The epidemiologic and economic burden of dengue in Singapore: A systematic review

Background Despite its well-regarded vector control program, Singapore remains susceptible to dengue epidemics. To assist evaluation of dengue interventions, we aimed to synthesize current data on the epidemiologic and economic burden of dengue in Singapore. Methodology We used multiple databases (PubMed, Embase, Cochrane, international/national repositories, surveillance) to search for published and gray literature (2000–2022). We included observational and cost studies, and two interventional studies, reporting Singapore-specific data on our co-primary outcomes, dengue incidence and dengue-related costs. Quality was assessed using the Newcastle-Ottawa Scale and an adapted cost-of-illness evaluation checklist. We performed a narrative synthesis and grouped studies according to reported outcomes and available stratified analyses. Findings In total, 333 reports (330 epidemiological, 3 economic) were included. Most published epidemiological studies (89%) and all economic studies were of good quality. All gray literature reports were from the Ministry of Health or National Environment Agency. Based predominantly on surveillance data, Singapore experienced multiple outbreaks in 2000–2021, attaining peak incidence rate in 2020 (621.1 cases/100,000 person-years). Stratified analyses revealed the highest incidence rates in DENV-2 and DENV-3 serotypes and the 15–44 age group. Among dengue cases, the risk of hospitalization has been highest in the ≥45-year-old age groups while the risks of dengue hemorrhagic fever and death have generally been low (both <1%) for the last decade. Our search yielded limited data on deaths by age, severity, and infection type (primary, secondary, post-secondary). Seroprevalence (dengue immunoglobulin G) increases with age but has remained <50% in the general population. Comprising 21–63% indirect costs, dengue-related total costs were higher in 2010–2020 (SGD 148 million) versus the preceding decade (SGD 58–110 million). Conclusion Despite abundant passive surveillance data, more stratified and up-to-date data on the epidemiologic and economic burden of dengue are warranted in Singapore to continuously assess prevention and management strategies.


Introduction
Dengue, the fastest spreading mosquito-borne viral disease in the world, has increased in global incidence by 30-fold between 1960 and 2010 [1,2].Annually, it now affects over 100 million individuals across the world [3].Apart from a high burden of disease, it also imposes an economic burden from direct and indirect costs, incurring an estimated USD 8.9 billion in total costs globally in 2013 [4].
The dengue virus (DENV) has four distinct serotypes (DENV-1 to -4) and is transmitted via the tropical/subtropical mosquito vectors Aedes aegypti and Aedes albopictus, whose extrinsic incubation periods are influenced by ambient temperature and the climate [2].Geographical areas with dengue transmission have expanded in recent years, mainly as a result of increased population growth rate, unplanned urbanization, inefficient mosquito control, climate change, and increased frequency of international travel from endemic to non-endemic regions [1,5,6].
Based on the World Health Organization (WHO) [2] regional classification, more than 70% of the world's dengue at-risk population live in only two regions, including the Western Pacific Region.Meanwhile, according to the Global Burden of Disease (GBD) [7] grouping, the high-income Asia Pacific region exhibited the highest annual increment in dengue incidence from 1990 to 2019.In the same 20-year period, it was also the fourth leading region in terms of rise in disability-adjusted life years (DALYs) from dengue [6].Singapore belongs to both these WHO-and GBD-defined regions.
The presence of mosquito vectors and appropriate environmental conditions for virus transmission makes dengue endemic in Singapore [8].Nevertheless, the dengue control program in Singapore is one of the most highly regarded globally, effectively reducing the force of infection of dengue in the background of a progressively aging Singaporean population [9].While vector control in Singapore applies an interagency "whole-of-government" approach, it primarily consists of environmental management, including more recent innovations such as (1) Wolbachia interventions, whereby eggs produced by male Wolbachia-infected mosquitoes and wild-type females do not hatch ("cytoplasmic incompatibility"); (2) the use of Aedes aegypti breeding percentages as risk indicators; and (3) risk mapping (e.g., spatiotemporal models) for targeted resource allocation [9].
Singapore's vector control initiatives are complemented by its disease surveillance program.Medical practitioners and laboratories in Singapore are mandated to report a case of dengue fever or dengue hemorrhagic fever (DHF) soon after diagnosis [8,9].Dengue is comprehensively monitored to understand not only its incidence by passive surveillance but also its serotype distribution by sentinel active surveillance.In the latter, random blood samples are shared for serotyping and genotyping to detect serotype shifts and provide an early warning of potential outbreaks [10].Although definitions of dengue outbreak are highly heterogenous in the literature [11], for this paper it is defined as 'a temporal anomaly in incidence rate from the expected number of cases for the period based on historical data'.
Despite dengue control efforts, Singapore has remained susceptible to epidemics, experiencing several outbreaks in the past decade [9].This phenomenon has been attributed to high-level mobility in and out of Singapore, as well as low herd immunity due to successful reduction of disease transmission [9].Meanwhile, the total costs of dengue are also exceptionally high in Singapore compared with other Southeast Asian countries, attributable to factors such as higher work absenteeism costs related to higher GDP per capita, and higher costs of hospitalization [12].The latter costs are borne by a combination of the individual, private or national income tax-based insurance and the government, with the ratio of contributions dependent on patient eligibility for government subsidies [13].Furthermore, the highly effective and rigorous publicly-funded vector control program implemented in Singapore comprises ~40-60% of total dengue-related costs [14].
Although the field of dengue vaccine production has advanced in recent years, challenges persist in terms of developing vaccines that are equally protective against all serotypes [15].To choose among these multiple available interventions for dengue control, up-to-date and comprehensive information about dengue epidemiology and costs in Singapore are needed.Hence, we undertook this systematic review to synthesize currently available data on the epidemiologic and economic burden of dengue in Singapore.

