由2009-2010年高雄市登革熱/登革出血熱流行病學探討環境、病媒蚊密度及公共衛生介入之影響

Abstract

2009年7月高雄市爆發第三型登革病毒的跨年流行，此波流行延燒至次年12月，共1,602本土登革確定病例。有鑑於往昔室內空間噴灑成效不彰，自2010年1月起，衛生局以藥劑噴霧罐取代之，並自3月進行全市目標60萬戶孳生源普查暨病媒蚊密度及大型孳生源調查。本研究目的是結合流行病學與空間統計，探查2009至2010年高雄市登革熱/登革出血熱的流行病學、環境條件及影響登革發生率的重要危險因子，並評估公共衛生介入對疫情控制之成效。 作法上，以2009至2010年高雄市的登革本土確定病例、大型孳生源控管及家戶病媒密度調查資料，藉由斯皮爾曼等級相關係數(Spearman’s rank correlation coefficient)、獨立樣本t檢定(independent Student’s t test)及邏輯斯迴歸(logistic regression)等統計模型分析登革發生率與其環境條件之關連性，再配合單/多變項的全域與區域空間自相關分析(univariate/multivariate global and local Moran’s I statistic)，找出與登革發生率及空間群聚相關之危險因子，以明瞭高雄市登革病例有、無里間不同的衛生環境條件對於登革發生率之影響。至於公共衛生介入的影響，是以皮爾森相關係數(Pearson’s correlation coefficient)及斯皮爾曼等級相關係數評估介入後多久(時間延遲)造成登革發生率與病媒密度之負相關。 自2009年1月至2010年12月，高雄市共有兩波流行，第一波自2009年7月31日至次年3月1日，共646例本土登革確定病例(包括9例登革出血熱，3例死亡)，其登革病例總發生率為萬分之4.23，而登革出血熱發生率為萬分之0.06，登革出血熱之致死率為33.3％(3/9)；第二波自2010年3月1日至12月31日止，共956位本土登革確定病例(包括6例登革出血熱，1例死亡)，其登革病例總發生率為萬分之6.25，登革出血熱發生率為萬分之0.04，登革出血熱之致死率為16.7 % (1/6)。 在環境因子、病媒蚊密度分析中，發現小型孳生源密度(以容器密度為度量) 較大型孳生源密度對登革發生率之相關性更大(rs= 0.52 vs. 0.16, p<0.01)；而容器指數較布氏指數、住戶指數及成蚊指數更能貼近流行現況(rs=0.49 vs. 0.37, 0.35, 0.20, p<0.01)，進一步單變項空間自相關分析後，發現登革發生率、人口密度及大、小型孳生源蚊密度、容器指數共四項因子不僅具有空間上的群聚現象(p<0.01)，且其熱點(hotspots)均在鼓山區及三民區重疊，顯示當地的高登革發生率，與環境因子息息相關。非屬於已知的大型孳生源(大型積水/誘蚊產卵地、以前列管地、空調停機冷卻水塔、破損髒亂空屋、資源/輪胎回收場)之病媒蚊其他孳生處，如雜草處等，與登革發生率具統計顯著相關(rs=0.188, p<0.01)。而13種小型孳生源和登革發生率均具有顯著正相關，且依序為馬桶水箱、桶缸甕盆、底盤、水塔表(rs=0.53, 0.45, 0.40, 0.40)；而和登革發生率最高者為無論室內外均為桶缸甕盆[rs=0.57(室內), 0.42(室外), p<0.01]，其次分別為室內花瓶(rs=0.44, p<001)及室外底盤(rs=0.30 p<0.01)。 比較各里有、無登革病例的環境差異，發現有病例的里在人口密度(25.95 vs. 18.44, p<0.01)、大型孳生源密度(59.85 vs. 47.94, p=0.02)、被清除孳生源的家戶比例(96.31 vs. 47.82, p<0.01)、小型孳生源密度(35.17 vs. 20.55, p<0.01)、容器指數(7.91/4.92, p<0.01)及埃及斑蚊成蚊比例(94.01 vs. 74.06, p<0.01)均較無病例的里高。利用多變項邏輯斯迴歸分析人口密度、小型孳生源密度、容器級數及大型孳生源密度後，得知人口密度高[校正後的勝算比(Adjusted Odds Ratio, ORadj) =9.70, 95%信賴區間(Confidence Interval, CI)=4.65-20.24, p<0.01]、小型孳生源密度高(ORadj=6.33, 95% CI=2.83-14.16, p<0.01)及容器級數高(ORadj=12.65, 95% CI= 6.11 ~ 26.29, p<0.01)三項均為登革的具統計差異之重要危險因子。最後，評估公共衛生介入，清除家戶孳生源在當週即與容器指數有負相關(-0.2)，且在四週後更具統計負相關 (-0.32, p<0.05)，且相關性在第五週達高峰(-0.52, p<0.01)。 本研究的結論是，環境因子(人口密度、大/小孳生源密度)和病媒密度(家戶指數、布氏指數、容器指數及埃及斑蚊比例)為登革發生率的重要危險因子，且具空間聚集。其中家戶小型孳生源與登革發生率的相關性較大型孳生源為高，然而大型孳生源具積水面積大、且不易清除的特性，對登革之影響不容小覷。人口密度高、小型孳生源密度多及容器級數高的地區較易發生登革流行。而在評估病媒密度時，容器指數較其他指數更能反映流行狀況。此外，日常的清除家戶孳生源及噴藥確實可控制病媒密度，但無法立竿見影。 埃及斑蚊比例與年俱增，未來應整合家戶與大型孳生源的兩種病媒監測系統，以徹底掌控當地病媒密度、提升介入實效、降低病媒抗藥性的機會，並配合強化病例監測系統及病例通報、診斷時效，構築完整的登革風險地圖，以因地制宜實施不同的公共衛生介入，有助於登革防治更能收事半功倍之成效。為預防高雄成為登革地方性流行地區，可將本研究結果與過去同為流行登革第三型病毒之2006及2009做比較、配合血清流行病學調查，以了解高雄市登革不顯性感染者比例及其空間分佈，並分析病毒流行模式與環境、宿主三者間之關聯性，建立登革預警指標，以及早阻斷傳播及登革出血熱的死亡病例。

An outbreak of dengue virus serotype 3 (DENV-3) occurred from July 2009 and extended to December 2010, with a total of 1,602 indigenous laboratory-confirmed dengue cases. In 2010, Public Health Bureau of Kaohsiung City Government in Taiwan firstly applied the aerosol cans to residents’ houses in January, conducted a large-breeding sites survey and a city-wide program of removing most containers aiming at 600,000 households since March. The overall objectives of this study were to integrate epidemiology and spatial statistics to understand the epidemiology of dengue/dengue hemorrhagic fever (DHF) in Kaohsiung City from January 2009 to December, to find out the environmental risk/protection factors