76646438

Embed Size (px)

Citation preview

  • 7/28/2019 76646438

    1/12

    Risk Analysis, Vol. 32, No. 6, 2012 DOI: 10.1111/j.1539-6924.2011.01740.x

    Modeling Exotic Highly Pathogenic Avian Influenza Virus

    Entrance Risk Through Air Passenger Violations

    Jyh-Mirn Lai,1 Yi-Ting Hwang,2 and Chin-Cheng Chou3,

    The highly pathogenic avian influenza virus (HPAIV) is able to survive in poultry productsand could be carried into a country by air travelers. An assessment model was constructedto estimate the probability of the exotic viable HPAIV entering Taiwan from two neigh-boring areas through poultry products carried illegally by air passengers at Taiwans mainairports. The entrance risk was evaluated based on HPAIV-related factors (the prevalenceand the incubation period of HPAIV; the manufacturing process of poultry products; andthe distributionstoragetransportation factor event) and the passenger event. Distributionfunctions were adopted to simulate the probabilities of each HPAIV factor. The odds of pas-sengers being intercepted with illegal poultry products were estimated by logistic regression.The Monte Carlo simulation established that the risk caused by HPAIV-related factors fromarea A was lower than area B, whereas the entrance risk by the passenger event from areaA was similar to area B. Sensitivity analysis showed that the incubation period of HPAIVand the interception of passenger violations were major determinants. Although the resultshowed viable HPAIV was unlikely to enter Taiwan through meat illegally carried by airpassengers, this low probability could be caused by incomplete animal disease data and mod-eling uncertainties. Considering the negative socioeconomic impacts of HPAIV outbreaks,

    strengthening airport quarantine measures is still necessary. This assessment provides a pro-file of HPAIV entrance risk through air travelers arriving from endemic areas and a feasibledirection for quarantine and public health measures.

    KEY WORDS: Air passenger; highly pathogenic avian influenza; illegal poultry products; riskassessment

    1. INTRODUCTION

    Highly pathogenic avian influenza virus(HPAIV) infection became a public health issue af-ter some strains caused human deaths.(1) An incident

    1Department of Veterinary Medicine, National Chiayi University,Chiayi 600, Taiwan.

    2Department of Statistics, National Taipei University, Taipei 104,Taiwan.

    3School of Veterinary Medicine, National Taiwan University,Taipei 106, Taiwan.

    Address correspondence to Chin-Cheng Chou, School of Vet-erinary Medicine, National Taiwan University, No. 1, Sec. 4,Roosevelt Road, Taipei 106, Taiwan; tel: +886233661292; fax:886223630495; [email protected].

    of HPAIV infection in chicken was estimated to havecaused Hong Kong a loss of 0.551.55% in its grossdomestic product.(2) Entry of HPAIV into a disease-free area occurs through the movement of migratorybirds and the legal and illegal international trade oflive bird and avian products.(3) Viable HPAIV hasbeen isolated from frozen duck meat shipped fromChina to Japan and to South Korea.(4,5) Thus, therisk of transmission caused by contaminated poultryproducts should be examined.(6)

    International air passengers play an importantrole in the introduction of exotic animal diseasesas the convenience of air travel and reduced traveltime have made it easier for viruses to spread.(7)

    1093 0272-4332/12/0100-1093$22.00/1 C 2011 Society for Risk Analysis

  • 7/28/2019 76646438

    2/12

    1094 Lai, Hwang, and Chou

    Food brought into a country by passengers couldhave been contaminated by HPAIV.(7) Althoughpassengers are not allowed to bring meat and ani-mal products into Taiwan, some quantities may es-cape detection by Customs.(8,9) The number of peo-

    ple traveling to Taiwan was 13.1 million in 2010.Among them, 4.8 million people traveled amongTaiwan, China, Hong Kong, and Macao, and0.9 million Southeast Asians traveled to Taiwan.(10)

    These above-mentioned countries around Taiwan allhave reported HPAIV cases.(11) Given that it onlytakes 10 particles of HPAIV to initiate the spreadof disease,(12) the possibility of the virus crossing theborder via illegal imported poultry products by inter-national passengers should not be ignored.(6)

    The improvement of virus detection, applica-tion of antiviral agents,(13) and the investigationof virus hosts and environmental risk factors(1416)

    are methods adopted to prevent or control poten-tial HPAIV outbreaks. Most models focus on inter-ventions after HPAIV outbreaks(17,18) through re-liable diagnostic tests and early treatment,(17) andby controlling disease transmission(19) to reducethe attack rate and subsequent damages. However,there has not been a quantitative risk assessmentmodel to evaluate the risk of HPAIV being broughtinto a country by air passengers via contaminatedpoultry products. Decisionmakers in Taiwan haveapplied the results of a foot-and-mouth diseaseimported risk assessment model in their commu-

    nication with the general public and trading part-ners.(20) This study aimed to develop a risk modelbased on the experience of Lin et al.(20) to eval-uate the entrance risk of HPAIV brought intoTaiwan through poultry products carried illegally byair passengers.

    2. MATERIALS AND METHODS

    2.1. Modeling Approach

    HPAIV in meat products was hypothesized to

    pass illegally through airports to a country after fourconsecutive events. First, the meat would have tocome from an infected animal, which was definedas N1. Second, the viable virus would have survivedthe processing of the meat, which was defined asN2. Third, HPAIV would have retained its infectivitythroughout the distributionstoragetransportationphase, which was referred to as the DST factor, orN3. This covered the period from when the meathad been processed, transported, and deposited into

    shops, to when the meat was purchased and car-ried into international airports by airline passengers.Finally, the contaminated meat would have beenbrought into Taiwan illegally without being inter-cepted by Customs. This event was defined as N4.

    (20)

    The term intercepted indicates that the passen-gers who carried poultry products were caught ei-ther by sniffer dogs or by Customs officers duringcustoms inspection. Assuming the quality and quan-tity of HPAIV particles in meat did not change,N1, N2, and N3 were considered as independent fac-tors. N4 was referred to as the passenger factor andwas unrelated to N1, N2, and N3. Hence, the riskof HPAIV being introduced from a specific areainto a given area through international airports couldbe formulated as P[HPAIV introduction|area] =P[N1N2N3N4|area]. Based on the indepen-dent assumption, the conditional probabilities ofN1, N2, N3, and N4 could be calculated sep-arately. The overall risk of HPAIV being in-troduced would equal the sum of the proba-bilities in different areas, which was given asI

    i=1 P[HPAIVdissemination|area = i ], where i wasthe total number of areas.