Methods
This review was conducted in line with the Cochrane Handbook for Systematic Reviews of Interventions [16], the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [17], and the Synthesis Without Meta-analysis reporting guidelines [18].

Criteria for study inclusion
We included studies conducted in Singapore that involved individuals of any age with suspected dengue, confirmed dengue, or potential previous dengue exposure.We excluded international studies from which data exclusive to Singapore could not be extracted.Studies with an observational (real-world) study design (e.g., case-control, cohort, cross-sectional, ecological, genomic) and cost studies (e.g., cost description, cost analysis, cost-utility analysis, cost-effectiveness analysis) were eligible for inclusion.Interventional studies (e.g., clinical trials) were considered if usable baseline data were available.No restrictions were placed on the type of intervention or comparator.Case reports and systematic reviews without meta-analyses were excluded.
In terms of outcome measures, we included studies that reported dengue incidence in Singapore (co-primary outcome of interest) or other outcomes pertaining to epidemiologic burden: serotype distribution, seroprevalence, hospitalization, mortality, complications, and the lengths of hospitalization, treatment, illness, or disability.For economic burden, we considered studies reporting dengue-related total costs (co-primary outcome of interest) or disaggregated costs (i.e., direct medical, direct non-medical, indirect societal).

Search methods
Using a predefined search strategy composed of a combination of controlled vocabulary and free-text terms (S1 and S2 Tables), we searched multiple electronic journal databases until 31 May 2022: Embase, MEDLINE via PubMed (1966-2022), and the Cochrane Library.Our search across journal databases was limited to publications in English and human studies.Additionally, to find both published data and gray literature, we searched several international and national research databases (S3 Table ), including dengue surveillance databases: WHO Bulletins and Newsletters for Dengue Situation Updates, Ministry of Health (MOH) Singapore, and Singapore's National Environment Agency (NEA).
A search was conducted separately for the epidemiologic burden (i.e., 'epidemiology search') and economic burden of dengue (i.e., 'costs search').To retrieve only the latest data, but also detect epidemiology trends over time and accommodate for long-term vector control and dengue prevention strategies, we set the search period to 2000-2022.Meanwhile, the search period for costs of dengue was restricted to 2009-2022 (the previous ~10 years), in accordance with the recommended approach to obtaining cost-of-illness data in systematic reviews, with the aim to increase generalizability of findings to current and future years [19].Duplicates between the epidemiology search and costs search were only counted once under the more appropriate search category.

