• Diseases
  • African Trypanosomiasis
  • Buruli Ulcer
  • Chagas Disease
  • Cholera
  • Dengue
  • Dracunculiasis
  • Emerging infectious diseases
  • Fascioliasis
  • Leishmaniasis
  • Leprosy
  • Lymphatic Filariasis
  • Malaria
  • Onchocerciasis
  • Schistosomiasis
  • Soil transmitted helminthiases
  • Trachoma
  • Tuberculosis
  • Yaws
  • Zoonoses
• Global health topics
  • Agriculture
  • Biotechnology
  • Climate change
  • Capacity building
  • Community health
  • Conflict
  • Diagnostics
  • Economics
  • Environment
  • Ethics
  • Equity
  • Funding
  • Gender
  • Health policy and systems
  • Human rights
  • Immunology
  • Innovation
  • Operations research
  • Pharmaceutical products
  • Poverty
  • Primary health care
  • Research priorities
  • Research translation
  • Social environment
  • Vaccines
  • Vectors
• News
• Reviews
• Editorial opinions
• Research articles
• Reports
• Virtual journals
• Profiles
• Policy & Strategy
• Funding opportunities
• Networking
• Resources
• Toolkits
  • Knowledge Translation (KT) Toolkit
  • Health Research Systems Analysis (HRSA) Toolkit

Communities of practice

• Disease Reference Groups
• Thematic Reference Groups

Dengue: Burden Of Disease And Costs Of Illness

30 Nov 2007

Source: WHO/TDR

Jose A. Suaya ¹ , Donald S. Shepard ¹ , Mark E Beatty ²

¹Schneider Institute for Health Policy, Heller School, Brandeis University, Waltham, MA, USA
²Pediatric Dengue Vaccine Initiative, Seoul, Republic of Korea

Abstract

Summary

This paper provides a framework for considering the global burden of illness attributable to dengue, and its cost, and describes challenges encountered in the estimation of these values. Major challenges include: lack of uniform application of the World Health Organization (WHO) case definition, limited capabilities and standards of dengue laboratories, limited accuracy of rapid tests, misdiagnosis, lack of uniform criteria to report cases of dengue to WHO, limited role of surveillance and reporting systems, under-reporting of fatal and non-fatal dengue, misclassification in reporting, limited public knowledge about major regions at risk and travellers. While the scientific literature contains few studies on the burden of dengue and cost of illness, available results suggest that the actual number of cases of dengue may range from 3 to 27 times the reported number. We propose a conceptual framework for research.

Working paper for the Scientific Working Group on Dengue Research, convened by the Special Programme for Research and Training in Tropical Diseases, Geneva, 1–5 October 2006

Full text source: Scientific Working Group, Report on Dengue, 1–5 October 2006, Geneva, Switzerland, Copyright © World Health Organization on behalf of the Special Programme for Research and Training in Tropical Diseases, 2007, http://www.who.int/tdr/publications/publications/swg_dengue_2.htm

Introduction

Dengue is a rapidly growing public health problem in tropical and subtropical countries where the majority of the world's population resides and is increasing most rapidly. However, most of these nations are economically disadvantaged and are faced with multiple public health problems, and therefore may not have the resources to combat the continued emergence of dengue. [1–3]

With approximately two billion people living in tropical and subtropical regions of the world, and an additional roughly 120 million people each year[4] travelling to these regions, a large share of the world's population is at risk of contracting dengue. Two estimates have suggested that between 50 and 100 million cases of dengue fever (DF) occur annually,[5–7] corresponding to an incidence rate of 2.5–5.0% of the two billion people worldwide at risk. These cases result in hundreds of thousands of hospitalizations, and about 20 000 deaths each year. [2] The spectrum of dengue infection ranges from asymptomatic infection to death. Clinical presentations of the febrile phase include a milder non-localizing fever syndrome, or influenza-like illness, and classic dengue, or break-bone fever. Immediately after the febrile phase, the disease may progress to the more severe but less common forms of disease, which include dengue haemorrhagic fever (DHF; a febrile illness followed by abnormally low platelet counts, egress of plasma into the pleural and abdominal cavities and haemorrhagic symptoms), and dengue shock syndrome (DSS; DHF with evidence of systemic hypoperfusion). Although death occurs rarely in the febrile phase, it is most commonly the result of hypoperfusion after the development of DHF. Between 250 000 and 500 000 people develop severe dengue each year [8].

Mildly symptomatic dengue is usually treated in an ambulatory care setting, while the more severe forms require inpatient management. Although more than 90% of patients who develop severe dengue have serological evidence of a previous dengue infection, it is not possible to predict which patients will progress to these more serious forms, complicating the triage and medical management of patients. Given the scope of the disease and the large numbers of persons with symptomatic infection, dengue infection may have a tremendous impact on the health-care systems of the countries involved and on household and labour economies, especially during epidemics.