    2.1.1. Risks of the Prevalence and the Incubation

    Period of HPAIV (N1)

    The incubation period of HPAIV was defined

    as the length of the time between exposure tothe pathogen and the first appearance of clinicalsigns.(21) The minimum, mode, and maximum daysof HPAIV incubation period were required to es-timate the pert distribution: Pert((minimum, mode,maximum)/365). The average number of fowls in-fected with HPAIV each year, denoted as xi, andthe average population of fowls at risk, denoted asni, could be used as parameters for the beta dis-tribution, as i = 1, 2, . . . , I. The distribution of theprevalence in a certain period for the ith area couldthen be expressed as Beta(xi + 1, ni xi + 1), and

    P(N1|area= i) could be generated from the followingdistribution:

    Beta(xi + 1, ni xi + 1)

    Pert((minimum, mode, maximum)/365), (1)

    where minimum, mode, and maximum are the corre-sponding measures for the distribution of the incuba-tion period.

  • 7/28/2019 76646438

    3/12

    Modeling Exotic HPAIV Entrance Risk Through Air Passenger Violations 1095

    Table I. The Quantities (in kg) of Illegal Poultry ProductsCarried by Passengers from Areas A and B, and Intercepted by

    Customs at Taiwans International Airports During July 2004 andJune 2006

    Type Ia Type IIb Type IIIc Total

    Area AChicken 151 303 1,826 2,280Duck 716 1,190 3,041 4,947Goose 0 252 423 645

    Area BChicken 136 7 359 502Duck 35 20 81 136

    aType I food refers to poultry products manufactured at or above70C, which are roasted, fried, boiled, or dried.bType II food refers to poultry products that are considered asHPAIV-risky, as they are processed below 70C.cType III food refers to poultry products whose processing meth-ods were not recorded.

    Note: Products are calculated separately according to the type ofrisks and meats.

    2.1.2. Risk of HPAIV Survival Postprocessing

    of Poultry Products (N2)

    The risk of HPAIV surviving postprocessing wasrelated to the level of contamination in meat andthe ability of HPAIV to survive postprocessing. Poul-try products confiscated by Customs were separatedinto three types (I, II, and III). Each type was fur-ther subdivided in relation to the type of meat (j),

    such as chicken (j = 1), duck (j = 2), and goose(j= 3). Type I category (Table I) contained thoseproducts with core temperatures during processingof 70C or above for a minimum period of2 minutes. These products were considered asHPAIV-safe due to the denaturing of the virus atthese temperatures.(20,21) Products that were mar-inated, smoked, or pan-fried were considered asHPAIV-risky because their core temperatures dur-ing processing were lower than 70C. These prod-ucts were grouped as Type II. Some of the confis-cated poultry products (5,730 of 8,510 kg) were not

    recorded properly and missing the details of theircooking methods or meat types. These products werecategorized as Type III as their risks could not beidentified through Customs inspection data.

    Since the virus could not survive in Type I prod-uct, only Type II and some of Type III products wereconsidered risky and their risks were estimated sep-arately. The probabilities of HPAIV surviving in ameat product from different areas after the manu-facturing process could be stated as P(N2|area i) and

    modified as:

    P[N2|area = i] =M

    j=1

    rijP[HPAIV dissemination

    during manufacturing | area = i, poultry =j],

    (2)

    where rij was the proportional weight of the jth type

    of poultry meat from the ith area andM

    j=1 rij= 1,where Mwas the total types of poultry meat from theith area. For a given area i, the distribution ofrij wasassumed to follow the Dirichlet distribution. Poultrymeats in this study were limited to chicken, duck, andgoose (j= 1 to 3).

    Products from areas A and B were calculatedseparately. Percentages and quantities were also cal-culated according to the type of meat and how it

    was processed, respectively. Because there was notenough information about Type III products, it wasassumed that Type III products contained both safeand HPAIV-risky items. It was also assumed that thepercentage of safe and risky products of this groupwas the same as the percentage of Type I (safe) andType II (risky) combination.

    Let Lijk denote the number of meat manufactur-ing methods (k) from the ith area and of the jth typeof poultry. The distribution of Pijk, i = 1, . . . , I; j=1, . . . , M; k= 1, . . . , Lij, was assumed to be a Dirich-let distribution and its parameters were the propor-

    tional weight of different meat products. Ten thou-sand samples generated from the Dirichlet distri-bution were used to estimate Pijk, i = 1, . . . , I; j=1, . . . , M; k= 1, . . . , Lij. Let the estimates be denotedas Pijk, i = 1, . . . , I; j= 1, . . . , M; k= 1, . . . , Lij. LetWIIIij denote the total weight of a Type III meat fromthe ith area and of the jth type of poultry. Then theweight of the kth manufacturing process could be ob-tained as QIIIijk = W

    IIIij

    Pijk.The total weight of Type II poultry products

    from the jth poultry in the ith area made by the kthmanufacturing process was denoted as QIIijk. Finally,it was assumed that the conditional distribution inEquation (2) was a beta distribution with parametersLij

    k=1QIIIijk and

    Lijk=1

    QIIijk.

    2.1.3. Risk of the DistributionStorage

    Transportation Factor Event (N3)

    The risk of HPAIV entering a country was de-pendent on HPAIV maintaining viability throughoutthe transportation and delivery of contaminated

  • 7/28/2019 76646438

    4/12

    1096 Lai, Hwang, and Chou

    meat. The DST factor included all possible stepsfrom processing of the meat for delivery to pur-chasing of the meat and carriage through air trans-port by consumers. Meat products were usuallykept chilled or frozen at a temperature between

    18 and 4

    C before they were sold to customers.Avian influenza virus maintained its viability be-low 4C chilled temperature.(22) It was assumedthat travelers did not throw away the goods theywere carrying during/after traveling. Let the sur-vival time of HPAIV and the traveling time be de-noted as S and T, respectively. The Pert distribu-tion was used to simulate factor T. The survivaltime was assumed to have an exponential distribu-tion with mean . Thus, based on the model con-structed for the above process, P(N3|meat) was cal-culated as the probability that the survival time of theHPAIV in poultry products was longer than the DSTtime of contaminated poultry products,(20) whichwas:

    P[S > ti | area = i] =

    ti

    1

    es/ds = e/ ti , (3)

    where ti denoted a possible transport time for a givenproduct from area i.

