This article has Open Peer Review reports available.
Spatio-temporal analysis on enterovirus cases through integrated surveillance in Taiwan
© Chan et al.; licensee BioMed Central Ltd. 2014
Received: 31 May 2013
Accepted: 5 January 2014
Published: 8 January 2014
Severe epidemics of enterovirus have occurred frequently in Malaysia, Singapore, Taiwan, Cambodia, and China, involving cases of pulmonary edema, hemorrhage and encephalitis, and an effective vaccine has not been available. The specific aim of this study was to understand the epidemiological characteristics of mild and severe enterovirus cases through integrated surveillance data.
All enterovirus cases in Taiwan over almost ten years from three main databases, including national notifiable diseases surveillance, sentinel physician surveillance and laboratory surveillance programs from July 1, 1999 to December 31, 2008 were analyzed. The Pearson’s correlation coefficient was applied for measuring the consistency of the trends in the cases between different surveillance systems. Cross correlation analysis in a time series model was applied for examining the capability to predict severe enterovirus infections. Poisson temporal, spatial and space-time scan statistics were used for identifying the most likely clusters of severe enterovirus outbreaks. The directional distribution method with two standard deviations of ellipse was applied to measure the size and the movement of the epidemic.
The secular trend showed that the number of severe EV cases peaked in 2008, and the number of mild EV cases was significantly correlated with that of severe ones occurring in the same week [r = 0.553, p < 0.01]. These severe EV cases showed significantly higher association with the weekly positive isolation rates of EV-71 than the mild cases [severe: 0.498, p < 0.01 vs. mild: 0.278, p < 0.01]. In a time series model, the increase of mild EV cases was the significant predictor for the occurrence of severe EV cases. The directional distribution showed that both the mild and severe EV cases spread extensively during the peak. Before the detected spatio-temporal clusters in June 2008, the mild cases had begun to rise since May 2008, and the outbreak spread from south to north.
Local public health professionals can monitor the temporal and spatial trends plus spatio-temporal clusters and isolation rate of EV-71 in mild and severe EV cases in a community when virus transmission is high, to provide early warning signals and to prevent subsequent severe epidemics.
KeywordsEnterovirus Epidemiologic characteristics Surveillance Taiwan
In July 2012, 54 children died of infections with enterovirus-71 (EV-71) in Cambodia . Before the laboratory results came back, the media called it a mystery disease, which made numerous Asian parents worried. In fact, severe epidemics of enterovirus have occurred frequently in Asia, including Malaysia , Singapore , Taiwan [4, 5] and China . The clinical severity varied from asymptomatic to mild symptoms [hand-foot-mouth disease (HFMD) and herpangina], severe pulmonary edema, hemorrhage and encephalitis . The EV-71 infections involved 6-29% asymptomatic infection, 13-32% non-specific viral syndrome, and 37%-40% HFMD/ herpangina symptoms . Among the non-polio enterovirus serotypes, EV-71, which has caused severe clinical illness and many fatal cases , and particularly a high risk of poor prognosis for children under one year of age , has become one of the most important public health concerns since the World Health Organization (WHO) launched the “global polio eradication initiative” program in 1988 . However, children with an asymptomatic and mild infection of EV-71 still can carry the virus to transmit to others . During 1998 ~ 2005, 37.98% (588/1,584) of severe EV cases in Taiwan had only encephalitis complications, and 22.8% (353/1,584) had encephalitis with pulmonary edema or hemorrhage . In other words, whether increasing numbers of mild EV cases would provide a possible sentinel signal of the early stage of an epidemic is worth investigating, particularly as a vaccine and more effective drugs for EV-71 have not been available [12, 13]. Therefore, using an integrated surveillance system to monitor the enterovirus activity and fully understanding the difference in epidemiological characteristics between mild and severe enterovirus cases will be the most important prevention and control measures in public health.
