Background

Dengue fever control in the tropical island of Bali in Indonesia carries important significance both nationally and globally, as it is one of the most endemic islands in Indonesia and a worldwide popular travel destination. Despite its importance, the spatial and temporal heterogeneity in dengue risk and factors associated with its variation in risk across the island has not been not well explored. This study was aimed to analyze for the first time the geographical and temporal patterns of the incidence of dengue and to quantify the role of environmental and social factors on the spatial heterogeneity of dengue incidence in Bali.

Methods

We analyzed retrospective dengue notification data at the sub-district level (Kecamatan) from January 2012 to December 2017 which obtained from the Indonesian Ministry of Health. Seasonality in notified dengue incidence was assessed by seasonal trend decomposition analysis with Loess (STL) smoothing. Crude standardized morbidity rates (SMRs) of dengue were calculated. Moran's I and local indicators of spatial autocorrelation (LISA) analysis were employed to assess spatial clustering and high-risk areas over the period studied. Bayesian spatial and temporal conditional autoregressive (CAR) modeling was performed to quantify the effects of rainfall, temperature, elevation, and population density on the spatial distribution of risk of dengue in Bali.

Results

Strong seasonality of dengue incidence was observed with most cases notified during January to May. Dengue incidence was spatially clustered during the period studied with high-risk kecamatans concentrated in the south of the island, but since 2014, the high-risk areas expanded toward the eastern part of the island. The best-fitted CAR model showed increased dengue risk in kecamatans with high total annual rainfall (relative risk (RR): 1.16 for each 1-mm increase in rainfall; 95% Credible interval (CrI): 1.03–1.31) and high population density (RR: 7.90 per 1000 people/sq.km increase; 95% CrI: 3.01–20.40). The RR of dengue was decreased in kecamatans with higher elevation (RR: 0.73 for each 1-m increase in elevation; 95% CrI: 0.55–0.98). No significant association was observed between dengue RR and year except in 2014, where the dengue RR was significantly lower (RR: 0.53; 95% CrI: 0.30–0.92) relative to 2012.

Conclusions

Dengue incidence was strongly seasonal and spatially clustered in Bali. High-risk areas were spread from kecamatans in Badung and Denpasar toward Karangasem and Klungkung. The spatial heterogeneity of dengue risk across Bali was influenced by rainfall, elevation, and population density. Surveillance and targeted intervention strategies should be prioritized in the high-risk kecamatans identified in this study to better control dengue transmission in this most touristic island in Indonesia. Local health authorities should recommend travelers to use personal protective measures, especially during the peak epidemic period, before visiting Bali.

中文翻译：

---

印度尼西亚巴厘岛登革热发病率的时空变化：一项生态学研究

背景

印度尼西亚热带岛屿巴厘岛的登革热控制在国内和全球都具有重要意义，因为它是印度尼西亚最流行的岛屿之一，也是全球热门的旅游目的地。尽管登革热很重要，但登革热风险的空间和时间异质性以及与全岛风险变化相关的因素尚未得到很好的探索。本研究旨在首次分析登革热发病率的地理和时间模式，并量化环境和社会因素对巴厘岛登革热发病率空间异质性的作用。

方法

我们分析了2012 年 1 月至 2017 年 12 月从印度尼西亚卫生部获得的街道级 ( Kecamatan ) 的回顾性登革热通报数据。通过使用黄土 (STL) 平滑的季节性趋势分解分析评估已通报的登革热发病率的季节性。计算了登革热的粗标准化发病率 (SMR)。Moran's I和局部空间自相关指标 (LISA) 分析用于评估研究期间的空间聚类和高风险区域。执行贝叶斯空间和时间条件自回归 (CAR) 模型以量化降雨、温度、海拔和人口密度对巴厘岛登革热风险空间分布的影响。

结果

大多数登革热病例是在 1 月至 5 月报告的，登革热发病率具有很强的季节性。在研究期间，登革热发病率在空间上聚集，高风险的kecamatans集中在该岛的南部，但自 2014 年以来，高风险地区向该岛的东部扩展。最适合的 CAR 模型显示，年总降雨量高的 kecamatans 登革热风险增加（相对风险 (RR)：降雨量每增加 1 毫米，1.16；95% 可信区间 (CrI)：1.03–1.31）和高人口密度（RR：每 1000 人/平方公里增加 7.90；95% CrI：3.01–20.40）。kecamatans的登革热 RR 降低海拔较高（海拔每增加 1 米，RR：0.73；95% CrI：0.55–0.98）。除 2014 年外，登革热 RR 与年份之间未观察到显着关联，2014 年登革热 RR 相对于 2012 年显着较低（RR：0.53；95% CrI：0.30–0.92）。

结论

巴厘岛的登革热发病率具有很强的季节性和空间聚集性。高风险地区从巴东和登巴萨的kecamatans向 Karangasem 和 Klungkung扩散。巴厘岛登革热风险的空间异质性受降雨量、海拔和人口密度的影响。在本研究中确定的高风险kecamatans中，应优先采取监测和有针对性的干预策略，以更好地控制登革热在印度尼西亚这个旅游胜地的传播。当地卫生部门应建议旅客在前往巴厘岛之前采取个人防护措施，尤其是在疫情高峰期。