    2.1.4. Risk of the Passenger Event (N4)

    The probability of passengers illegally carryingHPAIV-contaminated products was assumed to fol-low the Pert distribution. The equation was modifiedfrom Lin et al.(20) and was described briefly as fol-lows.

    Let D= 1 denote the event when a passengerwith prohibited poultry products was intercepted,and D= 0 denote the event of a passenger carrying il-legal poultry products who was not intercepted. Theodds of carrying poultry products illegally were de-

    noted by = P[D=1]P[D=0] . A seasonal pattern in the num-

    ber of passengers might be expected.(20) Let j, j=

    1, 2, . . . , 12 denote the jth month of the year andj = 0 otherwise. Let the method of Customs in-spection of passengers traveling from the ith areabe denoted as F, where F= 1 indicated that Cus-toms officers did the inspection and F= 0 meantthat detective dogs did the inspection. Using a logis-tic regression conditioning on the month and methodof inspection of passengers traveling from the itharea, the probability of passengers being interceptedwith prohibited poultry products could be obtained

    from:

    P[D = 1|j, j= 1, . . . , 12, F, area = ai ]

    =e0+

    11j=1 jj+12 F+13ai

    1+ e0+11j=1 jj+12 F+13ai

    .(4)

    From Equation (4), the probability of passengerscarrying poultry products illegally, could also be ob-tained by:

    P[j, j= 1, . . . , 11, F|ai ]

    =

    Bh=1

    P[(j, j= 1, . . . , 11, F) Bh|ai ],

    where B denoted the number of all possible combi-nations of values ofj, j= 1, . . . , 11 and Bh denoteda possible combination. The empirical distribution

    was used to estimate Equation (4). Replacing the es-timate of Equation (4) and j, j= 0, 1, . . . , 13, wecould get:

    P[D= 1|ai ] =B

    h=1

    P[D= 1|j, j= 1, . . . , 11, F, ai ]

    P[(j, j= 1, . . . , 11, F) Bh|ai ].

    Using the probability calculated as above, an es-timate ( ) of could then be obtained.

    As it was not possible to know how many pas-senger violations escaped interception, it was as-

    sumed that the odds of illegal poultry product in-fected by HPAIV was proportional to , that is,P[D=0,E= 1|area=ai ]P[D=0,E= 0|area=ai ]

    = (ai ), where (ai ) denoted the

    odds of having illegally carried poultry products fromarea i, was a positive constant and needed to begiven, and E denoted the event when the prohib-ited poultry product was infected (E = 1) or not in-fected (E = 0) by HPAIV. Then the probability ofthe event of the illegal poultry product infected byHPAIV could be obtained as:

    P[E = 1, D= 0|area =ai ] =(ai )

    1+ (ai )

    , (5)

    where the derivation was similar to that in Linet al.(20)

    Equation (5) was an increasing function of, which indicated the status of the intercep-tion at Customs of the airports. It was able tosimulate the values of probability P(N4|ai) by

    Pert( 1(ai )1+1(ai ) ,2(ai )

    1+2(ai ),

    3(ai )1+3(ai )

    ), where the multi-

    pliers i, i = 1, 2, 3 had to be specified.(20)

  • 7/28/2019 76646438

    5/12

    Modeling Exotic HPAIV Entrance Risk Through Air Passenger Violations 1097

    2.2. Data for Analysis

    2.2.1. General Data

    Passenger records were obtained from TaoyuanInternational Airport (TPE), located in the north-

    ern part of Taiwan, and Kaohsiung International Air-port (KHH), located in the southern part of Tai-wan. They contained the total number and detailsof passengers arriving at Taiwan between July 2004and June 2006. During this period, international airpassengers could only enter Taiwan through thesetwo airports. The information on arriving passen-gers was obtained from the Immigration Office andthe National Police Agency. There are many coun-tries and areas around Taiwan, such as China, HongKong, Japan, Korea, and other countries in South-east Asia. Our previous studies(8,23) showed that ahigher percentage of travelers from the Peoples Re-public of China and its special administrative regions,named as area A in this study, and from the South-east Asian countries, named as area B, were caughtfor carrying animal products illegally. People livingin area A are mainly Chinese who share similar hol-idays and culture. The cultures and customs of peo-ple from area B, however, are different. Thus, in thisstudy, it was set that I = 2 to distinguish areas Aand B.

    Variables such as the month in which the illegalpoultry products were intercepted, the nationality ofthe passenger who carried them, and whether they

    was detected by Customs officers or by sniffer dogswere provided by the Animal Quarantine Authorityof the Bureau of Animal and Plant Health Inspec-tion and Quarantine. The number of air passengersarrived at Taiwan was 9,310,011 between July 2004and June 2005, and 9,877,124 between July 2005 andJune 2006. During the first period (July 2004 to June2005), 1,331 passengers were intercepted for illegallycarrying poultry products. Among them, 720 camefrom area A and 370 came from area B. In the fol-lowing period (July 2005 to June 2006), 1,694 pas-sengers were intercepted. Among them, 1,002 came

    from area A and 527 came from area B. Between July2004 and June 2005, sniffer dogs caught 556 of thosepassengers inspected and Customs officers caught534 passengers. Between July 2005 and June 2006,845 of 959,048 passengers inspected by sniffer dogs,and 684 of those 5,226,326 passengers inspected byCustoms officers, were found to have carried illegalpoultry products.

