Background
Dengue is a mosquito-borne viral disease that has one of the world’s fastest growing burden [
1]. Despite substantial investments, existing vector control methods, such as insecticides, have proved insufficient to sustainably control dengue [
2]. Novel arbovirus vector control tools are needed, and a range of alternative approaches are currently under development to meet this need [
3,
4]. Mosquitoes infected with
Wolbachia, a naturally occurring bacterium, experience reduced rates of dengue virus (DENV) infection, and female mosquitoes can pass the bacterium on to the next generation, allowing
Wolbachia-infected mosquitoes to replace the wild-type population [
5]. Release of male mosquitoes infected with
Wolbachia can also be used for population suppression due to inviable mating with female
wild-type mosquitoes. Early releases of mosquitoes infected with the wMel
Wolbachia strain have shown promising replacement results, and suppression strategies with other strains are currently being tested in different countries around the world [
6‐
9].
An added advantage of a population replacement strategy is that
Wolbachia reduces replication of other arboviruses within the mosquito, including chikungunya, yellow fever and Zika viruses [
10,
11], and potentially offers the better longer-term strategy. Given such replacement programmes are self-sustaining, investment in a well-coordinated and properly monitored release campaign over 2 to 3 years could have many years of benefit. Existing releases at the local and city level have proven that
Wolbachia-infected mosquitoes can replace the wild-type
Aedes aegypti population and persist for at least 7 years’ post-release [
12]. Epidemiological evidence of effectiveness is also growing, and a cluster randomised controlled trial is currently underway in the city of Yogyakarta [
13]. The next phase of development for
Wolbachia will be to expand from single-site operations to coordinated sub-national roll-out.
As the most populous country in dengue-endemic South East Asia, Indonesia is consistently estimated to be among the three countries with the largest dengue burden [
14‐
16]. However, due to high rates of asymptomatic infection and symptoms which are not easily distinguishable from many other infections, the number of dengue cases is still highly uncertain. Accurate, contemporary estimates of the burden of dengue in Indonesia are needed to quantify the benefits of any scale-up in DENV control. Fully detailing how the economic and case burden of dengue is distributed over space, by disease severity and financial responsibility can help inform investment in new control tools. This is particularly important for diseases such as dengue where the burden is dominated by morbidity rather than mortality [
15]. Milder dengue cases are nearly always underreported [
17], and the costs of illness by various parties often hidden [
18]. When combined with model-based estimates of the impact of the intervention, burden estimates can be used to map where new interventions, such as
Wolbachia, are likely to have the biggest effect and can be used for evaluating eventual impact.
A major challenge to understanding the impact of interventions against DENV is an accurate estimation of baseline disease burden. Estimates of disease burden for specific settings are often scarce due to limited availability of data on the sub-clinical community-based burden of dengue including asymptomatic and mildly symptomatic cases. Efforts to estimate the burden of dengue can be categorised into either a bottom-up approach, where the primary focus is to estimate the total number of cases through community-based surveys for infection [
14], then divide into different levels of severity, or top-down approach where reported case numbers are multiplied by “expansion factors” to correct for underreporting [
16]. Multiple previous studies have estimated the burden of dengue in Indonesia [
14‐
16,
19‐
21] using a variety of data sources and methods, but it is difficult to assess consensus among them due to the differences in data sources, methods, case definitions and assumptions about transmission.
Three types of data are typically available for mapping the spatial distribution of dengue burden: occurrence (presence/absence), case incidence and seroprevalence (lifetime prevalence). Seroprevalence data contain the most information about long-term average burden in a given location, but few such surveys have been conducted, typically resulting in less information about the geographic variation. Occurrence data, on the other hand, is geographically ubiquitous, but many other factors determine how the presence of a disease translates into case numbers. Existing approaches to map dengue risk have been dominated by ecological niche modelling using occurrence data [
22‐
24] with a focus on mapping the distribution rather than the burden of dengue. Maps of reported dengue incidence at increasingly high spatial resolution are routinely used by ministries of health but are rarely combined with models to account for variations over time, reporting biases and quantification of uncertainty. Some attempts have been made to map seroprevalence data directly in areas with sufficient surveys [
25]. However, these contrasting approaches have never formally been compared to identify their strengths and weaknesses for mapping burden. There is also a lack of consensus on how useful extrapolating from data in other countries or transmission settings is for mapping burden in any one given country.
In the current absence of cluster randomised control trial results for
Wolbachia, estimates of effectiveness have been obtained by combining vector competence studies with mathematical models of DENV transmission [
26]. A range of DENV transmission models have been published and, despite some fundamental differences in their structures, consensus results about the effects of interventions can be drawn [
27]. Even with the imperfect reduction of DENV dissemination in the mosquito, substantial reductions in population-level burden can be achieved, even in very high-transmission settings [
26,
28,
29]. However, the critical relationship between baseline transmission intensity and
Wolbachia effectiveness is yet to be demonstrated in the field. Further, how control might be impacted by the highly heterogeneous transmission intensities routinely observed across small spatial scales [
30‐
32] remains unknown. It is possible that if the impact on transmission is small, this may just increase the average age of secondary, typically more severe, DENV infection to older more vulnerable age groups; thus a detailed consideration of DENV immunology is needed in such assessments.
Here, we produce the most up-to-date, detailed and robust estimates of the burden of dengue in Indonesia; map burden at a high spatial resolution throughout the country; and predict the effect of a widespread Wolbachia programme in different locations.
