Quantifying Arbovirus Disease and Transmission Risk at the Municipality
Total Page:16
File Type:pdf, Size:1020Kb
medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . 1 1 Title: Quantifying arbovirus disease and transmission risk at the municipality 2 level in the Dominican Republic: the inception of Rm 3 Short title: Epidemic Metrics for Municipalities 4 Rhys Kingston1, Isobel Routledge1, Samir Bhatt1, Leigh R Bowman1* 5 1. Department of Infectious Disease Epidemiology, Imperial College London, UK 6 *Corresponding author 7 [email protected] 8 9 NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 1 medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . 2 10 Abstract 11 Arboviruses remain a significant cause of morbidity, mortality and economic cost 12 across the global human population. Epidemics of arboviral disease, such as Zika 13 and dengue, also cause significant disruption to health services at local and national 14 levels. This study examined 2014-16 Zika and dengue epidemic data at the sub- 15 national level to characterise transmission across the Dominican Republic. 16 17 For each municipality, spatio-temporal mapping was used to characterise disease 18 burden, while data were age and sex standardised to quantify burden distributions 19 among the population. In separate analyses, time-ordered data were combined with 20 the underlying disease migration interval distribution to produce a network of likely 21 transmission chain events, displayed using transmission chain likelihood matrices. 22 Finally, municipal-specific reproduction numbers (Rm) were established using a 23 Wallinga-Teunis matrix. 24 25 Dengue and Zika epidemics peaked during weeks 39-52 of 2015 and weeks 14-27 of 26 2016 respectively. At the provincial level, dengue attack rates were high in 27 Hermanas Mirabal and San José de Ocoa (58.1 and 49.2 cases per 10,000 28 population respectively), compared with the Zika burden, which was highest in 29 Independencia and San José de Ocoa (21.2 and 13.4 cases per 10,000 population 30 respectively). Across municipalities, high disease burden was observed in Cotui (622 31 dengue cases per 10,000 population) and Jimani (32 Zika cases per 10,000 32 population). Municipal infector-infectee transmission likelihood matrices identified six 33 0% likelihood transmission events throughout the dengue epidemic and one 0% 34 likelihood transmission event during the Zika epidemic. Municipality reproduction 2 medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . 3 35 numbers (Rm) were consistently higher, and persisted for a greater duration during 36 the Zika epidemic (Rm = 1.0), than during the dengue epidemic (Rm = <1.0). 37 38 This research highlights the importance of disease surveillance in land-border 39 municipalities as an early warning for infectious disease transmission. It also 40 demonstrates that a high number of importation events are required to sustain 41 transmission in endemic settings, and vice versa for newly emerged diseases. The 42 inception of a novel epidemiological metric, Rm, reports transmission risk using 43 standardised spatial units, and can be used to identify high transmission risk 44 municipalities to better focus public health interventions for dengue, Zika, and other 45 infectious diseases. 46 3 medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . 4 47 Author Summary 48 49 Arboviruses remain a significant cause of morbidity, mortality and economic cost. 50 Between the years 2014-16, two large arbovirus outbreaks occurred in the 51 Dominican Republic. The first was a wave of dengue cases, followed by a large Zika 52 epidemic. Using various mathematical modelling and geospatial approaches, a 53 number of analyses were undertaken to both characterise the pattern of disease 54 transmission and identify high-burden municipalities. Throughout the process, a 55 novel metric was developed: the Rm. This parameter was used to identify the 56 transmission potential of any given municipality to surrounding municipalities, where 57 >1.0 is high transmission risk, and <1.0 is low transmission risk. This is useful as it 58 provides a standardised approach to determine where public health resources might 59 be focussed to better impact ongoing disease transmission. Additionally, analyses 60 demonstrated the importance of increased disease surveillance in municipalities that 61 share land borders with neighbouring countries, and how relatively few disease 62 importation events can spark and sustain an epidemic. 63 4 medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . 5 64 Introduction 65 Arboviruses are an informal name for a group of viruses transmitted by arthropods 66 such as ticks, mosquitoes and sand flies (1) - members of which include Rift Valley 67 Fever, Chikungunya and West Nile Virus (2). Arboviruses are commonly zoonotic, 68 and the cause of increasing human disease worldwide. In recent years, the 69 arboviruses Zika and dengue have afflicted millions via endemic and epidemic 70 transmission, in part due to relatively few, effective means of control (3). Indeed, 71 current estimates suggest that the global annual burden of dengue infections is 390 72 million, with 96 million manifesting clinically (4). Those at risk number 3.97 billion 73 across 128 countries worldwide (5). In the case of Zika, estimates of the global 74 burden are not yet available, however by the end of 2018, the Pan American Health 75 Organisation (PAHO) had reported 19,020 suspected cases of Zika, with 1,379 76 laboratory confirmed cases in Brazil alone (6). 77 78 Dengue and Zika are principally transmitted via Aedes mosquitoes. When a female 79 Aedes mosquito bites an infected human, the mosquito ingests a blood meal 80 containing the virus, at which point it enters the midgut, proliferates, and spreads to 81 the salivary glands. Once the mosquito bites another person, the cycle is complete 82 (7). However, vertical transmission can also occur, and while this is relatively rare for 83 dengue (8), such transmission is more common with Zika; indeed in a prospective 84 cohort, 26% of maternal cases resulted in vertical transmission to the unborn foetus 85 (9). Importantly, sexual transmission between humans is also a significant driver of 86 Zika epidemiology (10). 87 5 medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . 6 88 The basic reproduction number (R0) describes the average number of secondary 89 infections produced by a single infectious individual in a totally susceptible 90 population (11). Epidemics involving novel pathogens are best described using R0, 91 due to the absence of existing population immunity (12). By contrast, Reff is most 92 appropriate in endemic settings (11), when part of the population is already immune 93 (12). In the absence of field data, mathematical modelling is used to average the 94 expected number of new infections over all possible infected individuals. This idea 95 can be represented by a matrix where the reproduction number is recognised as the 96 dominant eigenvalue of an operator, which is linear for every pair of functions, and 97 can be calculated whilst considering other factors such as age stratification (13). 98 99 The simplest form of epidemiological modelling is mechanistic, which deploys 100 compartments with interconnected per capita rates to describe the movement of 101 individuals between disease states (14). This field has since been further expanded 102 to include network analysis. Wallinga and Teunis applied this approach (15) to 103 estimate both the serial interval distribution (16) and the Reff of Severe Acute 104 Respiratory Syndrome (SARS). In similar research, Routledge et al., 2018 also used 105 network-based analysis to predict malaria elimination time scales (17). Together, 106 these studies further developed mathematical modelling used to calculate individual 107 reproduction numbers (18) while building on the established Reed-Frost model of 108 epidemic transmission (19)(20). And while these approaches are powerful, they are 109 highly reliant on granular data to infer geospatial disease spread at fine scales, yet 110 these data are not always available 6 medRxiv preprint doi: https://doi.org/10.1101/2020.06.30.20143248; this version posted July 1, 2020.