    2.2.2. Data for Modeling the Prevalence and the

    Incubation Period of HPAIV (N1)

    Reports of fowls infected with HPAIV wereobtained from the World Organization for AnimalHealth (OIE).(11) Disease information of the reportseparates the number of animals in the new out-break into the number of susceptible, cases, deaths,destroyed, and slaughtered. Not all numbers werelisted on a report. Thus, the average number peryear of the combination of susceptible and the casenumbers from 2004 to 2006 were calculated to gen-erate the parameter xi. Figures of fowls at riskwere obtained from the Food Agriculture Organi-zation (FAO).(24) Hence, the parameters requiredfor the beta distribution x1 and n1 of area A were6,495,329 and 5,308,910,667, respectively.(11,24) Theparameters x2 and n2 of area B were 33,908,733 and

    1,977,941,667, respectively.According to the HPAI data released by the

    OIE(25,26) and the study of Tollis and Di Trani,(27)

    the minimum, mode, and maximum incubation pe-riods are 0, 3, and 21 days, respectively. Implement-ing these empirical data in Equation (1), P(N1|areai) could then be obtained.

    2.2.3. Data for Modeling the Risk of HPAIV

    Survival After the Manufacturing Process (N2)

    The information supplied by the Customs

    showed that poultry products confiscated weremainly roasted, fried, marinated, boiled, dried,smoked, or pan-fried. Only smoked, marinated,boiled, and pan-fried products were consideredas HPAIV-risky. Table I provides the quantityof intercepted poultry products according to thedegree of risk (HPAIV-safe, HPAIV-risky, andrisk-uncertain), type of meat (chicken, duck, orgoose), and the areas where the meat came from(area A or B).

    2.2.4. Data for Modeling the Risk of DST (N3)

    Based on the flight schedules at the two airports,the minimum, mode, and maximum travel times fromarea A to Taiwan were 2, 4, and 24 hours, and thosewere 2, 7, and 24 hours from area B to Taiwan, re-spectively.(28,29) These data were treated as parame-ters in the Pert distribution to simulated Tand let thesimulated value be denoted as ti , i = 1, 2.

  • 7/28/2019 76646438

    6/12

    1098 Lai, Hwang, and Chou

    Table II. The Parameter Estimates and the Corresponding Standard Errors (SEs) in the Logistic Regression and the Odds of PassengersBeing Intercepted at Taiwans International Airports, Listed by Month, Origin, and Method Used for Interception, per Year Between July

    2004 and June 2006

    Odds ( 103)

    Area A Area B

    Customs Sniffer Dogs Customs Sniffer DogsEstimates

    Month (SE) Yeara Yearb Yeara Yearb Yeara Yearb Yeara Yearb

    Jul. 0.3446 0.10 0.07 0.62 1.13 0.36 0.28 5.26 4.53(0.1188)

    Aug. 0.0427 0.14 0.11 0.55 0.92 0.26 0.41 5.20 4.88(0.1074)

    Sep. 0.0720 0.08 0.09 0.30 1.49 0.21 0.26 6.38 7.67(0.1101)

    Oct. 0.0146 0.12 0.09 0.40 2.81 0.29 0.30 5.50 7.13(0.1117)

    Nov. 0.2152 0.12 0.30 1.07 3.29 0.33 0.64 7.71 7.15(0.1183)

    Dec.c 0.14 0.33 1.03 3.32 0.26 0.76 5.84 8.66

    Jan. 0.3598 0.19 0.80 1.15 3.92 0.29 0.65 7.90 11.8(0.1010)

    Feb. 0.4165 0.17 0.29 2.15 4.44 0.28 0.59 7.23 12.4(0.0990)

    Mar. 0.4577 0.11 0.21 1.55 3.29 0.23 0.71 7.94 11.6(0.0969)

    Apr. 0.2387 0.11 0.29 1.41 1.73 0.34 0.85 8.60 9.07(0.1002)

    May 0.1647 0.09 0.23 1.21 1.92 0.30 0.55 6.25 7.01(0.1011)

    Jun. 0.3422 0.10 0.12 1.59 2.20 0.37 0.48 7.21 5.44(0.0990)

    aJuly 2004June 2005.bJuly 2005June 2006.cReference month.

    The parameters for the survival time were esti-mated as follows. The virus will lose its infectivityafter 30 min at 56C or after 1 day at 28C.(30) Infeces, the period can extend to up to 6 days. (31) Dr.Tsai, an avian disease expert at the National Tai-wan University, suggested that the survival time ofthe HPAIV in poultry products ranges between 4and 7 days. Then, it was reasonable to assume thatHPAIVs minimum, mode, and maximum survivaltimes were 4, 6, and 7 days, respectively. These num-

    bers were used as the parameters in the Pert distribu-tion to generate the possible values denoted as ofmean survival time. Hence, P(N3) was calculated.

    2.2.5. Data for the Risk of the Passenger Event (N4)

    The estimates j, j= 0, 1, . . . , 13 of j, j=0, 1, . . . , 13, in the logistic regression are shown inTable II. Using j, j= 0, 1, . . . , 13 the estimatedprobability P[D= 1|ai ] could be obtained based on

    Equation (4). Furthermore, we set 1 = 0.5, 2 = 1,and 3 = 2.

    The results of the empirical inspection proba-bilities in each month during the studied period areshown in Table III. The derived estimates P[E = 1,D = 0|area = ai] were Pert (1.13 10

    4 ( = 0.5),2.25 104 ( = 1), 4.51 104 ( = 2)) for area Aand Pert (1.28 104 ( = 0.5), 2.57 104 ( = 1),5.14 104 ( = 2)) for area B, respectively.

