Background: Dengue fever is the most common arboviral infection in humans, with viral transmissions occurring in more than 100 countries in tropical regions. A global strategy for dengue prevention and control was established more than 10 years ago. However, the factors that drive the transmission of the dengue virus and subsequent viral infection continue unabated. The largest dengue outbreaks in Taiwan since World War II occurred in two recent successive years: 2014 and 2015. Methods: We performed a systematic analysis to detect and recognize spatial and temporal clustering patterns of dengue incidence in geographical areas of Taiwan, using the map-based pattern recognition procedure and scan test. Our aim was to recognize geographical heterogeneity patterns of varying dengue incidence intensity and detect hierarchical incidence intensity clusters. Results: Using the map-based pattern recognition procedure, we identified and delineated two separate hierarchical dengue incidence intensity clusters that comprise multiple mutually adjacent geographical units with high dengue incidence rates. We also found that that dengue incidence tends to peak simultaneously and homogeneously among the neighboring geographic units with high rates in the same cluster. Conclusion: Beyond significance testing, this study is particularly desired by and useful for health authorities who require optimal characteristics of disease incidence patterns on maps and over time. Among the integrated components for effective prevention and control of dengue and dengue hemorrhagic fever are active surveillance and community-based integrated mosquito control, for which this study provides valuable inferences. Effective dengue prevention and control programs in Taiwan are critical, and have the added benefit of controlling the potential emergence of Zika. Keywords: Dengue, Hierarchical, Spatial clustering, Temporal clustering, Zika Background dengue virus serotypes (DENV1–DENV4) now circulate The global emergence and resurgence of epidemic arbo- in Asia, the Americas, and Africa . Compared with viruses such as dengue and Zika have been dramatic in other tropical infectious diseases, dengue has a relatively recent years. Dengue fever is the most common arbo- low mortality; however the large scale of human suffer- viral infection in humans, with viral transmission occur- ing and economic resources used for dengue prevention ring in more than 100 countries in tropical regions. It is and control makes it a major global public health prob- estimated that 390 million dengue infections occur an- lem [1, 5, 6]. There are several factors that contribute to nually, of which 50–100 million cases have apparent the increased frequency and magnitude of dengue fever clinical manifestations [1–3]. The geographical areas in and the emergence of dengue hemorrhagic fever, a se- which transmission of the dengue virus is common have vere form of the disease. The most important factors are been expanding over the past few decades and all four unprecedented growth of human population, unplanned and uncontrolled urbanization, a lack of effective vector * Correspondence: email@example.com control, and globalization [7, 8]. Department of Environmental and Occupational Health, College of The geographical extension of dengue viral transmis- Medicine, National Cheng Kung University, 1 University Road Tainan, 701 sion has followed the increased geographic distribution Tainan, Taiwan Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Lai et al. BMC Infectious Diseases (2018) 18:256 Page 2 of 11 and population densities of Aedes aegypti, the principal Using the scan test, we found that dengue incidence mosquito vector, which transmits dengue viruses in tended to peak simultaneously and homogeneously urban areas of the tropics . Even though great pro- among the neighboring geographic units with high rates gress has been made in dengue research, particularly in in the same cluster . identifying and treating dengue and understanding the structure and replication of the virus, we still do not Methods fully understand why most individuals do not have com- Study population plications while others experience a severe and fatal Dengue fever is a notifiable communicable disease in hemorrhagic disease. Many unanswered questions re- Taiwan. Information on dengue cases collected in Taiwan main regarding the virus-host