Study selection, data extraction, and quality assessment
Two reviewers independently screened titles and abstracts and assessed the full reports of potentially eligible studies, along with the dengue research and surveillance databases.A PRISMA flow diagram was used to detail the process of inclusion and exclusion.Using a structured data extraction form, two reviewers individually extracted data from the included reports and subsequently compared notes.Extracted data from journal publications were summarized in a table of characteristics of included studies.The extracted parameters included seroprevalence, serotype distribution, incidence, hospitalization, mortality, direct and indirect costs and expansion factors (EFs), the latter of which are used to account for underreporting of dengue cases when determining cost estimates.EF parameters are detailed in the S14 Table .The extracted parameters from dengue research and surveillance databases included incidence rates (IRs), serotype distribution, hospitalization rates and case fatality rates (CFRs).

Data synthesis
We performed a narrative synthesis and grouped the included studies based on reported outcomes: (1) each of the epidemiologic outcomes and (2) costs of illness.Epidemiological studies were further grouped according to stratified analyses by locality, age or age group, sex, disease severity, serotype, and infection type (i.e., past, primary, secondary, post-secondary).The literature search was not initially intended or designed to be conducted according to these subgroups; however, the findings are presented in this manner to facilitate robust interpretation.Findings were to be reported separately for studies published in peer-reviewed journals and for gray literature (i.e., not produced by commercial publishers), such as reports retrieved from surveillance databases [16].
Incidence data were summarized using the annual IR per 100,000 person-years.To produce annualized values, we combined weekly or monthly data from the same primary source or computed yearly averages, as appropriate.Imputation was performed to estimate the incidence of dengue by serotype distribution.The IRs were based on midyear population estimates from the Department of Statistics of Singapore [23].Hospitalization rates referred to the proportion of dengue cases hospitalized per year.CFRs were calculated from the reported number of deaths and cases per year.Other outcomes were summarized using metrics as reported in individual studies.All costs identified were standardized by conversion to 2021 Singapore Dollar (SGD) equivalent values inflated using the average Singapore consumer price index [24].
To present epidemiologic trends from the same primary source, we utilized bar, line, or scatter plots.When presenting findings from multiple primary sources, we used tables to summarize content.We used local polynomial regression fitting to produce the line graphs and Fisher's exact test or z-test to produce 95% confidence intervals (CIs) for CFRs.Statistical analyses were conducted in R 4.3.0[25].

Results of the search
Our search (Figs 1 and 2) yielded 1,366 records from electronic journal databases (703 records) and international/national databases (663 records).After removal of duplicates and screening of titles/abstracts, 61 published reports and all 663 gray literature reports underwent full-text screening.Finally, a total of 333 reports (31 published, 302 gray literature) were included in this review.Out of the 31 included published reports, 28 originated from the epidemiology search while three were acquired from the costs search [12,14,.All 302 reports included from gray literature were generated by the epidemiology search [55][56][57].
The published studies were conducted mostly among adults in a hospital or clinic setting.Most (n = 18) published epidemiological studies had a cross-sectional or retrospective cohort design (Table 1), whereas two out of three economic studies were cost-effectiveness analyses (CEAs) (Table 2).Seroprevalence was the most frequently reported outcome among the published epidemiological studies.Our co-primary outcomes of interest, i.e., dengue incidence and dengue-related costs, were reported in eight of the published epidemiological studies and in all three published economic studies, respectively.None of the included studies reported dengue complications and lengths of treatment, illness, or disability.The gray literature included in this review consisted of publicly available surveillance data from the MOH (n = 282) and NEA (n = 20).
In terms of methodological quality (Table 1), the majority of published epidemiological studies (25/28 = 89%) received the highest rating ('good').The remaining studies were of fair quality, most commonly owing to lack of comparability between outcome groups or nonreporting of the degree of uncertainty (e.g., no CIs).All three publications that were included from the costs search were of good quality (Table 2).

Subgroup analysis: by disease severity
Subgroup analysis of MOH surveillance data by disease severity (S6 Table) revealed IRs of 0.1-9.2cases per 100,000 person-years for DHF and 16.5-620.2cases per 100,000 person-years for dengue fever.Among cases, the risk of DHF ranged from 0.1% to 2.8% per year.It was

Subgroup analysis: by age group
Based on MOH surveillance data, the highest IRs were found in the following age groups: 15-24, 25-34, and 35-44 years (Fig 3B and S7 Table ).Meanwhile, the 0-4 and 5-14 age groups generally had the lowest IRs.Three published studies secondarily reported age-stratified IRs from MOH and found these similar trends [32,34,41].