    2.3. Analysis Tools

    The model used a Monte Carlo simulation torun 10,000 times in both the Modelrisk 4.1(32) andMicrosoft ExcelTM spreadsheet softwares. Sensitiv-ity analysis was used to evaluate the parameters inevents from N1 to N4. Factors whose rank scores(RC) were higher than 0.1 were evaluated. The spi-der graph and the figures in the tornado graphs

  • 7/28/2019 76646438

    7/12

    Modeling Exotic HPAIV Entrance Risk Through Air Passenger Violations 1099

    Table III. The Empirical Inspection Probabilities in Each MonthBetween July 2004 and June 2006, for Customs Officers and

    Sniffer Dogs, and for Products from Areas A and B, at TaiwansInternational Airports

    Empirical Probability (102)

    Customs Sniffer Dogs

    Month Area A Area B Area A Area B

    Jul. 8.40 8.53 0.81 1.14Aug. 8.19 8.73 0.84 1.20Sep. 7.18 7.37 0.84 0.85Oct. 7.53 6.81 0.81 1.17Nov. 6.39 6.10 0.95 1.08Dec. 6.62 6.41 0.86 1.08Jan. 6.53 6.05 1.28 1.01Feb. 7.28 7.93 0.98 1.49Mar. 7.13 6.64 1.28 1.72Apr. 8.02 6.92 1.30 1.57

    May 7.48 6.96 1.36 1.79Jun. 6.68 5.37 1.23 2.08

    supplied by Modelrisk 4.1 were used to evaluate theinfluence of those variables in this model.

    3. RESULTS

    Details of the Type III poultry products inter-cepted at the airports are shown in Table IV. Underthe validity of the logistic regression model in Equa-

    tion (4), P[D= 1|j, j= 1, . . . , 11, F, ai ] was esti-mated by:

    logit(P[D = 1|j, j= 1, . . . , 11, F, ai ])

    = 9.11+11

    h=1

    hh + 2.77F1.21ai . (6)

    As shown in Table II, the effect of the variable Fon Equation (6) was significant (p < 0.0001). The fac-tor of area was not a significant predictor. Based onthe residual analysis and the Pearson chi-square testfor the goodness of fit, the fitting model was shown to

    be appropriate (p < 0.0001).The odds of interception by sniffer dogs and Cus-toms officers ranged from 0.30 103 to 4.44 103

    and from 0.07 103 to 0.80 103, respectively,for passengers from area A; and from 4.53 103 to12.4 103 and 0.21 103 to 0.85 103, respec-tively, for passengers from area B, in each monthbetween July 2004 and June 2006 (Table IV). Theodds of passengers being intercepted by sniffer dogsfrom 2005 to 2006 were higher than those from

    Table IV. The Estimated Quantities (in kg) of Type III ProductsListed According to All Possible Manufacturing Methods, andthe Estimated Percentage of Each Manufacturing Technique

    Possible Processing Methods of PoultryProducts (Lijk)

    Area A (i = 1)Chicken (j= 1) Roasted Fried Marinated Boiled Dried

    0.063 0.039 0.663 0.019 0.216Quantities (kg) 115.07 71.38 1210.43 34.67 394.17SDa 10.44 8.31 20.17 5.83 17.58Duck (j= 2) Roasted Smoked Marinated Dried

    0.368 0.149 0.476 0.007Quantities (kg) 1119.53 453.18 1447.63 21.27SD 26.80 19.76 27.62 4.59Goose (j= 3)b Smoked Marinated Pan-Fried

    0.020 0.823 0.157Quantit ies (kg) 8.45 347.80 66.35SD 0.06 6.13 1.10

    Area B (i = 2)

    Chicken (j= 1) Fried Marinated Boiled Dried0.194 0.036 0.162 0.608

    Quantit ies (kg) 69.64 12.96 58.16 218.23SD 7.48 3.52 6.94 9.20Duck (j= 2) Roasted Marinated Dried

    0.633 0.361 0.006Quantities (kg) 51.30 29.19 0.49SD 4.33 4.33 0.70

    aStandard deviation.bAll types of goose meat were considered HPAIV-risky.

    2004 to 2005. The odds of intercepting passengers

    by sniffer dogs were about 6.99 times (e1.9439) higherthan of being intercepted by customs officers. Theprobabilities of intercepting passengers with illegalpoultry products in January, February, March,June, and July were significantly higher than inDecember (p < 0.005).

    The simulated values of survival probabilities ofHPAIV getting through the four consecutive eventsin areas A and B are presented in Table V. The lowerprevalence in area A causes the lower risk of N1,compared with that in area B. The risks of N2 andN3 in area A were slightly higher. The risks of N4 in

    both areas were similar. The values of the DST fac-tors ranged from 0.91 to 0.98 in area A, and 0.09 to0.14 in area B. The median probability of the HPAIVentering Taiwan via offending passengers from areaB (1.02 108) was 5.93 times higher than that fromarea A (1.72 109).

    The results of sensitivity analysis are shown inFig. 1. When the risk model of area A was eval-uated according to the factors rank scores (RC),the incubation period of HPAIV (RC = 0.91), the

  • 7/28/2019 76646438

    8/12

    1100 Lai, Hwang, and Chou

    Table V. Simulated Values for Probabilities of HPAI Viruses Surviving Through Illegal Air-Passenger-Carried Contaminated PoultryProducts to Enter Taiwan from Areas A and B Each Year

    Area Factors/Event/Probability 5th Percentile Median 95th Percentile

    A HPAI prevalence 1.22 103 1.22 103 1.22 103

    HPAI risk of ingredient (N2) 2.90

    10

    6

    1.62

    10

    5

    4.04

    10

    5

    Manufacture 3.46 101 4.93 101 6.33 101

    Manufacture and TSD (N3) 9.05 101 9.56 101 9.82 101

    HPAIV factor (N1 N2 N3) 1.29 106 7.51 106 1.99 105

    Passenger event (N4) 1.49 104 2.36 104 3.51 104

    Probability ( = 0.5) 3.12 1010 1.73 109 4.31 109

    Probability ( = 1) 6.24 1010 3.46 109 8.61 109

    Probability ( = 2) 1.25 1019 6.92 109 1.72 108

    Probability (AI|area A) 2.99 1010 1.72 109 5.11 109

    B HPAI prevalence 1.71 102 1.71 102 1.71 102

    HPAI risk of ingredient 4.06 105 2.28 104 5.66 104

    Manufacture 9.52 102 1.15 101 1.38 101

    Manufacture and TSD 8.73 102 1.11 101 1.38 101

    HPAIV factor 5.78 106 3.73 105 1.34 104

    Passenger event 1.71 104 2.83 104 4.04 104

    Probability ( = 0.5) 4.91 109 2.74 108 6.85 108Probability ( = 1) 9.82 109 5.49 108 1.37 107