interaction, immune path- since 1988 is publicly available through the Taiwan Cen- ology, and influence of genetic variation in the host and ters for Disease Control (http://www.cdc.gov.tw/english/ virus . index.aspx) and the Taiwan Government Open Data web- Taiwan is infested with both Ae. aegypti and Aedes site (http://data.gov.tw/en). This information includes the albopictus (a secondary mosquito vector), which trans- date an individual was diagnosed with dengue infection and mit dengue viruses. The two largest dengue outbreaks in his or her residence at diagnosis, place of infection, gender, Taiwan since World War II occurred recently, with and age. The study population used for this investigation is 15,492 autochthonous cases confirmed in 2014 and patients with laboratory confirmed autochthonous dengue 43,419 cases confirmed in 2015. Dengue cases nearly infection, which thus excludes imported cases of dengue. disappeared from the island of Taiwan for 40 years until The spectrum of clinical presentations of dengue infection an outbreak of 4389 cases occurred in 1988. In addition with any one of the 4 viral types is broad. Thus, laboratory to an outbreak of 5336 cases in 2002, a few small out- confirmation of dengue infection is crucial. Confirmed den- breaks occurred between 1989 and 2013. Before World gue viral infection in Taiwan is based on a positive diagno- War II, large dengue outbreaks were reported in 1915 sis from any one of 4 laboratory tests: virus isolation, and 1931 [10, 11]. nucleic acid amplification tests, antigen detection, and sero- Accurately recognizing geographical discrepancies and logical tests. Data on the place where the infection occurred heterogeneity in dengue incidence patterns and detect- are used in the analysis. If they are unavailable, the individ- ing the geographical areas in which the exposure to en- ual’s residence at diagnosis is used. Information on the daily vironmental or viral agents may be responsible for local climate variables, including temperature, rainfall, and intense dengue incidence will inform disease control and relative humidity, is available from Taiwan’s Central prevention efforts and provide important insights into Weather Bureau (https://www.cwb.gov.tw/eng/index.htm). the etiology of this disease. In this study, we used the Tainan and Kaohsiung are the two largest cities in the map-based pattern recognition procedure and scan test southern, tropical region of Taiwan. Kaohsiung is bigger to systematically explore geographical and temporal than Tainan in population and area, with 2.78 million clustering patterns of dengue incidence in an analysis of residents and 2952 km . Tainan has a population of 1.89 Taiwan’s dengue outbreaks in 2014 and 2015. The million and 2192 km . Ae. aegypti, is dispersed primarily map-based pattern recognition procedure is designed to in Tainan, Kaohsiung, and the area to the south of these recognize hierarchical incidence intensity patterns for cities. Ae. albopictus, has a widespread distribution some disease over geographical spaces by searching for throughout most of Taiwan. hierarchical (in intensity) clusters of mutually adjacent Figure 1 shows the yearly frequency distribution of areas with high rates . The procedure incorporates confirmed autochthonous dengue cases that occurred information about the intensity rank order into the or- between 1987 and 2016. In 2014, there were 15,492 con- dinary adjacency-based test statistic , which is de- firmed autochthonous cases in Taiwan, among which signed to analyze data from irregularly arranged and 97% (15,034 cases) occurred in Kaohsiung and < 1% (150 shaped geographic units like the irregular county bound- cases) occurred in Tainan. In 2015, 98% of the total aries within a US state. 43,419 confirmed autochthonous dengue cases occurred Our analysis of the largest Taiwan dengue outbreak in in Tainan and Kaohsiung combined, with 22,842 in Tai- 2015 showed that multiple geographic units with the nan and 19,746 in Kaohsiung. The 2015 Taiwan city highest rates of dengue incidence significantly aggre- (and county)-specific dengue incidence intensity distri- gated into 2 separate geographical areas located in bution is presented in Fig. 2. In 2015, the 3 places with Tainan and Kaohsiung in southern Taiwan. More im- the highest numbers of dengue cases per 100,000 per- portantly, we determined 3 distinct groups within these sons were Tainan, with a rate of 1212, Kaohsiung, with geographic units that had the highest dengue incidence 711, and Pingtung County, with 48. All other cities (and rates according to their intensity and delineated 2 separ- counties) had numbers less than 8. In 2016, there were ate clusters of hierarchical dengue incidence intensity. 