After the first nationwide epidemic of EV-71 occurred in Taiwan in 1998, there were another three cross-county epidemics in 2000–2001, 2004–2005, 2008–2009 . The question is, what are the important epidemiological characteristics that will be helpful in surveillance of EV-71 to minimize the severity of future epidemics? The specific aims of this study were: (1) to elucidate the spatio-temporal correlations between the mild and severe enterovirus cases through integrating the data of the three enterovirus-related surveillance systems, including the sentinel physician, national notifiable diseases and laboratory surveillance systems in Taiwan, (2) to find out the feasibility of establishing an early warning signal using the increasing numbers of mild EV-71 cases and their lag time periods to appearance of severe EV-71 cases, and (3) to evaluate the trends of severe EV-71 cases over a 9.5-year period for providing better recommendations on public health efforts in the future. With full understanding of the epidemiological characteristics, hopefully we can develop better measures and indicators from mild cases to provide early warning signals and thus minimizing subsequent numbers of severe cases.
Surveillance for enteroviruses
Three types of databases from the Taiwan Centers for Diseases Control (Tw-CDC), including national notifiable diseases surveillance, sentinel physician surveillance and laboratory surveillance from July 1, 1999 to December 31, 2008, were used in this study. Mild enterovirus infections including hand-foot-mouth disease or herpangina were reported by the sentinel physicians each week. For enterovirus infections with severe complications, the clinicians are required to report to Tw-CDC through the national notifiable diseases surveillance system within one week. The reporting definition of severe EV cases must meet one of two criteria: (1) patients with typical hand-foot-mouth disease (HFMD) or herpangina, or epidemiologically-linked enterovirus infected patients accompanied by myoclonic jerks or encephalitis, acute flaccid paralysis, acute hepatitis, myocarditis, or cardiopulmonary failure, and (2) infants under 3 months of age with myocarditis, hepatitis, encephalitis, thrombocytopenia, multiple organ dysfunction syndrome (MODS) or sepsis symptoms, excluding the infection of bacteria or other pathogens . For the sentinel-physician surveillance system, involving voluntary-based sentinel physicians in Taiwan, there were around 670 sentinel physicians from 510 clinics and 71 hospitals . The cases with HFMD or herpangina reported through this system were compiled on a weekly basis. For virological surveillance data, all those specimens that were collected by sentinel physicians for highly suspected EV cases and sent to the 12 regional contracted laboratories of the Tw-CDC were used to examine the viral types, including EV-71, coxasackievirus groups A/B, echovirus, and other enterovirus .
In this study, the temporal unit was the week, and the spatial unit was the city or county. To compare the trends between the mild and severe EV cases and between the isolation rates of different serotypes of EV and the number of severe EV cases, Pearson’s correlation coefficients were applied for measuring the consistency of the weekly data with the statistical software, SPSS (Version 18.0, SPSS Science, Chicago, IL). Because the data of sentinel physicians were collected on a weekly basis, the counts of the severe EV cases and virological results were also aggregated into a weekly basis for better comparison. Weekly numbers of severe cases and EV-positive isolation rates were used to compare their temporal trends, whereas the weekly incidence rates of severe EV-cases were employed to describe spatial distributions over time. A lag effect between mild and severe EV cases was taken into account to find out whether mild EV cases occurred earlier or later than severe EV cases, in order to determine whether mild cases might serve as early warning signals for severe cases. In addition, we applied cross-correlation analysis in a time series model to see whether there was any conditional correlation among the severe and mild EV cases and four types of EV isolation. An ARMA model (Autoregressive moving average model) was fit using the SAS Release 9.2 software (Cary, NC). The selection of the autoregressive (AR) and moving average (MA) was based on the minimum information criterion (MINIC) method .
