flu simulator The flu simulator takes an a gent based approach to simulate the daily travel of individuals, their social contacts, and resulting influenza transmission in the study area (Buffalo, NY) This approach involves stochastic simulation, discrete time steps, and spatially explicit representation of individual mobility. Each of t he 985,001 individuals is conceptualized as a modeling unit with a set of characteristics (e.g., age, occupation, in fection status, time and location of daily activities) and behaviors (e.g., traveling between locations for activities and having contact with other individuals). They are first grouped and a llocated to 400,870 household location s in 967 census block group s under the constraints of census data so that the modeled population matches the age and household structure of the real study area. These individuals are further modeled to travel between homes and 36,839 business locations (schools, workplaces, or serv ice places) to carry out their daily activities, such as working, sho pping, and recreation, which expose them to the risk of infection. The assignment of individuals to locations utilizes a business database released by ReferenceUSA Inc. and a travel surve y report from the Greater Buffalo Niagara regional transportation council. The implementation adopts a n algorithm developed by Mao and Bian [ 14 ] which is not a focus of this research As illustrated in Figure 2, e ach 3D a day. Vertical segments on a trajectory indicate a stay at a location for some time, while horizontal segments indicate t ravel from one location to another. The intersections between trajectories as circled in the map, imply that individuals meet at the same time and location, and may have contacts with one another that allow influenza transmission Since individuals travel over time and location, th is model represents a spatio temporally Figure 1. Structural desi gn of the simulation model. The model is composed of three major components, including a flu simulator, a cost calculator, and a set of mitigation strategies. The simulation involves several databases as inputs, and produces flu attack rates, flu costs, an d strategy costs for cost benefit analysis Figure 2. 3D travel trajectories of 10 individuals in a day T he X and Y axis represent a geographic space of the study area and the Z axis showing time.

PAGE 3

varying contact network woven by individual mobility The transmission of influenza from one individual to another is stochastically simulated based on such a network. Each individual is al lowed to take one of four infection status during a time period, namely, susceptible, latent, infectious, and recovered [ 15 ] The progress of infection status follows the natural h istory of influenza, including the latent, incubation, and infectious periods. During the infectious period, individuals being infected may manifest symptoms and become symptomatic. Details on influenza parameter setting are provided in Table 1. Table 1 K ey parameter s for simulating influenza spread Parameters Values Literature Basic reproduction Number R 0 1.3 1.4 [ 16 17 ] Infection rates per contact I age Children (under 5) : 0.1 Youth (6 17): 0.08 Adults (18 64) : 0.08 Senior (65+) : 0.09 Calibrated based on R 0 Likelihood of symptom manifestation 0.5 [ 18 ] Latent period 2 days [ 19 ] Incubation period 3 days [ 19 20 ] Infectious period Children: 7 days Youth : 4 days Adults: 4 days Senior: 4 days [ 19 21 ] To initiate the diseas e transmission, five infectious individuals are randomly seeded into the study area at the first day of simulation, which then lasts for 150 days. In each day, the model traces susceptible contacts of infectious individuals, and simulates the next generati on of infections using the Monte Carlo method. The model had been validated using the CDC weekly reports of confirmed cases in 2004 05 flu season in the study area [ 14 ] 2.2 Component 2: Flu Cost Calculator The flu simulator describe d above identifies symptomatic individuals (influenza cases) during every simulat ion day. For each influenza case, a flu cost calculator is triggered to value the mone tary costs of this case in response to influenza following a tree structured flowchart shown in Figure 3 Except for the first branch on the left in Figure 3 all other branches in the flow chart are determined