Quantifying the epidemiological and economic burden of dengue is key to formulating policy decisions on research priorities, prevention programmes, clinical training for management of the disease, and the introduction of new technology. Reliable diagnosis, testing, and reporting of dengue would allow better understanding of the true burden posed by dengue in a world of competing public health issues. Reports should capture seasonal and cyclical (annual) variations in disease incidence. Such reports can be used for multiple purposes. First, they provide the basis for comparisons of the burden of dengue in different geographical locations and time periods. Second, they help country-level, regional, and global public health authorities to make informed decisions on resource allocations. Decisions can be based on a comparison of the burden of dengue with that of other health problems. Accurate estimates of the magnitude of dengue can serve to justify donor funding, setting priorities for research, and accelerating the development of dengue vaccines. Third, such reports will serve as important baselines for assessing the impact of any intervention (e.g. a larvicide campaign or vaccine) that could alter the burden of dengue, and will also provide a key ingredient in cost-effectiveness analyses of a single or multiple interventions and technologies.

The prospect of introducing vaccines against dengue makes the importance of understanding the burden of dengue especially important. Substantial investments are required in preclinical research, human testing, manufacturing, distribution, etc. These activities will involve many stakeholders, from donors to manufacturers and governments to consumers. This paper seeks to make these decisions better informed.

The epidemiological burden of dengue

The burden of illness caused by dengue refers to the amount of disease imposed by dengue and measured using a set of epidemiological indicators, such as number of clinical cases classified by severity (DF, DHF, and DSS), duration of the illness episode, quality of life during the illness episode, case-fatality rate, and absolute number of deaths during a period of time. All of these epidemiological indicators can be combined into a single health indicator, such as quality-adjusted life years (QALY)[9] or disability-adjusted life years (DALY).[10] Internationally, DALYs are most often used. The burden imposed by dengue and the potential benefits of any intervention, such as vaccination, can then be expressed in terms of DALYs lost or saved and cost per DALY lost or saved.

Current knowledge of the burden of dengue

Dengue is endemic in all WHO Regions except the WHO European Region. [11] shows the global number of cases reported by year and the characteristic cyclical variations attributable to high epidemic and low post-epidemic years. In each year until 1976, fewer than 40 000 cases from fewer than 15 countries were reported annually to WHO. In 1974, WHO developed the first guidelines for the diagnostic and management of dengue. [8] Figure 1 shows that the lowest number of cases was reported in 1979 and the highest number in 2002, with 110 000 and 1 300 000 cases reported, respectively. Figure 2 shows the annual global number of deaths attributable to dengue and the number of countries reporting. There were similar cyclical variations in the number of deaths coinciding with epidemic and non-epidemic years. After 1977, the year with the lowest reported number of dengue deaths was 1978 (with 807 fatalities) and the year with the highest number was 1983 (with 6031 fatalities). The majority of the cases and deaths were reported from south-east Asia and the western Pacific.

Figure 1

Figure 2

A number of studies have attempted to estimate the burden of dengue in terms of DALYs. For example, the burden was estimated for Puerto Rico in 1984–1994.[12] Estimates were based on the numbers of cases and deaths reported to the national surveillance system, and on age-group expansion factors used to control for under-reporting. The authors estimated an average loss of 658 DALYs per million population per year, and concluded that this burden was comparable to that attributed to meningitis, hepatitis, malaria, tuberculosis, the childhood cluster of diseases (polio, measles, pertussis, diphtheria, and tetanus), or intestinal helminthiasis, and of the same order of magnitude as tuberculosis in Latin America and the Caribbean. A study in south-east Asia estimated a loss of 420 DALYs per million population per year, comparable to that of meningitis (390 DALYs per million population per year), twice the burden of hepatitis, and one third of the burden imposed by HIV/AIDS in the region.[13] A study for Thailand estimated that country's burden at a loss of 427 DALYs per million population per year for 2001.[14]

The Disease Control Priorities Project has recently published the global burden of disease for 2001 to 2003.[15] It estimated the global burden of dengue as 528 000 DALYs for the year 2001. This corresponds to a burden of 264 DALYs per million population per year for two billion people living worldwide in areas at risk of dengue.

Major gaps in knowledge of dengue burden

Current global estimations of the burden of dengue are considered to be uncertain because of a number of factors, discussed below. The under-reporting of dengue cases (both fatal and non-fatal) is probably the most important barrier to obtaining an accurate assessment.

Lack of uniform application of the WHO case definition

WHO has published guidelines for the diagnosis, classification, and management of dengue, which have been adapted by WHO regional offices.[16] Investigators have reported difficulties in applying the case definition because of its complexity and the limited ability to explain observed patterns of disease.[8,17–18] Because of these difficulties and the need to use categories relevant to their own needs for planning and management, some countries have instituted their own case definitions.8

Limited capabilities and standards of dengue laboratories

Tests for anti-dengue IgM antibodies are commercially available. Their accuracy may vary with the situation in which they are used. Tests for virus detection by cell culture or nucleic acid detection (e.g. polymerase chain reaction) require the capabilities of sophisticated research laboratories to produce test materials and perform quality testing.[19] Laboratory capabilities, infrastructure, technical expertise and research capacity need to be improved.[20]