    Probability ( = 2) 1.96 108 1.09 107 2.74 107

    Probability (AI|area B) 1.53 109 1.02 108 3.89 108

    Fig. 1. This spider graph of sensitivity analysis shows the risks from areas A (left) and B (right). Items 1 and 5 are the events of HPAIVsincubation period; item 2 is the percentage of smoked and marinated duck meats; item 3 is the percentage of risky meat. Items 4 and 8 arepassenger events (N4); item 6 is the percentage of marinated chicken meat; item 7 is the percentage of marinated duck meat.

    passenger event (RC = 0.30) and the factor of N2(RC = 0.22), including the risk from duck meat (RC= 0.20), were all highly related. The figures in the tor-nado chart showed that four factors had a RC higherthan 0.1, when the risk model of area B was evalu-ated. They were the incubation period (RC = 0.876),

    the N4 event (RC = 0.27), the risks from chickenmeat (RC= 0.47), and the risks from duck meat (RC= 0.17). A spider graph (Fig. 1) shows that the in-cubation period of HPAIV in both areas drags theoutcome dramatically. When the estimated value ofthe incubation period in area A lies in its 010%

  • 7/28/2019 76646438

    9/12

    Modeling Exotic HPAIV Entrance Risk Through Air Passenger Violations 1101

    range, the total mean risk is approximately 3.19 1010. When the incubation period lies in its90100%range, the total mean risk goes up to 4.80 109.Other factors change the outcome mildly.

    4. DISCUSSION

    This study aimed to develop an expandablemodel using the results of different risk analyses fromdifferent areas that would be able to simulate the riskof poultry products illegally carried by air passengers.Although different distributions in this study wereused to estimate the risk in only two areas, it is stillpossible to apply the same model for many more ar-eas. According to the parsimony theory, models withfewer parameters would allow for an easier establish-ment. Some negligible factors could be replaced withsuitable assumptions to keep the model more sta-ble. The above concepts are also considered in char-acterizing the survival patterns of HPAIV in prod-ucts, while the Weibull or the log-logistic regressiondistributions could be considered. Admittedly, theWeibull distribution could fit the curve better. How-ever, exponential distribution is still the basic form ofthe Weibull distribution, and requires fewer parame-ters.(33) Moreover, different distributions still presentalmost the same results using the current data. Thepurpose of using these assumptions in the calculationwas to keep the equation stable. Nevertheless, thestudy still supplies a general form to let users incor-

    porate more information into this model if it is nec-essary, such as available incubation times of HPAIVfor different avian species.

    The beta distribution is suitable to estimatethe uncertainty of the period-prevalence.(32) AreaBs prevalence is 14 times higher than area As(Table V). If the prevalence in areas A and B is thesame, the risk of area A will be 4.82 108, which isabout 14 times higher than the original outcome andabout 2.8 times higher than the outcome in area B.However, this result probably should be reevaluatedcarefully because the reporting of fowls with HPAIV

    clinical signs to the OIE is voluntary, and becausethe figure from area B was not for just one country,but the sum of the possible cases with HPAIV clin-ical signs from different countries. It is possible tosimulate the risk of individual countries with the cur-rent model. However, our previous study(20) showedthat data collected from two areas would be sufficientenough to establish this model.

    According to the animal disease notification sys-tem, no matter how many animals in a group were

    infected with HPAIV, the whole group should be re-ported altogether due to the highly contagious innatecharacter of HPAIV, and its total number will berecorded. Considering the potential risks to the pub-lic health and the limitations of detection methods,

    reporting only those infected cases will not be practi-cal and may underestimate the extent of the disease.Transparency in reporting is an important factor todetermine the risk accurately. However, some coun-tries only reported the number of suspected cases orthe number of animals destroyed. Although the re-sults of the sensitivity test show that prevalence is nota prime determinant factor, it can still increase therisks. Government can check the updated prevalenceto decide whether its necessary to raise the level ofsecurity in airports.

    The definition of the incubation period is statedin Section 2.1.1. It could be the time before the gov-ernment reports an outbreak to the OIE or whenthe government decides to take some actions, suchas to stamp out all susceptible birds. The influenceof the incubation period factor of HPAIV is signif-icant (Fig. 1). The high and low values in the fac-tor have dragged the outcome dramatically. Usu-ally, values of incubation period should be locatedaround the mean. Using the Pert distribution canincrease the sensitivity by up to four times whenthe values are mainly sitting around the mode, es-pecially when compared with the triangle distribu-tion.(32) Thus, the accurancy of the model increases.

    However, unless a country is given as much as21 days to prepare for an HPAIV outbreak, the riskof HPAIV crossing the border still exists. Gener-ally, H5N1 outbreaks were more likely to occur inwinter and early spring.(34) A more intensive inter-ception program before the potential outbreak sea-sons, such as increasing the in the model, should beuseful. H5N1 outbreaks also tend to match bird mi-gration patterns.(34) Sentinel ducks have been foundto be infected with influenza virus as early as six toeight weeks in an area.(35) Strengthening of wild birdssurveillance should be a possible way to minimize the

    risk and to increase the warning time.The value of N2 in area A is 4.3 times higherthan that in area B. This result shows the productsfrom area A are more likely to carry HPAIV thanthose originated from area B. It is probably becausepeople traveling from area A prefer to bring meatproducts, especially those highly HPAIV-risky oneslike duck (Table II). Different cultures often havedifferent cooking techniques and their own favoritefood. Maybe this is why travelers from area A are

  • 7/28/2019 76646438

    10/12

    1102 Lai, Hwang, and Chou

    more likely to bring duck products and visitors fromarea B tend to carry chicken products (Table I). Fac-tor N2 is considered to have a moderate influenceon the likelihood of HPAIV being brought in bypassengers (Fig. 1). As its risk ranges from 9.52

    10

    2 to 6.33 10

    1 (Table V), Customs may needto do more to lower the risk.

    The DST factor doesnt seem to have a signifi-cant effect in this model (Fig. 1). Because the rangeofN3 is narrow (Table V), its influence in this modelis very mild. The DST factor comprises three parts:distribution, storage, and transportation. However,the current model has indicated that separate equa-tions for each part may not be necessary. HPAIVcannot survive more than one day at 28C.(36) Thus,the longer the DST time is, the less likely the viruswill survive. Because the virus may die in the firsttwo stages, the overestimation of the risk in this stageshould be noted. However, it is suggested that Cus-toms officers should check the manufacture date toevaluate the risk of a product.