by probabilities adopted from the literatur e [ 22 23 ] as shown in Table 2. Influenza cases are separated into five age groups: under 5, 5 17, 18 49, 50 64, and over 65 years. According to an age specific probability each case is then categorized as either a high risk case or non high risk case. For either category, an influenza case may develop one of four possible outcomes: self care, outpatient, hospitalization, and death with probabilities estimated by Molinari et al. [ 22 ] from the literature, medical records, and reports. Those in the high risk ca tegory have greater chance to develop severe outcomes such as hospitalization and death [ 24 ] Table 2 Nationwide influenza parameters and distributions by age group, by risk and by health outcomes, all adopted from Molinari et al. [ 22 ] Influenza Parameters Age Group Mean Stan dard Deviation Likelihoods of high risk influenza cases 0 4 0.052 0.890 5 17 0.106 0.360 18 49 0.149 0.340 50 64 0.330 0.700 65+ 0.512 0.730 Pr (Outpatient visit| flu infection) Non High risk cases 0 4 0.455 0.098 5 17 0.318 0.061 18 49 0. 313 0.014 50 64 0.313 0.014 65+ 0.620 0.027 High risk cases 0 4 0.910 0.250 5 17 0.635 0.167 18 49 0.625 0.118 50 64 0.625 0.118 65+ 0.820 0.093 Pr (Hospitalization| flu infection) All risks 0 4 0.0141 0.0047 5 17 0.0006 0.0002 18 49 0.0 042 0.0014 50 64 0.0193 0.0064 65+ 0.0421 0.0140 Pr (Death| flu infection) All risks 0 4 0.00004 0.00001 5 17 0.00001 0.00000 18 49 0.00009 0.00003 50 64 0.00134 0.00045 65+ 0.01170 0.00390 Pr (Self care| flu infection) Non High risk 0 4 Pr (Self care| flu infection) =1 Pr (Outpatient visit| flu infection) Pr (Hospitalization| flu infection) Pr (Death| flu infection) 5 17 18 49 50 64 65+ High risk 0 4 Pr (Self care| flu infection) =1 Pr (Outpatient visit| flu infection) Pr (Hospitalization| flu infection) Pr (Death| flu infection) 5 17 18 49 50 64 65+ When the health outcome of an influenza case is determined, the direct and indirect costs associated with this outcome are estimated, and then summed into the tot al influenza costs to this individual (Figure 3) The direct costs come from the medical expenditure in response to influenza (e.g., hospitalizations, outpatient visits, and drug purchases), and vary over age groups, Figure 3 Workflow of flu cost calculator in the si mulation model to estimate flu costs to an influenza case The flu simulator first determines the sickness of an individual The cost calculator, then, uses the probabilities on branches to determine the associated health outcome and costs ( from Table 2 to 4)

PAGE 4

risk categories and outcomes. Details about the direct costs by age group, risk category, and health outcome are given in Table 3 Table 3 Direct costs of influenza by age group, by risk and by health outcome, all adopted from Molinari et al. [ 22 ] Cost per health outcome by age and risk group Medical (Direct) cost ($) Mean S .D. Distribution Self care All risks 0 4 3 2 Log normal 5 17 3 2 Log normal 18 49 3 2 Log normal 50 64 3 2 Log normal 65+ 2 2 Log normal Outpatient visit Non High risk 0 4 167 307 Log normal 5 17 95 258 Log normal 18 49 125 438 Log normal 50 64 150 766 Log normal 65+ 242 1,544 Log normal High risk 0 4 574 1,266 Log normal 5 17 649 1,492 Log normal 18 49 725 1,717 Log normal 50 64 733 1,307 Log normal 65+ 476 1,131 Log normal Hospitalization Non High risk 0 4 10,880 36,189 Log normal 5 17 15,014 86,804 Log normal 18 49 19,012 44, 636 Log normal 50 64 22,304 95,727 Log normal 65+ 11,451 23,128 Log normal High risk 0 4 81,596 123,626 Log normal 5 17 41,918 50,393 Log normal 18 49 47,722 85,644 Log normal 50 64 41,309 74,798 Log normal 65+ 16,750 32,091 Log normal Death Non High risk Mean S.D. Distribution 0 4 28,818 24,483 Log normal 5 17 28,818 24,483 Log normal 18 49 76,336 91,654 Log normal 50 64 118,575 333,879 Log normal 65+ 41,948 96,467 Log normal High risk 0 4 267,954 221,130 Log normal 5 17 267,954 221,130 Log normal 18 49 75,890 65,267 Log normal 50 64 118,842 345,973 Log normal 65+ 33,011 61,904 Log normal The indirect costs include the loss of productivity due to work/school absenteeism and due to death. The loss of productivity due to work/school absenteeism is calculated by multiplying the length of days absent from wo rk with the daily wage of a person ($145/day in Buffalo). For influenza cases ending with deaths, the productivity