380 confirmed autochthonous cases in Taiwan. Lai et al. BMC Infectious Diseases (2018) 18:256 Page 3 of 11 Fig. 1 Yearly Frequency Distribution of Confirmed Autochthonous Dengue Cases in 1987–2016 Map-based pattern recognition procedure for hierarchical areas with higher rates as high-risk areas at each step clusters of disease with the use of B. The method developed by Mantel  was generalized Therefore, the procedure provides the p-value of B by Cliff and Ord, who proposed the test statistic B = (1/ when the k top ranking areas among all areas under 2) Σω x x where x =1 if area i is a high-risk area for study are classified as high-risk areas for each k where k ij i j i some disease and 0 otherwise, and where ω =1 if areas =2, 3, 4….. The procedure can classify as many areas as ij i and j are mutually adjacent geographically and 0 other- high-risk areas as possible; however, it is unlikely that wise, ω = ω , ω =0 . The sum ranges over all pairs one would inquire about the possibility of clusters of ij ji ii of areas. It is an adjacency-based test statistic that mea- more than 20% high-risk areas. The main feature of the sures spatial autocorrelation for binary data and uses the procedure is to determine the hierarchical incidence in- distribution of the number of adjacencies of geographic tensity pattern through the distribution of p-values for k units. When high-risk areas tend to be geographically =2, 3, 4…, which will be illustrated in Results. adjacent to each other, the value of B tends to be large. Instead of relying on the assumptions associated with Using the test statistic B, one can test the null hypoth- the asymptotically normal distribution , we propose esis of the random allocation of high-risk areas over the to use simulation-based permutations using 1 million geographical region; that is, high-risk areas do not clus- replicates based on the exact district boundary map ter. Cliff and Ord derived the expressions for the mean under study to obtain the null distribution of B.The and variance of B under the assumptions of binomial basic geographic unit used in this report is a “district”, and hypergeometric distributions . which is the administratively defined subdivision of a city Instead of selecting a specific threshold rate of inci- in Taiwan, and which regularly reports health-related dence, the map-based pattern recognition procedure information to the city government through its health proposes to first list the areas under study in rank order department. There are 37 and 38 districts in Tainan and based on the disease intensity rates . It starts with Kaohsiung, respectively. The distribution was simulated classifying the 2 top ranking areas as high-risk areas and by randomly selecting exactly k districts among the 75 dis- calculates the value of B. Subsequently, the procedure tricts of Tainan and Kaohsiung combined 1 million times includes the area with the 3rd highest rate and the other and counting the number of the adjacent pairs appearing 2 areas with higher rates as high-risk areas and calcu- among the k districts for each of the 1 million replicates. lates the corresponding value of B. The p-value is the This process was applied for k =2, 3, 4… 14. Each of the probability that B is equal to or higher than the observed 13 distributions of B for k =2, 3, 4… 14 is given in Table 1. number of adjacencies involved between these 3 areas In this setting, at most 19% (14/75) of the 75 total districts with the highest disease intensity rates. The procedure are high-risk districts. With Table 1, we do not require the proceeds successively, including exactly one area with assumption of asymptotically normal distributions for B. high rate according to the rank order and the other The distribution of B would closely approximate a Poisson Lai et al. BMC Infectious Diseases (2018) 18:256 Page 4 of 11 Fig. 2 2015 Taiwan Dengue Incidence Intensity Distribution. The figure was generated by the Statistical Package R version 3.3.0  distribution for small values of k, because the values of the as it slides over the entire