Scientists at a recent WHO-sponsored meeting pointed out limitations in laboratory standards, quality control, and dengue serological diagnosis and virus isolation, as well as for reporting and information exchange in south-east Asian and western Pacific countries.[21] Similar considerations may apply to laboratories in the Americas as suggested by an evaluation of quality control of serological diagnostic tests in major national reference laboratories responsible for dengue surveillance and diagnosis in the region[22]. Although the majority of the participating laboratories had a high level of performance in detecting IgG and IgM antibodies to dengue, only 63% of 86 laboratories that received samples for testing between 1996 and 2001 decided to participate in quality control. This study also highlighted the challenges faced by participating laboratories, such as interruptions in the availability of antigens, the low sensitivity of testing for IgM antibody by enzyme-linked immunosorbent assay, and the lack of alternative tests and techniques for the diagnosis of dengue. There is no information about quality control for the isolation or identification of dengue virus. Regions with imported dengue, such as the USA and Europe, may face additional challenges. Since dengue is not common in these areas, laboratory testing for dengue may not be available and the disease may be overlooked as a cause of symptoms among ill travellers. For example, as part of an external quality assurance evaluation, 20 serum samples were sent in 2002 to 18 European participating laboratories to be tested for the presence of dengue virus-specific IgM and IgG antibodies.23 Laboratories reported concurrent and correct results for 71% of the IgG-positive samples and 89% of the IgG-negative samples. However, though 97% of the IgM-negative samples showed concurrent and correct results, only 58% of the IgM-positive samples had concurrent and correct results. These findings highlight the need for quality controls and improvements in testing for dengue in countries with imported dengue; worldwide laboratory capabilities and quality control are not adequate.

Limited accuracy of rapid tests

In a recent meta-analysis of 11 studies of rapid diagnostic assays for dengue, the authors evaluated the performance of an immunochromatographic test (ICT) manufactured by a leading company in its field. The meta-analysis showed that these assay had a wide spectrum of sensitivity (0.45 to 1.0) and specificity (0.57 to 1.0), suggesting that such tests may have limited accuracy[24]. Also, the cost of rapid tests is a barrier to their systematic use in developing countries. Little is known about the performance of other rapid diagnostic tests sold in the market.

Misdiagnosis

Despite the available clinical guidelines, dengue can be misdiagnosed (by under-diagnosis or over-diagnosis). Given the lack of specificity of the symptoms of dengue, clinicians can confuse dengue with other infections, such as influenza, enterococus, chikungunya, viral haemorrhagic fevers, leptospirosis, malaria, or typhoid[25]. Moreover, dengue infection as an undifferentiated febrile illness may represent a large proportion of all the symptomatic cases of dengue[26].

A study reviewing medical records for a period of 6 months in Laredo, Texas, in 1999 showed that only 50% of patients with clinical suspicion of dengue were diagnosed as such[27]. Misdiagnosis is more likely if other febrile diseases with similar clinical characteristics occur concomitantly. For example, in Barbados in 1995 and 1997, the majority of patients with suspected leptospirosis actually had dengue. Conversely, some of the cases of suspected dengue were actually leptospirosis[28]. In another study of an outbreak of dengue in Bangladesh, about 18% of cases of suspected dengue gave negative results in laboratory tests for dengue and proved to have leptospirosis[29]. In a study in a dengue-endemic province in Viet Nam, among 697 patients with acute undifferentiated fever visiting primary care facilities, for whom paired serum samples were collected, acute dengue was diagnosed in 33.6% cases[26].

Misdiagnosis can be influenced by treatment guidelines. For example, although WHO guidelines for the treatment of febrile children aged 2 months to 5 years are useful to ensure that children with fever and no alternative explanation are empirically treated for malaria, this guideline may contribute to misdiagnosis of dengue, particularly in areas of low malaria transmission or where physicians are not routinely using malaria smears to confirm the diagnosis[30,31]. In addition, drawing blood and sending it for testing for dengue may not be viewed by the health-service provider as a worthwhile expenditure of time or money. In the absence of rapid testing, the result of the test will not be available for days or weeks, and the provider may be left to diagnose the patient with a viral syndrome or fever of no known source and treat the patient empirically. Rapid testing may improve the situation, but the cost of the test may be a disincentive. Finally, even in those countries with adequate laboratory facilities, cross-reactivity between anti-dengue antibodies and antibodies to other flaviviruses (West Nile virus, St Louis Encephalitis) and the dynamics of the immune response to flavivirus (boosting of antibodies to the primary infecting flavivirus during a second flavivirus infection—‘original antigenic sin’) further contributes to the problem of confirming diagnosis of dengue in areas of the world where multiple flaviviruses exist.

Lack of uniform criteria to report cases of dengue to WHO

In the WHO Region of the Americas, cases of dengue are reported to WHO stratified by severity: DF and DHF. However, in the WHO South-East Asia Region and the Western Pacific Region, cases are reported without distinction of severity, though most reported cases involve hospitalization for DHF [11]. Additionally, while some countries only report cases of severe dengue, others report all cases, and still others report only confirmed suspected cases [32]. This lack of uniform reporting makes it difficult to perform meaningful international comparisons and aggregations.

Limited role of surveillance and reporting systems

Surveillance systems depend mainly on the capacity of the hospital to record, monitor, and report dengue statistics. Less information is obtained from clinics, and still more limited data are reported by private-sector medical practices. In south Asia, for example, notification is barely enforced and the number of cases of communicable diseases (such as dengue) dealt with in the private sector is usually unknown [33]. In the Americas, surveillance systems are also generally passive and considered to be ineffective in defining the full scope of transmission in a given community [34].