    The passenger event (N4) is a significant corre-lated factor (Fig. 1). The number of violations variessignificantly from month to month. Travelers behavedifferently in different months, especially in the pe-riods between January and March and from June toJuly. Chinese usually celebrate their New Year fes-tival between January and March; and the DragonBoat festival in June or July. These months are alsowhen people like to travel to Taiwan, and to visit

    friends with gifts, which often include delicacies fromtheir homeland. As this study has shown that nearlyall of the meat products should be considered asrisky, it is essential that travelers be banned from car-rying meat products into Taiwan.

    The multiplier describes the relationship be-tween the ratio of P[E = 1, D = 0|ai), P(E = 0, D =0|area i) and (ai). Value indicates the strength ofquarantine measures. It is also used to show whatcould be considered as proper types of measures atdifferent airports. KHH has a higher violation rateof passengers and its inspection rate of sniffer dogs

    is also higher than TPE. Because TPE receives morepassengers and flights than KHH, the number of snif-fer dogs in TPE may not be enough to maintain ef-fective inspection and quarantine. Since sniffer dogshave a better interception rate than Customs offi-cers,(23) increasing the number of sniffer dogs shouldbe encouraged.

    Based on the results estimated by the current ap-proach model, air passenger violations do not imposea substantial risk for the introduction of HPAIV into

    Taiwan. Prevention of HPAIV outbreak from otherpossible routes seems more important than the cur-rent assessed route.

    5. CONCLUSIONS

    The study indicated that the current quarantinemeasure is sufficient to prevent HPAIV entrance andif strictly enforced, these practices would be also ef-fective enough to keep the virus out of the border inthe future. Although the values of HPAIV entrancerisk seem relatively low, the zoonotic potential of theHPAIV of subtype H5N1 still cannot be ignored.(37)

    It is critical that the government continues strength-ening its border control to prevent the HPAIV fromentering. The general public should also be madeaware of poultry products from HPAIV-endemic ar-eas. Furthermore, as more incidents were recorded in

    some particular months, authorities should enforcestricter security inspections in or before these high-risk months. This study also provides a tool for othercountries in similar situations to implement properstrategies to avoid the introduction of the pathogenover their borders.

    ACKNOWLEGMENTS

    The authors are grateful for the valuable com-ments from the editors and referees. The authorswould like to thank the Bureau of Animal and Plant

    Health Inspection and Quarantine (95-AS 13.1.1BQ B3(2), 96-AS 14.1.1 BQ B3(1), 97-AS 14.1.1BQ B1(5)) and the National Science Council (NSC962313-B-002045-MY3) for funding the project,and the Immigration Office, National Police Agencyfor providing information on arriving passengers.They would also like to thank Professor Chin-TsangChiang and Ms. Tzu-Jung Huang from the Depart-ment of Mathematics of National Taiwan Universityfor their valuable suggestions as well.

    REFERENCES

    1. Capua I, Alexander DJ. Human health implications of avianinfluenza viruses and paramyxoviruses. European Journal ofClinical Microbiology and Infectious Diseases, 2004; 23(1):16.

    2. McLeod A, Morgan N, Prakash A, Hinrichs J, FAO (AGALand ESCB). Economic and social impacts of avian in-fluenza. 2010, Food and Agriculture Organization of theUnited Nations, 2010. Available at: http://www.fao.org/avianflu/documents/Economic-and-social- impacts-of-avian-influenza-Geneva.pdf, Accessed on Septemer 21, 2010.

    3. Alexander DJ, Manvell RJ, Irvine R, Londt BZ, Cox B,Ceeraz V, Banks J, Browna IH. Overview of incursions ofAsian H5N1 subtype highly pathogenic avian influenza virus

  • 7/28/2019 76646438

    11/12

    Modeling Exotic HPAIV Entrance Risk Through Air Passenger Violations 1103

    into Great Britain, 20052008. Avian Disease, 2010; 54(1Suppl):194200.

    4. Tumpey TM, Suarez DL, Perkins LE, Senne DA, Lee JG,Lee YJ, Mo IP, Sung HW, Swayne DE. Characterization of ahighly pathogenic H5N1 avian influenza A virus isolated fromduck meat. Journal of Virology, 2002; 76(12):63446355.

    5. Mase M, Eto M, Tanimura N, Imai K, Tsukamoto K,Horimoto T, Kawaoka Y, Yamaguchi S. Isolation of a geno-typically unique H5N1 influenza virus from duck meat im-ported into Japan from China. Virology, 2005; 339(1):101109.

    6. Harder TC, Teuffert J, Starick E, Gethmann J, Grund C,Fereidouni S, Durban M, Bogner KH, Neubauer-Juric A,Repper R, Hlinak A, Engelhardt A, Nockler A, SmietankaK, Minta Z, Kramer M, Gobig A, Metteneiter TC, ConrathsFJ, Beer M. Highly pathogenic avian influenza virus (H5N1)in frozen duck carcasses, Germany, 2007. Emerging InfectiousDisease, 2009; 15(2):272279.

    7. Ejaz R, Ahmed Z, Siddique N, Naeem K. Chicken meat as asource of avian influenza virus persistence and dissemination.International Journal of Poultry Science, 2007; 6(12):871874.

    8. Shih TH, Chou CC, Morley RS. Monte Carlo simulationof animal-product violations incurred by air passengers at

    an international airport in Taiwan. Preventive VeterinaryMedicine, 2005; 68(24):115122.9. Pharo HJ. Determination of the acceptable risk of introduc-

    tion of FMD virus in passenger luggage following the UK out-break in 2001. Available at: http://www.biosecurity.govt.nz/files/regs/imports/risk/paper-acceptable-risk-fmd.pdf, Ac-cessed on June 20, 2011.

    10. National Immigration Agency. [Statistical data]. Available at:http://www.immigration.gov.tw/mp.asp?mp = 1, Accessed onAugust 8, 2011.