period. The scan statistic is mean and variance are close, as shown in Table 1. the maximum number of events in a window (t, t + w), where w is the pre-determined window size as t takes on Scan test all values in a certain time frame. The model of the scan The scan test is frequently used to detect disease cluster- test that we applied here is based on the assumption of a ing over a temporal series and is structured to test for uniform distribution of events . Here, the scan test the largest cluster. The scan test employs a moving win- was used to test for clustering of dengue incidence and dow of pre-determined length and finds the maximum detect the date of the occurrence of maximum dengue number of cases of disease revealed through the window incidence in a district. Lai et al. BMC Infectious Diseases (2018) 18:256 Page 5 of 11 Table 1 Frequency Distributions of the Number of Adjacencies Simulated on the Basis of 1 Million Random Selections in Tainan and Kaohsiung Combined Test Statistic B 0123456789 10 11 12 13 14 15 1617181920212223 mean variance Number of risk districts 2 933,817 66,183 0.066 0.062 3 811,372 179,870 7051 1707 0.199 0.184 4 654,424 301,846 36,315 6830 464 121 0.397 0.361 5 487,271 390,927 97,453 20,434 3209 644 52 10 0.664 0.594 6 335,387 419,120 180,229 50,627 11,572 2565 423 66 10 1 0.994 0.877 7 209,934 387,122 258,564 101,421 31,711 8645 2042 455 85 15 6 1.395 1.215 8 121,202 310,762 301,587 167,122 67,449 22,739 6771 1762 480 98 23 4 1 1.857 1.593 9 63,531 219,245 297,010 224,774 118,538 50,464 18,105 5934 1788 482 99 21 8 1 2.384 2.014 10 30,153 136,056 247,281 252,633 175,657 92,771 41,346 15,969 5630 1762 548 146 37 8 2 1 2.653 2.482 11 12,838 74,469 177,840 239,690 213,280 143,620 78,475 36,271 15,089 5575 1965 652 174 45 15 1 1 3.646 2.985 12 4929 35,989 109,691 191,350 219,975 184,954 124,499 69,938 34,283 14,995 6014 2269 791 237 66 16 4 4.379 3.532 13 1719 15,343 59,144 131,075 189,560 200,876 165,069 111,753 65,553 33,651 15,590 6612 2592 972 340 105 31 13 1 1 5.169 4.108 14 491 5683 27,583 76,015 138,144 181,845 184,653 151,884 105,659 63,626 34,536 16,978 7799 3167 1250 464 142 60 15 4 1 1 6.035 4.721 Lai et al. BMC Infectious Diseases (2018) 18:256 Page 6 of 11 Results the difference of 152 between 908 and the next lowest 2015 Tainan and Kaohsiung dengue outbreak rate of 756 is largest in comparison with the difference Tainan and Kaohsiung are geographically adjacent to between the rates of any other two districts among the each other and, therefore, are considered as a single geo- upper spectrum of rates, except for the top 5 rates. graphical region. The rates, which were the numbers of There were exactly 14 districts with a rate equal to 908 dengue cases per 100,000 persons, ranged from 0 to or higher. In this analysis, at most 14 districts with the 4497 among the 75 districts. Fig. 3a presents the highest rates were classified as high-risk districts using district-specific dengue incidence intensity. We used a the map-based pattern recognition procedure. Fifty-five cut-off point of 908 in the rates to dichotomize because districts or 73% (= 55/75) had rates of 480 or lower. Fig. 3 a, b District-Specific Dengue Incidence a Intensity Map and b Intensity-Level Map in 2015 in Tainan and Kaohsiung. The figures were generated by the Statistical Package R version 3.3.0  Lai et al. BMC Infectious Diseases (2018) 18:256 Page 7 of 11 The 14 top ranking districts are listed by rank accord- of high-risk districts was 4 and the observed value of B ing to their rates in Table 2. With each district with a was 5, giving a p-value = Pr(B ≥ 5│k = 4) = 0.000121 (= high rate, Table 2 gives the rate, the observed value of B 121/1 million from Table 1), shown in the 5th row. The (= the observed number of adjacencies involved between p-value jumped to 0.000706 when we included the the district and districts with higher rates), and the Qianzhen district (5th in rank) as a high-risk district be- p-value (determined from the distribution in Table 1). cause the number of high-risk districts became 5 and When the (k =) 2 top ranking districts in Table 2 (West the observed value of B remained 5, as shown in the 6th Central and North districts) are classified as the row. The 4 top ranking districts are located in Tainan high-risk districts and the remaining 73 districts are not, while the Qianzhen district is in Kaohsiung. The next the