Underreporting of non-fatal dengue

There are a number of factors that contribute to the under-reporting of dengue. Firstly, notification of suspected dengue to public health authorities (communicable diseases units) is legally required in most of the affected countries, but rarely enforced. Since the results will often not be available for days or weeks after the visit that led to the sample collection, the results may be viewed as having little clinical value to the treating physician. Complicated reporting or lengthy requirements may be additional factors that reduce the motivation of health-care providers to routinely report cases, since reporting may not be of intrinsic value to the patient or busy health-care providers. Surveillance systems usually have logistic and budgetary constraints. Consequently, the reporting of dengue is likely to be fragmented, incomplete, inconsistent, and unreliable. Thus, under-reporting of dengue cases, and probably even of deaths attributable to dengue, is a major concern to be addressed.

Evidence of under-reporting is evident in studies conducted in Brazil, Indonesia, Puerto Rico, and Thailand. In Belo Horizonte, south-eastern Brazil, the level of reporting of hospitalized suspected cases of dengue was estimated to be 63% between 1997 and 2002 [35]. As the cases recorded in the reporting system were the more severe, the overall case-fatality rate may have been consequently overestimated. In Indonesia, the number of reported cases was compared with medical records of hospitalized DHF cases admitted in four major hospitals in Bandung during 1994. Only 31% of these cases were captured in the report [36]. Similar under-reporting was found in Puerto Rico, where only 28.4% of hospitalized cases of DHF were detected by any of the surveillance and reporting systems [37]. In another study, the same author tried to measure the burden of dengue in Puerto Rico during 1984–1994 [12]. To deal with the existing under-reporting, it was estimated that for every case of dengue reported among children, there were about 10 additional cases not reported. Among adults, it was estimated that for every case reported, 27 cases went unreported. In a recent study in Thailand, under-reporting was recognized and the true number was estimated as 10-fold the number reported [14].

Misclassification in reporting of dengue

Cases of dengue can be misclassified at the time of diagnosis because of the lack of familiarity of some medical providers with dengue as a disease, or difficulties with using the WHO classification system.8 Correct classification is important clinically, because death is associated with the more severe form of the disease. Correct classification is important epidemiologically because WHO has suggested that the dengue case-fatality rate can be computed by dividing the number of deaths by the number of cases of DHF [16]. If classification is not uniform, comparisons of case-fatality rate between countries can be misleading. The severity of dengue is also a predictor of the use of health-care services and of the costs of medical care [38]. In a study in Puerto Rico, only 17 DHF and 3 DSS cases were identified among 986 hospitalized cases of dengue reported via the surveillance system in 1990–1991. A review of the hospital records of those patients, however, found that 88 and 14 of them had a clinical diagnosis of DHF and DSS, respectively [39]. Reviews of medical records identified about five times more cases of severe dengue than were reported to the routine surveillance system. If appropriate allocation of resources to address the dengue problem is to occur, better recognition of the severity of dengue and improved reporting is needed. Another review of the medical records of patients with dengue during the 2002 epidemic in Taiwan, China, found that 71% of DHF patients were discharged without such a diagnosis [40]. A consequence of this misclassification is the under-reporting of severe dengue.

Underreporting of fatal dengue

Reports of deaths caused by dengue are intuitively assumed to be more accurate than reports for non-fatal cases. During the 1998 epidemic of dengue in Puerto Rico, there were 17 000 reported cases of dengue and 19 deaths for which dengue was confirmed or probable. For the same year, however, only five deaths attributable to dengue are shown on WHO DengueNet, the WHO-sponsored internet-based system for the global surveillance of DF and DHF [41]. This single time-point finding indicates a four-fold under-reporting of laboratory-positive dengue deaths. In addition, there were another 37 deaths for which dengue was initially suspected but could not be confirmed because the virus was identified by virus isolation or immunohistochemcial staining of tissue, the patient died before seroconverting, and no other explanation for the death was identified. However, dengue was ruled out in six of these cases [42]. An analysis of paired samples gathered during routine surveillance in Puerto Rico demonstrated that roughly half of the cases that were initially indeterminate based on testing of the acute-phase sample could be reclassified as confirmed owing to seroconversion identified by testing a convalescent sample (Garcia, unpublished data), suggesting that an additional 15 deaths suspected to be from dengue may actually may have been caused by dengue. If correct, this raises the under-reporting factor to seven in Puerto Rico. Difficulties in reporting and classification that occurred in Puerto Rico are likely to occur in other countries where dengue is endemic. Additional factors that may further interfere with confirmation and reporting of dengue deaths to WHO include political and economic disincentives raised by concerns regarding the possible impact on tourism.

Limited public knowledge from major regions at risk

Data on the transmission of dengue is limited for dengue-endemic regions that have a significant portion of the world's population—India, China, and sub-Saharan Africa.

India - One billion people (15% of the world's population) reside in India. India's population is twice that of south-east Asia, the region that currently reports the most dengue-related deaths. Despite comparable environmental risk conditions, the number of reported cases and deaths in India is only a fraction of that reported in south-east Asia,

In many regions of India, an increasing number of suspected cases of dengue are seropositive for IgM and IgG antibodies [43]. The existence of IgG antibodies in a patient demonstrates prior infection with dengue and an increased risk of the severe forms of the disease. Outbreaks of dengue are increasingly reported in rural areas, implying that the population at risk is increasing, since dengue is considered to be a predominantly urban disease [44–47].