    11. World Organization for Animal Health. Worldwide regionalrepresentation for 20042006. Available at: http://www.oie.int/downld/AVIAN%20INFLUENZA/A2004 AI.php,http://www.oie.int/downld/AVIAN%20INFLUENZA/A2005 AI.php, http://www.oie.int/downld/AVIAN%20INFLUENZA/A2006 AI.php, Accessed on October 17, 2010.

    12. Sabirovic M, Toth B, Coulson N. Veterinary risk assessment,

    Department for Environment Food and Rural Affairs, 2010.Available at: http://www.defra.gov.uk/foodfarm/ farmanimal/diseases/control/documents/captivebirds-quarantine.pdf, Ac-cessed on October 15, 2010.

    13. Zeitlin GA, Maslow MJ. Avian influenza. Current Allergyand Asthma Reports, 2006; 6(2):163170.

    14. Minh PQ, Morris RS, Schauer B, Stevenson M, Benschop J,Nam HV, Jackson R. Spatio-temporal epidemiology of highlypathogenic avian influenza outbreaks in the two deltas ofVietnam during 20032007. Preventive Veterinary Medicine,2009; 89(12):1624.

    15. Goutard F, Roger F, Guitian J, Balanc G, Argaw K, De-missie A, Soti V, Martin V, Pfeiffer D. Conceptual frame-work for avian influenza risk assessment in Africa: The caseof Ethiopia. Avian Disease, 2007; 51(1 Suppl):504506.

    16. Thomas ME, Bouma A, Ekker HM, Fonken AJM,

    Stegeman JA, Nielen M. Risk factors for the introduction ofhigh pathogenicity avian influenza virus into poultry farmsduring the epidemic in the Netherlands in 2003. PreventiveVeterinary Medicine, 2005; 69(12):111.

    17. Alexander ME, Moghadas SM, Rost G, Wu J. A delay differ-ential model for pandemic influenza with antiviral treatment.Bulletin of Mathematical Biology, 2008; 70(2):382397.

    18. Arino J, Brauer F, van der Driessche P, Watmough J, Wu J.Simple models for containment of a pandemic. Journal of theRoyal Society Interface, 2006; 3(8):453457.

    19. Wu JT, Riley S, Leung GM. Reducing the impact of the nextinfluenza pandemic using household-based public health in-terventions. PLoS Medicine, 2006; 3(9):e361.

    20. Lin XW, Chiang CT, Shih TH, Jiang YN, Chou CC. Foot-and-mouth disease entrance assessment model through air passen-ger violations. Risk Analysis, 2009; 29(4):601611.

    21. Anonymous. Definition of incubation period. Reportupdated 2001. Available at: http://www.medterms.com/script/main/art.asp?articlekey = 18956, Accessed on April 10,2011.

    22. WHO. Avian influenza: Food safety issues. 2011. Available at:http://www.who.int/foodsafety/micro/avian/en/index1.html,Accessed on October 5, 2011.

    23. Shih TH, Hsu LC, Chang CY, Chou YH. Violation riskof illegal animal products carried by air-passengers. TaiwanVeterinary Journal, 2006; 32(3):210216.

    24. Statistical Database (FAOSTAT). Food Agriculture Or-ganization. Global live animal production statistics. 20042006. Available at: http://faostat.fao.org/site/573/default.aspx#ancor, Accessed on November 10, 2010.

    25. World Organization for Animal Health. Animal disease dataabout highly pathogenic avian influenza. 2002. Available at:http://www.oie.int/eng/maladies/fiches/a A150.htm, Accessed

    on November 10, 2010.26. World Organization for Animal Health. TerrestrialAnimal Health Code, Chapter 10.4 Avian Influenza,2009. Available at: http://www.oie.int/eng/normes/Mcode/en chapitre 1.10.4.htm, Accessed on November 10, 2010.

    27. Tollis M, Di Trani L. Recent developments in avian influenzaresearch: Epidemiology and immunoprophylaxis. VeterinaryJournal, 2002; 164(3):202215.

    28. Taiwan Taoyuan International Airport. 2010. Availableat: http://www.taoyuanairport.gov.tw/english/index.jsp, Ac-cessed on November 15, 2010.

    29. Kaohsiung International Airport. 2010. Available at:http://www.kia.gov.tw/index.asp, Accessed on December 10,2010.

    30. Shahid MA, Abubakar M, Hameed S, Hassan S. Avian in-fluenza virus (H5N1): Effects of physico-chemical factorson its survival. Virology Journal, 2009; 6:38. Available at:

    http://www.virologyj.com/content/6/1/38.31. WHO. Highly pathogenic H5N1 avian influenza outbreaks

    in poultry and in humans: Food safety implications. Interna-tional Food Safety Authorities Network Information Note,No. 7, 2005.

    32. Vosesoftware. Vose Software. 2011. Available at: http://www.vosesoftware.com, Accessed on May 15, 2011.

    33. Anonymous. Weibull distribution. 2010. Available at:http://en.wikipedia.org/wiki/Weibull distribution, Accessedon November 10, 2010.

    34. Si Y, Skidmore AK, Wang T, de Boer WF, Debba P,Toxopeus AG, Li L, Prins HH. Spatio-temporal dynamicsof global H5N1 outbreaks match bird migration patterns.Geospatial Health, 2009; 4(1):6578.

    35. Halvorson DA, Kelleher CJ, Senne DA. Epizootiologyof avian influenza: Effect of season on incidence in

    sentinel ducks and domestic turkeys in Minnesota. Ap-plied and Environmental Microbiology, 1985; 49(4):914919.

    36. Center for Food Safety. Risk in brief avian influenza virusesand food safety, 2010. Available at: http://www.cfs.gov.hk/english/programme/programme rafs/programme rafs fm02 03.html, Accessed on November 10, 2010.

    37. Capua I, Alexander DJ. Animal and human health implica-tions of avian influenza infections. Bioscience Reports, 2007;27(6):359372.

  • 7/28/2019 76646438

    12/12

    Copyright of Risk Analysis: An International Journal is the property of Wiley-Blackwell and its content may

    not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

    permission. However, users may print, download, or email articles for individual use.