Simulation of Infectious Disease Spreading based on Agent Based Model in South Korea 1

Transcription

1 Vol.128 (Healthcare and Nursing 2016), pp Simulation of Infectious Disease Spreading based on Agent Based Model in South Korea 1 Jinhwa Jang 1, 2 1, 3*, Insung Ahn 1 Biomedical Prediction Technology Laboratory, Korea Institute of Science and Technology Information, 245 Daehak-ro, Yuseong-gu, Daejeon 34141, KOREA 2 Laboratory of Computational Biology and Bioinformatics, Institute of Public Health and Environment, Graduate School of Public Health, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, KOREA jhjang@kisti.re.kr 3 Dept. of Big Data Science, University of Science & Technology, 217 Gajeong-ro, Yuseong-gu, Daejeon 34113, KOREA Corresponding Author : isahn@kisti.re.kr Abstract. The world faced with emerge infectious disease including Middle East Respiratory Syndrome and Zika virus. These infectious diseases are associated with extremely rapid and complex propagation patterns due to diversified transportation increased international trades. Against this, we conducted study on predicting disease spread patterns using various public data including the GIS of South Korea. A simulation was performed using an agent-based model with highway traffic data and census data for each of the fourteen provinces. The findings indicated a correlation between the spread pattern of influenza during the 2009 influenza pandemic and some of the traffic movement patterns. Therefore, the possibility of using population movement through highway data as predictive parameter for the spread of infectious diseases was confirmed through a public open data based analysis. Keywords: Influenza Virus, Agent based model, Simulation, Public open data, Traffic data, Census data 1 Introduction There are several types of simulation modeling that can be used for predicting the spread of diseases, including discrete event modeling, agent-based modeling and system dynamics modeling [1]. An agent-based model defines the object, determines the behaviors for each agent, connects the agents, and exposes them to an interactive or 1 This research was supported by the Korea Institute of Science and Technology Information (K-16-L03-C02-S03) ISSN: ASTL Copyright 2016 SERSC

2 dynamic environment. Recently, agent-based models are increasingly applied for predicting the spread of various infectious diseases, such as influenza, severe acute respiratory syndrome (SARS), and measles. Combining the agent based model with a spacious model incorporating actual geological data can increase the accuracy of the simulation results [2]. South Korea experienced record high numbers of infections and deaths due to the spread of the Influenza virus type of H1N1 in 2009 [3]. Hence, the present study investigated the effect of highway traffic on the spread of influenza by examining data from 2009 to 2010, the time period recorded to have the highest number of H1N1 influenza infected patients. This study performed agent-based modeling using AnyLogic 7 Professional, XJ Technologies, St. Petersburg, Russian Federation [4]. The simulation was performed using an imported geographic information system (GIS) map of South Korea, traffic data in interchanges on highway of each province, and census data. 2 Materials and Methods 2.1 Epidemiological Compartment Modeling By including this data, this model can generate a more accurate prediction than the SIR model. The infection rate is calculated considering the potential for contact and susceptibility of infected individuals during a latency period. This study employed the SEIR model with the following expressions: μ refers to individual mortality; β refers to disease spread rate; γ refers to recovery rate, and its reciprocal (1/γ) refers to the mean infection time; σ refers to the rate of movement from the exposed class to the infectious class, and its reciprocal (1/σ) signifies the mean latency period [5]. Finally, for R 0, a variable in SEIR modeling we used the value of 1.53, which was calculated based on the data from the 2009 H1N1 influenza epidemic in the world [2]. 2.2 Epidemic Scenario using Transportation and Census Data We referred to the Korean Statistical Information Service ( to extract monthly and per-type-of-car traffic data in highway interchanges in South Korea in November 2009, which had the highest monthly incidence of H1N1 influenza, and November 2014 (313_DT_EX002_M_2009, 313_DT_EX002_M_2014), as well as the raw demographic data for each metropolitan Province and district, which were reported by the 2010 Population and Housing Census. The 313_DT_EX002_M_2009 data comprise 708,171 cases of traffic movement in 316 cities, and the 313_DT_EX002_M_2014 data comprise 998,165 cases of traffic movement in 323 cities. These data were saved in MySQL for data processing and parsing for each district. Partial sum data for each district were extracted using the sql. There were 316 locations of original departure and 323 destination locations. Then, we calculated the 54 Copyright 2016 SERSC