loss is estimated as the present value of lost earnings (PVLE), the projected earnings until T he average length s of work/school absenteeism and the PVLE by age group along with their statistical distribution s, are shown in Table 4 Table 4 Ind irect costs of influenza by age group, by risk and by health outcome, all adopted from Molinari et al. [ 22 ] Cost per health outcome by age a nd risk group Lost productivity (days) Mean Distribution Self care All risks 0 4 1.0 Poisson 5 17 0.5 Poisson 18 49 0.5 Poisson 50 64 0.5 Poisson 65+ 1.0 Poisson Outpatient visit Non High risk 0 4 1 Poisson 5 17 1 Poisson 18 49 1 Poisson 50 64 2 Poisson 65+ 3 Poisson High risk 0 4 6 Poisson 5 17 4 Poisson 18 49 2 Poisson 50 64 4 Poisson 65+ 7 Poisson Hospitalization Non High risk 0 4 8 Poisson 5 17 9 Poisson 18 49 12 Poisson 50 64 13 Poisson 65+ 13 Poisson High risk 0 4 31 Poisson 5 17 23 Poisson 18 49 21 Poisson 50 64 24 Poisson 65+ 18 Poisson Death Present value of lost earnings ($) Non High risk Mean S.D. Distribution 0 4 1,074,866 222,803 Log normal 5 17 1,276,012 900,934 L og normal 18 49 1,374,115 2,754,332 Log normal 50 64 521,083 1,588,835 Log normal 65+ 185,846 597,639 Log normal High risk 0 4 1,074,866 222,803 Log normal 5 17 1,276,012 900,934 Log normal 18 49 1,374,115 2,754,332 Log normal 50 64 521,083 1,588,835 Log normal 65+ 185,846 597,639 Log normal Since estimates in Table 3 and 4 chose 2003 as the base year this research inflated all the costs from 2003 to 2010 according to the consumer price index (CPI) of these two years. The CPI ratio between year 2010 and 2003 was set to 1.185 as reported by the

PAGE 5

Bureau of L abor After inflation, the costs of each influenza case can be valued The sum of costs to all influenza cases is the total costs of influenza to the study area, i.e., the FluCosts variable in Equation 1. 2.3 Component 3: Influenza mitigation strategies and their costs Table 5 shows the design of three mitigation strategies and their combinations. This research mainl y focuses on two social distancing strategies, namely, the workplace closure and travel restriction. A targeted antiviral prophylaxis (TAP) strategy is added to examine the cost effectiveness of combined strategies. The baseline scenario represents an infl uenza epidemic without any mitigation strategy, and is designed to compute FluCosts(0) in Equation 1. Table 5 Design of mitigation strategies and low high scenarios The wor kplace/ school closure strategy (WSC) shuts down a proportion of workplaces and schools where influenza cases are detected. Following Ferguson et al [ 3 ] workplaces and schools would be closed for 3 weeks once 1 case is detected. After reopened, they can close again if new cases occur. To account for the compliance issue, this research considers a low scenario (10%WSC) that closes 10% affe cted workplaces a long with 100% affected schools in a day, and a high scenario (30%WSC) that closes 30% affected workplaces a long with 100% affected schools. The costs of this strategy are measured as the loss of productivity due to work/school absenteeism For each closed place, the loss of productivity is the product of the number of employee s the daily wage per person ($145 on average), and the length of closure in days. The school absenteeism also causes loss of productivity of one parent because he/sh e needs to be absent from work to take care of children. The sum of productivity loss for all closed places is the total costs of this strategy, i.e., StrategyCost(i) in Equation 1. The travel restriction strategy (TR) aims to reduce the trips into and out of affected communities [ 3 25 ] Each of the 967 census block groups in the study area is treated as a community. The restriction is assum ed to last 2 weeks after the first case is detected in a community. Once the restriction is lifted, a community can be restricted again if new cases are identified. Following Germann et al. [ 26 ] a low level scenario (10% TR) restricts 10% trips into and out of all affected communities, while a high level scenario prohibits 50% trips (50% TR). The costs of this strategy come from the loss of productivity due to work/school absenteeism, because some pupil and staff cannot travel to schools and workplaces that are out of their residential communities The costs to a restricted indi vidual who cannot travel to workplace/school