Because the mild and severe EV cases were from different surveillance systems, the age definition of the EV cases we used was different. The mild EV cases were from sentinel surveillance, which had aggregated reported cases without age information. Thus, the age definition of mild EV was all ages. The severe EV cases were from the notifiable infectious disease system, which had complete age information. Because most of the high risk population of severe EV cases was children during the study period, we selected the cases aged equal or less than 14 years (99.7%, 1,512/1,517, median age = 21.5 months) for further cluster analysis. The cumulative incidence of the pediatric severe EV cases was calculated with the corresponding mid-year population from July 1999 to 2008. Most likely clusters with high incidence of severe EV cases were detected retrospectively using spatial statistic, temporal statistic, and space-time scan statistic implemented in SaTScan v.9.1.1.. The population data throughout the study period in each city or county were collected from Taiwan’s National Statistics website (http://ebas1.ebas.gov.tw/pxweb/Dialog/statfile9.asp). All the 9.5 years cases’ data were used, with a maximum cluster population size of 5% to minimize false clusters, and a maximum temporal window of one month to examine the temporal-cluster using the software of SaTScan. The analyses of data with case counts were carried out using the Poisson probability model (for a few pediatric EV cases among the child population) with 999 Monte Carlo replications to test for the presence of statistically significant spatial clusters , and choosing the parameter for no geographic overlapping clusters to avoid repeated counting. After identifying the space-time clusters of the severe EV cases, the corresponding data of the mild EV cases from sentinel physician surveillance were further analyzed for consistent temporally increasing trends or even earlier increasing trends. Space-Time permutation was applied for detecting the mild EV clusters due to the lack information of the population at risk . The directional distribution method was employed, with two standard deviations of the ellipse size weighted by either the mild EV cases per doctor or the number of severe EV cases . Then, we applied GIS software (ArcMap, version9.3; ESRI Inc.,Redlands, CA, USA) for visualization of all detected statistically significant clusters.
Age distribution of mild and severe EV cases
Temporal relationship between mild and severe EV cases
Temporal correlation between mild EV and severe EV cases in Taiwan by different years, from July 1, 1999 to December 31, 2008
July 1 ~ Dec. 31, 1999
The correlation among mild, severe EV and major four types of EV isolation rate
Severe EV Cases
Mild EV Cases
EV-71 Isolation Rate
CAV Isolation Rate
CBV Isolation Rate
ECHO Isolation Rate
Severe EV Cases
Mild EV Cases
EV-71 Isolation Rate
CAV Isolation Rate
CBV Isolation Rate
ECHO Isolation Rate
The cross-correlation analysis among severe EV cases, mild EV cases and four types of EV isolation rate
Mild EV cases
Mild EV cases
Spatial relationship between mild and severe EV Cases
To fully understand the temporal, spatial and tempo-spatial distributions of EV cases, we then monitored the trends in EV cases using these three methods separately. With the temporal scan method alone, the temporal cluster was only detected in June 2008, which had the highest number of monthly severe EV cases (n = 137) during the study period. In the further analysis, the number of the severe EV cases from April 2008 to June 2008 was 277, which was also the highest value for any three-month period during the study period.
The size and direction of movement during the EV epidemic in 2008
Ellipse size of the epidemics
Moving direction of the center (Y axis)
Epidemics of enterovirus have continued playing a major public health threat in the Asia-Pacific region . During the past decade, epidemics have also occurred in European countries, including Denmark , the United Kingdom , Hungary , France  and the Netherlands . The integrated information from different enterovirus surveillance systems (rather than from a single source) plus spatio-temporal analyses of epidemiological data in Taiwan might provide valuable experience for other countries. Enhanced surveillance and non-pharmaceutical public health policy such as school closure have been the major strategies implemented for preventing enterovirus epidemics, because effective vaccines and antiviral drugs have not been available. In this study, we have shown that integrated surveillance, including sentinel physician-based clinical surveillance, virological surveillance, and notifiable infectious disease surveillance, reflected not only the whole spectrum of EV cases from mild to severe but also the types of virus activity at the community level. With the spatial scan and spatio-temporal scan statistics, we found that central Taiwan and Penghu County, which is an island located 44 km offshore, were the locations of major clusters of severe EV cases. In the temporal pattern, the severe EV cases occurred either one week earlier than the mild EV cases or surged simultaneously, during the same week. However, taking the autoregressive effects and EV isolation rates together, severe EV cases could not significantly explain the temporal trend of mild EV cases. In contrast, mild EV cases might have better prediction capability even after controlling for the AR effects.