observed value of B is equal to 1, giving a p-value = relatively lower p-value of Pr(B ≥ 7│k = 6) = 0.000077 (= Pr(B ≥ 1│k = 2) = 0.066183 (= 66,183/1 million from (66 + 10 + 1)/1 million from Table 1) occurred by includ- Table 1), shown in the 3rd row of Table 2. Table 2 gives ing the Anping district (6th in rank) and the other 5 dis- the p-value when the k top ranking districts are classi- tricts with higher rates, leading to the number of fied as high-risk districts for k=2,3,4… 14. Low high-risk districts = 6 and the observed value of B =7, as p-values were found for each high-risk district listed in shown in the 7th row of Table 2. Table 2, except for the North district shown in the 3rd Correspondingly, we determined the 3 groups of dis- row. The null hypothesis of the randomization of k dis- tricts to use in constructing hierarchical clusters of mu- tricts with the highest rates was rejected at a nominal tually neighboring high-risk districts with different levels significance level of 0.05 for 3 ≤ k ≤ 14; that is, the k dis- of intensity using the map-based pattern recognition tricts with the highest rates showed significant clustering method . Level-1 districts are the 4 top ranking dis- based on B. We also provide the p-values in Table 2 tricts in Table 2 (West Central, North, South, and East based on the Poisson approximation for k ≤ 7. districts). Level-2 districts are Qianzhen and Anping, We note that lower p-values of B indicate high degrees which are respectively the 5th and 6th by rank. Level-3 of clustering, which conform to the adjacency-based def- districts are the 8 districts that rank from 7 to 14. When inition of a cluster [12, 13]. In Table 2, the p-values of the level-specific intensity is placed on the map, 2 hier- the 14 high-risk districts appear to be cycling over the archical dengue incidence intensity clusters clearly rates and are at their relative lowest at the points where emerge and are located in the urban areas of Tainan and the East and Anping districts enter the ranking. We ob- Kaohsiung, respectively, as shown in Fig. 3b. The first served a relatively low p-value of 0.000121 when we in- cluster geographically expands from the 4 Level-1 dis- cluded the East district (4th in rank) and the other 3 tricts to 7 mutually adjacent high-risk districts. This geo- districts with higher rates. In this scenario, the number graphical area displays the highest dengue incidence Table 2 Cluster Statistic for Districts with the High Rates in 2015 Tainan and Kaohsiung Combined b a City District Rate Rank Order Risk Level StatisticBP-value* P-value** 0 West Central 4497 1 1 0 North 4313 2 1 1 0.066183 0.063869 0 South 2785 3 1 2 0.008758 0.017360 0 East 1673 4 1 5 0.000121← 0.000059 1 Qianzhen 1415 5 2 5 0.000706 0.000621 0 Anping 1401 6 2 7 0.000077← 0.000080 1 Sanmin 1350 7 3 7 0.000561 0.000610 1 Lingya 1290 8 3 9 0.000126 – 1 Qianjin 1187 9 3 11 0.000030 – 0 Yongkang 1159 10 3 13 0.000011 – 1 Yancheng 1107 11 3 16 0.000001 – −6 1 Gushan 1000 12 3 18 < 10 – −6 0 Annan 984 13 3 22 < 10 – −6 1 Xinxing 908 14 3 25 < 10 – Number of Dengue Cases per 100,000 Persons 0 indicates Tainan; 1, Kaohsiung *P-value from Table 1 **P-value using Poisson Approximation Lai et al. BMC Infectious Diseases (2018) 18:256 Page 8 of 11 intensity, accounting for 50% of dengue cases. In com- Table 4 Cluster Statistic for Districts with the High Rates in 2014 Kaohsiung parison, the second cluster that consists of the other 7 high-risk districts explains 28% of dengue incidence. District Rate Rank Order StatisticBP-value* P-value** The scan test was used not only to determine the clus- Sanmin 1146 1 tering of the dengue incidence but also to identify the Qianzhen 1029 2 0 1 1 date of the occurrence of maximum incidence in a Xinxing 904 3 1 0.320838 0.299527 high-risk district. Using a window width of 7 days and Lingya 872 4 4 0.004213 0.006101 the time period from week 24 to week 52 of 2015, the Xiaogang 729 5 8 0.005743 0.007413 scan test was applied to each of the 14 top ranking dis- Qianjin 622 6 5 0.000371 – tricts. A very small p-value for the scan test was ob- tained for each of the 7 high-risk districts in Tainan, as Fengshan 615 7 11 0.000041 – shown in Table 3. More importantly, the date of occur- Number of Dengue Cases per 100,000 Persons **P-value from Supplemental Table 1 rence of the district-specific largest dengue cluster (max- **P-value using