are the product of the average daily wage ($145) and the length of being restricted in days. The total costs of this strategy, StrategyCost(i) in Equation 1, are the sum of costs to all restricted individuals in the study area. The TAP strategy identifies influenza cases every day, searches their household members (up to 10), and then targets antiviral drugs to all these individuals [ 26 ] If an individual takes antiviral drugs, the chance of being infected and infecting others can be reduced by 70% and 40%, respectively [ 27 28 ] Each targeted individual is simulated to takes 20 Capsules of Tamiflu for 10 days with a market price of $130 in total. Thus, the total cost of this strategy, StrategyCosts(i) is the total number of targeted individuals multiplied by $130. To account for limited health personnel, this research evaluates two scenarios. The low scenario assumes that only 30% of influenza cases (30%TAP) can be identified during a day, while the high scenario assumes 60% (60%TAP), both following the desi gn in Germann et al. [ 26 ] In addition to testing the three control strate gies individually, their combinations are also evaluated. A low level combination scenario (referred to as the Combined Low) includes all three strategies at their respective low levels. Likewise, a high level combination (referred to as the Combined High) contains all three strategies at high levels. The total costs of a combined strategy are the sum of costs from all individual strategies. The TAP strategy is assumed to be implemented at the time when the total number and last until the end of the epidemic. The workplace/school closure and travel restriction strategies are assumed to be applied 2.4 Cost benefit analysis Each scenario in Table 5 as a strategy i is incorporated into the flu simulator and the cost calculator to measure the FluCosts(i) and StrategyCosts(i) The simulation is performed 50 realizations per scenario to reduce randomness i n results. The averaged FluCosts(0) FluCosts(i) and StrategyCosts(i) of a strategy scenario are then plugged into Equation 1 to estimate the cost effectiveness ratio CER(i) The comparison between scenarios would suggest the most cost effective way to co ntrol influenza with minimum disruptions on socio economy. 2.5 Sensitivity analysis To account for the uncertainties of influenza outbreaks, all scenarios in Table 5 are further tested in a context of pandemic influenza. The methods are the same as describ ed above, but the basic productive number R 0 is set to 2.0 to represent a highly infective flu strain that may lead to pandemics. The infection rates per contact by age group are the calibrated to this R 0 for flu simulation, while all other model parameter s remain unchanged. 3 RESULTS 3.1 Cost effectiveness of mitigation strategies Without any intervention (the baseline scenario), the seasonal influenza ( R 0 =1.4 ) may cause sickness to an average of 18.6% population (Table 6 ). This reasonably falls within t he range of seasonal flu attack rate from 5~20% reported by the Center for Disease Control and Prevention (CDC) [ 21 ] The total economic burden can be amounted to $231.4 million in the study area. Among the four non pharmaceutical scenarios, the restriction of 10% travels ( TR10%) is not recommended, because it even increases the influenza attack rate to 20% and aggravates the Scenario Strategy Low High Baseline No Interventions No Interventions Wor kplace/School Closure (WSC) 100% schools +10% workplaces 100% schools+33% workplaces Travel restriction (TR) 10% trips 50% trips Targeted Antiviral Prophylaxis ( TAP ) 30% cases 60% cases Combined C ombin ation of all above C ombin ation of all above

PAGE 6

A possible reason is that a low level restriction of travels is not sufficient to pro hibit the penetration of influenza into other communities, but meanwhile it intensifies the frequency of household contacts, leading to more infections. The rest of three scenarios do produce mitigation effects on influenza outbreak. The WSC10% is the mos t expensive strategy with the highest costs per case a verted ( $37. 