In this study, the peak of the mild and the severe EV cases occurred almost the same week. In the years of 2000 and 2001, the increases of the mild EV cases occurred one to two weeks earlier than the increases of the severe EV cases. After 2002, the pattern changed to consistently high in the same week, or severe cases rising even much earlier. Several possible reasons might explain this phenomenon. First, the outbreak of SARS in 2003 made physicians in Taiwan more aware of unusual increases in case numbers of infectious diseases. Once severe EV cases were reported and announced through mass media, even at the early stage, it might have alerted the physicians to pay attention to additional suspected EV cases. Second, EV-71 was highly correlated with severe EV cases. This is consistent with the facts, that EV-71 was known to have high virulence, and pathogenicity in the central nervous system by inflammatory cytokines/chemokines [28, 29], severe pulmonary edema and heart failure . The early information on the positive isolation rate of EV-71 also can be helpful to predict the more likely occurrence of severe EV cases in the following one to two weeks.
With the space-time scan method, we found temporal clusters in June 2008, and spatial clusters in five cities and counties. One nearby island, Penghu, where the seashore is a famous summer tourist destination, had the highest local risk in 2000 and 2008. The possible reasons for this phenomenon might have been the smaller number of children amongst the population, lack of medical resources, and close transportation to another epidemic center, China.
The data in this study have three major limitations. Because of the de-identification between different released databases, direct linkage between the severe EV cases and the results of EV isolation was not feasible. Hence, it was hard to elucidate the correlation between local epidemics and the specific types of enterovirus circulated. Age-related data for mild EV incidence were not available in the sentinel surveillance database. Therefore, it was hard to differentiate among incidence in preschool children or school-aged children. Integration among different surveillance systems would be beneficial for disease surveillance. Public health surveillance and clinical surveillance could help detect aberrations at the early stage. Laboratory surveillance could help determine the severity of epidemics. In the future, utilizing community surveillance, hospital-based syndromic surveillance and National health insurance data, which cover 99% of the population in Taiwan, for monitoring age-specific enterovirus cases will offer the best chance to detect enterovirus activity with better sensitivity and timeliness at the local community level.
Local public health professionals can monitor the temporal and spatial trends plus spatio-temporal clusters and isolation rate of EV-71 in mild and severe EV cases in the community when virus transmission is high to provide early warning signals and to prevent subsequent severe epidemics. The increase of mild EV cases might be a possible predictor for the occurrence of severe EV cases with a time series model. In addition, comprehensive surveillance of school children might detect earlier signals and allow social distance intervention to minimize the size of spatial clustering.
The authors would like to express our sincere gratitude to public health professionals at local departments of health and central public health officials at the Centers for Diseases Control in Taiwan (Taiwan-CDC) for their effort in surveillance of enterovirus-related diseases. This study was supported by grants from the Center for Disease Control Taiwan (CDC/Taiwan) (DOH97-DC-1002) and National Science Council (NSC-102-2621-M-001-001).