Surveillance for dengue has been very limited in India and reporting to the central government has not been mandatory [48]. A recent study concerning the epidemic of dengue in Chennai in 2001 has suggested that the surveillance system was unlikely to generate proper information on the epidemiology of the disease [49]. In 2004, a WHO initiative called for promoting improvement of dengue surveillance as part of the Integrated Disease Surveillance Programme in India, strengthening laboratory networking and quality assurance, and reviewing case definitions [50]. Although improvements are being made, the current gaps in epidemiological data and surveillance mean that the burden of dengue in India is uncertain. However, dengue is recognized as one of the leading causes of death and hospitalization among children in India [51].

China - One billion three hundred million people, 20% of the world's population, live in China Roughly one fifth of China's land mass, including some of the more densely populated regions, lies in tropical climes where dengue transmission could occur all year round. Published reports on outbreaks of dengue detailed the re-emergence of dengue in the 1980s and 1990s [52–54]. However, since 2003, public data from WHO does not include cases in China [41], making the documentation of the current burden of dengue in this country very difficult.

Sub-Saharan Africa - The burden of dengue in Africa remains poorly understood. Travellers and military personnel visiting or stationed in Africa have been identified as having laboratory-confirmed dengue infections, indicating that the virus is circulating [55–56]. Several studies of seroprevalence and fever in sub-Saharan Africa have identified evidence for the presence of the dengue virus in many sub-Saharan countries, including Cameroon, Djibouti, Kenya, Senegal, the Sudan and Burkino Faso. These studies reported lower seroprevalence rates than those seen in other tropical countries, such Haiti, Brazil, or Thailand, ranging from 10% to 34% among persons tested [57–59]. As expected, higher levels of prevalence are noted among urban residents than rural residents [60–65]. In addition, periodic outbreaks of DF have been reported in the region [63–66]. Although genetic factors could provide some protection, without systematic surveillance and serosurveys with appropriate sample schemes to give a fair representation of the disease burden in the population, the past and current burden of dengue in Africa may remain poorly understood. Moreover, if dengue is an endemic problem in sub-Saharan Africa, more urbanization will only increase the burden of dengue.

Limited knowledge about dengue in travellers

According to the World Tourism Organization, in 2004, 125.4 million international tourists visited countries where they might be at risk for acquiring dengue infection [4]. Depending on the population studied and the laboratory methods used, serological evidence of recent dengue infection was found in between 7% and 45% of cases of febrile travellers returning from areas where dengue is endemic [67–69], confirming that dengue is an important cause of fever among returning travellers. The increasing number of cases of dengue creates a significant economic burden owing to working days lost. However, given the spectrum of clinical illness, not all patients may seek medical attention or receive diagnostic testing. As a result, under-reporting of dengue infection occurs even in developed countries. Moreover, of those patients who are diagnosed with dengue not all may be reported to public health authorities. For example, between 1 January 2001 and 31 December 2004 seven residents of the USA were diagnosed with dengue after returning from Thailand [56]. According to the World Tourism Organization, 2 012 077 USA tourists visited Thailand during the same period [70], giving a rate of 3.5 dengue infections per 1 million visitors to Thailand. However, among a prospective cohort of Dutch travellers, 0.7% of travellers returning from south-east Asia experienced symptomatic, laboratory-confirmed (anti-dengue IgM seroconversion) dengue infections [71]. If the risk of infection is similar for travellers from the Netherlands and the USA visiting south-east Asia, for each case reported to the United States Centers for Disease Control and Prevention there may be 5000 additional unreported clinical dengue infections. This is a conservative estimate since the USA, unlike the Netherlands, shares a border with a country where dengue is endemic—Mexico. As a result, USA residents have a higher potential exposure to dengue.

Personnel deployed in dengue-endemic areas during humanitarian emergencies and conflicts are at a higher risk of dengue infection than are regular travellers, since they usually live in areas without vector-control activities or air conditioning, and usually stay in those areas longer than do tourists. For example, during a 5-month deployment as part of the United Nations Mission in Haiti, 32% of 249 personnel with febrile illness had dengue [72].

The economic burden of dengue, or costS of illness

Terminology

Cost-of-illness calculations generally distinguish ‘direct’ and ‘indirect’ costs.

Direct costs are those within the health-care system. They comprise the cost of diagnosis, treatment and prevention of dengue. There are three major direct cost categories—medical care, surveillance and reporting, and prevention. The cost of medical care includes the cost for ambulatory and inpatient care. Surveillance and reporting costs take into account efforts by governments and international organizations to monitor and disseminate information about cases, outbreaks, and deaths. Prevention costs include activities to prevent dengue, such as vector control (e.g. inspections, management of disposables, use of larviciding and fumigation, education, media campaigns, and community mobilization).

Indirect costs are the economic value lost by households and society in general owing to illness and premature mortality of dengue patients and productivity losses of household members and friends affected.

The estimation of direct and indirect costs is complex because it must take into account different ‘payers’ or economic sectors (public sector, household, third party, employers, society), different levels of government (district, regional, national), different national government agencies (Ministries of Health, Education, Environment; the Armed Forces), and different international organizations (e.g. WHO, United Nations Development Programme).