Poisson Approximation imum dengue cases over any 7 consecutive days in a high-risk district) overlapped on 09/18–09/20 among 6 of the 7 adjacent high-risk districts, indicating that the with the highest rates. These 7 top ranking districts ac- 2015 Tainan dengue outbreak peaked almost simultan- count for 81% of the 2014 Kaohsiung dengue incidence. eously and homogeneously among the geographically The weekly distributions of the frequency and rate in neighboring high-risk districts. A similar phenomenon 2014 and 2015 in Kaohsiung are given in Additional file 2: was observed in Kaohsiung, as shown in Table 3. The Figure S1A. Applying the scan test to each of the 7 peaks of incidence for 5 high-risk districts were during a high-risk district in the same setting, Additional file 1: 2-week period of 11/12–11/25. Table S2 presents the results, indicating that 4 of the 7 high-risk districts had the date of occurrence of the 2014 Kaohsiung dengue outbreak district-specific largest dengue cluster during a period of We used the same approaches to study the pattern and 10/15–10/25. distribution of dengue incidence in 2014 Kaohsiung alone. Each of the 6 distributions of B for k =2, 3… 7is Effects of temperature, rainfall, and relative humidity shown in Additional file 1: Table S1. Correspondingly, Local weather affects dengue viral transmission and in- Table 4 gives the rate, observed value of B, and p-value fection . The effects of local climate variables on (determined from the distribution in Additional file dengue incidence were considered, including temperature, 1:Table S1) for each of the 7 top ranking districts. Small rainfall, and relative humidity. All 3 local climate vari- p-values were found by including 4 or more districts ables were similar in Tainan and Kaohsiung. This Table 3 Analysis of Scan Test for Each of 14 Risk Districts in 2015 Tainan and Kaohsiung District Rank Order Total Cases Statistic of Scan Test Date P-value Tainan −8 West Central 1 3485 694 9/14–9/20 1.14 × 10 −5 North 2 5724 891 8/31–9/06 2.23 × 10 −7 South 3 3502 637 9/15–9/21 2.93 × 10 − 5 East 4 3160 493 9/17–9/23 2.13 × 10 −7 Anping 6 916 166 9/14–9/20 3.55 × 10 − 6 Yongkang 10 2675 451 9/18–9/24 2.93 × 10 −5 Annan 13 1865 281 9/15–9/21 4.79 × 10 Kaohsiung − 6 Qianzhen 5 2726 473 11/13–11/19 1.28 × 10 −3 Sanmin 7 4673 596 11/12–11/18 1.01 × 10 −4 Lingya 8 2251 305 11/13–11/19 3.81 × 10 −2 Qianjin 9 325 33 09/08–09/14 1.84 × 10 −3 Yancheng 11 252 30 10/25–10/31 3.33 × 10 −3 Gushan 12 1369 173 11/14–11/20 1.20 × 10 −4 Xinxing 14 472 64 11/19–11/25 4.19 × 10 Lai et al. BMC Infectious Diseases (2018) 18:256 Page 9 of 11 finding is not surprising as the 2 cities are mutually difference in temperature and rainfall was observed in adjacent. As seen in Fig. 4a, b, relative humidity was 2015. The local weather did not seem to have an effect on consistently higher in Tainan than in Kaohsiung by at the difference in intensity between the 2 distinct hierarch- most 10% over the outbreak period, and no appreciable ical clusters. The Tainan dengue outbreak peaked on week Fig. 4 (A, B). a Weekly Dengue Incidence Distributions and b Weekly Information on Temperature, Relative Humidity, and Rainfall in 2015 Tainan and Kaohsiung Lai et al. BMC Infectious Diseases (2018) 18:256 Page 10 of 11 38 of 2015, with 3422 dengue cases; the Kaohsiung dengue The amount of efforts for the investigation should de- outbreak peaked on week 47, with 2571 cases. The corre- pend on this evidence . sponding rate in Tainan was 181 cases per 100,000 per- A global strategy for dengue prevention and control sons, which was nearly twice the rate of 93 cases per was established more than 10 years ago, and many ef- 100,000 persons on week 47 in Kaohsiung. While no ap- forts have been made to focus on 3 fundamental objec- preciable change in the local weather between 2014 and tives: surveillance for planning and response, reducing 2015 was observed in Kaohsiung, as shown in Additional the disease burden and changing behaviors to improve file 2: Figure S1B, poor environmental management for ef- vector control . However, the factors that drive den- fective integrated vector controls may be responsible for gue viral transmission and infection continue unabated, the worse dengue outbreak and later date of peak inci- and effective vector control remains elusive . dence occurred in 2015 Kaohsiung. In addition, the US Centers