- Severe complications of hand, foot and mouth disease (HFMD) caused by EV-71 in Cambodia – conclusion of the joint investigation. http://www.who.int/csr/don/2012_07_13/en/index.html,
- Chua KB, Kasri AR: Hand foot and mouth disease due to enterovirus 71 in Malaysia. Virol Sin. 2011, 26 (4): 221-228. 10.1007/s12250-011-3195-8.View ArticlePubMedGoogle Scholar
- Ang LW, Phoon MC, Wu Y, Cutter J, James L, Chow VT: The changing seroepidemiology of enterovirus 71 infection among children and adolescents in Singapore. BMC Infect Dis. 2011, 11: 270-10.1186/1471-2334-11-270.View ArticlePubMedPubMed CentralGoogle Scholar
- Lee MS, Chiang PS, Luo ST, Huang ML, Liou GY, Tsao KC, Lin TY: Incidence rates of enterovirus 71 infections in young children during a nationwide epidemic in Taiwan, 2008–09. PLoS Negl Trop Dis. 2012, 6 (2): e1476-10.1371/journal.pntd.0001476.View ArticlePubMedPubMed CentralGoogle Scholar
- Chang LY: Enterovirus 71 in Taiwan. Pediatr Neonatol. 2008, 49 (4): 103-112. 10.1016/S1875-9572(08)60023-6.View ArticlePubMedGoogle Scholar
- Mao LX, Wu B, Bao WX, Han FA, Xu L, Ge QJ, Yang J, Yuan ZH, Miao CH, Huang XX, et al: Epidemiology of hand, foot, and mouth disease and genotype characterization of Enterovirus 71 in Jiangsu, China. J Clin Virol. 2010, 49 (2): 100-104. 10.1016/j.jcv.2010.07.009.View ArticlePubMedGoogle Scholar
- Chen SC, Chang HL, Yan TR, Cheng YT, Chen KT: An eight-year study of epidemiologic features of enterovirus 71 infection in Taiwan. Am J Trop Med Hyg. 2007, 77 (1): 188-191.PubMedGoogle Scholar
- Lee MS, Tseng FC, Wang JR, Chi CY, Chong P, Su IJ: Challenges to licensure of enterovirus 71 vaccines. PLoS Negl Trop Dis. 2012, 6 (8): e1737-10.1371/journal.pntd.0001737.View ArticlePubMedPubMed CentralGoogle Scholar
- Ho M, Chen ER, Hsu KH, Twu SJ, Chen KT, Tsai SF, Wang JR, Shih SR: An epidemic of enterovirus 71 infection in Taiwan. Taiwan Enterovirus Epidemic Working Group. N Engl J Med. 1999, 341 (13): 929-935. 10.1056/NEJM199909233411301.View ArticlePubMedGoogle Scholar
- Dhole TN, Ayyagari A, Chowdhary R, Shakya AK, Shrivastav N, Datta T, Prakash V: Non-polio enteroviruses in acute flaccid paralysis children of India: vital assessment before polio eradication. J Paediatr Child Health. 2009, 45 (7–8): 409-413.View ArticlePubMedGoogle Scholar
- Chang LY, Tsao KC, Hsia SH, Shih SR, Huang CG, Chan WK, Hsu KH, Fang TY, Huang YC, Lin TY: Transmission and clinical features of enterovirus 71 infections in household contacts in Taiwan. JAMA. 2004, 291 (2): 222-227. 10.1001/jama.291.2.222.View ArticlePubMedGoogle Scholar
- Huang ML, Ho MS, Lee MS: Enterovirus 71 vaccine: when will it be available?. J Formos Med Assoc. 2011, 110 (7): 425-427. 10.1016/S0929-6646(11)60063-6.View ArticlePubMedGoogle Scholar
- Yang Y, Zhang L, Fan X, Qin C, Liu J: Antiviral effect of geraniin on human enterovirus 71 in vitro and in vivo. Bioorg Med Chem Lett. 2012, 22 (6): 2209-2211. 10.1016/j.bmcl.2012.01.102.View ArticlePubMedGoogle Scholar
- Taiwan CDC: Guidelines for Communicable Disease Surveillence. 2008, Taipei, Taiwan: Taiwan CDC (Centers for Disease Control), 45-47.Google Scholar
- Chien-Yu Lin C-YH, Wu T-SJ, Y-T Chu C-LK, Liao Y-J, Chang S-Y, Chiang P-S, Chu Y-C, King C-C: Epidemiology and Surveillance System of Human Influenza Viruses. Taiwan Epidemiol Bull. 2007, 23 (5): 254-272.Google Scholar
- Wang YC, Cheng HB, Chen HH, Liu CM, Chou CH, Sung FC: Circulating viruses associated with severe complicated enterovirus infection in Taiwan: a multi-year analysis. Pediatr Infect Dis J. 2010, 29 (4): 334-339.PubMedGoogle Scholar