The system of national health accounts (NHA) provides a framework for examining costs within the health sector (i.e. direct costs) [73]. This framework helps countries to assess the totality of financial resources available to the health sector (from the public, private, and donor sectors), to identify the financing agencies through which these funds flow, and to analyse how these funds are used (by type of provider, function, geographic region or population group). NHA also provides analysts and policy-makers with a tool that not only assists in the analysis of current use of resources, but also helps in the planning of future resource needs and tracking to determine whether resources are reaching the target population. In countries where dengue is endemic, NHA can help analyse current expenditure (public, private, and by donor) on treating dengue. In turn, this information can be used to analyse the cost-effectiveness of any new vaccine and understand who will derive the most benefit. Understanding patterns of health-care use and expenditure may contribute to the development of policies that will improve the allocation of resources to the poorer segments of society, who might not be able to pay for a vaccine or other dengue-related interventions. WHO published the Guide to producing national health accounts in 2003 [73].

The most important government activities related to dengue include vector control, educational activities, mass media programmes, and ambulatory and inpatient care. Knowledge about spending on these activities by district, regional or national governments is fragmented. Government-sponsored health-care activities include care at clinics and hospitals. In hospitals, patients can receive ambulatory care (outpatient department and emergency room) or inpatient care (general, intermediate, or intensive care). Information about the use of hospital services can be obtained by following a cohort of people for a given period of time (community-based study), or obtained from a hospital itself (facility-based study). Hospital costs include tests, drugs, supplies, health-care personnel and medical facilities. To estimate the hospital costs of dengue patients, two approaches can be adopted: micro- or macro-costing. Micro-costing consists of a detailed inventory of the different services available and used in the hospital, the quantity used and the unit cost for each of the services. Macro-costing estimates the average unit cost for each output (e.g. hospital day of care or emergency room visit) rather than cost for each of its components (each laboratory test, drug administered, or procedure carried out by medical personnel). Macro-costing is simpler than micro-costing, as it requires access only to the hospital annual budget or spending and its breakdown by departments, and the total number of output units (such as hospitalizations, average length of stay, outpatient visits, emergency room visits, etc).

Productivity losses and school absenteeism as a result of dengue infections have not been accurately evaluated in most countries. Similarly, care-seeking behaviour, household out-of-pocket spending on treatment for dengue, caregiver's time, and family and psychological disruption have not been systematically or consistently measured.

Current knowledge and gaps concerning economic burden

Data on costs (in US dollars) of treatment are limited to the impact of outbreaks in a few countries. A few examples follow.

Costs of dengue in Thailand

In Thailand, a cost study done on DHF in 1994 estimated the weighted average direct patient cost (including travel, food and lodging and opportunity) at US$ 63.60, plus US$ 45.56 borne by the government for routine service costs in hospitals, totalling US$ 109.16. The per-capita cost of vector control in Thailand in that year was US$ 0.081 [74] A more recent study calculated a similar cost per case of US$ 61 [14].

Costs of dengue in South-East Asia

In another study in south-east Asia that assessed the potential cost-effectiveness of a paediatric vaccine for dengue, it was estimated that the societal cost per case of treating dengue (including ambulatory visits, hospitalization, medications, travel expenses and parents' time seeking treatment) was US$ 139 for DHF, and US$ 4.29 for DF, with a baseline cost of treatment equal to US$ 99 per 1000 population per year. For comparison, the population-weighted average gross national income per capita was US$ 1083 for south-east Asia in 2001 [13].

The same study reported the cost of dengue vector control per capita in other Asian countries: US$ 0.015 in Indonesia in 1998, US$ 0.081 and US$ 0.188 in Thailand in 1994 and 1998, respectively, US$ 0.240 in Malaysia in 2002 and US$ 2.40 in Singapore in 2000. On the other hand, per-capita spending on vector control in 14 Latin American countries in 1997 ranged from US$ 0.020 to US$ 3.560, compared with US$ 0.140 to US$ 8.490 in 17 Caribbean islands in 1990.

Costs of dengue in Puerto Rico

According to a study on the impact of an outbreak of DF in rural Puerto Rico, the loss of income attributable to the disease,(either from illness or from loss of time in caring for ill family members) was estimated to be equal to US$ 305 per household or US$ 125 per person [75].

Worldwide summary by WHO

The UNICEF-UNDP-World Bank-WHO Special Programme for Research and Training in Tropical Diseases (TDR) has summarized the costs of epidemic outbreaks of DF/DHF in several countries [11]. In Cuba, for example, the cost per treated case is US$ 299 compared with US$ 44 in Nicaragua, and US$ 44 in Puerto Rico.

Facility-based studies in eight countries

With support from the Pediatric Dengue Vaccine Initiative, a team from Brandeis University is coordinating researchers in eight countries to implement facility-based studies to measure the socioeconomic impact of dengue on households and the local or national health system. These multi-country studies use a common protocol for data collection and analysis. Three of the countries are in south-east Asia (Cambodia, Malaysia, and Thailand), while the other five are in Central and South America (Brazil, El Salvador, Guatemala, Panama, and Venezuela). Each study identified treated cases of dengue via one or more health institutions in the country (hospital, clinic, national laboratory, or public insurance system). For hospitalized patients, the researchers abstracted data from the patient's medical record. One or two rounds of interviews were conducted with the patient or guardian (if the patient was a child). About 60% of patients were interviewed twice (generally once during treatment and again after recovery), and the remainder interviewed once. About 2000 patients from 62 health facilities (both public and private) were recruited. Of the eight studies, five cover adults and seven cover children.