Subgroup analysis: by serotype
Although the MOH did not present IRs stratified according to serotype, it reported the distribution of the four dengue serotypes from 2000 to 2022 (Fig 4A and S8 Table ).DENV-2 was the predominant circulating serotype in most of the 22-year period (2000-2003, 2007-2012, 2015-2020) while DENV-1 was predominant in the rest (2004)(2005)(2006)(2013)(2014).DENV-3 only gained prominence in 2021 and remained to be the predominant serotype in 2022, with an average distribution of 77.5%.

Subgroup analysis: by locality
A nested case-control study by Yung et al. (2016) [54] in five regions of Singapore between 2005 and 2013 showed that the number of dengue cases varied by region.Respectively, the number of cases was highest and lowest in the Central (204 cases) and Southeast (10 cases) regions (Fig 5).However, the authors found that the Northwest region, which lies near the city of Johor Bahru in Malaysia, had significantly lower odds of dengue on multivariate logistic regression analysis [54].Assessed using original and revised Newcastle-Ottawa Scales [20,21] for epidemiological studies and cost-of-illness evaluation checklist from Larg and Moss (2011) [22] for economic studies.

Hospitalization
Although the MOH did not routinely report hospitalization rates, a 2018 surveillance report indicated that 78-85% of yearly reported cases from 2000 to 2002 had been hospitalized or had received inpatient treatment [56].).Subsequently, owing to modifications of hospital admission protocols, hospitalization rates progressively declined to as low as 25.6% in 2014.The mean length of hospitalization among dengue cases over the study period was 3.8 days (range of 1-121 days).Elderly patients (>65 years old) showed significantly longer hospital stays, with an annual mean length of stay of 5-8 days, indicating more severe disease within this age group than others [28].

Subgroup analysis: by age group
In the same study, a subgroup analysis revealed that the decline in hospitalization rates occurred across all age groups, yet the risk of hospitalization was higher among older adults (�45 age groups) than among younger patients (Fig 3C and S10 Table).Furthermore, the length of hospitalization was significantly longer in the �65 age group (mean of 5-8 days) than in other age groups [28].
In contrast, a 2004 retrospective cohort study of hospitalized dengue patients in Tan Tock Seng Hospital found no significant difference between patients aged <60 years and those aged �60 years in terms of hospitalization duration (median of 4 days in both groups) [37].

Subgroup analysis: by infection type
Subgroup analysis of hospitalization by infection type was also available from a prospective cohort study of dengue suspects, which illustrated that the risk of hospitalization was higher with secondary infection than with primary infection (50% vs 20%) [42].

Mortality
Since 2002, the annual CFRs for dengue and their corresponding 95% CI upper bounds have all been less than 1% (Fig 3D and S11 Table ).More specifically, in the last 10 years (2010-2022), the CFRs and upper bounds have not exceeded 0.5%.As can be expected, the CIs were narrowest during epidemic years, when the number of cases was larger.Conversely, the wide CIs reflected the uncertainty of estimates for non-epidemic years, even if some CFRs were artefactually high.
Similar CFR findings were reported in five published studies, which also used secondary data from MOH surveillance [28,29,32,34,41].

Subgroup analysis: by disease severity
Mortality data stratified by disease severity (at presentation) were available from MOH surveillance until 2011 (S12 Table ), after which the MOH stopped reporting stratified data, possibly due to the 2009 changes in WHO case definitions [2].As expected, CFRs for dengue fever (0.1-0.3%) were generally lower than those for DHF, which had a peak of 10.9% in 2009.Three published studies also reported CFRs for DHF within these ranges.Koh et al. ( 2008) [32] and Ler et al. (2011) [34] secondarily reported MOH surveillance data, while Ong et al. (2007) [38] conducted a 2004 case-control study of patients in Tan Tock Seng Hospital.