2 K ), but it does not reduce infections remarkably. The WSC30% is quite effective in reducing the flu attack rate from 18.6% to 6.7%, and meanwhile its net costs are $349 million lower than t he WSC10%. A higher level of travel restriction (TR50%) seems to moderately contain the spread of influenza ( the attack rate drops 28%), but its high net costs ($1.65 billion) neutralize its control effectiveness. In summary, the workplace closure strate gy is more cost effective than the travel restriction. If the policy maker plans to choose a single strategy to control the flu outbreak, the WSC30% is a cost effective choice with $12.9K per case averted. This is particularly useful when no pharmaceutical treatments are available at early outbreaks of new influenza viruses. Table 6 Simulated economic costs of influenza control scenarios under R 0 =1.4 Scenario Di sease Attack R ate (%) Net Costs/Returns (Million $) Costs Effectiveness Ratio CER ($/case averted) Baseline 18.6 N/A N/A WSC10% 13.5 1866.3 37 1 51 WSC30% 6.70 1517.2 12,944 TR10% 20.01 345.4 N/A TR50% 13.3 1653.2 31 6 6 7 TAP30% 14.5 19.3 478 TAP6 0% 12.9 41.9 746 Combined Low 8.8 1742.3 18,048 Combined High 1.4 1796.6 10 60 4 *=FluCosts StrategyCosts. Positive values indicate net costs, while negative values are net returns. In contrast to the non pharmaceutical scenarios, two pharmaceutical s cenarios (TAP30% and TAP60%) could even save $400 800 for e ach averted case. This is not surprising because the closure of workplaces and schools causes the loss of productivity every day, and so does the travel restriction. If a workplace is closed for 21 days, the loss of productivity for a person can be amounted to $3,045, while treating this person with antiviral drugs only costs $130. However, these two TAP scenarios only produce 4~6 point drop in the disease attack rate, and neither could significantl y mitigate the influenza outbreak. It is also noteworthy that the TAP strategy requires a large stockpile of antiviral drugs, which may not be available at the early stage of influenza outbreaks. Therefore, the TAP strategy is suggested to be combined with social distancing strategies rather than deploying alone In the case that health resources allow a combined strategy, the high level combination of three individual strategies (Combined High) is recommended, because it almost eliminates the outbreak wit h only a little higher net costs than the Combined Low strategy. The costs per averted case are around $ 10 K the lowest among all scenarios. 3.2 Sensitivity analysis To test the sensitivity of recommendations made above, those control scenarios are also si mulated in a pandemic flu context ( R 0 =2.0). For a highly infective influenza strain, about 8% more people would feel sick due to infection, leading to a total flu costs of $331.1 million (Table 7 ). As a result, the costs per case averted are doubled or eve n tripled for most mitigation scenarios. In the case that antiviral prophylaxis was not available, the WSC30% would still be a cost effective option due to its relative lower costs per averted case ($35K) among individual scenarios. A low level (10%) restr iction of individual travels would again worsen the situation, and even the high level restriction produces few mitigation effects ( the attack rate only drops 14%). The WSC10% and TR50% produce similar mitigation effects, but both require a pretty high inv estment. The Combined_High scenario is the only one that reduces the attack rate under 10%. Meanwhile, the costs per averted case only increases 30% from $10.6 K for the seasonal flu ( R 0 =1.4) to $13.5K for the pandemic flu. This is the most effective and co st effective strategy to combat both seasonal and pandemic flu. Table 7 Simulated economic costs of influenza control scenarios under R 0 =2.0 *=FluCosts StrategyCosts Positive values indicate net returns, while negative values are net costs. 4 DISCUSSION In practice, health administrators need to balance the costs and benefits when choosing a mitigation strategy. This study shows that for a metropolitan area of 1 million people, an outbreak of seasonal influenza would cost $260 million, and approximately 8 10 times of such amount may need to totally invert its health outcomes. The pandemic influenza costs more, and requires more investments for mitigation. The results suggest that closing 30% affected workplaces every day could be a single cost effective s trategy to mitigate both seasonal and pandemic flu. If resources and manpower allow a combined strategy, the high level implementation of targeted antiviral prophylaxis