While analysis is still underway, data are feeding into efforts to work with ministries of health and other critical agencies in the collaborating countries to judge the potential benefit of dengue vaccines. For example, the investigators in Malaysia found that a hospitalized patient received an average of 10 visits from household members. Data from Thailand showed that family members invested the equivalent of 23 working days caring for one hospitalized patient. The Cambodian team documented substantial family disruption when a mother had to spend her day beside her sick child in the hospital while the grandmother provided food and other relatives cared for the other children at home. The breadth of involvement by household members illustrates that the impact of each case goes well beyond medical spending.

The data will allow careful comparisons to be made between the cost-of-illness for dengue and other diseases, such as rotavirus and pneumonia, which commonly cause hospitalization of children. The data suggest that the cost to the family is a severe burden and is as high as or greater than that of other diseases.

Models of disease burden and costs

Disease modelling is often used to estimate the burden of disease and costs at the country or regional (multi-country) level. A model for the economics of a disease comprises a group of mathematical relationships among disease states, such relationships being able to provide estimates of infections, clinical cases of varying levels of severity, treatments received, use of health resources, and costs. While an original study may seek to describe the burden of disease in one population at one time (e.g. a specific district), a model can potentially cover a large geographical area and span of time periods. Disease models typically incorporate data from a multitude of sources. By pooling studies, best estimates can be derived and refined to achieve internal consistency. For example, staff at WHO have developed a mathematical model of disease, ‘DISMOD’, to check the internal consistency of epidemiological estimates of incidence, prevalence, duration and case fatality [76]. Disease models not only describe the current or historical situation, but can also be used to project future situations under current policies, as well as the impact of new policies and technologies for prevention or treatment.

In 1993, Shepard & Halstead published a disease model to examine the benefits and cost-offsets of improved case management of dengue (medical treatment, environmental control, such as reduction of mosquito breeding sites, and the potential development of a vaccine) [77]. This study demonstrated that a model can be used not only to examine individual proposed policies, but also on any combination of proposed policies. A study on the cost-effectiveness of a potential paediatric tetravalent dengue vaccine [13] modelled the burden of disease attributable to dengue in south-east Asia in 2000, and projected how that burden would be reduced by the proposed vaccine. Figure 3 shows an updated version of the study's state-transition model of dengue infection and illness. The study's cost projections were based on estimated vaccine costs and projected savings in treatment costs. The net cost per 1000 population was estimated to be only US$ 17 (89% less than the gross cost). Also, the cost per DALY saved by a paediatric vaccine would be US$ 50, making the potential vaccine highly cost-effective.

Figure 3

Estimates of the costs of vector control

Several researchers have developed estimates of the costs of vector control programmes for dengue. In 1993, a team sponsored by WHO developed guidelines for assessing the cost-effectiveness of vector control, which included procedures for ascertaining the costs of such programmes [78]. They presented case studies on vector-control programmes for malaria and schistosomiasis, but not dengue. In the same year, another study reported the costs of Singapore's intensive vector-control programme [77].

Two of the authors of this paper recently developed a model for estimating the cost of vector-control programmes at a country level and applied the model to Malaysia [79]. The procedure identifies the two major components of a vector-control programme: inspections and fumigation. Aggregate costs were estimated by determining the volume of each activity per year (inspecting and fumigating premises and neighbourhoods near a location where dengue had been found) and the unit cost of each activity, and deriving the total cost. The study estimated the national cost of vector control in 2002 at US$ 5.8 million or US$ 0.24 per capita. Of this total, 74% of costs were attributed to inspection and 24% to fumigation.

Estimates of population-based costs

Building on selected facility-based studies and studies of the costs of vector control, the costs of hospitalized cases and vector-control activities were estimated for Malaysia using a NHA perspective [80]. The study found that these costs (which exclude ambulatory cases) to the health system in Malaysia was US$ 12.8 million or US$ 0.53 per capita, of which 54% was for treatment of illness and 46% was for vector control.

Standardization of protocols

Standardized protocols for the collection of epidemiological and cost data, for analysis and interpretation can make study results both complete and comparable across countries, as recently noted for another disease [81].

Research priorities

Conceptual framework

The challenge in estimating the epidemiological and economic burden of dengue can be encapsulated in an imaginary dialogue between the user and the producer of this information. The user of information, nicknamed ‘InfoNeed’, is the director of a programme or the developer of a policy around dengue. InfoNeed needs information for his policy-making and managerial responsibilities, such as planning or revising a control programme, developing a treatment plan, or considering the development or purchase of some new technology, such as a better diagnostic test or a dengue vaccine. The producer of information, nicknamed ‘InfoGive’, is an analyst with access to the scientific literature, public databases, and possibly additional data. InfoGive knows that research studies may have high precision but limited generalizability, while databases such as reported numbers of cases can be subject to under-reporting, misclassification, and other limitations discussed above. Data may be virtually absent for some regions of the world or times.