Subgroup analysis: by age group
The MOH did not report age-stratified deaths, but Lye et al. (2010) [37] found no statistically significant difference between the �60 and <60 age groups in terms of CFRs (0.0 vs 0.1%).
Four of these studies reported the results of nationwide serosurveys that determined the presence of dengue immunoglobulin G (IgG) to detect past infection (Fig 6 [26] conducted national health surveys.The studies (N = 3,293-4,152) demonstrated a generally decreasing trend in seroprevalence, from 59% in 2004 to 45.7% in 2017 [26,35,46,52].
Meanwhile, in two studies, IgG seroprevalence (35.0-55.9%)was determined upon hospital admission among patients with current suspected or confirmed dengue, representing past dengue infection [37,42].Two other studies determined the seroprevalence of dengue immunoglobulin M (IgM) to detect recent infection.In a 2007 study conducted among residents of dengue outbreak areas, seroprevalence of IgM or IgG was high (65.9%),as was the prevalence of inapparent dengue (78%) among those who were IgM-positive [51].Contrarily, a study among Chinese migrant workers living within a construction site during the 2002 outbreak showed a low IgM seroprevalence (10%) [44].

Subgroup analysis: by serotype
Subgroup analyses of seroprevalence by serotype were available from four published studies of varying cohort composition and study design (S13 Table) [33,35,48,50].The largest of these studies [35] showed a higher seroprevalence of antibodies against DENV-1 and -2 versus the other serotypes in 2009-2010, corresponding with the contemporaneously circulating dengue serotypes.Furthermore, as the age of subjects increased, the studies found a higher seroprevalence of antibodies against each serotype and a higher frequency of a multi-serotype pattern [33,35].

Economic burden
The three published economic studies varied in terms of component cost estimates, sources, and assumptions (Tables 3 and S14).Albeit not directly stated, a societal perspective was adopted in all three studies, based on their consideration of both costs borne by the patient (or family) and costs borne by the government as a payor [12,14,45].
Carrasco et al. ( 2011) [14] estimated the costs of dengue for 2000-2009 as part of a CEA of a potential dengue vaccination program.The study applied four approach permutations using a combination of two sets of methods: (1) human-capital versus friction-cost method and (2) constant versus age-dependent symptomatic rate.Among these approaches, the most conservative cost estimates were produced via the friction-cost method and a constant symptomatic rate.Direct costs (including costs of hospitalization or visits by ambulatory patients, in addition to transport costs) constituted a higher proportion of the total costs than indirect costs (such as average household service losses, cost of providing schooling, cost of absence from work; S14 Table ); such costs of productivity loss (indirect costs) comprised 21-29% of the total costs, depending on the approach.Additionally, the authors reported estimates with and without the costs of Singapore's contemporary vector control program (i.e., habitat reduction, insecticide spraying), which corresponded to SGD 68 million per year, according to their consultations with the NEA [14].
Meanwhile, Shepard et al. (2013) [12] performed a cost-of-illness study among countries in Southeast Asia, including Singapore.Using a human capital approach and the component costs listed by Carrasco et al. (2011) [14], the study produced probabilistic estimates of the direct and indirect costs per case of hospitalized and ambulatory dengue for 2001-2010 (S14 Table ).Compared with the study by Carrasco et al. (2011) [14], higher annual total costs were reported.Furthermore, indirect costs constituted a higher proportion of the total costs than direct costs (63% vs 37%) [12].
Lastly, Soh et al. (2021) [45] undertook a CEA of Wolbachia interventions, whereby the authors estimated dengue-related costs via the four approaches adopted by Carrasco et al. (2011) [14].Using the same, most conservative method, the study found that the average annual total costs for 2010-2020 were twice as high as those reported by Carrasco et al. (2011) [14] for the previous decade (2000-2009) [45].