and the two social distancing strategies (workplace closure and travel restriction) wou ld be the best strategy to combat influenza. The workplace closure and travel restriction strategies have been widely considered as alternative measures to contain emerging influenza viruses, particularly when vaccines and antiviral drugs are under develop ment [ 2 ] However, their cost effectiveness in large urbanized populations is poorly understood because few current models have explicitly considered the spatial location of workplaces and the travel behavior s of individuals between homes Scenario Disease Attack R ate (%) Net Costs/Returns (Million $) Costs Effectiveness Ratio CER ($/case averted) Baseline 26.4 N/A N/A WSC10% 23.4 2622.5 88,748 WSC30% 18.4 2750.9 34 ,910 TR10% 27.2 296.2 N/A TR50% 22.8 1775.9 50,082 TAP30% 22.9 5.9 171 TAP60% 21.4 17.6 357 Combined_ Low 19.8 2699.1 41,518 Combined_ High 6.2 2679.4 13,466

PAGE 7

and workplaces. T his study is the first attempt to evaluate the cost benefits of workplace closure and travel restriction strategies, and thus complement current understandings on social distancing strategies. Further, the mo del is implemented for a realistic urban population, rather than an US average population, and therefore the outcomes are more informative and appropriate to guide influenza containment in areas where the influenza is the most communicable. Although influe nza has been taken as an example, the proposed model can be easily extended to other emerging pathogens, such as the severe acute respiratory syndrome (SARS), by manipulating model parameters. Similar to any modeling analysis, this research has a number of limitations. First, the simulation model focuses on one US metropolitan area. It is possible that the model outcomes vary between cities with different demographics and standards of living. The interpretation of model outcomes should be limited to the stu dy area, or other similar areas. Nevertheless, the proposed model can be modified to fit other urban areas by changing input census, business, and travel survey data. Second, the model has not considered the active prevention of healthy individuals, assumi ng that they would not take any action to protect themselves unless being intervened. A coupled model of disease transmission and human preventive behavior could be considered to deal with this limit. Third, many model parameters used to calculate costs, such as the likelihood of being a high risk case and the length of work absenteeism, were estimated from a nationwide database by previous work, not specifically for the study area. The differences may bias the quantitative estimates, but may not be suffic ient to mislead the qualitative comparison between strategies. Last, a more comprehensive analysis of cost effectiveness should have included intangible costs, such as the loss of social values due to death and the long term effects of workplace/school clo sure. All these limitations warrant a future study. 5 REFERENCES [1] Coburn, B. J., Wagner, B. G. and Blower, S. Modeling influenza epidemics and pandemics: insights into the future of swine flu (H1N1). BMC Medicine 7 ( 2009), 1 8. [2 ] Dutta, A. The Effectiveness of Policies to Control a Human Influenza Pandemic: A Literature Review Social Science Research Network, City, 2008. [3] Ferguson, N. M., Cummings, D. A., Fraser, C., Cajka, J. C., Cooley, P. C. and Burke, D. S. Strategies for mitigating an influenza pandemic. Nature 442 ( 2006), 448 452. [4] Longini, I. M., Nizam, A., Xu, S., Ungchusak, K., Hanshaoworakul, W., Cummings, D. A. T. and Halloran, M. E. Containing pandemic influenza at the source. Science 309 ( 2005), 1083 1087. [5 ] Bridges, C. B., Thompson, W. W., Meltzer, M. I., Reeve, G. R., Talamonti, W. J., Cox, N. J., Lilac, H. A., Hall, H., Klimov, A. and Fukuda, K. Effectiveness and cost benefit of influenza vaccination of healthy working adults. JAMA: the journal of the Ame rican Medical Association 284 ( 2000), 1655 1663. [6] Yassi, A., Kettner, J., Hammond, G., Cheang, M. and McGill, M. Effectiveness and cost benefit of an influenza vaccination program for health care workers. The Canadian Journal of Infectious Diseases 2 ( 1991), 101. [7] Gasparini, R., Lucioni, C., Lai, P., Maggioni, P., Sticchi, L., Durando, P., Morelli, P., Comino, I., Calderisi, S. and Crovari, P. Cost