As the imaginary dialogue begins, InfoGive may plead that the data do not exist to answer the policy-maker's questions definitively. Continuing the dialogue, the policy-maker acknowledges the problem, but responds that the questions cannot wait until the ideal data become available. Policy-makers need guidance now. Understanding that need, InfoGive seeks to generate the most accurate answer possible based on existing data. For purposes of generating research priorities, this paper also seeks to identify the types of studies which could be done within a few years and with limited resources that would contribute most to strengthening the world's understanding of the burden of dengue.

To address the policy-maker's needs while acknowledging the limitations of existing data, it is helpful to generate a conceptual framework for the burden of dengue. Table 1 describes the three domains of epidemiological and economic burden. For each domain, it is important to describe quantities (numbers of surveillance and prevention activities, and numbers of cases treated) and aggregate costs. The last column, illnesses, comprises both epidemiological and economic burdens.

The cases of illness are the most complicated domain. In view of the earlier discussion in this paper about under-reporting and misdiagnosis, studying dengue illness reported via a surveillance system or diagnosed in a specific health facility is analogous to viewing an iceberg. At first sight, an analyst sees only the part above the water, yet 90% of the iceberg is hidden below the water. In the case of dengue, at first sight, the analyst may study only reported cases. The full burden of dengue includes a spectrum of types of services and reporting:

• Confirmed dengue, seen and correctly labelled by a health professional and confirmed by laboratory diagnosis.

• Suspected dengue, seen by a health provider and classified as suspected dengue, without laboratory confirmation. Note that the absence of laboratory confirmation may be due to many factors: the health provider did not order the test, the cost of testing is high, the time window was not appropriate, or, occasionally, the patient passed away before the biological marker would be applicable. Depending on the location of treatment and the completeness of the reporting system, the case may or may not be reported in the country's epidemiology system.

• Fever treated by a health provider, but attributed to an illness other than dengue. The patient did not receive a laboratory test for dengue because providers did not consider a diagnosis of dengue to be likely, the patient was diagnosed with a disease that is treatable and should not be ‘missed’ (e.g. malaria) as well as the factors above. Nevertheless, in aggregate these cases are numerous and some are likely to be dengue.

• Fever or other symptoms experienced by the patient and managed by self-treatment. The fear of dengue may cause these undifferentiated fevers to be treated more intensively (and perhaps hospitalized) than they would have otherwise if the risk of dengue infection were not present.

In addition, illness caused by the dengue virus falls within a spectrum of severity, ranging from asymptomatic infection to death. Figure 3, adapted from the authors' earlier model [13], shows this spectrum of the disease. Conceptually, the epidemiogical burden could be computed by estimating the number of patients who reach each stage in the diagram by assigning probabilities to each stage, their duration with dengue, and the associated loss in DALYs per case. Similarly, the economic burden can be obtained by multiplying the number of people at each stage by the cost to the health-care system per person at each stage and the other costs of illness per patient at each stage. The greatest practical challenge is obtaining the number of cases in each branch and their associated unit burdens.

Implications for the evaluation of vaccines

The literature as discussed above indicates how important it is that all stakeholders involved in the development of a dengue vaccine understand the burden of dengue, as well as the reasons why current data substantially underestimate the burden. To some extent, existing data can generate expansion factors that correct for under-diagnosis, misdiagnosis, and other limitations of existing information. More importantly, additional data are needed to fill the gaps. Studies are starting to focus on health facilities in which severe dengue is treated and concentrated, while population-based studies indicate the full spectrum of the disease. As dengue varies by locality, such studies are being conducted in many continents in which the disease is endemic. By linking burden to the cost of treatment and the loss of time and productivity, the economic burden of dengue is assessed together with the human burden. Donors such as the European Commission, the Pediatric Dengue Vaccine Initiative, and WHO are supporting work on these important gaps and contributing to the information base for vaccines and other approaches to controlling the disease.

Specific recommendations for research

Improve knowledge of the burden of dengue

• Standardize the reporting of cases among reporting counties to allow comparability.

• Improve documentation of reporting procedures and sites used by reporting countries to provide clarity in the application of the case definition.

• Develop and offer a system for quality assurance in laboratories in reporting countries, to maintain consistency and comparability in case reporting.

• Temporarily, and if possibly periodically, employ alternative but complimentary methods of case ascertainment in reporting countries to evaluate the effectiveness and representativeness of existing surveillance and reporting systems (e.g. population-based serosurveys and facility-based studies oriented towards capture-recapture methods).

• Promote comparative studies with other diseases to provide points of reference for competing public-health priorities.

Increase knowledge of costs

• Develop and apply standard methods for economic studies.

• Add economic components to epidemiological and clinical studies and to outbreak investigations.

• Measure the costs of vector-control programmes.

• Measure the costs of community efforts for vector control.

• Determine the cost-effectiveness of strategies for the prevention and management of dengue.

• Estimate the reduction in dengue burden, costs and cost-effectiveness of new diagnostic, preventive, and therapeutic technologies.

Generate data for under-studied regions and populations

• Initiate studies of epidemiological and economic burden in India, China, and Africa.

• Initiate studies of epidemiological and economic burden for international travellers.

Comments