3 difference in the total mean daily outbound traffic and total mean daily inbound traffic per province in November To examine the difference in the spread of influenza via traffic using the data from November 2014, during which there was no influenza epidemic, we performed the same data management and calculations with the November 2014 data as with the November 2009 data. We used IBM SPSS version 22.0 (IBM Corp., Armonk, NY, USA) to conduct a Pearson correlation analysis on the differences between the actual incidence rate by province in South Korea in 2009 and traffic inflow and outflow in 2009 as well as The computed correlation coefficients were used to analyze the correlation between traffic movement and number of influenza incidents. We used AnyLogic 7 Professional to simulate the spread of the disease; the SEIR model and number of inflow and outflow traffic movements according to the population number were applied to the South Korea GIS map to generate a graph for the results. 3 Results 3.1 Comparative Study between Incidence Rate and Traffic in 2009 and 2014 There was a difference between November 2009 and November 2014 in the values acquired by subtracting traffic outflow from inflow in highway interchanges for each province in South Korea. For example, the Gangwon Province had the highest traffic inflow and the North Gyeongsang Province had the highest traffic outflow in November 2009, while the South Jeolla Province had the highest traffic inflow and the Seoul Special Province had the highest traffic outflow in November The Pearson correlation analyses on these traffic data and the incidence ratio of influenza in South Korea in 2009 showed that influenza incidence in 2009 was positively correlated with traffic movement in 2009 (r = 0.336, p < 0.12) but negatively correlated with traffic movement in 2014 (r = , p < 0.21). In other words, the inflow of traffic in November 2009 was positively correlated with the incidence rate of influenza. Copyright 2016 SERSC 55

4 Fig. 1. Correlation between traffic and incidence in (Seoul (SU), Busan (BS), Incheon (IC), Daegu (DG), Gwangju (GJ), Daejeon (DJ), Ulsan (US), Gyeonggi-do(GG), Gangwon-do (GW), North Chungcheong-do (CB), South Chungcheong-do (CN), North Jeolla-do (JB), South Jeolla-do (JN), North Gyeongsang-do (GB), South Gyeongsang-do (GN)) 3.2 Simulation of Infectious Disease Spread in South Korea by Provinces We inputted the populations of each of the 14 provinces in South Korea with an infection probability of 0.1, contact per day of 5, and R 0 of 1.35 by province as initial values in the SEIR model to generate a graph of daily population. Furthermore, to apply traffic inflow and outflow data from November 2009, the differences between inflow and outflow traffic in highway interchanges per province were calculated and multiplied by 1.5 to reflect the final population for the simulation. In particular, the North Gyeongsang Province showed the most traffic movement, with 413,123 more cases of outflow than inflow; thus, 629,685 was subtracted from the population of the North Gyeongsang Province in the SEIR model, which generated smaller infectious and exposed values than those before the outflow of traffic. In fact, the North Gyeongsang Province had the lowest incidence rate of novel influenza in South Korea in In contrast, the Gangwon Province, which had the highest traffic inflow (396,452), had a higher incidence rate compared with other regions. The Gangwon Province had an incidence of 2,117.8 persons, which was greater than the 2009 national average incidence of novel influenza (1,538.1 persons). 56 Copyright 2016 SERSC

5 Fig. 2. Result of SEIR simulation in South Korea. A: Simulation of infectious disease spreading by 14 provinces, B: Graph indicated SEIR model using census data in Gyeongsangbuk-do, C: Graph indicated SEIR model using traffic and census data in Gyeongsangbuk-do. 4 Discussion and Conclusions This study compared highway traffic flow data based on the South Korea highway traffic volume statistics provided by the Korea Expressway Corporation and demographic data with the prevalence rate of H1N1 influenza during the 2009 influenza epidemic. In particular, we used the SEIR model and synced the data with South Korea s GIS to generate graphs representing the spread of disease per province. The results indicated that each province had very different inflow and outflow traffic values in highway interchanges in November 2009 and November We analyzed the correlation between these traffic data and the incidence rate and found that influenza incidence in November 2009 was positively correlated with November 2009 traffic data but negatively correlated with November 2014 traffic data, although insignificant. Based on these findings, we used AnyLogic 7, which uses the agent-based model, to sync the actual map of South Korea and applied the SEIR model with the census and traffic data for each of the 14 provinces. In particular, applying the SEIR model to the Gangwon Province, which had the highest traffic inflow in November 2009, showed increased numbers of both infectious and exposed persons. Indeed, the Gangwon Province had an incidence rate of 2,117.8 persons per 100,000 persons, which was the second-highest rate following the rate for Ulsan. In other words, we showed the potential for using traffic data to predict disease spread. Nevertheless, this study was limited by the use of only highway traffic data and neglecting to use other types of traffic Copyright 2016 SERSC 57

6 movement. To augment the accuracy of the prediction of disease spread patterns, highway traffic data as well as data regarding national territorial traffic and public transportation use should be included in the agent-based model to reflect more diverse traffic information variables. References 1. Sachs, J., Malaney, P.: "The economic and social burden of malaria." Nature. Vol.415, pp (2002) 2. Hwang, G.M., Mahoney, P.J., James, J.H., Lin, G.C., Berro, A.D., Keybl, M.A., Goedecke, D.M., Mathieu, J.J., Wilson, T.: A model-based tool to predict the propagation of infectious disease via airports. Travel Medicine and Infectious Disease. Vol.10, No.1. pp (2012) 3. Kim, Y.W., Yoon, S., Oh, I.: "The economic burden of the 2009 pandemic H1N1 influenza in Korea." Scandinavian journal of infectious diseases. Vol.45, No.5. pp (2013) 4. AnyLogic: Multimethod Simulation Software, (2000) 5. LI, M.Y., Muldowney, J.S.: Global Stability for the SEIR Model in Epidemiology Mathmatical Biosciences.Vol.125, pp (1995) 58 Copyright 2016 SERSC