benefit evaluation of influenza vaccination in the elderly in the Italian region of Liguria. Vaccine 2 ( 2002), B50 B54. [8] Luce, B. R., Zangwill, K. M., Palmer, C. S., Mendelman, P. M., Yan, L., Wolff, M. C., Cho, I., Marcy, S. M., Iacuzio, D. and Belshe, R. B. Cost effectiveness analysis of an intranasal influenza vaccine for the prevention of influenza in healthy children. Pediatrics 108 ( 2001), e24 e24. [9] Scuffham, P. A. and West, P. Economic evaluation of strategies for the control and management of influenza in Europe. Vaccine 20 ( 2002), 2562 2578. [10] Edmunds, W., Medley, G. and Nokes, D. Evalu ating the cost effectiveness of vaccination programmes: a dynamic perspective. Statistics in medicine 18 ( 1999), 3263 3282. [11] Sander, B., Nizam, A., Garrison Jr, L. P., Postma, M. J., Halloran, M. E. and Longini Jr, I. M. Economic evaluation of influen za pandemic mitigation strategies in the United States using a stochastic microsimulation transmission model. Value in Health 12 ( 2009), 226 233. [12] Perlroth, D. J., Glass, R. J., Davey, V. J., Cannon, D., Garber, A. M. and Owens, D. K. Health outcomes and costs of community mitigation strategies for an influenza pandemic in the United States. Clinical infectious diseases 50 ( 2010), 165 174. [13] Milne, G. J., Halder, N. and Kelso, J. K. The Cost Effectiveness of Pandemic Influenza Interventions: A Pand emic Severity Based Analysis. PloS one 8 ( 2013), e61504. [14] Mao, L. and Bian, L. Spatial temporal transmission of influenza and its health risks in an urbanized area. Computers, Environment and Urban Systems 34 ( 2010), 204 215. [15] Kermack, W. O. and McKendrick, A. G. A contribution to the mathematical theory of epidemics. Proceedings Royal Statistics Society 115 ( 1927), 700 721. [16] Mills, C. E., Robins, J. M. and Lipsitch, M. Transmissibility of 1918 pandemic influenza. Nature 432 ( 2004), 904 906. [17] Chowell, G., Miller, M. A. and Viboud, C. Seasonal influenza in the United States, France, and Australia: transmission and prospects for control. Epidemiology and Infection 136 ( 2007), 852 864. [18] Ferguson, N., Cummings, D. A. T., Cauchemez, S., F raser, C., Riley, S., Meeyai, A., Iamsirithaworn, S. and Burke, D. S. Strategies for containing an emerging influenza pandemic in Southeast Asia. Nature 437 ( 2005), 209 214. [19] Heymann, D. L. Control of Communicable Diseases Manual American Public Heal th Association, Washington, DC, 2004. [20] Halloran, M. E., Ferguson, N. M., Eubank, S., Ira M. Longini, J., Cummings, D. A. T., Lewis, B., Xu, S., Fraser, C., Vullikanti, A., Germann, T. C., Wagener, D., Beckman, R., Kadau, K., Barrett, C., Macken, C. A., Burke, D. S. and Cooley, P. Modeling targeted layered containment of an influenza pandemic in the United States. Proceedings of the National Academy of Sciences 105 ( 2008), 4639 4644.

PAGE 8

[21] CDC Key Facts About Seasonal Influenza (Flu) Center for Disease Control and Prevention, City, 2008. [22] Molinari, N. A. M., Ortega Sanchez, I. R., Messonnier, M. L., Thompson, W. W., Wortley, P. M., Weintraub, E. and Bridges, C. B. The annual impact of seasonal influenza in the US: measuring disease burden and costs. Vaccine 25 ( 2007), 5086 5096. [23] Meltzer, M. I., Cox, N. J. and Fukuda, K. The economic impact of pandemic influenza in the United States: priorities for intervention. Emerging infectious diseases 5 ( 1999), 659 671. [24] Fiore, A. E., Shay, D. K., Habe r, P., Iskander, J., Uyeki, T., Mootrey, G., Bresee, J. S. and Cox, N. J. Prevention and control of influenza: Recommendations of the Advisory Committee on Immunization Practices (ACIP). Morbidity and Mortality Weekly Report 56 ( 2007), 1 54. [25] Camitz, M. and Liljeros, F. The effect of travel restrictions on the spread of a moderately contagious disease. BMC medicine 4 ( 2006), 32. [26] Germann, T. C., Kadau, K., Longini Jr, I. M. and Macken, C. A. Mitigation strategie s for pandemic influenza in the United States. Proceedings of the National Academy of Sciences 103 ( 2006), 5935 5940. [27] Longini, I., Halloran, M. E., Nizam, A. and Yang, Y. Containing pandemic influenza with antiviral agents. American Journal of Epidem iology 159 ( 2004), 623 633. [28] Hayden, F. G. Perspectives on antiviral use during pandemic influenza. Philosophical Transactions of the Royal Society of London. Series B 356 ( 2001), 1877 1884.