contributing to increasing/decreasing dengue cases, and to evaluate the effectiveness of different intervention strategies and their time lags. We collected the information of large-breeding sites, all data of different mosquito indices and numbers of laboratory-confirmed dengue cases from 2009 to 2010, and used general statistic (Speraman’s rank correlation coefficient, independent Student’s t-test, logistic regression) to find out the risk factors of dengue in order to understand the impact of different environment situation in Lis with or without dengue cases. Then, spatial statistic methods (univariate/ multivariate global Moran’s I, univariate/ multivariate Anselin Local Moran's I) were conducted to understand the distributions of risk factors of dengue, the spatial autocorrelations and the hot/cold spots of environmental factors, vectors and dengue incidence Furthermore, we evaluated the time lags between different public health interventions and dengue incidence by Spearman’s rank correlation coefficient (eg. negative correlation between dengue incidence and mosquito density). From January 2009 to December 2010, an epidemic of dengue occurred with two waves in Kaohsiung City. The first wave was from July 31st 2009 to March 1st 2010, with totally 646 confirmed indigenous dengue cases [including 9 DHF cases and 3 deaths, case fatality rate (CFR) of 33% (3/9), dengue incidence of 4.23 per 10,000 persons, DHF incidence of 0.06 per 10,000 persons]. The second wave was from March 1st 2010 to December 31st 2010, with totally 956 confirmed indigenous dengue cases [including 6 DHF cases and 1 deaths, case fatality rate (CFR) of 16.7% (1/6), dengue incidence of 6.24 per 10,000 persons, DHF incidence of 0.04 per 10,000 persons]. Data analysis between mosquito density and environment showed that the density of small-breeding sites (measured as “container density”) had higher correlation with dengue incidence than the density of large-breeding sites (rs= 0.52 vs. 0.16, p<0.01), and container index was more sensitive to dengue incidence than house index, Breteau index and adult mosquito index (rs= 0.49 vs. 0.37, 0.35, 0.20, p<0.01). In univariate spatial autocorrelation analyses (univariate Moran’s I and univariate Local Indicator Spatial Analysis), dengue incidence, population density, and the density of large/small-breeding sites, and container index had spatial clusters and their hotspots overlapped in Gushan and SanMin Districts. It indicated that environmental factors were closely correlated to high dengue incidence. The other types of large-breeding sites excluding basements with cumulated water, ovitrap-ponds, past listed locations with mosquito breeding sites , air-conditioning cooling water towers, vacant and dirty houses waiting for repairing, resource/tire recycle factories), such as wasteland, was significantly correlated to dengue incidence (rs=0.188, p<0.01). In contrast with large-breeding sites, totally 13 types of containers had significantly positive correlations with dengue incidence, the ranking of the first four high correlation coefficients were toilet tanks, jars, pots and water gauges (rs= 0.53, 0.45, 0.40, 0.40). In addition, indoor and outdoor jars had the highest correlations with dengue incidence (indoor: rs= 0.57, p<0.01; outdoor: rs= 0.42, p<0.01), indoor flower vases (rs= 0.44, p<0.01) and outdoor pots (rs= 0.304, p<0.01) ranked the second, respectively. The comparisons between Lis with/without dengue cases, the Lis with dengue cases had higher means of population density (25.95 vs. 18.44, p<0.01), density of large-breeding sites (59.85 vs. 47.94, p=0.02), the proportion of households with removing containers (96.31 vs. 47.82, p<0.01), density of small-breeding sites (35.17 vs. 20.55, p<0.01), container index (7.91 vs. 4.92, p<0.01) and the proportion of field adult Aedes aegypti (94.01 vs. 74.06, p<0.01) than the Lis without dengue cases. In multiple logistic regression analyses, the adjusted odds ratios (ORadj) of high population density (ORadj= 9.70, 95%CI=4.65-20.24, p<0.01), high small-breeding sites (ORadj= 6.33, 95%CI=2.83-14.16, p<0.01) and high levels of container index (ORadj= 12.65, 95%CI= 6.11-26.29, p<0.01) were the important risk factors of dengue with statistical significance. Finally, evaluations of public health interventions demonstrated that the week with removing household containers had a timely negative correlation with container index in that week. The negative correlation became statistically significant after 4 weeks (-0.32, p<0.05) and such a correlation peaked at 5 weeks after the removal household breeding sites (-0.52, p<0.01). In summary, environmental factors (population density, the densities of large-/small-breeding sites) and vector density (house index, Breteau index, container index and the proportion of field adult Aedes aegypti) were the important risk factors of dengue with spatial clusters in particular areas. In addition, density of small-breeding sites had higher correlation with dengue incidence but cannot neglect the impact of large-breeding sites covering large areas with difficulties in removal. The areas with higher population density, more density of small-breeding sites, and greater container index had higher likelihood to have dengue epidemics. Container index was the most sensitive to reflect epidemic status among all available mosquito indices studied. In fact, the routine activities of removing containers and spraying did control vector density certainly, but their effectiveness cannot be observed immediately. With the trends in increasing proportion of Aedes aegypti in Kaohsiung City, future efforts should target at integrating two vector surveillance systems (household container and large-breeding sites) to control the vector density, elevate public health effectiveness and reduce insecticide resistance. With the aid from strengthening surveillance system and timely reporting/diagnosis, we will be able to construct the risk maps of dengue for better effectiveness in control. In order to prevent dengue become endemic in Kaohsiung City, future efforts should involve: (1) the comparison of this study with the past DENV-3 epidemics in 2006 and 2009, (2) investigating the spatial distributions of asymptomatically infected dengue cases through seroepidemiological studies, and (3) establishment of pre-epidemic warning index to interrupt dengue transmission and minimize DHF cases, after fully understanding the triad interrelationships between agent, host and environment.