Discussion
In this paper, we produce comprehensive estimates of the burden of dengue in Indonesia and find that a large proportion of cases self-manage their own disease (64%, 5.0 million) or are treated in outpatient departments (22%, 1.7 million). We use multiple mapping methods and data sources to show that the spatial distribution of dengue risk is heterogeneous even in an endemic country such as Indonesia. The highly urbanised nature of the population means that 14.7% of the national burden is concentrated in just 10 cities. Finally, we show that a nationwide Wolbachia campaign could (over the long term) avert a significant proportion of burden (86.2%, UI 36.2–99.9%) with elimination predicted in low transmission settings.
The high spatial concentration of dengue burden in cities, in highly urbanised countries such as Indonesia, presents opportunities for targeted control strategies. In particular, Wolbachia, which is deployed on a per-km2 basis, could offer major scaling advantages over vaccines, which are deployed on a per-person basis, in areas with high population density. The large number of people covered by a focal Wolbachia programme has the potential to outweigh the reduced efficacy of the intervention in these high transmission settings, and formal cost-effectiveness analysis is needed to compare the investment cases between urban and rural areas.
This work adds to a growing body of evidence that the majority of the burden of dengue is attributable to morbidity rather than mortality [
14,
15,
19,
52]. The large number of self-limiting mild infections contributes more to DALY burden than the small number of infections that result in severe or fatal manifestations. Many of these mild cases do not seek treatment, are not clinically diagnosable and thus do not have any opportunity to be reported in routine health statistics. These results can be used to assess the hidden economic burden of the disease and to estimate the cost-effectiveness of interventions for dengue [
16,
27]. Our results also suggest that only 12% (UI 7–45%) of hospitalised cases are reported. While lower than the regional average (42%) [
17], underreporting of dengue is not unusual and may occur for a variety of reasons including lack of reporting in the private sector, misdiagnosis and limited coverage of the surveillance system [
53].
A key limitation of our analysis is the wide uncertainty intervals for our final estimates of burden, and thus predicted efficacy of
Wolbachia. This arises due to the limited quantity and variable quality of datasets detailing the treatment-seeking behaviour for dengue [
17], reliability of diagnosis and underreporting of identified cases. In this study, we chose to ensemble different burden estimation methods with equal weighting due to different data sources and methodological approaches challenging any formal assessment of quality or comparativeness. Initiatives such as the WHO burden estimation toolkit [
53] aim to provide guidance to countries on how to conduct burden estimation for dengue and aim to generate more standardised and internationally comparable data for dengue burden estimation. Additionally, while using the national SUSENAS survey to estimate the treatment-seeking rates was a great strength due to its sample size and comprehensive design, it did require assuming that treatment seeking for fever is comparable to treatment seeking for dengue. As fever is one of the milder symptoms of dengue [
54], this may have underestimated rates of seeking care [
55].
Different data sources suggest different spatial distributions of dengue risk. This is partly because each data source has strengths and weaknesses for measuring different aspects of dengue’s distribution (summarised in Additional file
1: Table S11) [
23]. Occurrence data is most informative about the extent of transmission, incidence about temporal variation and seroprevalence about long-term risk of infection. Occurrence and incidence data may also be subject to spatial reporting bias, e.g. higher probability of reporting in urban areas, which may lead us to overestimate the concentration of risk in high-density areas. We tried to overcome this by using notifications of other infectious diseases (which are also subject to the same biassed sampling frame) as background points, and the relative influence statistics (Additional file
1: Table S9) and covariate effects plots (Additional file
1: Figure S6) do not suggest simple univariate drivers of dengue’s distribution in Indonesia. Disease mapping frameworks have been suggested that would enable simultaneous joint inference of the distribution and observation bias of multiple rare diseases and could improve occurrence maps for diseases that share similar characteristics but limited data [
56]. Future work will attempt to more formally define relationships between occurrence, incidence and seroprevalence data and their relationship with burden to enable joint inference that accounts for the accuracies, sensitivities and biases in each data source [
57].
Our mathematical model assumed a stable prevalence of
Wolbachia in the wild
Aedes population and only focussed on the long-term stable-state effectiveness. With the high levels of herd immunity currently present in Indonesia, it is possible that elimination would temporarily be achieved even in high transmission intensity areas and short-term impact would generally likely be higher than predicted here [
58]. Our analysis of vector competence data only compared dissemination rates to the mosquito salivary glands in lab-reared (not-field caught) mosquitoes. Effectiveness may be higher in the field due to the effect field conditions impose on the mosquito immune system and the availability of nutritional resources [
51]. Due to the lack of available vector competence data, we were only able to model the reduction in transmission due to one strain of
Wolbachia (wMel) and one vector species (
Ae. aegypti).
Ae. albopictus, a known secondary DENV vector, is also present in Indonesia, although it typically has a more rural distribution and its role in sustaining dengue transmission in this setting remains unclear [
59]. Different
Wolbachia strains also vary in their DENV-blocking dynamics, their effects on mosquito longevity and can be affected by local conditions, e.g. temperature [
60], meaning further reductions in DENV transmission may be possible. Finally, our modelling comparison exercise only used the parameter estimates from each of the models, not the model structures themselves, which may include additional uncertainty and provide further insights into the effectiveness of
Wolbachia and its variation across transmission intensity. Our current estimates are in agreement with earlier work suggesting elimination is achievable in low transmission intensity but not high transmission intensity environments [
26]. This raises the possibility that
Wolbachia may need to be combined with a range of other dengue control tools in high endemicity environments. The key strength of this analysis is that it is the most detailed analysis of Indonesia’s dengue burden to date. We combine multiple modelling and mapping approaches with multiple datasets and fully propagate uncertainty at each step through to our final results.
Future work will include pairing these burden estimates and impact predictions with economic data on the costs of dengue illness and of deploying Wolbachia in different areas. This will allow estimates of the cost-effectiveness of Wolbachia programmes and estimates of how it varies throughout Indonesia that can be used to quantify the costs and benefits of future investments in wide-scale releases and inform different release strategies.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.