- Box GEP, Jenkins GM, Reinsel GC: Time Series Analysis: Forecasting and Control. 1994, Englewood Cliffs: NJ: Prentice HallGoogle Scholar
- Kulldorff M, Athas WF, Feurer EJ, Miller BA, Key CR: Evaluating cluster alarms: a space-time scan statistic and brain cancer in Los Alamos, New Mexico. Am J Public Health. 1998, 88 (9): 1377-1380. 10.2105/AJPH.88.9.1377.View ArticlePubMedPubMed CentralGoogle Scholar
- Kulldorff M: A spatial scan statistic. Commun Stat-Theor M. 1997, 26 (6): 1481-1496. 10.1080/03610929708831995.View ArticleGoogle Scholar
- Kulldorff M, Heffernan R, Hartman J, Assuncao R, Mostashari F: A space-time permutation scan statistic for disease outbreak detection. PLoS Med. 2005, 2 (3): e59-10.1371/journal.pmed.0020059.View ArticlePubMedPubMed CentralGoogle Scholar
- Shobugawa Y, Wiafe SA, Saito R, Suzuki T, Inaida S, Taniguchi K, Suzuki H: Novel measurement of spreading pattern of influenza epidemic by using weighted standard distance method: retrospective spatial statistical study of influenza, Japan, 1999–2009. Int J Health Geogr. 2012, 11: 20-10.1186/1476-072X-11-20. doi:10.1186/1476-072X-11-20View ArticlePubMedPubMed CentralGoogle Scholar
- Solomon T, Lewthwaite P, Perera D, Cardosa MJ, McMinn P, Ooi MH: Virology, epidemiology, pathogenesis, and control of enterovirus 71. Lancet Infect Dis. 2010, 10 (11): 778-790. 10.1016/S1473-3099(10)70194-8.View ArticlePubMedGoogle Scholar
- Badran SA, Midgley S, Andersen P, Bottiger B: Clinical and virological features of enterovirus 71 infections in Denmark, 2005 to 2008. Scand J Infect Dis. 2011, 43 (8): 642-648. 10.3109/00365548.2011.577094.View ArticlePubMedGoogle Scholar
- Bible JM, Iturriza-Gomara M, Megson B, Brown D, Pantelidis P, Earl P, Bendig J, Tong CY: Molecular epidemiology of human enterovirus 71 in the United Kingdom from 1998 to 2006. J Clin Microbiol. 2008, 46 (10): 3192-3200. 10.1128/JCM.00628-08.View ArticlePubMedPubMed CentralGoogle Scholar
- Pankovics P, Boros A, Szabo H, Szekely G, Gyurkovits K, Reuter G: Human enterovirus 109 (EV109) in acute paediatric respiratory disease in Hungary. Acta Microbiol Immunol Hung. 2012, 59 (2): 285-290. 10.1556/AMicr.59.2012.2.13.View ArticlePubMedGoogle Scholar
- Tan CY, Gonfrier G, Ninove L, Zandotti C, Dubot-Peres A, de Lamballerie X, Charrel RN: Screening and detection of human enterovirus 71 infection by a real-time RT-PCR assay in Marseille, France, 2009–2011. Clin Microbiol Infect. 2012, 18 (4): E77-E80. 10.1111/j.1469-0691.2012.03769.x.View ArticlePubMedGoogle Scholar
- van der Sanden S, van Eek J, Martin DP, van der Avoort H, Vennema H, Koopmans M: Detection of recombination breakpoints in the genomes of human enterovirus 71 strains isolated in the Netherlands in epidemic and non-epidemic years, 1963–2010. J M Epidemiol Evol Genet Infect Dis. 2011, 11 (5): 886-894. 10.1016/j.meegid.2011.02.011.View ArticleGoogle Scholar
- Ryu WS, Kang B, Hong J, Hwang S, Kim A, Kim J, Cheon DS: Enterovirus 71 infection with central nervous system involvement, South Korea. Emerg Infect Dis. 2010, 16 (11): 1764-1766. 10.3201/eid1611.100104.View ArticlePubMedPubMed CentralGoogle Scholar
- Wang SM, Lei HY, Liu CC: Cytokine immunopathogenesis of enterovirus 71 brain stem encephalitis. Clin Dev Immunol. 2012, 2012: 876241-PubMedPubMed CentralGoogle Scholar
- Wu JM, Wang JN, Tsai YC, Liu CC, Huang CC, Chen YJ, Yeh TF: Cardiopulmonary manifestations of fulminant enterovirus 71 infection. Pediatrics. 2002, 109 (2): e26-10.1542/peds.109.2.e26.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2458/14/11/prepub
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.