for Disease Control (1990) issued a set of guidelines for investigating clusters of Discussion health events. According to the guidelines, the four Historically, Tainan and Kaohsiung experienced the stages are (1) initial contact with and response to the in- worst dengue incidence in large dengue outbreaks in dividual who reported the cluster; (2) a preliminary as- Taiwan [10, 11]. One major reason is that most areas of sessment, including evaluations of whether an excess has Tainan and Kaohsiung are infested with the principal occurred; (3) a formal feasibility study; and (4) a full vector, Ae. aegypti. In the analysis of the 2015 dengue etiologic investigation . This study provides valuable outbreak, the 4 Level-1 districts had high population information and inference in the second and third stage density, respectively the 2nd, 1st, 6th, and 3rd by rank in of the guidelines. We acknowledge some limitations of population density in Tainan. This small area experi- this study, including (1) this investigation is observa- enced extraordinarily high dengue incidence, explaining tional by nature and the exact cause effects cannot be 37% of the dengue incidence in Tainan and Kaohsiung concluded, and (2) the data on the location at which the combined. The 7 districts with the highest dengue in- infection occurred are missing for many individuals, for cidence rates in 2015 Kaohsiung were also among the those the individual’s residence at diagnosis is used in districts with the highest population density in Kaohsiung. the analysis. This indicates that dengue viruses have adapted to the do- The Zika virus essentially has the same epidemiology mesticated Ae. Aegypti and most transmission occurs in and mosquito vectors in urban areas as dengue and is and around the domestic environment in Tainan and following the same path of global spread via competent Kaohsiung. In addition, it is possible that poor physical en- mosquito vectors . The potential for the Zika virus vironments in Tainan and Kaohsiung could be contribut- to emerge in Taiwan is great due to increased air travel. ing factors for recent dengue outbreaks and our results of Thirteen imported cases were reported in Taiwan in analysis call for better environmental management for in- 2016. tegrated vector controls to reduce chance of dengue out- breaks in these regions. We note that the dengue Conclusion incidence outbreak appeared to be initiated earlier and Beyond significance testing for disease clustering, our in- more concentrated geographically in Tainan than in vestigation of dengue incidence distribution over spatial Kaohsiung, shown in Figs. 4a and 3a, b. They may explain and temporal series is desired by and useful for health why the dengue outbreak occurred earlier and appeared authorities who require optimal characteristics and pat- to rise and fall more rapidly in Tainan. Similar climate terns of disease incidence on maps and in a temporal might also be yet another reason for the dengue outbreaks series for effective prevention and control programs. Ef- in Tainan and Kaohsiung. fective prevention and control programs for dengue in Because dengue incidence rates vary substantially by Taiwan are critical, and have the added benefit of con- districts and because we attempt to accurately recognize trolling the potential emergence of Zika. geographical heterogeneity patterns of varying dengue incidence intensity, the map-based pattern recognition Additional files procedure is used and provides important epidemiologic pattern analysis. Our investigation exactly delineates the Additional file 1 Table S1 Frequency Distributions of the Number of 2 tight clusters, which are distinct in location, intensity, Adjacencies Simulated on the Basis of 1 Million Random Selections in and date of peak incidence. As stressed by Kulldorff Kaohsiung. Table S2 Analysis of Scan Test for Each of 7 Risk Districts in 2014 Kaohsiung. (DOCX 17 kb) (2001), p-values should be used as an indicator concern- Additional file 2 Figure S1 Weekly Dengue Incidence Distributions and ing the evidence for true spatial or space-time clusters Weekly Information on Temperature, Relative Humidity, and Rainfall in rather than maintaining a strict cut-off for the p-value to 2014 and 2015 Kaohsiung. (JPG 89 kb) decide whether to investigate detected clusters or not. Lai et al. BMC Infectious Diseases (2018) 18:256 Page 11 of 11 Funding 14. Wallenstein S, Neff N. An approximation for the distribution of the scan The research presented in this manuscript was partially supported by the statistic. Stat Med. 1987;6(2):197–207. Taiwan Ministry of Science and Technology grants MOST 104–2118-M-006- 15. Mantel N. The detection of disease clustering and a generalized regression 006 and 105–2118-M-006-008 to C.C.W. approach. Cancer Res. 1967;27(2):209–20. 