Discussion
MOH surveillance was the source of most epidemiologic data shown in this review.In terms of epidemiologic burden, Singapore experienced five large epidemics in the past 20 years, interspersed with periods of endemic dengue [9], and reached peak IR during the latest outbreak in 2020 amidst the COVID-19 pandemic [59].Our subgroup analyses revealed that dengue IRs were highest for the serotypes DENV-2 (previously) and DENV-3 (currently) and in the 15-44 age group.For the past 10 years, the risk of DHF has been low (<1%).In terms of economic burden, the included studies demonstrated higher dengue-related costs in the past decade than in the immediately preceding decade (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009), corresponding with an increased dengue burden compared with the preceding decade [45].Indirect costs of dengue contributed greatly to high total costs.Additionally, we found that in recent years, hospitalization occurred in about one-third of cases and more frequently among adults aged �45 years.Nevertheless, mortality from dengue has been consistently low (CFR <1%) for the last 20 years.Lastly, the most recent seroprevalence data illustrated that dengue IgG, although increasing with age, was present in <50% of the general population and has been falling over time.Based on this review, research gaps remain in Singapore in terms of stratified epidemiologic data, such as incidence by age group, which may be added to routine MOH surveillance and reporting to develop targeted interventions for dengue control.As with the study by Ang et al. (2019) [28], national hospital-claims databases could be used to acquire hospitalization rates and CFRs according to disease severity, especially if out-of-hospital deaths are assumed to be absent.Meanwhile, seroprevalence according to serotype could be obtained through incorporation into the design of national serosurveys, as done by Low et al. (2015) [35], or the use of residual sera, as done by Ang et al. (2015) [26,27], although the cost of doing so may be high.
CFRs stratified by infection type (primary, secondary, post-secondary) are also currently missing from the local literature.The risk of severe dengue may be higher in secondary versus primary infection owing to antibody-dependent enhancement, whereby viral replication is increased from priming with pre-existing antibodies to one serotype [60,61].Alternatively, the risk of severe disease in post-secondary infection may be no different, or even lower, than in primary or secondary infection [62,63].To further understand these nuances, more prospective cohort studies such as that by Rouers et al. (2021) [42] are warranted.In addition, evaluating the impact of interventions for dengue control will require updating of some epidemiologic data, including dengue hospitalization rates, length of hospitalization, and national seroprevalence.Relatedly, EFs in the identified economic studies were estimated using data that are now >10 years old.Hence, studies that estimate EFs are essential, as performed by Tan et al. (2019) [46], who estimated an EF of 6 for 2014-2017.
Akin with epidemiologic data, stratification of costs is presently lacking in local literature (e.g., by age group, by epidemic versus non-epidemic year, by societal or governmental perspective).For instance, the mixed cost estimates by Soh et al. (2021) [45] were based on an assortment of costs from public hospitals and private outpatient facilities, as well as subsidized and out-of-pocket costs.Claims databases, surveys of clinics and hospitals, and line-item costing are some possible sources of more stratified data.Proper costing would also involve all agencies, institutions, and personnel in the costs [64].Furthermore, the macroeconomic burden of dengue in Singapore, including its impact on tourism and other public sectors (e.g., education, trade, domestic services), remains to be elucidated as it has been in Brazil, where dengue is similarly endemic and the macroeconomic impact is considerable [65].While the existing cost-of-illness studies have discussed the microeconomic impact of dengue in Singapore in terms of average household services lost, average productivity loss per absent day at work, DALY's, etc. [12,14,45], less adequately described in the literature is the economic impact of dengue on household consumption potential of the affected patient owing to an increase in healthcare expenditure.
The available evidence has some limitations, such as those concerning MOH surveillance data.Underreporting remains a possibility in passive surveillance.Patients with inapparent or mild dengue infection may not seek care amidst their potential for viral transmission [64], and health-seeking behavior may change over time, such as during the COVID-19 pandemic.Merging multiple data sources (surveillance, cohort studies, community-based surveys) could address this issue [64].Nevertheless, some evidence points to improving dengue reporting practices in Singapore.The reduction in dengue seroprevalence despite an increasing incidence suggests an "unmasking" of infections from better reporting [46], rather than a genuine increase in incident infections.The decline in risk of DHF may also be explained by higher ascertainment rates of milder dengue disease, which align with improving case notification in Singapore [46].
Data from published epidemiologic literature were of good quality but similarly limited.We found scarce primary data while most secondary data were aggregated.Without the benefit of individual patient data, the capacity for subgroup analysis or synthesis is restricted.Additionally, only a few prospective studies were identified.
A number of contextual features are distinct to Singapore.The enduring vulnerability of Singapore to dengue epidemics is potentially a result of low and falling herd immunity and persistent viral transmission [66].In turn, low herd immunity is caused by historic falls in the force of infection, which had begun to decline in Singapore in the 1990s [46].Presently, Singapore has the lowest force of infection among endemic countries in Asia [67].Some have attributed this phenomenon to its aging population, as more older individuals represent more immunity against dengue, but it is an unlikely primary cause in light of decreasing seroprevalence in Singapore [46].The more probable reason is Singapore's vector control program, which has effectively lowered transmission risk [9,46,67].Paradoxically, the resulting low herd immunity decreases the effectiveness of vector control programs, because the risk of infection is higher even if the number of vectors is lower [46].Notwithstanding vector control, persistent transmission of all dengue serotypes occurs in Singapore, whereas across-type interactions contribute to the emergence of outbreaks [68].Coupled with climate change, increased travel to and from Singapore, as measured through air passenger flow, can lead to DENV lineage movement, long-run local transmission, and seeding of outbreaks [9,69,70].
Our review process was limited by the absence of a meta-analysis.The variability in reported outcomes among identified studies, limited number of primary sources, and inconsistent groupings across included studies were factors that inhibited a quantitative synthesis.Moreover, for purposes of summarization, we performed some transformation of data (e.g., averaging, standardization, imputation) that potentially introduced information bias.Another important limitation is that most of the literature predates the deployment of novel technologies for vector control, namely the Wolbachia program and the use of mass Gravitrapping for targeted vector control [9], as well as the licensing of the first dengue vaccine [71].

Conclusion
Singapore, where dengue is endemic, has seen a dramatic rise in dengue incidence rates over the past two decades but, simultaneously, a notably low force of infection, which is likely the result of effective and nationally funded comprehensive dengue control strategies inclusive of active inter-epidemic epidemiological surveillance that is largely unseen elsewhere in the region [9,66,72,73].However, as for other countries of similar climate where dengue is endemic, there is room for improvement with respect to government-led surveillance and reporting measures [72,73].In our review of 20 years' worth of data in Singapore, we identified that despite improved data collection from mandatory passive surveillance, there remains a need for sufficiently detailed and up-to-date data on the epidemiologic and economic burden of dengue in Singapore.Owing to the complexities of dengue in terms of disease transmission and pathophysiology, stratified analyses by age, severity, infection type, and serotype are indispensable, and we would strongly recommend the Ministry of Health to consider implementation of surveillance and reporting measures (e.g., national serosurveys to identify seroprevalence according to serotype) to capture and publish such subgroup data for subsequent analysis.A robust reporting system can ultimately provide the data that can be used to accurately determine the cost-effectiveness of interventions for dengue control and focus research and development efforts accordingly.

Fig 3 .
Fig 3. Incidence and severity of dengue in Singapore, 2000 onwards.Panel A, Incidence rate of dengue from 2000 to 2022.Panel B, Relative incidence rate of dengue per age group from 2000 to 2018.Reference group: 35-44 years.Panel C, Hospitalization rate, or proportion of dengue cases hospitalized per year, overall and stratified by age group from 2003 to 2017.Panel D, Dengue case fatality rate from 2000 to 2022 with 95% confidence intervals.Data for panels A and B were retrieved from Singapore Ministry of Health [55, 56]; panel C, from Ang 2019 [28]; panel D, from the Singapore Ministry of Health and National Environment Agency [55-57].p/y, person-years.https://doi.org/10.1371/journal.pntd.0012240.g003

Table 1 .
[22]21]ued) Sample size not applicable to MOH-notified cases, which represent outcome groups, not base populations.Assessed using original and revised Newcastle-Ottawa Scales[20,21]for epidemiological studies and cost-of-illness evaluation checklist from Larg and Moss (2011)[22]for economic studies.

Table 2 . Characteristics and quality of included publications: Economic studies. Study identifier Study design Data collection period Population (sample size) a,b Outcome of interest Quality rating c,d
a Sample size with usable data for this review.b Sample size not applicable to MOH-notified cases, which represent outcome groups, not base populations.c

Table 3 . Characteristics and results of published studies reporting the total costs of illness from dengue in Singapore. Study identifier Period of estimation Method Mean aggregated cost to Singapore per year, million 2021 SGD (95% CI)
[12,14,45]capital method values lost time or premature death using an individual's gross earnings, *calculated from gross domestic product per capita and the total years of premature life lost based on the discounted, weighted life expectancy using WHO life tables for each country.The friction cost method assumes job absenteeism or death lead to productivity losses, which in both studies is assumed to be 30 days for fatalities and to last as long as symptoms in non-fatal cases, and productivity losses are offset according to the elasticity of annual labor time versus labor productivity[12,14,45].Costs originally published in 2010 USD were converted to 2010 SGD and standardized to 2021 values using the ratio of 2021 to 2010 average consumer price indices; costs were rounded to the nearest million.CI, confidence interval;-, not reported. #