16. Ad C, Jk O. Spatial Autocorrelation. London: Pion Press; 1973. Availability of data and materials 17. Banu S, Guo Y, Hu W, Dale P, Js M, Mengersen K, Tong S. Impacts of El Niño Information on dengue cases collected in Taiwan since 1988 is publicly southern oscillation and Indian Ocean dipole on dengue incidence in available through the Taiwan Centers for Disease Control (http:// Bangladesh. Sci Rep. 2015;5:16105. www.cdc.gov.tw/english/index.aspx) and the Taiwan Government Open Data 18. Kulldorff M. Prospective time periodic geographical disease surveillance website (http://data.gov.tw/en). using a scan statistic. Journal Of The Royal Statistical Society: Series A (Statistics In Society). 2001;164(1):61–72. Authors’ contributions 19. Centers For Disease Controls. Guidelines for investigating clusters of health CCW and CHC designed the study and participated in project conception. events. MMWR. 1990;39:1–23. WTL performed analyses. WTL and HH participated in data reformatting and 20. Musso D, Dj G. Zika Virus. Clin Microbiol Rev. 2016;29(3):487–524. management. CHC, RBC and SS contributed in discussion of results and 21. R: A Language And Environment For Statistical Computing. 3.3.0 Edn. revision of the original manuscript. CCW drafted the manuscript. All authors Vienna: R Foundation For Statistical Computing; 2016. read and approved the final manuscript. Ethics approval and consent to participate Not applicable. Competing interests The authors declare that they have no competing interests. Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Author details Department of Statistics, College of Management, National Cheng Kung University, Tainan, Taiwan. Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan. School of Chinese Medicine, China Medical University, Taichung, Taiwan. Department of Environmental and Occupational Health, College of Medicine, National Cheng Kung University, 1 University Road Tainan, 701 Tainan, Taiwan. Departments of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA. Departments of Epidemiology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA. Received: 30 June 2017 Accepted: 23 May 2018 References 1. Jd S, Ds S, Ea U, Ya H, Le C, Oj B, Si H, Bedi N, Im B, Ca C-O. The global burden of dengue: an analysis from the global burden of disease study 2013. Lancet Infect Dis. 2016;16:712–23. 2. Wilder-Smith A, Dj G. Dengue vaccines at a crossroad. Science. 2015; 350(6261):626–7. 3. Bhatt S, Pw G, Oj B, Jp M, Aw F, Cl M, Jm D, Js B, Ag H, Sankoh O. The Global Distribution And Burden Of Dengue. Nature. 2013;496(7446):504–7. 4. Mg G, Sb H,Artsob H,BuchyP,FarrarJ,Dj G,Hunsperger E, Kroeger A, HsM, Martínez E. Dengue: a continuing global threat. Nat Rev Microbiol. 2010;8:S7–S16. 5. Stahl H-C, Vm B, Ht T, Gozzer E, Skewes R, Mahendradhata Y, Runge- Ranzinger S, Kroeger a, Farlow a. cost of dengue outbreaks: literature review and country case studies. BMC Public Health. 2013;13(1):1. 6. Sj T. Preventing dengue—is the possibility now a reality? N Engl J Med. 2015;372(2):172–3. 7. Dj G. Dengue and dengue hemorrhagic fever. Clin Microbiol Rev. 1998; 11(3):480–96. 8. Gubler Dj, Ooi Ee, Vasudevan S, Farrar J: Dengue and dengue hemorrhagic fever: Centre for agriculture and Biosciences International; 2014. https:// www.cabi.org/bookshop/book/9781845939649. 9. Als J, Sn A, Dj G. Barriers to preclinical investigations of anti-dengue immunity and dengue pathogenesis. Nat Rev Microbiol. 2013;11(6):420–6. 10. Akashi K. A dengue epidemic in the Tainan District of Taiwan in 1931. Taiwan No Ikai. 1932;31:767. 11. Koizumi M, Yamaguchi K, Tonomura K. Dengue Fever. Nisshin Igaku. 1916;6:955–1004. 12. Rc G, Kc W, Pwc J. Search for hierarchical clusters of disease: spatial patterns of sudden infant death syndrome. Soc Sci Med. 1981;15(D):287–93. 13. Cliff Ad, Ord Jk: Spatial processes: Models & Applications: Taylor & Francis; 1981.
BMC Infectious Diseases – Springer Journals
Published: Jun 4, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera