Acknowledgements

First and foremost, I want to extend my most sincere graduate to my dissertation advisor Maureen Lichtveld for the opportunity to pursue my PhD at Tulane and her guidance through the past five years. Her mentorship was paramount in providing me with a well-rounded understanding of public health and the tools to develop strong research methodology skills. Dr. Lichtveld encouraged me to work independently and often taught me the best lessons by placing me in a situation in which I had to make, and justify, my own decisions. Aside for the professional relationship, I would like to thank Dr. Lichtveld for her friendship.

She always listened to my ‘coming-of-age’ problems and gave emotional support in times of challenge. I will always cherish the hot Surinamese evenings drinking fresh juice and eating spicy food while listening to calypso music with you.

Dr. Oberhelman, you were also instrumental in my coming to Tulane. Dr.

Oberhelman’s efforts in funding the Transdisciplinary doctoral scholars program created the unique opportunity of combining tropical medicine, environmental health, and community health in my education and research. His door was always open for me and he took the time to help me deal with research setbacks and to guide me through the dissertation process. Thank you for believing in me since the very first day.

Dr. Shankar, thank you for being so patient with me. Dr. Shankar pushed me further into biostatistics than I thought I had the capabilities of managing and in the process she taught me that you can always do more. Every time I met with her I was greeted with a smile and a supportive attitude that I hope I can emulate

2 towards others. Within my own department, I had the fortune to work with Dr.

Svendsen. Dr. Svendsen joined the faculty my second year of PhD and since he started he has always supported me and my peers in all aspects of our academic lives. Thank you for your guidance, friendliness and even opening up your house for our very culturally rich dinner. When I first signed up for the medical entomology course a student told me Dr. Wesson was the entomology guru.

After taking her class, I am convinced she is! Dr. Wesson is the type of person that makes you wonder if you could ever teach her anything because she is always one step ahead of you yet she shares her knowledge in an unassuming and reassuring manner. Thank you for being so warm and kind to me and for always being excited about my research even when I failed to see the point.

Every now and then you meet someone we wish would be a bigger part of our life. For me, that someone is Dr. Jessurun. He is everything a doctor should be and without his intellectual curiosity this research topic would never have come into fruition. The short time I spent with you was didactic and fun and I wish I had more opportunities to learn from you.

In Suriname, I was privileged to work with two great women: Dr. Helene Hiwat and Astracia Warner. Their efforts and eagerness to advance science in

Suriname made my research possible. Thank you for opening the doors of the

Bureau of Public Health to me and welcoming me to Suriname. In the field, I counted with incredible backing of the environmental inspectors who not only helped me with my research but took care of me while showing me the warmth of the Surinamese people. Suriname would have been much harder to navigate

3 hadn’t it been for Dr. Hawkins’ advice and friendship. Thanksgiving and the

Super Bowl would have been too boring without him and Kate!

In New Orleans, the Department of Global Environmental Health Sciences made my 5 years at Tulane fly by. Dr. Wickliffe’s candor and research insight were always a breath of fresh air when I was stuck with a problem. Dr. McCaskill’s unique perspective always added dimension to every argument creating much more research awareness in me. Dr. Wilson, I’ve enjoyed watching you grow professionally and become a parent, thank you for teaching us how to power through! My past colleagues and friends, Ben and Amy, I am so lucky to have your friendship. Euridice, I miss our discussion about everything very much! My office mates, and partners in crime, Lekan and Kyle, what can I say… it’s been a wonderful journey and I am glad I got to share it with you. And Dev, thanks Bro.

This dissertation is dedicated to my parents. They have always been and will always be my lifeline. My Papa and Mama have made me into the person I am today and given me all their support and love I need to achieve my best. Thank you for supporting me through the good, the bad and the ugly. Peter, my brother, you add color to my life. Thank you for sharing the good, getting me through the bad and covering up the ugly :P

To my grandparents, I miss you very much.

And to my husband, without whom my life would not be complete, I am grateful for every day I get to spend with you. Can’t wait to grow old together!

Hartelijk dank,

Diana.

4

Table of Contents Abbreviation List 6

1. Abstract 7

2. Background and Significance 8

Literature Review 11

Hypotheses and Research Questions 27

3. Materials and Methods 29

4. Spatial distribution of epidemiological cases of dengue fever in Suriname,

2001-2011 45

5. Characterization of Aedes aegypti breeding sites in Paramaribo, Suriname:

a comparison between a high and a low rate cluster of cases of dengue66

6. A retrospective analysis of dengue cases in Suriname: implications for

treatment and prevention in a LMIC 90

7. Main Results Summary 102

8. Discussion 123

9. Conclusion and Recommendations 136

10. Appendix 141

11. References 151

5

Abbreviation list

ABS Algemeen Bureau voor de Statisiek (General Bureau for Statistics) AZP Academische Ziekenhuis Paramaribo (Academic Hospital Paramaribo) AZPL Academische Ziekenhuis Paramaribo Laboratorium (Academic Hospital Paramaribo Laboratory) BOG Bureau Openbare Gezondheidzorg (Bureau of Public Health) CAREC Caribbean Epidemiology Center CARICOM Caribbean Community CI Confidence Interval CPDB Cash Programma Dengue Bestrijding (Urgent Dengue Control Program) DENV Dengue Virus DF Dengue Fever DHF Dengue Hemorrhagic Fever DSS Dengue Shock Syndrome ENSO El Niṅo Southern Oscillation GIS Geographic Information Systems GPS Global Positioning System HRC High rate cluster Ig Immunoglobulin IRB Internal Review Board LDS Long dry season LRC Low rate cluster LRS Long rain season MDS Meteorologische Dienst Suriname (Meteorological Service Suriname) NS Non-structural protein OR Odds Ratio PAHO Pan American Health Organization RGD Regionale Gezondheidsdienst (Regional Health Service) RNA Ribonucleic acid RT-PCR Reverse Transcription Polymerase Chain Reaction SDS Short dry season SRS Short rain season SVZ Sint Vincentius Ziekenhuis (Saint Vincent Hospital) WHO World Health Organization

6

1. Abstract

The dengue virus (DENV) is a flavivirus capable of causing severe disease in infected humans. DENV is most commonly transmitted by the Aedes aegypti

(genus Aedes, subgenus Stegomyia) mosquito, which is uniquely domestic and thrives in urban settings. Upon infection with DENV, humans can develop mild disease, known as dengue fever (DF), or progress into the severe forms of the disease, known as dengue hemorrhagic fever (DHF) or dengue shock syndrome

(DSS). High levels of urbanization and tropical climate conditions have facilitated the rapid emergence of dengue in Suriname, where dengue has become hyperendemic. The last confirmed dengue outbreak in 2012 exhausted national healthcare capabilities and prompted the Ministry of Health to open up an emergency hospital and implement citywide vector control measures in

Paramaribo. These practices however were in general not preventive in nature.

This research focuses on assessing the relative contribution and the relationship of environmental and demographic risk factors that potentially contribute to the development of DHF in Suriname in at risk populations. Epidemiological data were utilized to map historically high transmission areas of DF and DHF.

Collection and integration of environmental entomological data allowed for determining how influential such factors are to transmission. The results obtained were used to inform best surveillance practices and shift the focus of dengue transmission from an ad hoc intervention to a more prevention-oriented approach.

7

2. Background and Significance

Suriname is located in the Northeastern region of the South American continent.

It shares a border with French Guyana to the East, Guyana to the West, Brazil to the South and the Atlantic Ocean to the North. In 1995, the country became a member of the Caribbean Community (CARICOM) and is classified as an upper- middle income country by the World Bank. Suriname has a tropical climate marked by two wet seasons: a long rainy season (LRS) from late April to August and a short rain season (SRS) from December to February. Approximately 80% of Suriname is covered by tropical rain forest, which is sparsely inhabited. Most of the population lives in the northern lower-coastal area concentrating in the country’s capital of Paramaribo. Currently, more than 70% of the total population of 529,000 lives in an urban area compared to 1975 when half of the population lived rurally. Following this continuous upward urbanization trend, it is estimated that by 2050 more than 80% of the population will be concentrated in an urban area (The World Bank, 2012; UNDESA, 2012). The increasing level of urbanization and the warm temperatures characteristic of the tropics has facilitated the fast emergence and persistence of some tropical diseases in

Suriname, in particular dengue fever (DF).

Over the last 50 years, the incidence of dengue fever worldwide has increased by 50-fold placing 2.5 billion people who live in dengue-endemic regions at risk of contracting the disease. An estimated 50 – 100 million cases of DF occur worldwide every year (WHO, 2015). Historically, the burden of DF disease has been in the Southeast Asia and Western Pacific regions where most of the DF

8 research and intervention activities have taken place. The region of the Americas has remained understudied. In this region, most of the DF control and intervention strategies have been left to state governments. In this regard dengue has not always been the highest public health priority and as a consequence it has often been under-resourced. This has led to a steady increase in the region’s incidence of DF and dengue hemorrhagic fever (DHF) cases since the first DF outbreak occurred in the Caribbean in 1963 (PAHO, 1997). The total DF cases incidence for the region of the Americas went from 16.4/100,000 in the 1980’s and 35.7/100,000 in the 1990’s to 71.5/100,000 in the 2000’s. In addition, the incidence of the more severe form of the disease, DHF, increased from

0.2/100,000 in the 1980’s and 0.8/100,000 in the 1990’s to 1.7/100,000 in the

2000’s. The percentage of DHF as a percentage of total dengue cases increased from 1.3% and 2.1% to 2.3%, in the 1980’s, 1990’s and 2000’s respectively (San

Martin et al., 2010). In 2013, 2.35 million cases of dengue were reported in the

Americas out of which 37,687 were classified as severe dengue (WHO, 2015).

As in almost all of the countries of the Americas region, dengue has become hyperendemic in Suriname.

The first DF outbreak in Suriname occurred in 1981 when an estimated 10% of the population of Paramaribo suffered from DF-like symptoms and a total of 22 cases of DF were laboratory confirmed (PAHO, 1997). An annual resurgence of

DHF began in 1997 (WHO, 2013). The last registered dengue fever outbreak occurred towards the end of the short rainy season at the beginning of 2012. In

January 2012, a high number of suspected DF cases were admitted to hospitals

9 in Suriname, particularly in the districts of Paramaribo, Wanica and Nieuw

Nickerie. More than 300 cases were reported to the Academic Hospital

Paramaribo (Academische Ziekenhuis Paramaribo - AZP). According to local media, the outbreak quickly exhausted national dengue diagnostic and hospital capabilities (AFP, 2012). The Surinamese Bureau for Public Health (Bureau

Openbare Gezondheidzorg - BOG) responded to the outbreak by opening up an emergency hospital to manage the high number of dengue fever cases and by trying to control mosquito populations in the cities (Stabroek News, 2012). Such actions however are not preventive in nature but are geared towards disease control in response to a DF or DHF outbreak. Furthermore, the Chikungunya virus was introduced in the Caribbean in December 2013 reaching Suriname in

June 2014. By January 2015, more than 1210 confirmed cases and one death have been reported (PAHO 2015) in Suriname. This demonstrates that, like dengue, other mosquito-borne viruses can become endemic within the

Surinamese population if preventive methods are not used in a resource efficient and effective manner.

To date, no comprehensive study has attempted to determine the relative contribution and examine the relationship of environmental and demographic risk factors that contribute to DF and DHF/DSS. This study examined the demographic, clinical and environmental risk factors that contribute to dengue outbreaks and the development of severe dengue complication in the

Surinamese population. The findings were used to develop a framework to advance dengue prevention, early detection, treatment and disease outcome

10 while bolstering dengue surveillance and prevention practices based on local- evidence based information.

2.1. Literature Review

Suriname is characterized by an ethnic diversity profile unlike any other country in Latin America. Such diversity is represented in Table 1, which depicts the household population in 2012 of different ethnic groups in the two most urban and populated districts, Paramaribo and Wanica, of Suriname. The main languages are Dutch, Sranan Tongo and Sarnami and most tourism and

commercial traffic comes from the

Caribbean, The Netherlands and China.

The ethnic, religious and linguistic

components of the Surinamese society

contribute to its unique cultural diversity.

This sets the society and economy apart

from its Spanish, French, English and

Portuguese speaking regional neighbors

making it inadequate to extrapolate DF Table 1. Total and percentage population by ethnicity (ABS 2014) study findings from Latin American and Caribbean countries to Suriname. Due to the country’s uniqueness, this dengue fever study that engaged the disciplines of environmental public health, medicine, laboratory sciences, and ecosystem research in a transdisciplinary fashion.

11

Dengue in Suriname

The BOG has recorded all the dengue fever cases since 1981. Table 2 represents the annual registered dengue and dengue hemorrhagic fever / dengue shock syndrome (DHF/DSS) cases from 1981 to 2014. The incidence rate of total dengue cases (DF and DHF/DSS) has increased as outbreaks are becoming more frequent and affect a larger population. In addition, the DHF cases as a percentage of DF cases is becoming larger with every outbreak as seen in 2009 when more than half of the cases reported presented severe disease. This indicates that a higher percentage of dengue-infected people who become symptomatic are developing the severe forms of the disease.

In 1995, a systematic collection of serotype circulation began throughout the

Americas. Co-circulation of serotypes can increase the risk of severe disease due to the biological characteristics of transmission of the dengue-causing virus.

At the beginning of the twentieth century, only one serotype, dengue virus 1

(DENV1), circulated in the Americas. In 1953, the American subtype of DENV-2 was identified in Trinidad, making it the first American originated DENV genotype

(Allicock et al., 2012). This is of particular importance because the severity of illness has also been associated with the DENV genotype in which some strains of a serotype have shown to be more virulent than others (Guzman et al., 2010).

Thus, it is necessary to fully understand the genetic diversity and spatial- geographic transmission histories of the four DENV genotypes. Since 1953, all 4 serotypes are prevalent in the Americas as well as in Suriname. PAHO performed an in-depth study to chronologically document when each serotype

12 was introduced into different countries. In Suriname, DENV-1 was the first documented serotype introduced in 1978 and subsequently DENV-4 was reported in 1981 (Wilson & Chen, 2002). In 2005, Carrington et al. analyzed the phylogeography of DENV2 (subtype III) and DENV4 in the Americas. Their results showed shortly upon introduction of a serotype, epidemics of dengue in the local population correlated with rapid increase of DENV lineage numbers

(Carrington et al. 2005). The phylogeography studies also shed light on the dissemination patterns each genotype follows. Dispersal history reconstruction demonstrates the strongest dissemination links of DENV2 between the Lesser

Antilles islands and Suriname suggesting that language and economic relationships influence the diffusion of DENVs (Carrington et al., 2005, Allicock et al, 2012, Mir et al, 2014).

Table 2. Annual registered dengue (DF and DHF/DSS) and DHF/DSS cases, total incidence and DHF as a percent of total DF cases from 1981 to 2014

Yearab Dengue DHF/ Incidence (DHF/ Yearab Dengue DHF/ Incidence (DHF/ DSS ratec Dengue) DSS ratec Dengue) X 100 X 100 1981 22 0 6.01 0 1998 1140 11 251.1 0.96 1982 25 3 6.79 12 1999 695 0 151.08 0 1983 0 0 0 0 2000 1073 4 257.31 0.37 1984 0 0 0 0 2001 760 12 181.38 1.6 1985 0 0 0 0 2002 1104 23 263.48 2.13 1986 64 0 16.79 0 2003 285 1 68.02 0.35 1987 1 0 0.26 0 2004 375 7 89.5 1.87 1988 5 0 1.27 0 2005 2853 141 680.91 4.94 1989 4 0 1 0 2006 285 32 68.02 11.23 1990 16 0 3.94 0 2007 41 0 9.79 0 1991 40 0 9.68 0 2008 24 12 5.73 50 1992 24 0 5.47 0 2009 120 69 6.83 57.5 1993 171 7 40.33 4.09 2010 113 20 27.56 17.7 1994 75 1 17.44 1.33 2011 409 24 33.49 5.8

13

1995 344 0 79.08 0 2012 664 183 77.83 27.5 1996 677 0 153.51 0 2013 21 0 1.85 0 1997 90 0 20.13 0 2014 19 0 0.74 0 a 1981 to 1994 reported by the DengueNet database of the World Health Organization (WHO, 2013) b 1995 to 2014 report to from the Ministry of Health of Suriname in the Number of Reported Cases of Dengue and Dengue Hemorrhagic Fever (PAHO, 2009) c Incidence per 100,000 population. Total annual population based on The World Bank database (The World Bank, 2012)

The (DENV) is a pathogenic virus that belongs to the genus flavivirus (family flaviviridae). It is a positive, single-stranded RNA virus with an 11 kb large genome that encodes three structural proteins (capsid, membrane and envelope glycoproteins) and seven non-structural proteins (NS1, NS2A, NS2B, NS3,

NS4A, NS4B and NS5) (Rodenhuis-Zybert et al., 2010). There are 4 different types of DENV (DENV type 1 through 4, DENV1-4) capable of causing DF in humans. DENV1-4 share 65% of their genome each resulting in a distinguishable serological strain (serotypes 1-4) (Guzman et al., 2010). Upon primary infection with DENV1-4, the human innate immune system targets the virus envelope proteins with immunoglobulin (Ig) M antibodies until DENV specific (usually against NS1) IgG antibodies are produced (Rodenhuis-Zybert et al., 2010).

However, the immune response is different upon secondary infection with a different serotype. In fact, heterotypic immunity has been identified as a risk factor for the onset of DHF/DSS (Halstead & O’Rourke 1977, Guzman et al 2013,

Simmons et al., 2012) and it has been documented that introduction of a new serotype into a population increases the incidence and number of DHF/DSS cases during a dengue outbreak (Diaz-Quijano & Waldman, 2012). Thus, it is essential to know which DENV serotypes are circulating within a population.

14

Table 3 depicts the circulating serotypes found in Suriname from 1995 to 2014.

Over the past 15 years, all four serotypes have circulated within the population.

Table 3. Circulating DENV serotypes in Suriname from 1995 to 2011.

* * *

4

Year 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 201

a 1 1 1 1 1 2 2 2 2 4 - - 1 2 2 3 3 2 3 2 2 2 - 4 2 4 4 3 4

Serotypes a As reported by the PAHO and the Caribbean Epidemiology Center (PAHO ,2009; CAREC, 2012) * Data not available but not an indication that there was no dengue virus circulating in Suriname at the time

When a person presents DF-like symptoms in a policlinic or hospital in Suriname, a serum sample is collected and analyzed via virus culture or IgM and/or IgG detection assays to diagnose DF. Clinical symptoms and a positive serological test indicate a probable case of DF. In addition, in 2011 the AZP Laboratory

(AZPL) introduced reverse transcriptase polymerase chain reaction (RT-PCR) to detect the DENV nucleic acid in patients’ serum. This diagnostic test allows for a more specific and sensitive detection of the DENV and DENV serotype, thus enhancing dengue diagnostic practices and surveillance in the country. Different

DENV serotypes (Rico-Hesse et al., 1997; Vaugh et al., 2000), the sequential order in which different DENV serotypes are encountered within a population

(Burke et al., 1988; Thien et al., 1997; Nisalak et al., 2003) and the time elapsed between primary and secondary infections (Guzman et al., 2002a; Diaz-Quijano

& Alves Waldman, 2012; Montoya et al, 2013) are associated with increased pathogenicity and the development of DHF/DSS. Consequently, serotypes have to be especially considered in DENV endemic regions as these can assist in

15 forecasting whether a mild, moderate or severe disease outbreak will occur within a population. Such risk factors underscore the importance of continuous, long-term vector control to identify circulating serotypes in mosquitos before humans become infected and avoid having an increasingly larger population that is susceptible to developing DHF/DSS.

Demographic Factors

Dengue is hyperendemic in Southeast Asia and the Americas but the disease pattern is different in each region. The cumulative incidence of DHF/DSS is 18 fold greater in Southeast Asia compared to the Americas. The severe forms of the disease are also manifested more in infants and young children in Southeast

Asia compared to the Americas (Halstead, 2006). The differences in DF disease patterns highlight the need to further investigate the demographic and clinical variations of dengue pathogenicity within different populations. DF and DHF/DSS studies performed in Southeast Asia and the Americas have identified additional individual risk factors that can contribute to the development of disease.

Established risk factors that contribute to the development of severe dengue disease include race (Chiewslip, et al 1981; Bravo et al 1987; de la Sierra et al.,

2006), gender and age (Guzman et al., 2002b; Halstead et al., 2002). For this reason, it is imperative to analyze whether race, gender and age are covariants in the prediction of more severe forms of dengue in Suriname especially due to the unique demographic composition of this country.

16

The population’s immunological status is likely to be a contributing factor for the increase in dengue fever and DHF/DSS incidence increase over the past 30 years in Suriname (Table 1). Sequential heterotypic DENV infections elicit a different, more consequential antibody response, referred to as antibody- dependent enhancement (ADE), in humans. This antibody response can enhance virus infectivity and patients are more likely to progress into a critical

(DHF/DSS) phase of the disease characterized by plasma leakage, thrombocytopenia, leukopenia, organ impairment and even death (Halstead,

1983; WHO, 2009). The antibody response is heightened depending on which

DENV serotype caused the infection (Vaughn et al., 2000; Rico-Hesse et al.,

1997; Burke et al., 1988) and the time elapsed between primary and secondary infection (Guzman et al., 2002; Diaz Quijano et al., 2012; Montoya et al 2013).

When DENV was first introduced in Suriname in 1980, the general population was immunologically naïve. Since then, however, all four DENV serotypes have been introduced making the current population immunologically sensitized and placing a higher proportion of the population at a higher risk of presenting severe dengue upon infection with DENV.

Clinical Management

Clinical management and how clinical services are delivered affect the progression of DF into DHF/DSS and death. Health care accessibility is managed by the Ministry of Health (MoH).The biggest health care provider is the government, which through the Regional Health Services (Regionale

Gezondheidsdienst – RGD) and the Medical Mission (Medische Zending – MZ)

17 delivers primary services to the poor and near poor. Access to health care is also available via the State Health Insurance, private firms or self-pays. The State

Health Insurance covers 35% of the population, mostly civil servants, while the

MoH covers 42% of the population. The rest of the population is covered by private health insurance (PAHO, 2002). For secondary care, Suriname has 6 hospitals with a total of 1,687 beds (ABS, 2012). The MoH subsidizes and coordinates through the BOG 43 RGD clinics in the coastal region and there are

57 Medical Mission clinics in the interior to provide primary care to the poor or near poor (Coronie, 2007, MZ 2012, RGD 2015). Due to the endemicity of the dengue, most health care practitioners and physicians are familiar with the symptoms of DF and DHF/DSS. Experience with recognizing the early symptoms of the disease is essential in curbing the progression of DF into DHF/DSS and reducing DHF/DSS mortality in individuals. Dengue infection presents itself in a heterogeneous fashion (Figure 1a). DF is often asymptomatic or it can present mild symptoms including a high fever, headache, muscle and joint pains, vomiting with or without a rash. This febrile phase can also be accompanied by more severe symptoms such as petechiae, hepatomegaly and a continuous drop of white blood cell count resulting in leukopenia. The individual can progress into a critical phase in which progressive leukopenia and increased capillary permeability leads to shock. During this phase, it is imperative to provide fluid replacement therapy to prevent organ impairment and death (WHO, 2009). This therapy is usually the sole intervention for DHF/DSS since its implementation in

1975 (WHO, 1997). Few studies have compared the efficacy of different fluids

18 but Wills et al. (2005) demonstrated that isotonic crystalloid solution remains the cheapest and most effective course of treatment for most DHF and DSS patients.

The implementation of fluid therapy decreases the DHF/DSS mortality from 20-

30% to less than 1% (WHO, 2012). Thus, as with any other disease that is treatable, early recognition is essential to decrease mortality (Figure 1b). A decrease in DHF/DSS mortality has been observed in Suriname since the emergence of the first cases in the 1980’s (Dr. A. Jessurun, personal communication, July, 2012). Under Dr. Jessurun’s leadership, the Saint

Vincentius Hospital (Sint Vincentius Ziekenhuis - SVZ) in Suriname has maintained a medical record database of all dengue cases since 1980. The database includes information on demographic and clinical risk factors that were analyzed to assess their contribution for DF and DHF/DSS within the patient population. The information was also used to elucidate factors that affect the length and cost of hospitalization of dengue cases and to develop a framework to inform dengue diagnostic and surveillance practices in clinical settings in

Suriname.

Figure 1. Dengue presents itself as a spectrum of disease, A) upon infection with the dengue virus a person can remain asymptomatic or develop dengue shock syndrome (DSS) if no treatment is provided. B) fluid therapy can decrease DSS mortality to less than 1% thus early disease recognition is essential to decrease mortality

19

Environmental Risk Factors and Disease Transmission

Aedes aegypti is the main vector of DENV transmission to humans in Suriname.

It is believed to have been introduced into the Western Hemisphere between the

XV and XVII century from West Africa after which it spread to most countries in the Americas. However, this mosquito was successfully eradicated from the

Americas by mid-twentieth century (Pinto Severo, 1955) but reinfestation occurred after vector reduction and eradication campaigns were halted (“The

Feasibility of eradicating Aedes aegypti in the Americas,” 1997). Aedes aegypti accelerated resurgence was possible due to its evolutionary adaptation to the urban environment. This mosquito is anthropophagic and breeds in natural or man-made containers where fresh water accumulates. After a blood meal, the female mosquito oviposits up to a hundred eggs at a time in natural or man-made fresh water containers and it takes only seven days for adult mosquitos to emerge (Service, 2008). Increased urbanization has augmented the number of potential Aedes aegypti breeding sites especially in moist tropical regions where correlations between mosquito larval densities and temperature, relative humidity and average precipitation have been identified (Hopp & Foley, 2003). Other breeding site magnification factors in tropical regions, especially in developing countries, include irregular garbage collection and unreliable clean water supply.

This is particularly a problem in Paramaribo were systematic garbage collection has only been recently implemented allowing garbage to easily accumulate in yards or sidewalks. In addition, due to irregular water supply in certain areas of the city, people cut into clean water-supply lines to retrieve water (Hiwat, 2012a).

20

This generates more intentionally and non-intentionally man-made water containers resulting in additional Aedes aegypti breeding sites in densely populated neighborhoods.

There are other mosquitoes competent of transmitting arboviruses to humans or animals in Paramaribo. These include members of the Culicidae family, including

Culex, Mansonia, Aedes, Anopheles, Wyeomyia and Psorophora (De Haas &

Arron-Leeuwin, 1975). The importance of these different mosquitoes is related to the impact their larvae can have on Aedes agypti while competing for aquatic resources (Juliano & Lounibos, 2005). Aedes albopictus, an invasive container- breeding mosquito, has been documented to decrease the number of local

Aedes aegypti when introduced into a community (O’Meara et al., 1995). This invasive species has spread to many North and South American countries over the past thirty years. Aedes albopictus was introduced into Brazil in the 1980’s, however, Suriname remains uninfested (Juliano & Lounibos, 2005; Benedict et al., 2007). There is no consensus why Aedes albopictus has not invaded

Suriname as the country’s climate and ecology provide a suitable niche.

However, the potential introduction of Aedes albopictus into Suriname could be affected by changes in climate, increase in trade with countries in which the species is established and human encroachment activities in the Amazonian forest that separates Brazil and Suriname. It is possible that the introduction of

Aedes albopictus could affect the Aedes agypti population numbers.

Seasonal variations can limit or increase the number of Aedes aegypti available breeding sites. Thus, the incidence and severity of vector-borne diseases such

21 as DF tend to vary with environmental conditions such as temperature, relative humidity and average precipitation (Altizer et al., 2006). In many countries, DF outbreaks typically follow the rainy season (Schreiber, 2001). Towards the end of the rainy season, fresh water containers support an increased emergence of additional adult mosquitos. This is a particular problem in densely populated cities where precipitation intensifies the virus-vector-host interaction where the

DENV is endemic (Diaz-Quijano et al., 2008).

Additionally, other meteorological factors, such as temperatures and humidity affect the Aedes aegypti gonothropic cycle, larval survival and DENV replication speed (Focks et al., 1995; Parker & Holman, 2012; Hales et al., 2002) affecting the quantity of infected mosquitoes present at any given time. Different simulation models at local and global levels have identified relationships between climate-induced variations in mosquitos and cases of DF and DHF in tropical regions (Hopp & Foley, 2003; Focks et al., 1995; Hales et al., 2002). These models illustrate the relationship and the impact of climatologic risk factors such as temperature, precipitation and rainfall on DF transmission. As the effects of climate change in tropical countries become more apparent, this relationship may become more prominent.

Future climate projections indicate a global temperature rise of 1 to 3.5°C by

2100 (Githeko et al., 2000). The effects of such an increment are expected to be even more severe in the tropics due to the effects of El Niño-Southern Oscillation

(ENSO). In ENSO years, temperatures are on average 0.5°C higher and severe droughts and floods are more likely to occur throughout the American region.

22

Furthermore, epidemiological associations between ENSO and dengue epidemics have been demonstrated in South America, Southeast Asia and the

Pacific (Gagnon et al 2001; Kovats et al., 2003, Ferreira 2014). In Suriname,

ENSO causes warmer temperatures and below-average rainfall. A 2001 correlation analysis revealed a statistical significant relationship between ENSO events and both of Suriname’s 1986 and 1992 dengue epidemics (Gagnon et al.,

2001). The below-average rainfall prompts people to store water in containers especially in areas where access to clean water is unreliable or non-existent.

Thus, climatologic variations do not only affect the biology of Aedes aegypti and the DENV but it also prompts changes in human behavior that increases their probability of contracting a dengue infection. In Suriname, the peak of DF cases occurred towards the end of the short rainy season. Public health practitioners and physicians have observationally associated seasonality with the extent and severity of a DF outbreak. To our knowledge, there have been no comprehensive investigations in Suriname correlating the historical average monthly rainfall, temperature or relative humidity to DF and DHF/DSS incidence.

Spatio-Temporal Visualization of Dengue Cases

Innovative spatio-temporal visualization techniques, such as geographic information system (GIS), are becoming more commonplace in efforts to implement better surveillance and control techniques against vector-related diseases. For DF, GIS provides visual patterns of case distribution and vector infestation and elucidates spatial relationships between risk factors and disease.

The software can be utilized to identify DF and DHF/DSS in spatial clustering,

23 how they emerge and if those cases are related to environmental or demographic drivers overtime (Duncombe et al., 2012). Additional factors that influence vector density and dengue transmission, such as temperature and precipitation (de Wet et al., 2001), has to be examined to obtain a more comprehensive depiction of

DF and DHF outbreak patterns overtime. Surveillance maps should also consider the immunological status of the population to forecast the risk the introduction of a specific DENV serotype would have on the population. Unfortunately, the DF and DHF case diagnostic serology test does not include the DENV serotype in most developing countries.

Mapping epidemiological data to ascertain dengue tranmission is often preferred due to the difficulty of collecting adult Aedes mosquitos. For this same reason, surveillance often revolves around the larval and pupal stages of Aedes aegytpi

(Eisen & Lozano-Fuentes, 2009). Immature mosquito indices, such as the

Breteau Index and the House Index, however, do not translate to the risk of transmission as well as late-instar larvae, pupae or adult mosquito densities do

(Focks & Chadee, 1997; PAHO, 1994; Sanchez et al., 2006; Morrison et al.,

2008). In fact, pupae indices that measure the number of pupae per person in a community, and the identification of pupae productivity in different classes of containers better reflect the risk for dengue transmission in a community (Focks et al., 2000; Focks & Alexander, 2006; Barrera et al., 2006). Thus, initial steps for mapping epidemiologic as well as entomologic data are necessary to identify areas that need more urgent attention.

Current Prevention Practices in Suriname

24

A variety of established management practices exist as methods of vector control. These range from modifying or manipulating the environment (mosquito- proofing of water-storage containers) to chemical control (using adulticides and larvicides) (WHO, 2009). However, before any vector control measures can be implemented, it is necessary to establish a sound vector surveillance system to evaluate the success of interventions aimed at reducing Aedes aegypti populations. In 2000, the BOG, supported by the Rotary Clubs in Suriname and the Pan American Health Organization (PAHO), put together a ‘Plan of Action to

Control Dengue and Dengue Epidemics in Suriname’ under which an Urgent

Dengue Control Program (Cash Programma Dengue Bestrijding - CPDB) was created. The CPDB includes recommendations to combine epidemiological and entomological data to inform potential control activities against the 3 life cycle stages (egg, larval and adult) of Aedes aegypti. Furthermore, it presents stepwise strategies for the endemic DF phase when isolated cases of the severe form of the disease (DHF/DSS) arise, when DF clusters are identified, when an outbreak occurs in a district and when there is an epidemic (BOG, 2000). The strategies include environmental and chemical control methods, medical resources allocation, information dissemination and education to the affected population and health care practitioners, and surveillance and monitoring techniques that have to be implemented at all times. Since the creation of the

CPDB no structural base for larval and mosquito control or data collection were implemented. Most of the vector control response strategies occur on an ad-hoc basis and are carried out by environmental health inspectors under the

25

Environmental Inspection division of the BOG (A. Warner, personal communication, January, 2013). However, following the 2012 dengue epidemic in Suriname, the Entomology Division of the BOG conducted two Aedes aegypti larval surveillance studies in Albina and in Marienburg (Hiwat, 2012a; Hiwat,

2012b). In Albina and Marienburg, the house index and the breteaux index were found to be above the PAHO threshold categorizing both towns as high transmission risk for dengue (Appendix 1). Additionally, the larva indices demonstrated that certain containers, such as water storage containers including water drums and buckets, were the most productive breeding sites for Aedes mosquitos (Hiwat et al 2013). These results are the first attempt to research

Aedes aegypti populations and dengue risk transmission within Suriname and highlight the importance of breeding site reduction activities in decreasing the risk of a future dengue epidemic.

Successful transmission of dengue depends on four essential components: the human host, the adequate environment for the pathogen, the dengue virus, to develop, and the vector. These four components are influences by external factors that can affect transmission as well as the rate of transmission

(Supplement Figure 1) and the development of DHF (Supplement Figure 2). This study integrated dengue demographic and clinical data into a GIS platform that was used to identify clusters of dengue in Suriname. To further elucidate the environmental factors that influence transmission of dengue in Suriname, entomological surveillance of Aedes aegypti pupae were conducted in the identified dengue clusters. The results from this study elucidate different factors

26 that influence the human host, the environment, the dengue virus and the Aedes aegypti vector in the transmission of disease in Suriname.

2.2. Hypotheses & Research Questions

The goal of this study is to shift the focus from a dengue fever ad-hoc response approach to a more transdisciplinary prevention and surveillance system. We used reported dengue cases and demographic, clinical and environmental data to generate risk maps. These maps were used to conduct environmental and entomological surveys in Paramaribo to identify dengue transmission risk factors and help improve dengue surveillance and prevention strategies in Suriname.

Research Questions

1. What are the demographic, clinical and environmental risk factors that

contribute to dengue outbreaks and the development of severe dengue

complications in Suriname?

2. How can these factors aid in developing a cumulative and transdisciplinary

risk framework to advance dengue prevention, early detection, treatment and

disease outcome while bolstering surveillance?

27

Hypotheses and Aims

Hypothesis 1: There is a statistical significant association between demographic and clinical factors and the progression of dengue fever into the severe forms of the disease in Suriname

Aim 1. 1 Characterize the incidence, prevalence and severity of dengue in

Suriname from 2001 to 2012

Aim 1. 2. Examine the environmental, demographic and clinical factors that influence the development of the severe form of dengue in Suriname

Hypothesis 2: Spatio-temporal trends and relationships of factors influencing dengue incidence and severity will identify disease hotspots in Suriname

Aim 2.1. Integrate clinical and demographic information of all dengue cases reported in Suriname since 2001 into a GIS platform

Aim 2.2. Identify spatial and temporal historic clusters of DF and DHF from 2001 to 2012

Hypothesis 3: Prediction models and dengue disease cluster analysis can identify effective prevention methods and risk reduction strategies for dengue in

Suriname

Aim 3.1. Characterize dengue-related entomological and environmental data in previously identified high transmission risk

Aim 3.2. Design effective, low-cost dengue intervention activities based on local evidence-based information to reduce the risk of dengue transmission in

Suriname

28

3. Materials and Methods

Since 2001, the BOG has documented suspected, probable and confirmed cases of DF reported to the country’s hospitals and regional clinics. Upon recognition of

DF, doctors and nurses in clinics and hospitals are obliged to fill out a Dengue

Case Investigation Report. This report includes demographic information about the patient, symptoms, diagnosis and the serology test and results. These reports are sent to the BOG and entered into a dengue database. The database also includes the source, clinic or hospital, and the patient’s address.

Data Sources

Data related to clinical cases of DF were also obtained from SVZ. In addition to reporting these cases to the BOG, the hospital also maintains an electronic database for all pediatric and adult patients diagnosed with a dengue since 2001.

In 2011, SVZ began sending serum samples of probable dengue fever patients to the AZP Laboratory (AZPL) to undergo RT-PCR analysis for dengue virus detection. The AZPL has compiled in an electronic database all the serum samples tested including the hospital or clinic source, the outcome of the test and the serotype of DENV detected. The AZLP has provided this electronic database for the purpose of this research. Additionally, from the Meteorological Service

Suriname (Meteorologische Dienst Suriname- MDS) and the General Bureau of

Statistics (Algemeen Bureau voor Statistiek – ABS), we obtained historical climatologic data and census data, respectively. The content of each of the datasets used for the purpose of this research is depicted in Table 5.

29

Table 5. Dataset description by source

Source Name Year Demographic Clinical / Environmental AZPL Academisch Ziekenhuis 2011-2012 -Name* -Onset date Paramaribo Laboratorium -Date of Birth -Physician (Academic Hospital -Age -RT-PCR result Paramaribo Laboratory) -Gender -Type

BOG Bureau Openbare 2001-2012 -Identifier -Source Gezondheidzorg -Date of Birth -Physician (Bureau of Public Health) -Age -Onset date -Gender -Admission date -Ethnic Group -Discharge date -Address -Death -Lab Result -DHF / DSS -Type

SVZ Saint Vincentus 2001-2012 -Name* -Physician Ziekenhuis -Date of Birth -Admission date (Saint Vincent Hospital) -Age -Discharge date -Gender -Initial diagnosis -Ethnic Group -Lab result -Address -Type

MDS Meteorologische Dienst 2000-2012 -Monthly temperature Suriname (average, min/max) (Meteorological Service -Monthly relative Suriname) humidity -Monthly precipitation

ABS Algemeen Bureau voor 2000-2012 -Population data Statistiek (General -Street codes Bureau of Statistics)

The BOG assigns each case a unique identification code to protect patient confidentiality. The AZPL and SVZ databases include patients’ names. In order to protect patient confidentiality, each case was assigned a unique code

(identifier) and all addresses were coded in concordance to the ABS codebook.

Suriname is divided into 10 administrative levels known as districts and each district is divided into resorts. There are a total of 62 resorts. The ABS has assigned a unique numerical or alphabetical code to each district and resort,

30 respectively, and every street within each resort has also been coded. Neither of the three dengue case datasets were coded using the ABS coding system.

Therefore, each address was referenced back to the ABS street code database in order to uniformly assign each case within the BOG, SVZ and AZPL database with the appropriate ABS code. In this manner, the databases employed for the purpose of this research were stripped from all identifiers and patient confidentiality was protected. All data were kept in an encrypted computer property of Tulane University.

Human Subject Protection

The study protocol was approved by the Tulane University School of Public

Health and Tropical Medicine Institutional Review Boards (IRB) and the Ministry of Health Suriname.

Case definition

Diagnostic tools cannot be used in DF prevention but allow for the differentiation among suspected, probable and confirmed cases of DF. However, diagnostic tools such as RT-PCR can play a role in surveillance to effectively and expeditiously identify which serotype is present in infected patients during an outbreak. For the purpose of this study, the criteria in Table 6 were used to differentiate among suspected, probable and confirmed cases of DF and DHF.

31

Table 6. Case classification

Definition Criteria for DF

Suspected -Acute febrile illness -Two or more of the following: headache, retro-orbital pain, myalgia, arthralgia, rash, hemorrhagic manifestations, leucopenia

Probable -Clinical symptoms

-Compatible serology (reciprocal hemagglutination-inhibition antibody titer > 1280, comparable IgG EAI titer or positive IgM antibody test in serum specimen

Confirmed -DENV isolation in cell culture -RT-PCR detection of nucleic acid in serum

-Clinical symptoms of DHF/DSS (thrombocytopenia / platelet count <100,000/μL)

Hospitals and clinics in Suriname follow the WHO dengue guidelines for diagnosis. The BOG includes all diagnostic test and serology result information in their database which allowed for the classification of a dengue case as probable based on laboratory testing. Similarly, the SVZ follows WHO recommendations of diagnosis and inputs only cases which are clinically compatible with the definition of DF. Thus, all cases in the SVZ database were classified as suspected based on clinical presentation. Patients with compatible serology are classified as probable. Thus, both the BOG and SVZ DF cases were treated as suspected or probable and only classified as confirmed when records showed the DENV was isolated or detected in the patient’s serum. Moreover, due to the unique symptom presentation of DHF/DSS, clinical diagnosis of DHF was used to confirm dengue only for the BOG dataset.

32

Data Analysis

Aim 1. 1. Characterize the incidence, prevalence and severity of dengue in

Suriname from 2001 to 2012

Due to the manner in which the cases were reported to the BOG, the BOG database was treated as a predetermined sample of all suspected cases of DF reported in Suriname between 2001 and 2012. The frequency distribution of cases for the entire country from 2001 to 2012 was calculated and cases were classified according to the criteria outlined in Table 6. The areas in which this study was based on consist of 6 districts (Paramaribo, Wanica, Commenwijne,

Saramaca, Para and Nickerie) in Suriname. These districts accounted for 85% of all BOG reported DF cases from 2001 to 2012. Census data pertaining to piped water availability, number of dwellings and health care center distribution were used to describe each district. For each district, and when possible for each resort, the frequency of probable and confirmed cases by demographic factors was calculated. The DF and DHF incidence, prevalence and mortality were calculated by year for each district and resort.

Aim 1. 2. Examine the demographic and clinical factors that influence the development of the severe form of dengue in Suriname

To identify factors that contribute to the development of DHF in Suriname the

BOG, SVZ and AZPL databases were used. For univariate and multivariate analyses, the case inclusion and exclusion criteria were:

33

- Inclusion: All probable and confirmed dengue cases reported in a clinic or

hospital between 2001 and 2012

- Exclusion: Suspected dengue cases and cases for which no date of birth

or address was documented

For the BOG and SVZ

database, separate univariate

and multivariate analyses were

performed. For each database,

the data here screened to

identify missing or excess data

as well as outliers. Before any

statistical analysis was

Figure 2. Data cleaning process conducted, it was determined how to treat missing data and how to correct for excess data and outliers. The databases were edited and the process was documented to decrease bias or errors during data input (Figure 2).

BOG database: The frequency of probable and confirmed cases of dengue stratified by demographics was calculated. The annual incidence was calculated using ABS census data and further stratified by district. A Chi Square test was used to test for the significance of categorical predictors in association with onset of severe dengue (DHF). A p-value < 0.05 was considered statistically significant.

34

For marginal frequencies that were small (N < 5), the Fisher’s “exact” test was used to determine significance.

A multivariate analysis was conducted to test for the association of selected predictors (such as age, ethnicity, and gender) on the development of DHF was conducted. A binomial logistic regression model, given the categorical nature of the predictor variables and binary outcome of disease ([Y = 0]: no DHF; [Y = 1]:

DHF), was used to measure the effects of variables on predicting the outcome of

DHF. The model contained interaction variables that affected the magnitude of the outcome. The Maximum Likelihood method and likelihood ratio test were used to test for the joint significance and interactions at α = 0.5. Odds ratios (OR) and their 95% confidence intervals (CIs) were estimated. The data were also analyzed for over and under dispersion (Goodness of Fit test) before testing the statistical significance of the overall model and the individual predictors.

SVZ and AZPL databases: After screening and editing both datasets, the names, date of birth and date of hospitalization of dengue cases presenting at SVZ in

2011 and 2012 were cross-referenced with the AZPL dataset. The frequency of suspected, probable and confirmed cases, stratified by age, gender and ethnicity was calculated. A chi-square test with only the confirmed cases was used to determine if there was a difference in frequency of DHF cases compared to DF.

To determine whether there is an association between demographic characteristics and dengue diagnosis, demographic characteristics and the length of hospitalization, and dengue diagnosis and length of hospitalization we

35 performed a special type of structural equation modeling (SEM) known as path analysis. This type of structural equation modeling (SEM) can be modified to handle categorical variables. Each categorical variable was coded into dichotomous variables to obtained odds ratios (OR) in the following manner: gender – male / female, race – African / Asian descent, case classification – laboratory tested / not laboratory tested, season – rainy / dry, and dengue diagnosis – DF / DHF. Continous variables, hospitalization time and age, remained unchanged. A Yule’s transformation of OR into the Q-metric was employed to approximate Pearson’s correlation coefficients between the dichotomous variables. The correlation matrix was analyzed by SEM using

IBM®SPSS®Amos 22.00. The Χ2 test of absolute model fit was calculated as well as tests of relative fit (such as comparative fit index (CFI) and the root mean square error of approximation (RMSEA)) used to compare the specified model fit to the independent model fit. Maximum likelihood (ML) estimation was used.

Furthermore, the total and the mean length of hospitalization (in days – continuous variable) for suspected, probable and confirmed cases were obtain to make inferences about diagnostic test and hospitalization time and DHF. For this reason, DHF case classification was based on diagnostic test result and not only on clinical symptoms. The Kolomogorov-Smirnov and Shapiro-Wilk test determined that the hospitalization length (in days) data was positively skewed and not normally distributed. Thus, medians and inter-quartile ranges (IQR) were calculated and the non-parametric tests (Mann-Whitney U-test and the Kruskal-

Wallis H test) were used to compare length and cost of hospitalization between

36 case classification means. Furthermore, to identify variability between case classifications, we performed a pair-wise comparison with Bonferroni correction.

Data analysis was conducted using SPSS® Statistics v19.

Spatial Analysis

Demographic, clinical and environmental risk factors influence the progression of

DF and DHF. These risk factors also contribute to the scale of DF outbreak in

Suriname. Thus, it is imperative to identify which factors exhibit patterns that correlate with the spatial and temporal distribution of DF and DHF cases. Maps to display and analyze dengue epidemiological data were used to implement a more systematic surveillance and to develop more targeted intervention strategies for Suriname.

Aim 2.1. Integrate clinical and demographic information of all dengue cases reported in Suriname since 2001 into a GIS platform

To determine the spatial distribution of DF cases across the coastal districts of

Suriname, we examined patterns of case distribution using Geographic

Information System (GIS). As indicated above, the BOG database contained all reported cases of DF in Suriname since 2001. All reported cases were given a unique identifier by the BOG and their addresses were coded following the ABS codebook. This codebook was made available by the BOG for the purpose of this research. We obtained Suriname basemaps depicting spatial limits of different administrative levels, such as country, district, and resort. The geodatabase also included a layer with all the streets in Suriname. However, the geodatabase

37 street codes were not coded according to the ABS codebook. Therefore, the street names of all probable and confirmed dengue cases were cross-referenced to the attribute table of the street layer by street, resort and district. The street level data layer included highways, major roads, minor roads and side-street information depicted as lines but no parcel boundaries were available. Thus, each matched dengue case was located to the street midpoint. All cases for which an address code or a source was not documented were excluded. The probable and confirmed dengue cases were exported into a single point-data layer. The corresponding attribute table included all the clinical and demographic information depicted in Table 2 pertinent to each case allowing us to select by different attributes. The dengue case layer and Suriname basemaps were used for the spatial analysis.

Aim 2.2. Identify spatial and temporal historic clusters of DF and DHF from

2001 to 2012

Epidemiological maps were generated at the district and resort level. Census data for all districts and resorts were used to cumulative incidence of DF and

DHF by corresponding year. The population density of districts and resorts was estimated using annual population growth rates for the years for which no census data were available. The case and population numbers were used to measure spatial correlation and to identify global clusters in the data following the decision tree depicted in Figure 3.

38

Figure 3. Spatial analysis to identify global and local clusters.

To measure the overall clustering of the data, we generated a Moran’s I as a statistical measurement for global spatial clustering. Upon confirmation of global clustering (z>1.96, p<0.05) SaTScan™ was used to determine the value of the

Local Moran’s I and to identify where and when spatial and temporal clusters of

DF and DHF occurred in Suriname. A local Moran’s I close to -1 or close to +1 will indicate negative or positive spatial autocorrelation, respectively, allowing the identification of low and high transmission areas of dengue. A spatial autocorrelation with a p-value < 0.05 was considered statistically significant. If, however, no global clusters had been identified in the spatial data, Poisson regression analysis would have been used to test for spatial random distribution of probable and confirmed DF and DHF cases (Ho: Dengue fever cases are

39 randomly distributed over space and time; Ha: Dengue fever cases are not randomly distributed over space and time). A p-value < 0.05 will be considered statistically significant.

Furthermore, upon cluster identification, we created neighborhood polygon layers to guide the environmental surveys. All layers were created and visualized using

ArcGIS v10.2 (www.esri.com).

Environmental Surveys

Aedes aegypti is a uniquely domestic vector that thrives in urban environments.

Vector control activities have proven to be the best strategy to decrease dengue transmission of disease when a vaccine or chemotherapy is absent (Scott &

Morrison, 2010). However, due to the complex interaction of the environment, the pathogen, the human host and mosquito vector, each community faces different dengue transmission dynamics that are heavily influenced by risk factors discussed in this study. In Suriname, as in other countries with endemic dengue, activities to decrease the transmission of disease need to be based on local epidemiological and environmental conditions. GIS technology can be used as an evidence-based decision-making tool to tailor interventions to communities’ needs. However, to better understand the local transmission of disease patterns, it is necessary to include entomological data and serotype information into surveillance strategies in Suriname. GIS is not currently used as a dengue surveillance tool in Suriname but the BOG has purchased 3 Global Positioning

Systems (GPSs) and has trained 12 employees in the use of these receivers (A.

40

Warner, personal communication, March, 2013). The BOG recently received funding for the GIS software platform and training. The research hereby presented elucidates how GIS and statistical modeling can be used to guide public health decision-making using evidence-based research to achieve quantitative goals.

Aim 3.1. Characterize dengue-related entomological and environmental data in previously identified high transmission risk areas

To gain further insight into local conditions, we identified high and low transmission areas based on epidemiological information. In conjunction with the

BOG, we conducted surveys to assess the entomological and environmental situation within the high and low transmission areas. The identified clusters were divided into neighborhoods. Within each neighborhood, the number of blocks and streets were calculated. Currently, public and private premises within the clusters have not been mapped therefore we were not able to determine the number of units (houses or premises) within each neighborhood. Therefore, a random selection of streets within each neighborhood was selected for sampling. Every other house on selected streets were surveyed. A geographical position system

(GPS) handheld device was used to obtain the coordinates of all surveyed houses. The BOG environmental inspector asked for verbal permission to inspect the outside of the house and the yard (inside, second floors and roof gutters were not inspected) for Ae. aegypti breeding sites. If verbal consent was not given, or the house was empty, the next house was sampled for each of the selected streets. The condition of the houses and yard and the accessibility to piped-water

41 was recorded. For each house, the type of water-holding containers will be classified according to descriptive characteristics (such as size, material, location, disposability) in order to identify which containers were the most productive breeding sites. Moreover, Aedes aegypti larval density and pupae count were documented for every positive container. Even though Ae. aegypti larvae and pupae are visible to the naked eye and easily distinguished from other mosquito species, pupae were collected in the field using a dipper or a pipette and brought back to the Central Laboratory Suriname for emergence and species identification. The information collected was analyzed to determine entomological indices (such as the pupae/person ratio, Breteau Index, Container

Index, House Index, and Dispersion Index). The GPS coordinates with corresponding survey information was visualized in a GIS platform. The survey results were used to identify extremely productive containers that could enable more targeted intervention activities. Data collection in both high and low dengue transmission areas also clarified which local factors affect the transmission of this vector-borne disease in Suriname. This was the first attempt to map entomological and environmental data that potentially influences the rate of dengue transmission in Suriname.

Aim 3.2. Propose effective, low-cost dengue intervention activities based on local evidence-based information to reduce the risk of dengue transmission in

Suriname

Surveillance is essential to decrease the transmission of dengue in Suriname.

Currently, clinical and demographic information of suspected, probable and

42 confirmed cases of DF and DHF are systematically reported to the BOG. Hereby, we demonstrated that historical cases of dengue can help identify spatial and temporal patterns of dengue transmission in Suriname. Additional information, such as the entomological and environmental data collected and incorporated into a GIS platform helped to further elucidate factors associated with higher transmission of dengue. The inclusions of epidemiological, environmental and entomological data into a real-time surveillance system can inform the most efficient and effective strategy to shift dengue control to prevention oriented approach and contribute to the development of more dynamic epidemiological models for dengue.

Limitations

For the purpose of this research, DF case classification followed the criteria described in Figure 4. The criteria are based on WHO recommendations, which are currently under revision due to the heterogeneous manner in which of DF presents clinically (Kalayanarooj 2011, Hadinegoro, 2012). Thus, there is no consensus in the literature regarding a standard guideline to distinguish between suspected, probable and confirmed DF cases. In Suriname, health practitioners diagnose DF in patients following the WHO guidelines (WHO, 2009) and only send suspected DF cases to clinics and hospitals in the country. Therefore, all patients in the BOG and SVZ database are classified as suspected cases of DF.

This predetermined sample of DF affected the statistical analyses and our interpretation of the results. Furthermore, no information on comorbidities, treatment or laboratory results (non-diagnostic) were available for the analyses.

43

The environmental surveys were limited by time and resources. Unfortunately, adult mosquitos were not collected due to unavailability of an aspirator. The inside or second floor of houses, roofs and roof gutters were not inspected limiting the ability to recognize all breeding sites. No schools or other type of premises were inspected. Furthermore, a delay in the time it takes for the BOG to collect and input all reported cases of dengue hindered our ability to overlap DF and DHF cases with the areas surveyed because information on dengue cases reported between November 2013 and February 2014 has not yet been released by the BOG.

44

4. Spatial distribution of epidemiological cases of dengue fever in Suriname,

2001 – 2011.

Accepted for publication in the West Indies Medical Journal

Title

Spatial distribution of epidemiological cases of dengue fever in Suriname, 2001-2012.

D Hamer1, M Lichtveld1

1Department of Global Environmental Health Sciences, Tulane University School of Public Health and Tropical Medicine, New Orleans, USA.

Short title

Distribution of dengue fever cases in Suriname

Key words dengue fever, spatial distribution, Suriname

Synopsis

Retrospective epidemiological data was used to characterize the frequency, incidence and severity of dengue fever and to identify clusters of disease in Suriname. The results illustrate the need to identify risk factors that influence the transmission of dengue in Suriname.

45

Abstract

Objective: To characterize the frequency, incidence and severity of dengue fever in Suriname and to detect historic clusters of disease by integrating epidemiological data into a spatial visualization platform.

Methods: The frequency, incidence and severity of all reported dengue fever (DF) and dengue hemorrhagic fever (DHF) cases in Suriname from 2001 to 2012 were calculated and stratified by demographic factors. Using a geographic information systems (GIS) platform, we visualized the distribution of DF cases and used Moran’s I to detect autocorrelation. Furthermore, a retrospective spatial Poisson probability model was used to identify local clusters of DF within Suriname. Local clusters were divided into neighborhoods and individual DF cases were mapped to the street level.

Results: In Suriname, cases of DF emerge in cyclical patterns (3 to 5 years) with seasonal peaks following the short and the long rainy season. Chi-square analysis indicated a statistically significant (p<0.05) difference between age group, ethnicity and district and the onset of DHF. The spatial analysis detected spatial autocorrelation and four statistically significant (p<0.05) clusters were identified in the two most populated districts of Paramaribo and Wanica.

Conclusion: In Suriname, identification of demographic and environmental risk factors that contribute to the development of DHF is essential to determine how preventive action can be more effectively allocated. The integration of epidemiological data into a GIS platform allowed for the identification of historic epidemiological clusters of dengue which will be used to guide environmental health studies in Suriname.

46

INTRODUCTION

Over the last 50 years, the incidence of dengue fever (DF) worldwide has increased by

50-fold and 2.5 billion people currently live in dengue-endemic regions. An estimated 50 million cases of DF occur worldwide every year (1). Currently, dengue is hyperendemic in most of the countries of the World Health Organization’s Region of the Americas, including Suriname (2). Over the past 30 years, an increased level of urbanization and successful re-infestation of Aedes aegypti, the primary vector of dengue transmission, have contributed to the rapid spread of dengue in Suriname (3). The first endemic cases of DF occurred in 1981, with a total of 22 laboratory confirmed cases and an estimated

10% of the population of Paramaribo suffering from DF-like symptoms (4). An annual resurgence of DF and DHF began in 1997 (5) and the last registered dengue outbreak occurred towards the end of the rainy season in 2012. This annual resurgence of DF and

DHF cases is influenced by demographic and environmental factors that have not been previously studied in Suriname. Thus, it is imperative to identify which factors exhibit patterns associated with the spatial and temporal distribution of DF and DHF. This study aims to 1) characterize the frequency, incidence and severity of dengue cases in Suriname from 2001 to 2012; and 2) integrate clinical and demographic information of all dengue cases to identify historic clusters of DF in Suriname.

The dengue virus (DENV) is a pathogenic human virus that belongs to the genus

Flavivirus (family Flaviviridae). Four different types of DENV (DENV type 1 through 4,

DENV1-4), which result in distinguishable serological strains, are capable of causing DF in humans (6). DENV-1 became the first documented serotype circulating in Suriname and subsequently DENV-4 was reported in 1981 (7). DENV4 was reported in Suriname

47 in 1994, followed by DENV1 in 1998, DENV2 in 1999 and DENV3 in 2001 (8, 9).

Moreover, different serotypes have co-circulated within the population: DENV1, DENV2 and DENV3 were isolated from the patients during the largest dengue outbreak in

Suriname in 2005.

The Surinamese Bureau of Public Health (Bureau Openbare Gezondheidzorg – BOG), under the Ministry of Health, is the national institute in charge of dengue surveillance and control. In 2001, epidemiological surveillance of dengue began by having hospitals and regional clinics report all cases of DF and DHF to the BOG. The reported dengue cases includes demographic (age, gender, ethnicity) and clinical (date of onset, hospital admission and discharge and laboratory test results) information. The surveillance and vector control practices of the BOG are essential in curbing the disease. Additionally, the application of innovative spatio-temporal visualization techniques, such as geographic information system (GIS), is becoming more commonplace in efforts to implement better surveillance and control techniques against vector-related diseases. GIS software can be utilized not only to identify if cases of DF and DHF cluster, but also to visualize how they emerge overtime and if those cases are also related to environmental or demographic drivers (10). Thus, integrating GIS technology to improve disease surveillance enables more efficient public health planning by identifying target areas in which more resources should be invested.

The objective of this study is to complete the initial steps for the integration of epidemiological data into a spatial visualization platform to map dengue risk areas in

Suriname. The findings will be used to inform an interdisciplinary study that integrates epidemiological data to guide environmental health research in Suriname.

48

METHODS

Data Sources & Case Classification

Dengue is a reportable disease in Suriname. Since 2001, all suspected cases of DF reported to the BOG have been assigned a unique reference number and complied into a database that includes demographic and clinical information and an address coded according to the Surinamese General Bureau of Statistics (ABS) codebook. Researchers did not have access to these personal identifiers. The database was screened and DF cases were selected for analysis based on positive serology test or on virus/nucleic acid isolation, respectively. DHF cases were selected based on clinical diagnosis (according to the WHO criteria) when no diagnostic test was recorded. However, due to the generic nature of DF symptoms, cases of DF based solely on clinical diagnosis were classified as suspected and excluded from the analysis. In this study, cases of DHF are not counted as cases of DF therefore the term ‘dengue cases’ refers to both cases of DF and DHF. The study protocol was approval by the Tulane University School of Public Health and

Tropical Medicine Institutional Review Board (IRB) and the Ministry of Health

Suriname.

Data analysis: The annual frequency of total cases of dengue, DF and DHF for the entire country from 2001 through 2012 were calculated. Annual dengue severity was measured as the number of DHF cases compared to total cases of dengue. Frequency distributions were also calculated by age group, gender, ethnicity and district. These categorical variables were used to compare the frequency of DF v. DHF by Chi-square test using

SPSS® Statistics v 19. Suriname is divided into 10 subdivisions also known as districts.

Annualized dengue incidence rates were obtained using mid-year population estimates

49 based on 2004 and 2012 ABS census information. District incidence rates were obtained using total dengue cases and district population estimates from 2002 through 2012 (no case by district data is available for 2001).

Spatial analysis

To identify global patterns of dengue cases we calculated Moran’s I as a global index for spatial autocorrelation using GeoDa™ (http://geodacenter.asu.edu). Cases were weight according to distance and the number of nearest neighborhoods was set to 4. For the randomization, 999 permutations were selected. To detect local clusters of probable cases of dengue, we deployed the Kuldorff’s spatial scan statistics using SatScan software version 9.1.1. (www.satscan.org). We used a retrospective spatial Poisson probability model for discrete data to detect significant high rate (hotspots) and low rate (coldspot) clusters with geographic overlap. The number of replications was set to 999 times with the maximum spatial cluster size set at 20%. The most likely clusters (primary clusters) and other clusters (secondary clusters) that were detected within Suriname were visualized using ArcGIS v10.2 (http://www.esri.com). The relative risk (rr), log- likelihood ratio (LLR) and p-value for each displayed cluster was recorded.

Neighborhood borders within the Paramaribo district were drawn and exported as a polygon shapefile and the district’s streets were added in as a layer. There were no neighborhood borders available for other districts. The address codes provided in the

BOG dengue database allowed each case of dengue to be localized to a street within a resort. To protect patient privacy, the BOG address codes provide the street name and the two intersecting streets for each case of dengue but no house number. Therefore, the cases of dengue were mapped half-way on the street of the two intersecting streets.

50

RESULTS

There were a total of 2393 cases of dengue out of which 366 progressed into DHF reported to the BOG between 2001 through 2012. Figure 1 illustrates a temporal pattern of emergence in cases of dengue since 2001 with a higher proportion of annual cases during the months of August to October and December to February. In addition, the incidence of dengue peaked in during the 2005 epidemic with 16.4 cases per 10,000 people (Table 1). To date, 2005 has the greatest incidence of dengue cases but only 11% of all cases progressed into DHF. During the 2009 epidemic, 38% of cases progressed into DHF but the number of people affected was much lower at 5.1 cases per 10,000

(Table 1). Therefore, the 2009 epidemic was much worse in severity compared to 2005.

The frequency of dengue cases, DF and DHF stratified by age group, gender and ethnicity are shown in Table 2. There is no variation in severity between genders.

Severity is higher in certain age groups (>1, 1-14 and 45-59) and ethnicities (Hindustani and Javanese). Chi-square analysis indicates no significant difference in cases of DF and

DHF by gender. However, the difference between DF and DHF cases was statistically significant for age group and ethnicity (p<0.05) (Table 2) indicating certain demographic factors influence the progression into DHF within our study population.

Global and local clusters

The frequency, incidence and severity of dengue cases from 2002 to 2012 are shown in

Table 3. The districts of Nickerie and Coronie and Marowijne and Brokopondo were grouped together for chi-square analysis. Frequency, incidence and severity were highest in the two most urbanized districts of Paramaribo and Wanica (Table 3). Furthermore,

Global Moran’s I suggest that there is spatial autocorrelation, or clustering, of dengue

51 cases (Moran’s I = 0.052, z > 1.96, p = 0.001) at the resort level within Suriname. To identify were the population had the highest relative risk (rr) of dengue cases in

Suriname, we identified the two most likely (highest likelihood ratio-LLR) and statistically significant (p<0.05) clusters (also referred to as hotspots) by resort between

2002 to 2012 with a maximum spatial cluster size of the total population set at 20%

(Table 4). Additionally, the two lowest relative risk (rr) clusters (or coldspots) were identified using the same procedure. The most likely hotspots and coldspots were classified as primary while all other identified clusters were classified as secondary. The relative risk (rr), log likelihood ratio (LLR) and p-value are displayed on Table 4. The centroid and radius of the clusters was used to visualize the clusters in Suriname using

ArcGIS v10.2. Both primary and secondary hotspots and coldspots were located in resorts within districts Paramaribo and Wanica (Figure 2). The resorts within the hotspots included Blauwgrond, Rainville and Munder located on the northern side of district

Paramaribo while Koewarasan and Latour made up the coldspots in a more southern region of this district. The resorts that are closer to each other tend to have similar annualized incidence rates. Neighborhood boundaries, streets and cases of dengue within the two primary clusters better illustrate the dispersion of cases in the identified high and low epidemiological rate areas (Figure 2).

DISCUSSION

In Suriname, the number of cases of dengue peak following the end of the long and short rainy season in August and February, respectively. Since 2001, dengue epidemics have emerged every 3-4 years, which has also been documented for other countries (11). The cyclical emergence of dengue is influenced by climate conditions and climatological

52 phenomena such as El Nino-Southern Oscillation as is evident by the large 2005 epidemic in Suriname (12 – 14). In addition to an increase in dengue cases, frequency analysis indicated that the severity of dengue has increased since 2001; most notably during the smaller but more severe epidemics of 2009 and 2012. An increase in severity is expected because since 2001, all four serotypes of dengue have circulated or co- circulated within the population (Figure 1). Unfortunately, there is no record of the dengue serotype circulating during the 2009 epidemic. Such information is important because the introduction of a new serotype into a non-naïve population increases the incidence and number of DHF cases during a dengue outbreak (15). Overtime the immune status of the population changes as more people become infected with one of the four dengue viruses increasing their susceptibility of developing DHF. An increase in cases of DHF strains the healthcare system as more people require hospitalization during a dengue epidemic. This was evident during the 2012 epidemic which exhausted national healthcare capabilities and prompted the Ministry of Health to open up an emergency hospital. Additionally, the serotypes (16, 17) the sequential order in which the serotypes are encountered within a population (18 – 20) and the time elapsed between primary and secondary infections (15, 21) are associated with increased pathogenicity and the development of DHF. Therefore it can be expected that as more people become exposed to the dengue virus the disease dynamics within the Surinamese population will change.

These changes require a better characterization of the demographic factors that influence the development of DHF in Suriname.

Our results indicate that certain ethnicities (Javanese and Caucasians) and age groups (1-

14) occur in a disproportionate number of cases of DHF. These demographic

53 characteristics have been identified as contributing risk factors for the development of

DHF (22 – 24). The Surinamese population is characterized by an ethnic diversity profile unlike any other country in South America. Therefore, these findings are the first step to examine demographic risk factors that contribute to the development of DHF in within such a unique demographic composition.

Innovative spatio-temporal visualization techniques, such as GIS, are becoming more commonplace in efforts to implement better surveillance and control techniques against dengue fever. These techniques have been used to identify DF and DHF spatial clustering, where clusters emerge and how demographic factors change over space and time (10). Here, the results of the global spatial analysis revealed that there was significant clustering of cases of dengue in Suriname. The identified hotspots are located in areas where there is a historically higher relative risk of dengue compared to the rest of the population. Similarly, the coldspots are located in areas where the relative risk of dengue is much lower.

The data are limited to cases of dengue from hospitals and clinics in Suriname reported to the BOG between 2001 and 2012. Asymptomatic or flu-like DF cases are frequently underreported potentially causing dengue incidence to be underestimated. Additionally, serology tests have a low specificity and sensitivity to acute dengue infection or distinguish between primary and secondary infections of dengue (25, 26). Therefore, data do not provide information on previous dengue infections, which are an important determinant in the development of DHF. However, the data analysis does provide epidemiological knowledge that can be used as a tool to design resource- efficient environmental and entomological studies. Such studies would provide more information

54 about the local dispersion of the Aedes aegypti mosquito and dengue virus transmission.

This study represents the first steps to identify demographic and environmental factors that influence the transmission of dengue fever and the development of DHF in

Suriname.

55

REFERENCES

1. World Health Organization. Dengue guidelines for diagnosis, treatment, prevention

and control New Edition. Geneva: WHO Press; 2009.

2. San Martin JL, Brathwaite O, Zambrano B, Solorzano JO, Bouckenooghe A, Dayan

GH et al. The epidemiology of dengue in the Americas over the last three decades: A

worrisome reality. Am J Trop Med Hyg 2010; 82: 128 – 35.

3. Brathwaite DO, San Martin JL, Montoya RH, del Diego J, Zambrano B, Dayan GH.

The History of Dengue Outbreaks in the Americas. Am J Trop Med Hyg 2012; 87:

584 – 98.

4. Pinheiro FP. Dengue in the Americas 1980 – 1987. Epidemiological Bulletin 1989;

10: 1 – 8.

5. World Health Organization. DengueNet [Internet]. Washington (DC): World Health

Organization [cited 2013 Feb 2]. Available from

http://apps.who.int/globalatlas/default.asp

6. Guzman MG, Halstead SB, Artsob H, Buchy P, Farrar J, Gubler DJ et al. Dengue: A

continuing global threat. Nat Rev Microbiol 2010; 8: S7 – 16.

7. Wilson ME, Chen LH. (2002). Dengue in the Americas. Dengue Bulletin, 26, 44-61.

8. Pan American Health Organization. Number of reported cases of dengue and dengue

hemorrhagic fever (DHF), Region of the Americas (by country and subregion)

[Internet]. Washington (DC): Regional Offices for the Americas of the World Health

56

Organization [cited 2014 Feb 2]. Available from:

http://amro.who.int/english/ad/dpc/cd/dengue-cases-2008.htm

9. Caribbean Epidemiology Centre. Report on communicable diseases for weeks 1 - 8,

2012. CAREC Surveillance Report 2012; 32: 1 – 28.

10. Duncombe J, Clements A, Hu W, Weinstein P, Ritchie S, Espino FE. Geographical

information systems for dengue surveillance. Am J Trop Med Hyg 2012; 86: 753 –

55.

11. Halide H, Ridd P. A predictive model for Dengue Hemorrhagic Fever epidemics. Int

J Environ Heal R 2008; 18: 253 – 65.

12. Kovats RS, Bouma MJ, Hajat S, Worrall E, Haines A. El Niṅo and health. Lancet

2003; 362: 1481 – 1489.

13. Gagnon AS, Bush ABG, Smoyer-Tomic KE. Dengue epidemics and the El Nino

Southern Oscillation. Clim Res 2001; 19: 35 – 43.

14. Hopp MJ, Foley JA. Wolrdwide fluctuations in dengue fever cases related to climate

variability. Clim Res 2003; 25: 85 – 94.

15. Diaz-Quijano FA, Waldman EA. Factors associated with dengue mortality in Latin

America and the Caribbean, 1995-2009: An ecological study. Am J Trop Med Hyg

2012; 86: 328 – 34.

16. Rico-Hesse R, Harrison LM, Salas RA, Tovar D, Nisalak A, Ramos C et al. Origins

of dengue type 2 viruses associated with increased pathogenicity in the Americas.

Virology 1997; 230: 244 – 51.

57

17. Vaughn DW, Green S, Kalayanarooj S, Innis BL, Nimmannitya S, Suntayakorn S et

al.(2000). Dengue viremia titer, antibody response pattern, and virus serotype

correlate with disease severity. J Infect Dis 2000; 181: 2 – 9.

18. Burke DS, Nisalak A, Johnson DE, Scott RM. A prospective study of dengue

infections in Bangkok. Am J Trop Med Hyg 1988; 38: 172 – 80.

19. Thein S, Aung MM, Shwe TN, Aye M, Zaw A, Aye K et al. Risk factors in dengue

shock syndrome. Am J Trop Med Hyg 1997; 56: 566 – 72.

20. Nisalak A, Endy TP, Nimmannitya S, Kalayanarooj S, Thisayakorn U, Scott RM et

al. Serotype-specific dengue virus circulation and dengue disease in Bangkok,

Thailand from 1973 to 1999. Am J Trop Med Hyg 2003; 68: 191 – 202.

21. Guzman MG, Kouri G, Valdes L, Bravo J, Vazquez S, Halstead SB. Enhanced

severity of secondary dengue-2 infections: Death rates in 1981 and 1997 Cuban

outbreaks. Rev Panam Salud Publ 2002; 11: 223 – 227.

22. de la Sierra B, Kouri G, Guzman MG. Race: A risk factor for dengue hemorrhagic

fever. Arch Virol 2007; 152: 533 – 42.

23. Guzman MG, Kouri G, Bravo J, Valdes L, Vazquez S, Halstead SB. Effect of age on

outcome of secondary dengue 2 infections. Int J Infect Dis 2002; 6: 118 – 24.

24. Halstead SB, Lan NT, Myint TT, Shwe TN, Nisalak A, Kalyanarooj S et al. Dengue

hemorrhagic fever in infants: Research opportunities ignored. Emerg Infect Dis

2002; 8: 1474 – 89.

58

25. Chaterij S, Allen JC, Chow A, Leo YS, Ooi EE. Evaluation of the NS1 rapid test and

the WHO dengue classification schemes for use as bedside diagnosis of acute dengue

fever in adults. Am J Trop Med Hyg 2011; 84: 224 – 228.

26. Souza VA, Tateno AF, Oliveira RR, Dominques RB, Araujo ES, Kuster GW, Pannuti

CS. Sensitivity and specificity of three ELISA-based assays for discriminating

primary from secondary acute dengue virus infection. J Clin Virol 2007; 39: 230 –

33.

59

ACKNOWLEDGEMENTS

This study is in collaboration with the Bureau of Public Health Suriname. A special thanks to Dr. Beatrix Jubithana, epidemiologist of the Bureau of Public Health, for her assistance with this project.

60

TABLES

Table 1: Annualized cases of dengue, DF and DHF

Year Dengue DF DHF Incidence* Severity (%) Serotype N (%) N (%) N (%) Total 2399 (100) 2033 (100) 366 (100) 3.94 15.2 2001 65 (2.7) 56 (2.7) 9 (2.4) 1.4 13.8 3 2002 81 (3.4) 77 (3.8) 4 (1.1) 1.7 4.9 3 2003 22 (0.9) 22 (1.1) 0 (-) 0.5 0 2 2004 56 (2.3) 53 (2.6) 3 (0.8) 1.1 5.3 3 2005 816 (34.0) 723 (35.6) 93 (25.4) 16.4 11.4 2, 3 2006 182 (7.6) 166 (8.2) 16 (4.4) 3.6 8.8 2 2007 45 (1.9) 38 (1.9) 7 (1.9) 0.9 15.5 2 2008 91 (3.8) 70 (3.4) 21 (5.7) 1.4 23.1 -** 2009 292 (12.2) 179 (8.8) 113 (30.9) 5.1 38.7 -** 2010 141 (5.9) 121 (5.9) 20 (5.5) 2.5 14.2 1, 2, 4 2011 157 (6.5) 135 (6.6) 22 (6.0) 2.8 14.0 2, 4 2012 451 (18.8) 393 (19.3) 58 (19.3) 7.7 12.9 1, 2, 4 *per 10,000 people **No serotype information available

Table 2: Frequency and severity of dengue, DF and DHF cases by demographics

Dengue DF DHF Severity

N (%) N (%) (%) (%) Total 2393 (100) 2027 (100) 366 (100) 15.3 Age*

>1 85 (3.5) 70 (3.4) 15 (4.1) 17.6 1-14 675 (28.2) 552 (27.2) 123 (33.6) 18.2 15-29 679 (28.4) 598 (29.5) 81 (22.1) 11.9 30-44 504 (21.1) 435 (21.5) 69 (18.8) 13.7 45-59 302 (12.6) 242 (11.9) 60 (16.4) 19.9 60+ 144 (6.0) 127 (6.3) 17 (4.6) 11.8 Gender

Male 1357 (56.7) 1147 (56.6) 210 (57.4) 15.5 Female 1028 (42.9) 873 (43.1) 155 (42.3) 15.1 Ethnicity*

Creole 295 (12.3) 256 (12.6) 39 (10.6) 13.2 Hindustani 922 (38.5) 788 (38.8) 134 (36.6) 14.5 Javanese 389 (16.2) 314 (15.5) 75 (20.5) 19.3

61

Chinese 259 (10.8) 226 (11.1) 33 (9.0) 12.7 European 42 (1.7) 30 (1.5) 12 (3.3) 28.6 Indigenous 58 (2.4) 51 (2.5) 7 (1.9) 12.1 Maroon 59 (2.5) 49 (2.4) 10 (2.7) 16.9 *Chi-square test p<0.05

Table 3: Frequency, severity and incidence of dengue, DF and DHF cases by district

Dengue DF DHF Severity

N (%) N (%) N (%) (%) Incidence** Total* 2393 (100) 2027 (100) 366 (100) 15.3 3.9

District***

Paramaribo 1286 (57.7) 1085 (57.6) 201 (58.4) 15.6 4.7

Wanica 481 (21.6) 388 (20.6) 93 (27.0) 19.3 4.8

Para 64 (2.9) 54 (2.8) 10 (2.9) 15.6 3.0

Commewijne 109 (4.9) 92 (4.9) 17 (4.9) 15.6 3.8

Saramaca 68 (3.0) 57 (3.0) 11 (3.2) 16.2 3.8 Nickerie/ 176 (7.9) 172 (9.2) 4 (1.2) 2.3 4.0 Coronie Marowijne/ 27 (1.2) 23 (1.2) 4 (1.2) 14.8 0.8 Brokopondo Sipaliwi 17 (0.7) 13 (0.7) 4 (1.2) 23.5 0.4 *No information by district available for 2001 **per 10,000 person-year ***Chi-square test p<0.05

Table 4: Cluster analysis of dengue cases in Suriname

Relative risk Likelihood Cluster Rate Cluster center (rr) ratio (LLR) p-value

Primary High Blauwgrond 26.03 150.6 0.000 Secondary High Munder 19.9 93.4 0.000 Primary Low Latour 0.18 110.9 0.000 Secondary Low Koewarasan 0.41 38.6 0.000

62

FIGURES

Figure 1:

63

Figure 2:

64

FIGURE LEGENDS

Figure 1: Annual frequency of dengue cases by month. Dengue cases reported between

January 2001 and December 2012 were stratified by month. There is a temporal pattern of emergence of dengue cases with the number of cases peaking after both the short and long rainy season in February and October, respectively. The biggest dengue outbreak occurred in 2005.

Figure 2: Spatial distribution of dengue cases for Paramaribo & Wanica, Suriname. Four clusters, one primary (red circle) and one secondary (orange circle) hotspot and one primary (blue circle) and one secondary (green circle) coldspots were detected within

Paramaribo and Wanica districts based on dengue cases distribution from 2002 to 2012.

Only dengue cases within the resorts of the primary hotspot (Blauwgrond and Rainville) and primary coldspot (Latour) are displayed on the map. Dots indicate a single case of dengue.

65

5. Characterization of Aedes aegypti breeding sites in Paramaribo, Suriname: a

comparison between a high and a low rate cluster of cases of dengue

Submission for publication and authorship designation pending authors’

review to the American Journal of Tropical Medicine

HAMER AND OTHERS Aedes aegypti breeding sites in Paramaribo, Suriname

Characterization of Aedes aegypti breeding sites in Paramaribo, Suriname: a comparison between a high and a low rate cluster of cases of dengue

Diana Hamer1, Hélène Hiwat2, Justin K. Davis3, Richard A. Oberhelman4, Arti Shankar5, Eric Svendsen1, Dawn M. Wesson3, Maureen Y. Lichtveld1

1Department of Global Environmental Health Sciences, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA; 2Department of Entomology, Bureau of Public Health, Paramaribo, Suriname; 3Department of Tropical Medicine, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA; 4Department of Global Community Health and Behavioral Sciences, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA; 5Department of Biostatistics and Bioinformatics, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA.

Abstract:

Address correspondence to Diana Hamer, Department of Global Environmental Health Sciences, Tulane University School of Public Health and Tropical Medicine, 1440 Canal St, Suite 2100, New Orleans, LA 70112. E-mail: [email protected]

66

Introduction

Aedes aegypti is the primary vector of dengue virus (DENV) transmission to humans in the Americas. It has been postulated that the vector was first introduced into the Western

Hemisphere between the 15th and 17th century from West Africa and subsequently spread to most countries in the Americas. During the mid-twentieth century, the vector was successfully eradicated following aggressive, Pan-American eradication campaigns

(Pinto Severo, 1955) but reinfestation occurred after these campaigns were halted (“The

Feasibility of eradicating Aedes aegypti in the Americas,” 1997). Ae. aegypti resurgence was possible due to its evolutionary adaptation to the urban environment. Urbanization has increased the number of potential Ae. aegypti breeding sites especially in moist tropical regions where climatological conditions enhance and determine its geographic distribution and range (Hopp & Foley, 2003). Dengue fever outbreaks (DF) typically follow the rainy season. Seasonal variations, such as temperature, relative humidity and average precipitation can increase the number of Ae. aegypti available breeding sites influencing the transmission of DF (Altizer et al., 2006). Particularly in densely populated cities where there is an excess availability of man-made water-holding containers that can support the emergence of additional adult mosquitos (Diaz-Quijano et al., 2008; Schreiber, 2001). Such containers intensify the dengue virus - Ae. aegypti - host interaction in endemic regions across Latin America and Caribbean. Over the past three decades, dengue infections and mortality in these regions have significantly increased

(Diaz Quijano & Alves Waldman, 2012). The factors that contribute to the availability of containers and the transmission of dengue in these regions, however, remain understudied, particularly in Suriname.

67

Entomological indices have been used as epidemiologic indicators of dengue transmission. Although useful, immature mosquito indices, such as the Breteau Index

(BI), the House Index (HI), and the Container Index (CI), however, do not inform the risk of transmission as well as adult mosquito densities, such as adults per persons, do (Focks

& Chadee, 1997; PAHO, 1994; Sanchez et al., 2006; Morrison et al., 2008). Surveillance often focuses on immature surveys due to the difficulty in collecting adult mosquitos

(Eisen & Lozano-Fuentes, 2009). Ae. aegypti larval stages, however, are hard to collect and quantify during surveys. In fact, pupae surveys are more effective at assessing dengue transmission because it is easier to count absolute numbers and they have a lower mortality compared to earlier stages of Ae. aegypti (Focks & Chadee, 1997; Knox et al.,

2010)). Therefore, indices based on pupal surveys better reflect the risk of dengue transmission in a community (Focks et al., 2000). Additionally, pupae surveys are used to identify the most productive types of containers that are epidemiologically important for

Ae. aegypti reduction and control activities (Focks & Alexander 2006, Barrera et. al.

2006).

In Suriname, vector control measures include perifocal spraying for adults and Bacillus thuringiensis israelensis (Bti) for larvae. A previous study based on larval indices demonstrated that certain containers, water storage containers including water drums and buckets, were the most productive breeding sites for Aedes mosquitos (Hiwat et al 2013).

However, additional entomologic surveillance data based on pupal and adult Ae. aegypti surveys are necessary to further elucidate the environmental factors that influence the transmission of dengue. As in many Caribbean countries, dengue is endemic and on the rise in Suriname, with more widespread and more severe outbreaks documented in recent

68 years (WHO 2013). Recently chikungunya, alos transmitted by Ae. aegypti, was introduced into the country. Therefore, entomological and environmental research is necessary to develop efficient and effective vector control strategies to reduce Ae. aegypti populations and mitigate the transmission of disease.

Previously, we integrated all probable cases of dengue fever reported to the Surinamese

Bureau of Public Health (BOG) between 2002 and 2012 into a spatial visualization platform and identified two clusters of disease in Paramaribo, the capital city. The northern cluster (or high rate cluster –HRC) had an increased incidence of dengue cases compared to the rest of Suriname. Similarly, the southern cluster (or low rate cluster –

LRC) had the lowest incidence of dengue cases (Hamer & Lichtveld 2014). The two clusters indicate that even within the same city, each neighborhood or community may face different dengue transmission dynamics that are heavily influenced by the complex interaction of the environment, the virus, the human host, and the mosquito.

To determine what factors influence the transmission of dengue in Suriname, we performed detailed entomological and environmental surveys in a sample of houses within both clusters (HRC and LRC). Ae. aegypti pupal surveys were performed to identify and compare breeding sites within and between clusters as well as elucidate which environmental factors support a higher number of dengue cases within the HRC or mitigate cases in the LRC. To our knowledge, this is the first attempt to use epidemiological information to guide entomological and environmental studies on dengue transmission in Suriname.

69

Methods

Data Collection

Suriname has a short rain season (SRS) from December to February and a long rain season (LRS) from April to August. For the purpose of this research, surveys were conducted before the SRS (pre-SRS) (November 2013) and during the SRS (February

2014) in northern and southern neighborhoods of Paramaribo (5°52’N, 55°10’W). Mean monthly temperature and total monthly rainfall for November and February were

28.03°C and 89.20mm, and 27.70°C and 195.40mm, respectively. Paramaribo is an urban district with a total population of 240,924 (Surinamese Census 2012). The city is divided into 12 neighborhoods (known as resorts). Resort characteristics (population demographics, number of households, access to electricity, and piped water) were obtained from the General Bureau for Statistics (ABS 2012). From 2001 to 2012, there have been 1,088 cases of dengue fever, 201 cases of dengue hemorrhagic fever / dengue shock syndrome (DHF/DSS), and 26 dengue-related deaths reported in Paramaribo. All cases of dengue were mapped and incorporated into a geographic information system

(GIS). Areas with high and low rate of dengue cases (HRC and LRC respectively) were identified following cluster analysis as described in Hamer & Lichtveld 2014. Within each cluster (HRC and LRC), streets were randomly selected for surveying. For each selected street, every other house was visited to conduct the surveys. Under the guidelines of the Suriname the BOG, a BOG inspector requested adult verbal consent to survey the surrounding area of the house but the inside, upper levels or roof gutters of the house were not inspected. When consent was not available or the house was not occupied, the next house was selected for surveying. Location was determined using a

70 global positioning system (GPS). The access to piped-water was defined according to the

ABS as houses with indoor water plumbing or outdoor water plumbing within 200 meters of the house. A standardized classification scale developed by the BOG was used to assess the condition of the house and yard surveyed. ‘Good’ was given to houses in proper condition that were well-kept with no damage to the structure. ‘Medium’ was given to houses that were not well-kept and had some damage to the structure. While

‘bad’’ was given to houses that were in unsanitary living condition and had severe damage to the structure. Similarly, yards were rated on the same scale. A classification of

‘good’ was given to well-kept yards that did not have accumulated trash. A classification of ‘medium’ was given to yards that were not well-kept and / or had accumulated trash.

While a ‘bad’ was given to neglected yards that had an accumulation of unsanitary trash.

Additionally, during the February surveys only, the BOG inspector asked consenting adult how many people lived in the house, how many years they lived in the surveyed house and if any children lived in the house.

All outdoor containers, dry and water-holding containers, on the premise were inspected.

The water holding containers were classified into 8 categories: garden attributes (plant pots, plant plates, wheel barrels), water storing (buckets, drums), discarded household items (furniture, appliances, tires), garbage (discarded cans, food containers), household items (pots, pans, coolers), recreation (swimming pool, boat, fountain), structural components (septic tank, well), and natural habits (tree holes, leafs). For each water- holding container, descriptive characteristics were recorded. These characteristics included usage status (in-use or not in-use), material (plastic, glass, metal or other), capacity (<1L, 1 – 15L, 15 – 50L or >50L), circumference (<5cm, 5 – 30cm, 30 – 60cm

71 or >60 cm), insecticide / larvicide presence, cover status, location (shaded, partially shaded, sun), and biological control (frogs, turtles, fish) presence. The presence of larva and density of larva (0 – 10, 11 – 50, 50 – 100, >100) was recorded. The total pupae numbers were counted and recorded when present as a measure of adult productivity

(Focks & Chadee 1997; Knox et al 2010). During the February surveys, pupae were collected using a pipette or a dipper for larger containers. Pupae were taken back to the laboratory for adult mosquito emergence and species identification. The Ae. aegypti male to female ratio was calculated. It was not feasible to collect all pupae especially in larger containers that could not be drained due to the household necessity for clean water.

However, for each container, the percentage of pupae identified as Ae. aegypti was used as an estimation of total Ae. aegypti pupae present.

This study was approved by the Tulane University School of Public Health and Tropical

Medicine Institutional Review Boards and the Ministry of Health Suriname.

Data Analysis

Household characteristics and piped water availability between the two clusters and between survey months (November and February) were transformed into contingency tables (p<0.05). Positive identification of pupae were used to calculate the container index, household index, and Breteau index. The most productive container category was identified using the dispersion index (Focks and Alexander 2006) for both clusters and survey months. Furthermore, Fisher’s exact test of association for count data (and chi- square analysis when the exact method was computationally infeasible) was used to test for association of Ae. aegypti pupae presence among container categories and container characteristics (significance p<0.0045 based on Bonferroni correction). A post-hoc test

72 was conducted to identify differences between category and container characteristics and estimate odds ratios.

Results

The surveys took place within the previously identified high rate cluster (HRC) and low rate clusters (LRC) (Figure 1). The HRC covered most of resort Blauwgrond. The LRC centered in resort Latour and covered the south of resort Flora and north east of resort

Pontbuiten. The percentage of households with access to electricity, piped water, population, population density, and ethnic composition are summarized in Table 1. A total of 536 houses, 242 were surveyed during the pre-SRS with 136 houses located within the HRC and 106 within the LRC (Figure 2). Within the HRC, all houses surveyed had access to piped water while 26 houses (36%) did not in the LRC. Furthermore, within the LRC, houses without access to piped water were more frequently found to be positive for pupae (p<0.05) than houses with access during the pre-SRS. During both seasons, there was an association between house condition and presence of pupae in the HRC but this association is not seen in the LRC. During the SRS, a total of 294 houses were surveyed: 179 in the HRC and 115 in the LRC. Similarly to the pre-SRS, almost all houses in the HRC had access to piped water while 24 houses (or 26%) in the LRC did not. During the SRS, fewer houses were positive for pupae in the LRC compared to the

HRC. However, fewer houses were positive for Ae. aegytpi pupae in the LRC, but not the

HRC, compared to the same clusters during the pre-SRS. Moreover, the frequency of pupae positive houses in the LRC was not statistically associated with access to piped water in the SRS. We used the number of people per household as a proxy for population density within the clusters surveyed. The HRC has on average fewer people per

73 household compared to the LRC (p<0.05) and there was no difference in the average number of years people lived in the house surveyed between clusters. Households with children were more likely in the LRC (p<0.05).

There was no difference in the number of houses positive for larva breeding sites between the HRC and LRC during either the pre-SRS or SRS (p=0.570 and p=0.352, respectively). The number of houses in the LRC positive for pupae is higher compared to the HRC (p<0.05) during the pre-SRS. However, there is no difference between clusters and the presence of pupae during the SRS (Table 2). To gain further insight into breeding sites (pupae presence) we calculated common entomological indices (Table 3). These indices, container index (CI), house index (HI), and Breteau index (BI) are useful to equate differences in Aedes aegypti presence between seasons and clusters. All three indices were higher pre-SRS. The HRC had lower CI, HI, and BI during the pre-SRS but these indices were higher for the HRC during the SRS compared to the LRC. The average number of pupae per household was significantly higher during the pre-SRS (p<0.05).

During the pre-SRS, the average number of pupae per household was higher in the LRC

(p<0.05) but not during the SRS where no statistical difference was detected between the

HRC and LRC (p=0.308). We calculated the dispersion index to assess which where the most productive breeding sites of Aedes aegytpi. The dispersion index (DI) was similar for both seasons but, during the pre-SRS, was lowest in the LRC where most breeding sites were encountered in water storage containers (Figure 2A). Water storage containers are also the most frequently encountered followed by household use containers in both clusters during both seasons (Table 4). Furthermore, compared to all other categories, the most productive breeding sites were water storage containers . A high DI indicates that

74 breeding sites are more evenly dispersed among a larger category of containers. We found that breeding sites in the HRC pre-SRS dispersed over a few container categories; however, in the SRS Aedes aegypti dispersion became more concentrated in the HRC and more dispersed in the LRC .

A chi-square analysis was used to determine whether certain container categories or characteristics are associated with presence of Ae. aegypti pupae. Among categories, water-storage containers were the most frequent breeding sites and discarded household items had the highest odds ratio (OR) of being positive for pupae followed by garden attributes (Table 5). However, container category contributes minimally to explain abundance of pupae (R2=0.15). Certain characteristics were associated with higher odds for the presence of pupae such as the container usage, disposability, cover status, the material, the circumference, and the water source (Table 5). An increased circumference was associated with an increase in the odds of pupae presence.

Discussion

The results of this study indicate that there are seasonal differences in Aedes aegypti breeding site availability for two epidemiological clusters in Paramaribo. The environmental surveys were conducted before (November) and after (February) the SRS.

Aedes aegypti population dynamics are influenced by climatic factors (Hopp & Foley

2003). Warmer temperatures and heavy precipitation hasten and lengthen dengue fever outbreaks (Halide & Ridd, 2008; Parker and Holman 2012; Fan et al 2013). In Suriname, cases of dengue peak in August and February after the long and the short rainy season, respectively (Hamer & Lichtveld, 2014) and possible association between dengue

75 outbreaks and the El Nino Southern Oscillation have been documented (Gagnon et al

2001; Kovats et al 2003). Furthermore, we compared a cluster where there is a historically higher relative risk of dengue (HRC) to a cluster with much lower relative risk (LRC) (Hamer & Lichtveld 2014). The high ERC is located within the Blauwgrond neighborhood. This neighborhood has a lower population density and almost all houses surveyed had available piped water (Table 1). Most roads within Blauwgrond are paved, garbage collection services are available and the neighborhood is relatively affluent (field observations). In contrast, the neighborhoods that compromised the LRC have a much higher population density and not everyone has access to piped water, particularly in

Pontbuiten. The roads in these neighborhoods were often unpaved and there was no regular garbage collection service. However, the relative risk of dengue in the LRC was much lower even though lack of piped water and high population density contribute to the occurrence of dengue outbreaks (Cummings et al 2004; Braga et al 2010; Schmidt et al 2011). Interestingly, the two clusters differ in demographic composition. The percentage of South East Asian decent people is much higher in the HRC compared to the LRC that has a higher percentage of maroons. Presumed risk factors for severe dengue include a higher genetic susceptibility in South East Asians and Caucasians compared to Blacks (Chiewslip, et al 1981, Bravo et al 1987; de la Sierra et al 2006). The demographic composition of the clusters differs in that about 40% of the neighborhood

(Blauwgrond) composing the HRC is of Southeast Asian (Hindustani or Javanese) descent while the population of the two largest neighborhoods (Pontbuiten and Latour) within the LRC are predominantly (about 70%) of African (Maroon or Creole) descent.

Therefore, we can infer that the higher number of recorded dengue cases within the HRC

76 could be due to having a larger percent of the population that is genetically more susceptible to severe dengue. Furthermore, the population within the HRC (neighborhood

Blauwgrond) is relatively more affluent, has more access to piped water, and paved roads compared to the LRC. These factors potentially influence access to health and health seeking behaviors contributing to a higher number of recorded cases of dengue. The differences in demographic composition, socio-economic status, and infrastructure

(paved roads, piped water) could in part explain the finding that the less densely populated neighborhood within the HRC presents a higher number of reported cases of dengue compared to much more densely populated neighborhoods within the LRC.

During the SRS entomological surveys, we found a higher number of pupae positive houses and containers for the LRC. The pupal indices used to estimate risk of dengue transmission (Sanchez et al 2006; PAHO 1994; Focks et al 2000) are useful to compare the presence of Ae. aegytpi breeding sites between the two clusters but fail to explain the high concentration of dengue reported cases in the HRC. However, the findings are reversed during the SRS in which the pupae indices are higher in the HRC compared to the LRC. An explanation for these results is that during the drier month of November, people in the LRC that have no or intermittent access to piped water, depend more on water-storage containers to fulfill their needs. However, the heavy precipitations in

February constantly replenish water-storage containers limiting the number of containers needed (Figure 2A). We observed a similar change in behavior in the HRC between seasons. During the SRS, people tend less to their yards due to the heavy rains and small water storage containers such as buckets become opportunistic breeding sites.

Furthermore, we expected to find more pupae positive containers during the SRS

77 compared to the pre-SRS because of the increase in dengue cases after the rain season.

This expectation was observed when comparing larva positive containers (data not shown) but not for pupae positive containers regardless of the higher number of water holding containers identified during the SRS (Figure 4). An explanation for the lower number of pupae identified during the SRS is that heavy rains can lead to short-term breeding sites. Thus, rather than becoming long-term, focal breeding sites more containers have the potential of supporting less pupae for a brief period but at a higher rate during the SRS.

The previously identified clusters used in this analysis are based on historic cases of dengue reported to the Suriname Bureau of Public Health from 2001 to 2012 (Hamer &

Lichtveld, 2014). However, we were not able to obtain the number of dengue cases reported for the areas inspected during the survey months due to the time between dengue cases are recognized in clinics and hospitals in Suriname, the date the health practitioner sends the case information to the BOG and BOG reports it. Furthermore, we designed our sampling based on the number of streets within each cluster. Ideally, the spatial distribution of households within the clusters would have enhanced sampling and spatial analysis however, such maps are not yet available for Paramaribo. The analysis was based on solely on pupae counts and no adult mosquitos were collected. Adult mosquito surveillance, however, is a better proxy for dengue transmission but additional resources

(workforce and equipment such as aspirators) are a constraint for health departments in developing countries. We were also unable to survey inside the houses or roof gutters, identified as Ae. aegypti breeding reservoirs particularly during the dry season (Gustave et al 2014), limiting our ability to identify all potential breeding sites.

78

This study attempts to identify environmental risk factors that contribute to the transmission of dengue in Paramaribo, Suriname. To our knowledge, this is the first time that epidemiological data were used to guide environmental surveys in Suriname and our findings explain some of the differences in Aedes aegytpi breeding site availability between neighborhoods. In addition to the dengue virus, Aedes aegypti also spreads the

Chikungunya virus. Chikungunya was first introduced into Suriname in June 2014 and by

January 2015 more than 1210 confirmed cases and one death have been reported (PAHO

2015). Therefore, it is important to gain insight into environmental risk factors that contribute to proliferation of Aedes aegypti breeding sites in order to guide preventing control methods in a resource efficient and effective manner.

79

References

1. Pinto Severo O, 1955. Eradication of the Aedes aegypti mosquito from the

Americas. Yellow Fever, a Symposium in Commemoration of Carlos Juan Finlay

6: 39 – 59.

2. Pan American Health Organization, 1997. The feasibility of eradicating Aedes

aegypti in the Americas. Rev PanamSalud Pública 1: 381 – 388.

3. Hopp MJ, Foley JA, 2003. Worldwide fluctuations in dengue fever cases related

to climate variability. Clim Res 25: 85 – 94.

4. Altizer S, Dobson A, Hosseini P, Hudson P, Pascual M, Rohani P, 2006.

Seasonality and the dynamics of infectious diseases. Ecol Lett 9: 467 – 484.

5. Díaz-Quijano FA, González-Rangel AL, Gómez-Capacho A, Espíndola-Gómez

R, Martínez-Vega, RA, Villar-Centeno LA, 2008. Rainfall and acute febrile

síndrome in a dengue-endemic area. Rev Salud Pública 10: 250 – 259.

6. Schreiber KV, 2001. An investigation of relationships between climate and

dengue using a water budgeting technique. Int J Biometeorol 45: 81 – 89.

7. Diaz-Quijano FA, Waldman EA, 2012. Factors associated with dengue mortality

in Latin America and the Caribbean, 1995 – 2009: an ecological study. Am J Trop

Med Hyg 86: 328 – 334.

8. Focks DA, Chadee DD, 1997. Pupal survey: an epidemiologically significant

surveillance method for Aedes aegypti: an example using data from Trinidad. Am

J Trop Med Hyg, 56: 159 – 167.

80

9. Pan American Health Organization, 1994. Dengue and dengue hemorrhagic fever

in the Americas: guidelines for prevention and control. Washington DC: PAHO

Scientific Publication, 1 – 98.

10. Sanchez L, Vanlerberghe V, Alfonso L, Marquetti MC, Guzman MG, Bisset J,

van der Stuyft P, 2006. Aedes aegypti larval indices and risk for dengue

epidemics. Emerg Infect Dis 12: 800 – 806.

11. Morrison AC, Zielinski-Gutierrez E, Scott TW, Rosenberg R, 2008. Defining

challenges and proposing solutions for control of the virus vector Aedes aegypti.

PLoS Med 5, 362 – 366.

12. Eisen L, Lozano-Fuentes S, 2009. Use of mapping and spatial and space-time

modeling approaches in operational control of Aedes aegypti and dengue. PLoS

Negl Trop Dis 3: 1 – 7.

13. Knox TB, Nguyen YT, Vu NS, Kay BH, Ryan PA, 2010. Quantitative

relationships between immature and emergent adult Aedes aegypti populations in

water storage container habitats. J Med Entomol 47: 748 – 758.

14. Focks DA, Brenner RJ, Hayes J, Daniels E, 2000. Transmission thresholds for

dengue in terms of Aedes aegypti pupae per person with discussion of their utility

in source reduction efforts. J Trop Med Hyg 62: 11 – 18.

15. Focks DA, Alexander N, Villegas E, 2006. Multicountry study of Aedes aegypti

pupal productivity survey methodology: Findings and recommendations. Geneva:

World Health Organization, 1 – 48.

81

16. Barrera R, Amador M, Clark GG, 2006. Use of the pupal survey technique for

measuring Aedes aegypti productivity in Puerto Rico. Am J Trop Med Hyg 74:

290 – 302.

17. Hiwat H, Doerdjan K, Kerpens M, Samjhawan A, Sardjoe P, Soekhoe T, 2013.

Importance of domestic water coutnainers as Aedes aegypti breeding sites in

Suriname, implications for dengue control. Acad J Sur 4: 403 – 407.

18. Hamer & LIchtveld 2014

19. Bureau of Statistics Suriname, 2012. Statistical yearbook 2011 (No. 288).

Paramaribo: Algemeen Bureau voor de Statistiek.

20. Halide H, Ridd P, 2008. A predictive model for dengue hemorrhagic fever

epidemics. Int J Environ Health Res 18: 253 – 265.

21. Parker D, Holman D, 2012. Event history analysis of dengue fever epidemic and

inter-epidemic spells in Barbados, Brazil, and Thailand. Int J Infect Dis 16: e793

– e798.

22. Fan J, Lin H, Wang C, Bai L, Yang S, Chu C, Yang W, Lui Q, 2013. Identifying

the high-risk areas and associated meteorological factors of dengue transmission

in Guangdong Province, China from 2005 to 2011. Epidemiol Infect 142: 634 –

643.

23. Gagnon AS, Bush ABG, Smoyer-Tomic KE, 2001. Dengue epidemics and the El

Nino southern oscillation. Clim Res 19: 35 – 43.

24. Kovats RS, Bouma MJ, Hajat S, Worrall E, Haines A, 2003. El Niṅo and health.

Lancet 362: 1481 – 1489.

82

25. Cummings DA, Irizarry RA, Huang NE, Endy TP, Nisalak A, Ungchusak K,

Burke DS, 2004. Travelling waves in the occurrence of dengue hemorrhagic fever

in Thailand. Nature 427: 344 – 347.

26. Braga C, Luna CF, Martelli CM, de Souza WV, Cordeiro MT, Alexander N, de

Albuquerque MF, Junior JC, Marques ET, 2010. Seroprevalence and risk factors

for dengue infectious in socio-economically distinct areas of Recife, Brazil. Acta

Trop 113: 234 – 240.

27. Schmidt WP, Suzuki M, Thiem VD, White RG, Tsuzuki A, Yoshida LM, Yanai

H, Hague U, Tho le H, Anh DD, Ariyoshi K, 2011. Population density, water

supply, and the risk of dengue fever in Vietnam: cohort study and spatial analysis.

PLoS Med 8: e1001082.

28. Chiewslip P, Scott RM, Bhamarapravati N, 1981. Histocompatibility antigens and

dengue hemorrhagic fever. Am J Trop Med Hyg 30: 1100 – 1105.

29. Bravo JR, Guzman MG, Kouri GP, 1987. Why dengue hemorrhagic fever in

Cuba? 1. Individual risk factors for dengue hemorrhagic fever / dengue shock

syndrome (DHF/DSS). Trans R Soc Trop Med Hyg 81: 816 – 820.

30. Sierra BC, Kouri, G, Guzman MG, 2007. Race: a risk factor for dengue

hemorrhagic fever. Arch Virol 152: 533 – 542.

31. Bowman LR, Runge-Ranzinger S, McCall PJ, 2014. Assessing the relationship

between vector indices and dengue transmission: a systematic review of the

evidence. PLoS Negl Trop Dis 8: 1 – 11.

32. Pan American Health Organization, 2014. Number of reported cases of

chikungunya fever in the Americas, by country or territory, 2013 – 2014 EW 45.

83

Available at: http://www.paho.org/hq/index.php?option=com_topics&view=rdmore&tt=PAHO

%2FWHO+Data%2C+Maps+and+Statistics&id=5927&tp=&Itemid=40931&lang

=en. Accessed November 12, 2014.

84

Figure 1. Map of the two clusters (HRC and LRC) where environmental surveys were conducted in Paramaribo.

(HRC = high rate cluster, LRC = low rate cluster, SRS = short rain season)

85

Figure 2. Total houses surveyed within each cluster (HRC and LRC) during the pre-SRS and SRS.

(SRS = short rain season, HRC = high rate cluster, LRC = low rate cluster)

Table 1. Demographic characteristics of populations within the high (HRC) and low (LRC) rate cluster

Neighborhood Cluster % % Total Pop. Maroon Creole Hindustani Javanese electricity piped Pop. Density % % % % water* Blauwgrond High 97 96.5 31483 732.16 2.8 19.9 16.0 24.7 Flora Low 97 97.7 19538 4884.5 2.7 44.5 17.6 2.9 Latour Low 86 77.4 29526 4921.0 48.7 22.7 13.3 3.6 Pontbuiten Low 91 72.8 23211 3868.5 51.4 20.0 14.4 2.4 *piped water accessibility within 200 meters of the household (ABS 2012)

86

Table 2. Household characteristics by cluster and season surveyed and identified positive for pupae breeding sites.

November (Pre- February (SRS) SRS) HRC LRC HRC LRC (pupae+) (pupae+) (pupae+) (pupae+) HOUSEHOLD CHARACTERISTICS Houses 136 (43) 106 (48) 179 (48) 115 (25) surveyed Piped water Yes 136 (43) 80 (29) 176 (46) 91 (19) No 0 26 (19) 3 (2) 24 (6) p-value* 0.0291 .1940 0.9056 House condition Good 96 (23) 27 (8) 128 (29) 43 (12) Medium 34 (16) 55 (26) 44 (14) 53 (10) Bad 6 (4) 24 (14) 7 (5) 19 (3) p-value* .0063 .1155 0.0142 0.4745 Yard condition Good 70 (17) 13 (4) 66 (15) 20 (3) Medium 42 (12) 47 (20) 81 (24) 44 (11) Bad 24 (14) 46 (24) 32 (9) 51 (11) p-value* 0.0097 0.3644 0.6598 0.7248 People X x 3.85 5.76 household p-value** 0.00 Ave. Years in X x 19.69 19.31 house p-value** 0.907 Children Yes X x 75 70 present* No X x 100 44 PRESENCE of IMMATURES Larva Yes 72 60 85 61 No 64 46 94 54 p-value*** 0.570 0.352 Pupae Yes 43 48* 48 25 No 93 58 131 90 p-value*** 0.029 0.306 (HRC = high rate cluster, LRC = low rate cluster, SRS = short rain season) * Fisher’s exact test for count data, within cluster and season **t-test p<0.05, within season *** Chi-square analysis p<0.05, within season

87

Table 3.Entomological indices before and during the short rain season (SRS) for both clusters

November (pre SRS) February (SRS) Both HRC LRC Both HRC LRC All CI pupae 16.53% 13.47% 20.51% 7.73% 9.21% 5.64% 14.70% HI pupae 37.60% 31.61% 45.28% 24.83% 26.81% 21.74% 30.59% BI pupae 0.68 0.56 0.84 0.27 0.42 0.29 0.66 Pupae / 6.09* 4.68 7.90* 2.90 3.17 2.48 5.72 household Pupae / - - - 0.64 0.84 0.43 - person (HRC = high rate cluster, LRC = low rate cluster, SRS = short rain season) CI = container index; HI = house index; BI = Breteau index

Table 4. Container category frequencies and pupae breeding sites by cluster and season.

Pre-SRS SRS Both HRC LRC Both HRC LRC (pupae+) (pupae+) (pupae+) (pupae+) (pupae+) (pupae+) Garden attributes 68 (13) 62 (12) 6 (1) 90 (8) 84 (8) 6 (0) Water-storage 524 (82) 254 (28) 288 (54) 729 (45) 377 (45) 352 (20) Discarded household 69 (41) 42 (18) 27 (23) 84 (19) 50 (14) 34 (5) items Garbage 21 (4) 15 (1) 6 (3) 37 (4) 29 (2) 8 (2) Household use 208 (13) 128 (7) 80 (6) 298 (11) 181 (7) 117 (4) Recreation 17 (2) 15 (2) 2 (0) 28 (0) 25 (0) 3 (0) House structure 63 (8) 39 (6) 24 (2) 130 (2) 88 (0) 42 (2) Natural habitats 10 (2) 9 (2) 1 (0) 14 (0) 13 (0) 1 (0) Total 998 (241) 564 (76) 434 (165) 1410 (109) 825 (76) 585 (33) Dispersion Index 3.99 5.01 2.96 3.42 3.23 3.27 (HRC = high rate cluster, LRC = low rate cluster, SRS = short rain season)

88

Table 5. Odds ratio for positive pupae containers by category and characteristics

OR CI p-value Container Water-storage 1.00 0.60 – 2.55 <0.0045 Category Garden attributes 1.30 0.60 – 2.55 Discarded household items 4.51 2.65 – 7.59 Garbage 1.06 0.30 – 2.86 Household items 0.40 0.20 – 0.73 Recreation 0.41 0.02 – 2.09 Structural components 0.33 0.11 – 0.77 Natural habits 0.95 0.05 – 5.46 Usage Not in Use 1.00 0.30 – 0.65 <0.0045 In Use 0.45 0.30 – 0.65 Material Plastic 1.00 <0.0045 Glass 0.26 0.08 – 0.59 Metal 1.28 0.69 – 2.24 Other 0.47 0.19 – 0.85 Capacity <1L 1.00 0.005 1 - 15L 2.02 1.35-3.04 15 - 50L 1.27 0.87-1.88 >50L 1.27 0.89-1.83 Circumference <5cm 1.00 <0.0045 5 - 30 cm 8.65 3.97-22.7 30 - 60 cm 9.20 4.37-23.7 >60 cm 6.07 2.86-15.7 Disposable No 1.00 <0.0045 Yes 1.72 1.15 – 2.64 Cover Uncovered 1.00 <0.0045 Covered 0.18 0.05 – 0.48 Water Source Rainfall 1.00 <0.0045 Manual 0.09 0.03 – 0.22 Location Shaded 1.00 0.1704 Partially 1.34 0.97-1.83 Sun 1.18 0.88-1.59 Biological No 0.7113 Yes Larva No 1.00 <0.0001 Yes 743 237-4499 Larval Density 0 -10 1.00 <0.0001 11 - 50 31.7 22.2-45.6 50 - 100 51.5 33.7-79.6 >100 36.5 13.5-100.5 OR = Odds ratio; 99.8% CI = 99.8% Confidence Interval; p-value = probability of Chi-square test.

89

6. A retrospective analysis of dengue cases: implications for treatment and prevention in a LMIC

Accepted for publication to International Journal of Tropical Disease and Hygiene A retrospective analysis of dengue cases in Suriname: implications for treatment and prevention in a Lower Middle Income Country (LMIC)

Diana Hamer1, Arti E. R. Jessurun2, Manodj Hindori2, John Codrington3, Jimmy Roosblad3, Maureen Lichtveld1

1Tulane University School of Public Health and Tropical Medicine, New Orleans, LA 2St. Vincentius Hospital, Paramaribo, Suriname 3Academic Hospital Paramaribo, Paramaribo, Suriname

ABSTRACT: Aims: To describe hospitalized dengue cases and characterize the hospitalization length and cost of dengue based on clinical and laboratory classification in a primary hospital in Paramaribo, Suriname. Study design: A retrospective study was conducted to identify patients at higher risk of dengue hemorrhagic fever (DHF) and to compare the length and cost of hospitalization by dengue classification and dengue severity. Methodology: We analyzed 2800 dengue cases hospitalized between 2001 to 2012. All dengue cases were spatially visualized using a geographic information system (GIS).Dengue cases were stratified by demographic characteristics and classified as suspected, probable and confirmed. This classification was used to compare length and cost of hospitalization. Results:The majority of hospitalized dengue cases, 50.1%, were ethnic Hindustani or Javansese (Southeast Asian descent). Dengue was laboratory confirmed in a 188 cases between 2001 and 2012. However, ethnicity is not associated with progression into DHF in hospitalized cases with a confirmed diagnosis of dengue. When comparing length of hospitalization, suspected dengue cases stayed on average longer hospitalized (7.81 days) than probable (6.65 days) and confirmed cases (6.29 days). In contrast, confirmed cases had the highest cost of hospitalization (3100 Surinamese dollars – SRD) compared to suspected (2766 SRD) and confirmed (2157 SRD) cases. Suspected and probable dengue fever cases had longer hospital stays compared to the more severe DHF. There is a difference in the length and cost of hospitalization among suspected, probable and confirmed dengue cases and dengue fever cases have longer hospitalization terms than DHF for suspected and probable cases.

Conclusion:This study contributed to the limited evidence of the demographic characteristics and the economic burden of dengue in Suriname. There is a need to standardize and increase diagnosis capabilities to improve surveillance and treatment of dengue while reducing hospitalization costs in Suriname.

Keywords: Dengue, Cost of dengue, Suriname, GIS

90

1. INTRODUCTION

Dengue has become hyperendemic in Suriname. The distribution of dengue is heavily influenced by degree of urbanization and meteorological factors. Currently, 2.5 billion people live in endemic regions and an estimated 50 – 390 million dengue infections occur yearly [1,2]. There are four different types of dengue virus (DENV type 1 – 4) capable of causing dengue fever (DF) or the more severe forms of the disease (dengue hemorrhagic fever and dengue shock syndrome – DHF and DSS) in humans [3]. Co-circulation of serotypes within a population increases the risk of a secondary heterologous infection, which is the most important risk factor for DHF and DSS [4]. Furthermore, the dengue serotype [5,6], the sequential order in which different serotypes are encountered within a population [7-9] and the time elapsed between primary and secondary dengue infections[10-12] are associated with increased pathogenicity and the development of DHF and DSS.

In Suriname, DENV-1 was the first documented serotype introduced in 1978 and subsequently DENV-4 was first reported in 1981 [13]. Shortly upon introduction of a serotype, epidemics of dengue in the local population correlated with rapid increase of DENV lineage numbers [14]. Dispersal histories reconstruction demonstrate the strongest dissemination links of DENV between the Lesser Antilles islands and Suriname [15] suggesting that language and economic relationships influence the diffusion of DENVs [14]. When DENV was first introduced in Suriname in 1980, the general population was immunologically naïve. However, after all four DENV serotypes were introduced, the current population is immunologically sensitized and a higher proportion of the population is more likely to show severe manifestations upon infection with DENV. Due to the endemicity of the DENV, most health care practitioners and physicians are familiar with the symptoms of DF and DHF/DSS. Dengue infections are often asymptomatic or it can present mild symptoms including a high fever, headache, muscle and joint pains, vomiting or rash. This febrile phase can also be accompanied by more severe symptoms such as petechiae, hepatomegaly and a continuous drop of white blood cell count resulting in leukopenia and even shock [2]. Thus, dengue morbidity can come at a significant cost (both directly and indirectly) to the infected person and can exert a high cost on society by exhausting national healthcare capabilities.

During the last confirmed dengue outbreak in 2012 the Surinamese Ministry of Health (MoH) was prompted to open an emergency hospital to accommodate all suspected dengue cases [16].The result was an unplanned allocation of workforce (heath care practitioners) and workplace (hospital beds, diagnostic capabilities) resources to contain the dengue epidemic at a significant financial cost. Inadequate surveillance of dengue leads to clinical mismanagement during epidemic and non-epidemic periods. As exemplified in Suriname, during a dengue epidemic cases of dengue become grossly over-reported when diagnostic resources become strained and patients are hospitalized based solely on symptoms in an attempt to curb the epidemic resulting in social disruption, lost tourism and lost work and school productivity. In non-epidemic times, however, dengue is under-reported due to a lack of good surveillance resulting in an underestimation of the burden of disease and limiting effective prevention practices [17].In an study of the economic burden of dengue in the Americas, the estimated average direct medical cost per dengue case (in 2010 US$) in Suriname was $92 and $463 for ambulatory and hospitalized cases, respectively [18]. These cost are much higher when indirect medical and non- medical cost were included. Addressing the rising direct and indirect cost of dengue in Suriname is a priority exemplified in Selck et al, which classified Suriname in the highest dengue expenditure per thousand population quintile from a 108 country cost analysis[19].

This study describes hospitalized dengue cases in a private hospital in Suriname that adheres to the WHO’s 1997 classification of dengue fever (DF), dengue hemorrhagic fever (DHF) and dengue shock syndrome (DSS). The objective of this study is to identify patients at higher risk of DHF and or DSS.All case were hospitalized in the Sint Vincentius Ziekenhuis (SVZ – Saint Vincent Hospital), a private hospital located in the capital Paramaribo. The hospital has 186 beds, with 17 specialists, and a staff of about 600 from which more than half are nurses. The hospital

91 was founded in 1916. Gradually, SVZ services have expanded with modern clinical facitlities. In 2014, the hospital expanded its services with a new 24-hour emergency unit providing services to a large number of patients in Paramaribo.

We characterized the frequency of hospitalization of suspected, probable and confirmed dengue cases and stratified by demographic characteristics from 2001 to 2012. All cases were spatially distributed using a geographic information system (GIS) to facilitate visualization and analysis. Furthermore, hospitalization and diagnostic practices were used to compare the length and cost of hospitalization by dengue classification. The findings will illustrate the need to improve dengue surveillance, diagnostic and prevention practices in Suriname.

2. METHODOLODY

This study took place in Paramaribo (5°52’N, 55°10’W), the capital city of Suriname. Paramaribo has a tropical climate defined by a short and a long rainy season (SRS and LRS – from December to January and from April to August, respectively) and a short and long dry season (SDS and LDS – from February to March and September to November, respectively). The Bureau of Public Health (BOG), under the Ministry of Health (MoH), is in charge of vector control and dengue epidemiological surveillance. Suriname has 6 hospitals with a total of 1,687 beds [20]. The MoH subsidizes and coordinates through the BOG 56 Regional Health Service clinics in the coastal region and the Medical Mission clinics in the interior to provide primary care to the poor or near poor [21]. Of the six hospitals, four, two private and two public, are located in Paramaribo [20]. One hospital, SintVincentius Hospital (SVZ), was involved in this study.

Since 2001, SVZ has kept a digital database of all in-patient cases of dengue fever and DHF/DSS. All patients admitted with an initial diagnosis of suspected dengue fever are given a unique reference code. Additionally, demographic information, current address, hospital admission and discharge date, and test results are recorded. An initial dengue diagnosis is based on symptomology following the WHO criteria of two or more of the following: headache, retro- orbital pain, myalgia, arthralgia, rash, hemorrhagic manifestations and/or leucopenia. Blood serum is collected fordetection of dengue IgG or IgM antibodies using the QuickTest™ Dengue IgG/IgM Plasma/Serum (Orgenics Ltd.). In 2011, the Academic Hospital Paramaribo (AZP) introduced reverse transcription polymerase chain reaction (RT-PCR) for the detection of the dengue virus.

Blood samples from SVZ suspected dengue cases were analyzed by the AZP laboratory for molecular diagnosis.The Medical Laboratory of AZP is the largest clinical laboratory of the country, conducting hematological, serological, bacteriological and clinical chemistry research. A cross-reference of SVZ patients using name and date of birth was performed to identify RT-PCR results from the AZP laboratory database. For the purpose of this study, a patient was classified as ‘suspected’ when the QuickTest™ and RT-PCR were negative. A compatible serology (positive QuickTest™ result) classified the patient as ‘probable’ and a positive RT-PCR detection of the dengue virus in the serum was used to classify a case as ‘confirmed’. Dengue cases for which a diagnostic test was performed but the result was not documented was excluded from our analysis.

From 2001 to 2012, there were a total of 2822 dengue-suspected hospitalization. The addresses of all hospitalized cases (suspected, probable and confirmed) were coded according to the Suriname General Bureau of Statistics (ABS) codebook.Suriname is divided into ten districts. Each district is further divided into resorts (similar to a municipality). Digital maps of the Suriname’s districts and resorts were used to incorporate all dengue cases with an identifiable address into a geographic information system (GIS). A cluster analysis using Anselin Local Moran’s I was used toidentify spatial clustering of dengue cases from 2001-2012 using resorts as the basemap. All dengue cases were visualized and cluster analysis was performed using ArcGIS v10.2 (http://www.esri.com).

92

All hospitalized cases (suspected, probable and confirmed) were stratified according to gender, race, age, death, dengue serotype and season. A chi-square analysis to test for the association between patient demographic characteristics and DHF/DSS was performed using only confirmed cases.

Dengue cases were stratified by year and the percentage of cases tested for dengue (either by QuickTest™ or, since 20111, RT-PCR) was calculated. We counted the amount of hospitalization days for each dengue case and obtained the total and mean yearly length of hospitalization. Furthermore, daily hospitalization costs (not including diagnostic tests or treatment) were used to estimate the total and mean yearly hospitalization cost of dengue cases. The mean length and cost of hospitalization was stratified by case classification. The Kolmogorov-Smirnov and Shapiro-Wilk test determined that the hospitalization length data was not normally distributed (p<0.0001). Therefore, the medians and inter-quartile ranges (IQR) were calculated and a non- parametric Kruskal-Wallis test was used to compare the hospital length to the three case classifications with a significance level of ≤0.05. We followed up with a Mann-Whitney test on pairs of groups with a Bonferroni correction. A Furthermore, a Mann-Whitney U-test was used to understand whether suspected, probable or confirmed dengue cases median length and cost of hospitalization differ by severity (dengue fever and DHF/DSS).

3. RESULTS

We performed a retrospective analysis of all dengue cases hospitalized at SVZ from 2001 to 2012. A total of 2822 cases were admitted with an initial diagnosis of dengue fever of which 22 were excluded because there was no documentation of the dengue diagnostic test result. For more than half of all hospitalized cases (1637 or 58.0%) there was no recorded test and 464 cases (16.4%) had a negative QuickTest™ or RT-PCR result. Therefore, 2101 dengue cases (2007 dengue fever and 94 DHF/DSS) were classified as suspected. Of the remaining 25% of cases, 511 (18.10%) were classified as probable and 188 (6.7%) as confirmed (Figure 1).

Hospitalized dengue cases 2800 (100%)

Suspected Probable Confirmed 2101 511 188 (75.0%) (18.25%) (6.75%)

Dengue DHS/DSS Dengue DHF/DSS Dengue DHF/DSS Fever 94 Fever 59 Fever 24 2007 452 164

Fig. 1.SVZ hospitalized dengue cases by classification and dengue fever (DF) and dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS) from 2001 to 2012

Of the 2800 dengue cases, 2477 had an address that was coded and incorporated into ArcGIS (Figure 2A). The Global Moran’s I (I = 0.809, z>1.96, P=.05) indicated spatial autocorrelation of dengue cases from 2001 to 2012. Using the local Moran’s I, the resorts of dengue cases admitted to SVZ were Blauwgrond, Munder, Rainville, Welgelegen, Tammenga, Flora and Latour (Figure 2B). Based on the geographic distribution, the majority of hospitalized cases come from the resort, or a neighboring resort, in which SVZ is located (Blauwgrond).

93

Fig. 2.a) shows the spatial distribution of all hospitalized dengue cases from 2001 to 2012; Figure; b) shows the spatial distribution of cases in Paramaribo (by age groups and severity) highlighting the resorts with highest case density surrounding the hospital from 2001 to 2012.

Table 1. Frequencies of hospitalized cases by demographic and case classification

Suspected Probable Confirmed Total

Gender Male 1047 277 114 1438 Female 1054 234 74 1362 Race Creole 356 50 22 428 Hindustani 598 156 48 802 Javanese 464 102 34 600 Chinese 218 106 29 353 European 43 12 6 61 Indian 34 6 7 47 Maroon 19 3 2 24 Other/Unknown* 369 76 40 485 Age ≤14 617 205 61 883 >14 1484 306 127 1917 Death No 2069 505 188 2562 Yes 32 6 0 38 Serotype 2 - - 170 170 3 - - 6 6 4 - - 8 8

94

Season SRS 351 101 60 512 SDS 502 68 60 630 LRS 781 163 62 1006 LDS 467 179 6 652 *Other/Unknown – includes individuals who self-reported race is not included in the 7 specified, or individuals for which race was not reported SRS – short rain season, SDS – short dry season, LRS – long rain season, LDS – long dry season

Table 2. Frequencies of hospitalized cases by gender and age stratified by ethnicity, and demographic stratified by severity (DF and DHF/DSS)

Ethnicity N Creole Hindustani Javanese Chinese P-value* Gender Male 1140 215 413 297 215 0.01 Female 1062 217 395 309 141 Age ≤14 655 127 260 193 75 <0.001 >14 1547 305 548 413 281 Severity N DF DHF/DSS p-value Gender Male 114 100 14 0.80 Female 74 64 10 Race Creole 22 20 2 0.70 Hindustani 48 43 5 Javanese 34 28 6 Chinese 29 26 3 Age <14 61 44 17 <0.001 >14 127 120 7 Other Serotype 1 0 0 0 2 170 146 24 - 3 6 6 0 4 8 8 0 Season SRS 60 48 12 0.16** SDS 60 54 6 LRS 62 56 6 LDS 6 6 0 *Chi-square test p<0.05 **LDS not included in Chi-square analysis DF – dengue fever, DHF/DSS – dengue hemorrhagic fever/dengue shock syndrome SRS – short rain season, SDS – short dry season, LRS – long rain season, LDS – long dry season.

Dengue cases by classification were further stratified by gender, race, age, serotype and season (Table 1). The majority of hospitalizations due to dengue fever occurred in ethnic Hindustani or Javanese, in adults, and during the LDS. Furthermore, since the implementation of RT-PCR dengue diagnosis in 2011 only serotypes 2, 3, and 4 have been detected in hospitalized cases. Only 38 deaths in patients hospitalized with an initial diagnosis of dengue fever were recorded between 2001 and 2012, with none of the deaths having been had a positive laboratory confirmed dengue diagnostic. Dengue infections can manifest with non-specific symptoms making clinical diagnosis unreliable. Furthermore, rapid tests vary in their sensitivity and specificity. Therefore,

95 only confirmed cases of dengue were used to compare demographic characteristics by ethnicity and severity (frequency of DF compared to DHF/DSS cases) (Table 2). In the hospitalized confirmed dengue cases, Chinese males are more frequently hospitalized compared to females. However, no significant association was detected between ethnicity and severity (DHF/DSS) of dengue. Similarly, Chinese adults (>14) are more likely than Chinese children to be hospitalized (p>0.05). Even though when comparing confirmed dengue cases by severity, we observed a statistically higher frequency of children presenting with DHF/DSS compared to adults (P=.05). There was no association between season (excluding the LDS for Chi-square analysis) and disease severity (P=.05), which indicates that seasonal variations can have an effect on the number of dengue cases (Table 1) but does not influence the progression of severe disease.

Table 3.Frequency of hospitalized cases, dengue diagnostic test, and total and mean hospitalization length and cost stratified by year.

Mean Hospital Mean IgG/IgM IgM/I RT- Length Length cost Hospital Hospital % Year N RT-PCR gG PCR of Stay of Stay SVZ Cost cost Tested NEG POS POS (days) (days/ (SRD/ (SRD) (SRD/ case) day) case)

2001 220 17 6 0 10.5 1775 8.08 - - -

2002 197 2 7 0 4.6 1590 8.07 - - -

2003 101 4 2 0 5.9 770 7.78 - - -

2004 115 10 9 0 16.5 1119 9.73 - - -

2005 507 143 236 0 74.7 4278 8.47 - - -

2006 198 67 49 0 58.6 1735 8.81 - - -

2007 114 43 11 0 47.4 1074 9.50 - - -

2008 171 32 28 0 35.1 1431 8.37 275 393525 2301.31

2009 312 20 50 0 22.4 2053 6.58 350 718550 2303.05

2010 189 6 17 0 12.2 1277 6.79 350 446950 2364.81

2011 177 46 45 25 65.5 1129 6.41 450 508050 2870.33

2012 521 74 51 163 55.3 3318 6.37 500 1659000 3184.26

Total 2822 464 511 188 41.2 21550 7.66 - - -

IgG/IgM – Immunoglobulin G/Immunoglobulin M, RT-PCR – reverse transcription polymerase chain reaction, NEG – negative, POS – positive, SVZ – Sint Vincentius Ziekenhuis (Saint Vincent Hospital) SRD – Suriname dollars

All cases of dengue were stratified by year. The largest outbreak of dengue fever was recorded in 2005 with an estimated infection of 10% of the population affected by the disease. Between 2005 and 2010, the percentage of hospitalized cases tested for dengue infection increased compared to 2001 – 2004 but remained below 50% except for 2006, the year following the large outbreak. The introduction of RT-PCR diagnosis technology in the AZP laboratory allowed for more dengue diagnostic analysis to be performed and increased the percentage of total (either with

96

QuickTest™ and RT-PCR) cases tested above 50% for 2011 and 2012 (Table 3). However, the total percentage of dengue cases tested remains low at 41%.

Table 4.Length (in days) and cost (in SRD) of hospitalization stratified by dengue classification.

Length of Hospitalization (2001-2012) N Median (IQR) Mean (STD) Chi-Square‡ p-value Suspected 2094 6 (5-9) 7.81 (5.31) 17.33 <0.001* Probable 510 6 (5-8) 6.65 (3.55) Confirmed 188 6 (5-7) 6.29 (2.81) Total 2792 6 (4-8) 7.50 (4.92) Cost of hospitalization (2008-2012) Suspected 981 2450 (1650-3150) 2766.33 (1852.17) 66.43 <0.001* Probable 191 1925 (1400-2500) 2157.59 (1067.85) Confirmed 188 3000 (2500-3500) 3100.26 (1395.29) Total 1360 2450 (1650-3150) 2724.10 (1720.00) ‡Kruskal-Wallis H-test SRD – Surinamese dollars

Table 5.Mann-Whitney U-test of differences in mean rank of hospitalization length and cost by classification.

Length of Hospitalization in days / person (2001-2012) N Mean Rank Z-score p-value Direction Suspected 2101 1326.42 -3.309 0.001* S>P Probable 510 1204.3 Suspected 2101 1153.5 -2.92 0.003* S>C Confirmed 188 1007.84 Probable 510 352.89 -0.737 0.461 - Confirmed 188 340.32 Cost of Hospitalization in SRD / person (2008-2012) Suspected 981 605.61 -4.38 <0.001* S>P Probable 191 488.37 Suspected 981 559.67 -5.86 <0.001* S

Table 3 depicts the total and mean hospitalization length for all dengue cases (suspected, probable and confirmed) between 2001 and 2012. A total of 29,559 hospitalization days were recorded for all 2822 hospitalized cases with a mean of 10.47 days/case. Since 2008, the ‘hotel’ hospitalization cost has gone up with a total expenditure of 4,525,275.00 Surinamese dollars (SRD) for the 1362 patients admitted into SVZ with an initial dengue diagnosis between 2001 and 2012.On average, each year 3322.52 SRD was spent per patient.These costs, however, do not include nor reflect changes over the years in cost for diagnosis, treatment or medical consults. We calculated the mean and median length and cost of hospitalization for the three dengue

97 classifications. Suspected cases, dengue cases for which there was either no diagnostic test performed or laboratory result came back negative, stayed on average longer hospitalized compared to probable and confirmed cases (Table 4 & 5). However, confirmed dengue cases cost was the highest. This higher cost per confirmed case reflects the higher daily hospital cost in 2011 and 2012, which, coincidentally, corresponds with the introduction of RT-PCR diagnosis that provides a confirmation of a case of dengue. This indicates that the higher cost per confirmed dengue case compared to suspected and probable cases is not related to a longer hospitalization stay but rather to an increase in the daily cost of hospitalization. Furthermore, we observed a significant difference between the length and cost of hospitalization for suspected and probable cases with DF compared to those with DHF (P=.05). However, no differences were detected for confirmed dengue cases (Table 6). Such difference could be attributed a change in the clinical manage of dengue once it has been confirmed by a laboratory test.

Table 6.Mann-Whitney U-test of differences in mean rank hospitalization lenght and costs by classification stratified by severity.

Length of Hospitalization in days per person (2001-2012) N Mean Mean Rank Z-score p-value Direction Suspected DF 2000 7.88 1058.01 -3.68 <0.001 DF>DHF DHF 94 6.44 823.98 Probable DF 451 6.88 268.70 -5.63 <0.001 DF>DHF DHF 59 4.86 154.63 Confirmed DF 164 6.41 97.06 -1.709 0.08 - DHF 24 5.46 77.00 Cost of Hospitalization in SRD per person (2008-2012) Suspected DF 2000 2797.30 498.90 -2.93 0.003 DF>DHF DHF 94 2417.50 401.99 Probable DF 451 2245.86 101.12 2.27 0.023 DF

4. DISCUSSION

To our knowledge, this is the first study to explore demographic characteristics and hospitalization practices of dengue cases in a hospital in Suriname. Most of the dengue cases (75%) admitted to SVZ did not have a laboratory confirmed dengue diagnosis. Since the DENV was first introduced into the population, the annual number of dengue cases has steadily increased with two of the largest outbreaks, in 2005 and 2012, taking place in the last ten years [16,22,23]. Both outbreaks exhausted national healthcare capabilities and in 2012 even prompted the MoH to open an emergency hospital to treat the high number of dengue cases. This upward trend is of concern. Most of the Surinamese population (70%) lives in an urban area [24]. This urbanization and the climate conditions characteristic of the tropics facilitates and sustains the

98 emergence of DF. Additionally, the effects of climate change could cause more frequent and more severe dengue outbreaks [25-27]. As in Southeast Asia where dengue has been endemic since the 1950s, specific demographic characteristics, such as race [28], gender and age [29,30]are established risk factors in the development of DHF/DSS. Therefore, as the DENV becomes established in Suriname the local population becomes immunologically sensitized changing the dengue disease dynamics. As indicated by the hospitalization information on confirmed dengue cases, children are more likely to present DHF/DSS compared to adults. However, due to the low percentage of cases confirmed by a diagnostic laboratory dengue test, we can only draw limited conclusions about demographic risk factors for severe dengue in the SVZ patient population.

From a health policy perspective, the length and cost of hospitalization analyses in this study highlight the need to increase diagnostic capabilities in this hospital as well as other hospital and clinics in Suriname. The 2005 outbreak increased national awareness of dengue increasing the annual percentage of symptomatic cases tested for the disease. The adoption of RT-PCR technology at the AZP has enabled national diagnostic capabilities allowing both private and public hospitals to confirm a dengue case in a timelier manner. Since the implementation of RT- PCR technology, SVZ has been able to confirm more than 25% of the dengue hospitalized cases and combined with in-house serology test, thelaboratory tested cases increased above 50% in 2011 and 2012. The increase in diagnostic capabilities can improve the clinical recognition and treatment of dengue. However, there has been an increased in the mean hospitalization cost per case of dengue even though the length of hospitalization has decreased mostly due to increasing cost of hospitalizations. The hospitalization cost in this study is a gross underestimation of the cost per case of dengue because we did not include cost of diagnosis, treatment and provider fees. Previous cost studies indicate that due to the care DF and DHF patients require are mostly supportive (intravenous fluids), cost models assume a premium between 20% to 50% above hospitalization cost for developing countries to include all direct medical costs [18,31,32] Furthermore, the burden of dengue is not only measured in direct medical cost (ambulatory and hospitalization cost), but should also include indirect medical costs (any expense related to seeking medical care) and non-medical costs (lost of income, school or job absentee time). Thus, even though the hospitalization cost here presented are an underestimation, they are a partial reflection of the economic burden dengue has on the Surinamese society.

Length and cost of hospitalization differ among suspected, probable and confirmed cases. On average, confirmed dengue cases have the shortest hospitalization length but sinceconfirmation began in 2011, it is possible that the shorter hospitalization terms are due to better clinical practice. Interestingly, when only comparing length of hospitalization by severity of disease, DHF cases have on average a shorter hospitalization term compared to DF for suspected, probable but not confirmed cases. Such discrepancies illustrate the need to implement and standardize diagnostic practices that will enhance clinical management. Moreover, the differentiation between DF and severe dengue (DHF and DSS) has not been precise prompting the WHO to change the classification of dengue to better reflect different levels of severity [22]. The new classification enables an improved standardization of clinical management but its implementation does increase the workload and requires dengue confirmatory tests [33,34].

Standardization of clinical and diagnostic practices of dengue would also strengthen surveillance and research practices in Suriname. As mentioned earlier, patterns of dengue disease differ across geographical regions (Southeast Asia v The Americas), age and race [30,35]. Suriname is a country with a unique ethnic population composed of Hindustani, Javanese, Creole, Maroon, Chinese and Amerindian people. This diverse profile is unlike any other in Latin America presenting the opportunity to examine the relationship between known risk factors, such as race, and severe dengue. Currently, dengue is hyperendemic in Suriname but the disease is relatively new compared to Southeast Asia. Studies have demonstrated that the longer dengue has been endemic in a region the more severe outbreaks become in different age groups with a population [10,11,29] Therefore, enhancing diagnostic practices would strengthen national dengue

99 surveillance, strengthen our understanding of dengue disease pathogenesis, and enable the country to inform dengue prevention and control practices based on local research.

There are several limitations to this study associated with the use of retrospective data collected for medical, rather than research, purposes. First, the secondary database utilized did not provide information on patients’ comorbidities, symptomology, time of initial onset of symptoms, or course of treatment. We were not able to evaluate additional medical costs, for diagnosis and treatment, nor that patients’ indirect and non-medical cost associated with the dengue. Finally, selection bias is possible because our patient population is from only one of the six hospitals in Suriname.

CONCLUSIONS

Since 2001, the majority of cases hospitalized for dengue were people of Southeast Asian descent. However, ethnicity is not associated with DHF in our hospitalized population. Children were more likely to develop DHF compared to adults. Thus, further research is necessary to which demographic characteristics increase the risk of DHF in Suriname. The percentage of hospitalized dengue cases tested using serological diagnosis has increased since 2001. Diagnostic practices were further enhanced by the introduction of RT-PCR technology. There is a need to standardize and increase diagnosis capabilities to improve surveillance and treatment of dengue while reducing hospitalization costs in Suriname.

ETHICAL APPROVAL This study was approved by the Tulane University School of Public Health and Tropical Medicine Institutional Review Boards and the Ministry of Health Suriname Ethics Board.

ACKNOWLEDGEMENTS

We would like to thank Arti Shankar (Tulane University) for data analysis support

CONFLICT OF INTEREST

Authors indicate that no conflict of interest exist

REFERENCES

1. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, et al.The global distribution and burden of dengue.Nature. 2013;496(7446): 504-507. 2. World Health Organization. Dengue guidelines for Diagnosis, treatment, prevention and control (New Edition edition). Geneva: WHO Press; 2009. 3. World Health Organization. Dengue haemorrhagic fever: diagnosis, treatment, prevention and control. 2nd edition. Generva: WHO; 1997 4. Halstead SB. Observations related to pathogenesis of dengue hemorrhagic fever. VI. Hypothesis and discussion. Yale J Biol Med.1970;42(5): 350-362. 5. Rico-Hesse R, Harrison LM, Salas RA, Tovar D, Nisalak A, Ramos C, et al.Origins of dengue type 2 viruses associated with increased pathogenicity in the Americas.Virology. 1997;230(2): 244-251. 6. Vaughn DW, Green S, Kalayanarooj S, Innis BL, Nimmannitya S, Suntayakorn S, et al. Dengue viremia titer, antibody response pattern, and virus serotype correlate with disease severity.J Infect Dis. 2000;181(1): 2-9. 7. Burke DS, Nisalak A, Johnson DE, Scott RM. A prospective study of dengue infections in Bangkok.Am J Trop Med Hyg.1988;38(1): 172-180. 8. Thein S, Aung MM, Shwe TN, Aye M, Zaw A, Aye K, et al. Risk factors in dengue shock syndrome.Am J Trop Med Hyg. 1997;56(5): 566-572. 9. Nisalak A, Endy TP, Nimmannitya S, Kalayanarooj S, Thisayakorn U, Scott RM, et al. Serotype-specific dengue virus circulation and dengue disease in Bangkok, Thailand from 1973 to 1999.Am J Trop Med Hyg. 2003;68(2): 191-202.

100

10. Guzman MG, Kouri G, Valdes L, Bravo J, Vazquez S, Halstead SB. Enhanced severity of secondary dengue-2 infections: Death rates in 1981 and 1997 Cuban outbreaks.RevPaname Salud Publica. 2002;11(4): 223-227. 11. Diaz-Quijano FA, Waldman EA. Factors associated with dengue mortality in Latin America and the Caribbean, 1995-2009: An ecological study.Am J Trop Med Hyg. 2012;86(2): 328-334. 12. Montoya M, Gresh L, Mercado JC, Williams KL, Vargas MJ, Gutierrez G, Kuan G, Gordon A, Balmaseda A, Harris E. Symptomatic versus inapparent outcome in repeat dengue virus infections is influenced by the time interval between infections and study year. PLoSNegld Trop Dis. 2013;7(8): e2357 13. Wilson ME, Chen LH. Dengue in the Americas.Dengue Bulletin. 2012;26: 44-61. 14. Carrington CV, Foster JE, Pybus OG, Bennett SN, Holmes EC. Invasion and maintenance of dengue virus type 2 and type 4 in the Americas. J Virol. 2005;79(29): 14680-14687. 15. Allicock OM, Lemey P, Tatem AJ, Pybus OG, Bennett SN, Mueller BA, et al. Phylogeography and population dynamics of dengue viruses in the Americas. Mol Bio Evol. 2012;29(6): 1533- 1543. 16. Suriname hit with dengue epidemic, health ministry says [Internet]. Suriname: Medical Express (AFP); [updated 2012 Jan 25; cited 2014 Dec 7]. Available from: http://medicalxpress.com/news/2012-01-suriname-dengue-epidemic-health-ministry.html 17. Gubler DJ. Epidemic dengue/dengue hemorrhagicfever as a publichealth, social and economicproblem in the 21st century. TrendsMicriobiol. 2002;10(2): 100-103. 18. Shepard DS, Laurent C, Halasa YA, Zambrano B, Dayan GH. Economic impact of dengue illness in the Americas.Am J Trop Med Hyg. 2011;84(2): 200-207. 19. Selck FW, Adalja AA, Boddie CR. An estimate of the global health care and lost productivity costs of dengue. Vector Borne Zoonotic Dis. 2014;14(11): 824-826. 20. Algemeen Bureau voor de Statistiek (ABS).Statistical yearbook 2011 No. 288. Paramaribo: Algemeen Bureau voor de Statistiek; 2012. 21. Coronie N. Health in the Americas: Suriname.Health in the Americas. 2007;2: 640. 22. DengueNet [Internet]. Geneva: World Health Organization (WHO); [updated 2013 Feb 02; cited 2012 May 5]. Available from http://apps.who.int/globalatlas/default.asp 23.Hamer D, Lichtveld MY.Spatial distribution of epidemiological cases of dengue fever in Suriname, 2001 – 2011. West Indian Med J. 2015; (In press) 23. Suriname [Internet]. Washington D.C.: The World Bank; [updated 2013 Feb 02; cited 2014 Dec 7] Available from http://data.worldbank.org/country/suriname 24. Focks DA, Daniels E, Haile DG, Keesling JE. A simulation model of the epidemiology of urban dengue fever: Literature analysis, model development, preliminary validation, and samples of simulation results. Am J Trop Med Hyg. 1995;53(5): 489-506. 25. Hales S, de Wet N, Maindonald J, Woodward A. Potential effect of population and climate changes on global distribution of dengue fever: An empirical model.Lancet. 2012;360(9336): 830- 834. 26. Hopp MJ, Foley JA. Worldwide fluctuations in dengue fever cases related to climate variability.Climate Res. 2013;25: 85-94. 27. de la C Sierra B, Kouri G, Guzman MG. Race: A risk factor for dengue hemorrhagic fever.Arch Virol.2007;152(3): 533-542. 28. Guzman MG, Kouri G, Bravo J, Valdes L, Vazquez S, Halstead SB. Effect of age on outcome of secondary dengue 2 infections.Int J Infect Dis. 2002;6(2): 118-124. 29. Halstead SB, Lan NT, Myint TT, Shwe TN, Nisalak A, Kalyanarooj S, et al. Dengue hemorrhagic fever in infants: Research opportunities ignored.Emerg Infect Dis. 2002;8(12): 1474- 1479. 30. Suaya JA, Shepard DS, Siqueira JB, Martelli CT, Lum LC, Tan LH, et al. Cost of dengue cases in eight countries in the Americas and Asia: a prospective study. Am J Trop Med Hyg. 2009;80(5): 846-855. 31. Dengue cost model methods [Internet]. Baltimore (MD): UPMC Center for Health Security; [updated 2013 March; cited 2015 Feb 20]. Available from http://www.idcostcalc.org/contents/dengue/cost-model.html 32.Kalayanarooj S. Clinical Manifestations and Management of Dengue/DHF/DSS. Trop Med Health. 2011;39(Suppl. 4): 83-87.

101

33. Hadinegoro SR. The revised WHO dengue case classification: does the system need to be modified? PediartInt Child Health. 2012;32(Suppl. 1): 33-38. 34. Halstead SB. Dengue in the Americas and Southeast Asia: Do they differ?RevPanam Salud Publica. 2006;20(6): 407-415.

102

7. Main Results Summary

Within this section, the main findings of the dissertation research are summarized. Results from the previous sections and unpublished results are presented according to the hypotheses and aims framework.

Hypothesis 1 – There is a statistical association between demographic and clinical factors and the progression of dengue fever into the severe forms of the disease in Suriname

Aim 1.1 Characterize the incidence, prevalence and severity of dengue in

Suriname from 2001 to 2012.

In 2001, the BOG began collecting information from hospitals and clinics in

Suriname on all the reported dengue cases. In Figure 1 Section 4, the total monthly frequency of dengue (including suspected, probable and confirmed cases) is shown by year. Dengue epidemic occurred in 3 to 5 year cycles with the largest outbreak documented in 2005. Furthermore, frequency of cases was stratified by DF and DHF and the incidence was calculated using total population census data (Table 1 Section 4). Even though 2005 had the highest number of reported DF cases, the 2009 outbreak was more severe due to the high number of DHF reported (Table 1). During both outbreaks, DENV serotype 2 was detected in the population.

The frequency, incidence and severity of dengue were stratified by demographic characteristics to elucidate if certain populations are more affected by this infectious disease. Severity was estimated as the percent of number of cases of

103

DHF divided by the total cases of dengue reported for that demographic group

between 2001 and 2012. There is little difference between age groups and

gender, however, Chinese have a higher incidence and severity compared to

other ethnicities (Table 2 Section 4 & Table1).

Incidence* Incidence* Incidence* Year Dengue DF DHF 2001 1.38 1.19 0.19 2002 1.70 1.62 0.08 2003 0.46 0.46 0.00 2004 1.14 1.08 0.06 2005 16.37 14.50 1.87 2006 3.61 3.29 0.32 2007 0.88 0.75 0.14 2008 1.76 1.35 0.41 2009 5.57 3.42 2.16 2010 2.65 2.28 0.38 2011 2.91 2.50 0.41 2012 8.44 7.36 1.09 Age 0-14 4.67 3.82 0.85 15-29 4.68 4.17 0.57 30-44 4.11 3.55 0.56 45-59 4.09 3.27 0.81 60+ 2.93 2.58 0.35 Gender Male 4.86 4.11 0.75 Female 3.74 3.17 0.56 Ethnicity** Creole 34.73 30.14 4.59 Hindustani 62.11 53.08 9.03 Javanese 52.59 42.45 10.14 Chinese 328.47 286.62 41.85 Caucasian 251.95 179.96 71.99 Ameri-Indian 28.51 25.07 3.44 Maroon 5.02 4.17 0.85 District Paramaribo 4.68 3.95 0.73 Wanica 4.78 3.86 0.92 Para 2.99 2.52 0.47

104

Commewijne 3.84 3.24 0.60 Saramaca 3.75 3.14 0.61 Nickere/Coronie 3.75 3.83 0.09 Marowijne/Brokopondo 0.79 0.67 0.12 Sipaliwi 0.45 0.34 0.10 Table 1. Incidence of dengue, dengue fever (DF) and dengue hemorrhagic fever (DHF) by year, age group, gender, ethnicity and district. *Incidence is the total cases of dengue adjusted for total population 2002 through 2012 (10,000 per person-year). **Census data by ethnicity only reported for 2012 (10,000 per person)

Aim 1.2 Examine the environmental, demographic and clinical factors that

influence the development of the severe form of dengue in Suriname.

BOG reported dengue cases:

The univariate analysis (using Chi-square or Fisher’s Exact Tests) in Section 4

was performed to further determine if there is an association between

environmental, demographic and clinical factors and DHF. A difference between

DF and DHF cases by age group, ethnicity and district was detected at a

significant level (p<0.05). To determine if there was a correlation between

demographics and DHF, seven variables were selected for the logistic

regression. Interaction terms ethnicity and age group, gender and age group, and

age group and ethnicity were not significant indicating the effects of these factors

were independent. The final model included the interaction terms and age, case

classification, district and length of hospitalization to be associated with DHF

(Table 1). According to the Maximum Likelihood ratio the model was significant

as a whole (P<0.05) with a good adjustment quality (Hosmer-Lemeshow test

p=0.353). With a 0.05 criterion of statistical significance, age group, case

definition and district had significant partial effects. The odds ratio (OR) for age

groups indicates that, holding all other variables constant, children age 3 to 14

105 are 2.8 times more likely to be reported as having DHF. For case classification, the inverse OR indicates that confirmed cases are 76% and 90% more likely to be reported with DHF compared to suspected and probable, respectively.

Similarly, suspected and probable cases are more likely to be reported as having

DHF compared to the confirmed cases, however, the difference is negligible

(OR= 0.013 and OR=0.011, respectively). Similarly, the Nickerie/Coronie districts is 17% less likely to have reported DHF (inverse OR = 1/0.057). Even though the model fits, additional variables (more demographic information on the patient population such as comorbidities, clinical and laboratory test results, primary or secondary dengue infection) and a larger sample size would strengthen the model’s predictive power.

Wald p- B OR 95% CI Χ2 value Symptoms to admission 0.031 3.621 0.057 1.031 (0.999-1.965) Hospitalization Time -0.044 3.681 0.055 .957 (0.916-1.001) Age Group 0-2 0.169 0.102 0.749 1.184 (0.421-3.332) 3-14 1.041 6.401 0.011 2.831 (1.264-6.341) 15-39 0.160 0.161 0.688 1.174 (0.537-2.565) 40-59 0.410 0.954 0.329 1.506 (0.662-3.427) 60+ 1.000

Gender Male 0.057 0.094 0.759 1.059 (0.736-1.522) Female 1.000

Ethnicity Creole -0.224 0.156 0.693 0.800 (0.264-2.424) Hindustani -0.034 0.004 0.950 0.967 (0.334-2.803) Javanese 0.315 0.320 0.572 1.370 (0.460-4.076) Chinese -0.168 0.086 0.769 0.845 (0.275-2.600) Maroon -0.400 0.278 0.598 0.671 (0.152-2.960) Ameri-Indian -0.239 0.104 0.748 0.788 (0.184-3.367) European 1.000

Case Definition Suspected -4.359 404.072 <0.000 0.013 (0.008-0.020)

106

Probable -4.516 306.540 <0.000 0.011 (0.007-0.018) Confirmed 1.000

District Paramaribo -1.498 2.659 0.103 0.224 (0.037-1.353) Wanica -1.031 1.226 0.268 0.357 (0.058-2.211) Para -1.242 1.498 0.221 0.289 (0.040-2.111) Commewijne -1.240 1.589 0.207 0.289 (0.042-1.990) Saramaca -1.157 1.234 0.267 0.314 (0.041-2.422) Nickerie/Coronie -2.868 7.202 0.007 0.057 (0.007-0.461) Marowijne/Brokopondo -2.660 3.198 0.074 0.070 (0.004-1.291) Sipalwili 1.000

Table 1. Logistic regression predicting DHF from clinical and demographic factors

A seasonal trend analysis of monthly dengue cases was performed using the

BOG dataset to adjust for seasonality between 2001 and 2012. The year was divided into four seasons: short rain season (SRS – December and January), short dry season (SDS – February and March), long rain season (LRS – April to

August), and long dry season (LDS -- September to November). A total of 48 time points were used in the analysis. The results for the seasonal Kendal Tests were:

Z τ p-value* slope

Dengue cases 3.642 0.411 0.0357 -1.833

*Adjusted for correlation among seasons

Y = -1.833 + 3.667xYear

Thus, on average there is an increase of approximately 4 cases of dengue per year (2001-2014) adjusting for season. An ANOVA test indicated that there was no difference in the mean dengue cases between seasons. A summary of dengue cases by season from 2001 to 2012 indicated that most dengue cases

(36%) occurred during the LRS, followed by the LDS (24%), the SDS (22.5% and the SRS (17.3%).

107

SVZ reported dengue cases:

This dataset is a subset of the BOG database for St. Vincentius Hospital (SVZ), which is one of the biggest, private hospitals in the capital Paramaribo known for its pediatric care. All hospitalized cases of dengue between 2001 and 2012 were classified as suspected, probable and confirmed based on a laboratory diagnosis. The classified cases were stratified by demographic variables (Table 1

Section 6). Due to the extensive use of RT-PCR diagnostic technology since

2011, it was possible to obtain laboratory confirmation of DF and DHF. Overall,

Chinese males were more likely to be admitted into SVZ as a confirmed dengue case compared to Chinese women and Chinese adults were also disproportionately affected compared to their children counterparts (Table 2

Manuscript 2). Chi-square analyses were performed to detect associations between demographic and clinical variables and DHF in confirmed cases only. A difference in the frequency of DHF between adults and children was detected

(p<0.05).

The PATH allowed us to further investigate the association between demographic characteristics on both dengue diagnosis and hospitalization time.

Below is an illustrative model of the relationship among dengue diagnosis and hospitalization time that fits the data (Χ2=14.63, Df=8, p=0.067). Significant unstandardized coefficients (p<0.001) are highlighted in yellow.

108

This model highlights the relationship between the dengue diagnosis and hospitalization time as shown by the statistically significant unstandardized regression coefficients. The standardized coefficients reveal a relatively strong relationship between age and hospitalization time, season and dengue diagnosis, and race and dengue diagnosis. The measurement part of the model is low for dengue diagnosis (R2=0.045) and for hospitalization time (R2=0.112) indicating that the model accounts for little of the variance in the measured items. The CFI and RMSEA corroborate the model’s fit (CFI=0.985, RMSEA=0.017).

Since the compilation of this database began, the percentage of admitted dengue cases tested using serological diagnosis has increased. A large dengue outbreak in 2005 created national awareness of dengue and intensified the diagnostic practices. The diagnostic practices were further enhanced by the introduction of

RT-PCR technology in 2011 as seen with an increase over 50% of cases tested in 2011 and 2012 (Table 3 Manuscript 3). To highlight the importance of dengue laboratory diagnosis, we calculated and compared (with a Kruskal-Wallis H-Test and a Mann-Whitney U-Test) hospitalization times and costs by case

109 classification (Table 5 & 6 & 7, Manuscript 3). Differences were detected (at p<0.05) in time and cost of hospitalization among classifications, with suspected cases having a longer hospitalization term compared to probable and confirmed cases. Furthermore, suspected and probable cases of DF had a longer hospitalization term than cases of DHF. While no difference was detected when comparing hospitalization time and cost of confirmed cases of DF to DHF.

Hypothesis 2 – Spatio-temporal and relationships of factors influencing dengue incidence and severity will identify disease hotspots in Suriname

Aim 2.1. Integrate clinical and demographic information of all dengue cases reported in Suriname since 2001 into a GIS platform

After every reported dengue case (from the national BOG database) address was coded, cross-referenced and a total of 3894 dengue cases were integrated into a GIS platform. The country-wide map for all reported dengue cases depicts every case as a point which is colored blue for DF and red for DHF/DSS (Figure

1). Figure 2 and 3 represents all the cases in the capital Paramaribo and

Nickerie, respectively.

110

111

112

The location of the six hospitals and regional clinics is represented with an ‘H’.

Figure 4 represents a density map of the DF incidence (4a), DHF incidence (4b) and severity (4c). The layers for these maps were created using all the case information included in the BOG database and census population data thus each layer has a corresponding attribute table that allows for the selection and visualization of cases by each feature (age, gender, classification, etc.).

Aim 2.2. Identify spatial and temporal historic clusters of DF and DHF from 2001 to 2012

Using dengue frequency data, a global spatial autocorrelation of dengue cases was detected in Suriname. Three high rate and one low rate cluster were identified in Paramaribo (Figure 2 Section 4). Using a street layer with all the major highways, streets and side streets of Suriname and a hard-copy map that

113 contained neighborhood information, a neighborhood (polygons) layer was created (Figure 5). Table 2 summarizes the resorts and neighborhoods identified within each the primary high rate and primary low rate cluster.

Cluster Resorts Neighborhoods Number of streets Number sampled

Morgenstond 110 14 Clevia 27 3 Leonsberg 22 3 Citrus Plantage 65 8 Geyersvlijt 54 8 Flamingo Park 12 2 Monplaisir 23 3 Primary Tweekinderen 47 6 Blauwgrond* High rate Uiten Thuis 15 2 Maretraite 4 46 6 Maretraite 5 51 6 Maretraite 6 36 4 Elizabethshof 41 5 Tourtonne 4 24 3 Tourtonne 5 42 5 Tourtonne 6 39 5

Aurora 17 2 Primary Pontbuiten* Winitiwai 4 - Low Rate Pontbuiten 51 8

Menckendam 21 3 Latour 24 3 Latour* Kamta 20 3 Ephraimszegen 128 18

Flora* Half-Flora 57 8

Table 2. Resorts, neighborhoods and number of streets that fall within each cluster. *Incomplete list of neighborhoods because not all neighborhoods fall within a cluster.

The clusters’ neighborhoods and streets were used in the sampling design for the environmental and entomological surveys.

114

115

Hypothesis 3 – Prediction models and dengue disease cluster analysis can identify effective prevention and risk reduction strategies for dengue in Suriname

Aim 3.1.Characterize dengue-related entomological and environmental data in previously identified high transmission risk

For our sampling, we followed a systematic random sampling design. The high and low rate cluster where the surveys took place were different sizes. We were not able to obtain population census or household information by neighborhood.

Therefore, the number of streets within each cluster (N) and neighborhood (n) were used to calculate the sampling interval k (k = N/n). The high rate cluster had

933 streets while the low rate cluster had 688. For each, the neighborhood with the highest number of streets was used to calculate ‘k’. For each neighborhood, a list of streets was generated and the first street was randomly selected. Within the high rate cluster neighborhoods every 8th and within the low rate cluster neighborhoods every 7th street was selected for sampling. The total number of streets sampled per neighborhood is depicted in Table 2. Before surveying started, a map for each neighborhood with the selected streets highlighted was created (Supplemental Figure 4). On each selected street, every other house was selected for sampling and the geographic location recorded with a GPS device (Figure 1 Section 5).

As described in Section 5, environmental and entomological surveys took place in November 2013 and February 2014. The monthly weather data was used to determine that there was no statistically significant (p<0.05) correlation between

116 monthly average temperature, monthly average precipitation, monthly total precipitation, monthly average relative humidity and the monthly number of dengue cases (Table 3).

Correlations Std. Mean 1 2 3 4 5 Dev 0.010 0.025 0.068 0.060 1 Dengue cases 45.95 53.57 1 p=0.911 p=0.708 p=0.420 p=0.479 Ave. Temperature -0.707 -0.566 -0.563 2 27.70 0.76 1 (degrees) p<0.001 p<0.001 p<0.001 Ave. Relative 0.721 0.725 3 77.62 3.80 1 Humidity (%) p<0.001 p<0.001 Ave. Precipitation 0.998 4 6.24 3.66 1 (mm) p<0.001 Total Precipitation 5 189.67 111.05 1 (mm) Table 3. Descriptive statistics and Pearson’s correlation matrix for monthly temperature, relative humidity, and precipitation averages and precipitation and dengue cases.

As mentioned above, Suriname is marked by the four weather seasons: depicted in Table 4.

Ave. Total. Ave. Relative Ave. Total Season Months Precipitation* Precipitation* Humidity* Temperature* Dengue‡ SRS Dec. – Feb. 6.23 192.81 79.38 27.09 1437 SDS Mar. – Apr. 5.94 175.42 75.59 27.26 1173 LRS May – Aug. 7.94 242.26 79.09 27.59 2478 LDS Sept. – Nov. 3.63 109.52 75.09 28.58 1529 (SRS – short rain season, SDS – short dry season, LRS – long rain season, LDS – long dry season). *ANOVA P<0.05 ‡Kruskal-Wallis H-Test Table 4. Average and total precipitation, average relative humidity, average temperature and total dengue cases by season

Average and total precipitation, average relative humidity, and average temperature were normally distributed and there was a statistically significant difference in the means between the fours season for all. The total number of dengue cases were not normally distributed and no difference detected in the

117 seasonal median (p=0.395) (Table 4). Furthermore, no significant difference was detected between seasonal total dengue cases and the seasonal meteorological averages (Table 5). Dengue simulation models at local and global levels have demonstrated that there is a relationship between climate-induced variations in mosquitos and cases of dengue in tropical regions (Hopp & Foley, 2003; Focks et al., 1995; Hales et al., 2002) influenced by the biology and oviposition behavior of Ae. aegypti (Service, 2008).

Correlations Std. Mean 1 2 3 4 5 Dev 0.096 0.149 0.193 0.185 1 Dengue cases 137.85 147.40 1 p=0.549 p=0.351 p=0.189 p=0.209 Ave. Temperature -0.610 -0.472 -0.476 2 27.64 0.67 1 (degrees) p<0.001 p<0.001 p<0.001 Ave. Relative 0.700 0.714 3 77.37 2.94 1 Humidity (%) p<0.001 p<0.001 Ave. Precipitation 0.999 4 5.95 2.84 1 (mm) p<0.001 Total Precipitation 5 180.84 85.96 1 (mm) Table 5. Descriptive statistics and Pearson’s correlation matrix for seasonal temperature, relative humidity, and precipitation averages and precipitation and dengue cases.

Therefore, the surveys were conducted at the end of the LDS (November 2013) and the end of the SRS (February 2014). Two questionnaires were used to record house and yard characteristics and container descriptives. The first questionnaire was the BOG’s Aedes Larvae Surveillance Survey modified to include pupae count information (Aedes Larvae/Pupae Surveillance Survey –

Supplemental Figure 5).This questionnaire was used to describe and categorize positively identified mosquito breeding site containers using the classification scheme below:

118

Category 1. Garden attributes (plant pot, plant plates, wheel barrel, watering can) 2. Water storage containers (buckets, drums, barrels) 3. Discarded household items (refrigerator, electric apparatus, toilets, furniture) 4. Garbage (discarded cups, cans, tires, bicycle, car parts) 5. Household use containers (plastic cups, small Tupperware, kettles, pans) 6. Recreation (swimming pool, boat, fountain, fish tank) 7. House structures (septic tank, drainage points) 8. Natural habitats (tree holes, bromelias)

A second surveillance tool was developed to quickly record container characteristics in the field (Supplemental Figure 6). In total, 536 houses and 2408 containers were surveyed. Section 5 Figure 1 displays the locations of all the surveyed households within each cluster. Furthermore, we identified differences in the population density and ethnic composition of the neighborhoods between the two clusters. Overall, houses and yards classified as being in bad condition were more likely to be positive for Ae. aegypti larva in the low rate cluster during both seasons (Table 1 & Table 2 Section 5). The entomological indices were used to compare abundance of breeding site between the two clusters during both seasons (Table 3 & 4 Section 5). Unexpectedly, our field observations indicated that during the short rain seasons, changes of peoples’ behavior in both the LRC and the HRC affected the total number of containers available as breeding sites as well as the number of positively identified Ae. aegypti breeding sites. These findings have significant implications for vector control programs in

Paramaribo. Moreover, there was an association detected between the presence of Ae. Aegypti pupae and water-holding containers usage status, material, capacity, circumference, disposability, cover status and water source (p<0.01).

119

Aim 3.2.Design effective, low-cost dengue intervention activities based on local evidence-based information to reduce the risk of dengue transmission in

Suriname

In 2000, the BOG created the ‘Plan of Action to Control Dengue and Dengue

Epidemics in Suriname’ issuing recommendation on how to combine epidemiological and entomological data to inform control activities against the three lifecycle stages (eggs, larva and adult) of Ae. aegypti. Under this plan, the

Urgent Dengue Control Program was formed to implement stepwise strategies to deal with cases of severe dengue, dengue clusters and epidemics (BOG 2000).

However, the research described in this dissertation indicates that no effective plan of action has been implemented. Moreover, no structural base for larval and mosquito control or for dengue collection has been implemented.

Currently, we found that vector control response strategies occur on an ad-hoc basis. Health practitioners or the general public report cases of dengue or mosquito infestation to the BOG prompting environmental inspectors to visit the affected premises and control for larva using Bti. However, while conducting the environmental surveys, people often indicated calling the BOG to report mosquito infestations but no inspection was carried out. The BOG, like many governmental public health organizations has budgetary constrain, is understaffed, and in charge of all public health challenges, not only dengue. However, progress is being made in the right direction. Right before the November 2013 environmental surveys began, the BOG environmental health inspectors underwent a one-week training to search, identify and collect mosquito eggs, larva and pupae.

120

Furthermore, GIS software and GPS devices were purchased to better document environmental inspections.

The mosquito-surveillance training and the implementation of GIS technology are instrumental in developing low-cost dengue intervention activities. In February

2014, the Director of the Environmental Inspection department at the BOG provided an opportunity to demonstrate how the data collected during the environmental surveys could be quickly incorporated into a GIS platform to visualize the situation on the ground. Together with two environmental inspectors, we modified both surveillance tools (the Aedes Larvae/Pupae

Surveillance Survey and the Container Characteristic Survey) to integrate as a

GIS spreadsheet that could be accessed on the GPS devices in the field. In this manner, during routine (non-dengue related) inspections, the inspectors can include breeding site information by simply pulling up the spreadsheet in their

GPSs. Thus, for the first time in Suriname, mosquito breeding site information is collected as part of routine environmental inspections rather than based on the reporting of dengue cases or mosquito infestations. However, any future conclusions drawn from breeding site identification will be incomplete as long as the mosquitos’ DENV infection status is not tested.

An essential component for any local evidence-based strategy is whether DENV and which serotype is circulating in the mosquito population. Upon mosquito breeding site recognition during an inspection, trained inspectors could collect a random sample of eggs, larva or pupae to bring back to the Central Laboratory for emergence and species identification. These are low-cost activities that

121 confirm the presence of Ae. aegypti but to check for the DENV requires expensive equipment and extensive knowledge. Collected mosquitos can be frozen and sent to national or international laboratories for dengue confirmation.

This has to be done in a timely manner because dengue epidemics usually result from the introduction of a new DENV serotype that circulates within the local mosquito population before transmission to humans occurs. Thus allowing for a more effective and proactive vector-control, dengue case identification and education strategy.

In addition to real-time vector information, real-time dengue case surveillance is necessary to reduce the risk of dengue transmission. In Suriname, hospitals and clinics report all their suspected dengue cases to the BOG. However, while visiting some hospitals, we realized that there was no standardized practice on how to report the cases. All are required to fill out the same dengue form, however, some hospitals send each report individually, while others wait until the end of the week to mail all the reports or until enough have accumulated. Thus, there is a lag from when a case is diagnosed with dengue until it is reported to the BOG. The reports are not digital and once the BOG receives the forms, manual input into the national dengue database is not done the same day.

Standardization and digitalization of dengue case reporting would significantly improve surveillance in the country. However, these are not low-cost measures and their effective implementation is beyond the scope of the research presented.

122

8. Discussion

This is the first study to explore demographic, clinical and environmental factors that influence the transmission of dengue in Suriname. Taking advantage of existing dengue databases, we performed a retrospective analysis of clinical and demographic data to examine the national dengue situation. In addition to the epidemiological data, population and climatological information were integrated to create a more robust picture of dengue transmission. These data, local knowledge from Surinamese health practitioners, and input from environmental health researchers, epidemiologists, biostatisticians, medical and entomological experts was used to guide the field surveys described in this dissertation.

Aim 1.1.Characterize the incidence, prevalence, and severity of dengue in

Suriname from 2001 to 2012

Digital records of dengue cases to the BOG exist from 2001 onward. The first cases of DF and DHF were recognized in the country in the early 1980s and for the next twenty years, few cases were detected regularly throughout the population and incidence remained low. However, in 2005 an unprecedented dengue outbreak is estimated to have affected 1 in every 10 people in Suriname creating disease awareness among the public, healthcare professionals and the government. During this outbreak, DENV 1, 2, and 3 were detected circulating in the population. The DENV 2 was identified to be the Asian subtype, a genotype associated with higher virulence (Rico-Hesse et al 1997, Carrington, et al 2005).

2005 was also an El Nino year accentuating climatological conditions that could favor the transmission of dengue (Kovats et al 2003, Hopp & Foley 2003,

123

Ferreira 2014)Thus, DENV 2 Asian subtype and El Nino are two factors that could have contributed to this up-scale in cases.

To date, dengue outbreaks have not reached the magnitude of the 2005 but more severe outbreaks have occurred. In 2009, 38% of all reported dengue cases were classified as DHF. Unfortunately, there is no information on the

DENV serotype for that year. It is understood that the introduction of a new serotype into a population often results in large outbreaks. The heterogenic immunity to DENV serotypes is short-lived (Carrington et al., 2005). Thus, overtime more people are likely to become sensitized to severe disease upon a secondary dengue infection changing the population’s immunological status

(Halstead & O’Rourke, 1977, Guzman et al., 2002a, Montoya et al., 2013). The incidence and severity of the 2012 dengue outbreak was much lower compared to 2005 and 2009, respectively. This outbreak did strain healthcare resources and an additional hospital had to be opened to accommodate dengue-infected patients (AFP, 2012, Stabroek News, 2012). The scale of the response to the

2012 outbreak could be due to better dengue recognition and diagnosis practices and the government’s quick resolve to take action to control the spread of dengue. Moreover, it illustrates the lack of preparedness the healthcare and governmental community would have in confronting an epidemic with the magnitude of 2005 or severity of 2009.

The characterization of dengue in Suriname using the national database from the

BOG also highlights that certain demographic are more affected by the disease.

Incidence of DHF is higher in children under 14 and males. Furthermore,

124 ethnicities are disproportionately affected: Caucasians and Chinese exhibited much higher incidence of DF and DHF compared to others. Both these ethnicities have been documented as having a higher proclivity to severe forms of dengue.

Caucasians exhibit a higher cross-reactivity upon a secondary dengue infection

(de la Sierra et al 2006, Restrepo et al., 2008). Research into single nucleotide polymorphisms (SNPs) of genetic variants in Asian groups is ongoing (Chang et al., 2013) in efforts to find genetic differences that explain the documented severity of dengue in Asia (Halstead, 2006, Huy et al., 2013). On the contrary,

African ancestry has been identified to protect against severe forms of dengue

(Blanton et al., 2008, Chacon-Duque et al., 2014). This genetic variation is important when considering the ethnic composition of Suriname as it could explain some of the differences in DHF incidence between districts. The two most populated districts, Paramaribo and Wanica, tend to have a larger

Southeast Asian population as well as a concentration of Caucasians (mostly

Dutch and Belgian tourists) and Chinese compared to other more rural districts.

Aim 1.2 Examine the environmental, demographic and clinical factors that influence the development of the severe form of dengue in Suriname.

Univariate and multivariate analyses using both the BOG and SVZ further exemplified which environmental, demographic and clinical factors are associated with DHF. In both the BOG and SVZ dengue cases samples, children under the age of 14 were associated a higher frequency of DHF. Age has been identified as a risk factor for severe dengue (Guzman et al., 2002b, Halstead et al., 2002, Hammond et al., 2005). However, other known risk factors such as

125 ethnicity were not found to be statistically associated with DHF in our best fit model. The BOG model has a good fit but is not predictive enough to conclusively predict which factors affect progression into DHF. Additional information, such as comorbidities, viral load, DENV serotype, platelet count, previous dengue infections, etc. would strengthen the predictive power of our model. This clinical information could further explain the progression of dengue disease in Surinamese patients but does not necessarily explain the difference in

DHF reporting. For example, in the model the odds of DHF in Nickerie/Coronie are significantly lower compared to other districts. The BOG database contains all the reported cases of dengue thus the discrepancy could also be influenced by extrinsic factors such as the population’s access to care, health seeking behaviors, local beliefs and practices. A comprehensive dengue predictive model has to include extensive demographic and clinical data as well as confounding information on population access to health, local mobility, non-medical treatments and people’s perceptions and beliefs on the causes and treatment of

DF. The regression analyses are a first attempt at understanding and identifying factors within the Surinamese population that affect the severity of dengue and demonstrate the need for more accurate reporting.

The SVZ database analysis findings also highlight the importance of good diagnostic practices in the treatment of DF. In 2011, the AZP laboratory adopted

RT-PRC technology to confirm a dengue diagnosis. This laboratory detection of the DENV has vastly improved in-country diagnostic capabilities and enabled healthcare practitioners to implement the best course of treatment. This is

126 reflected in the shorter hospitalization time of confirmed dengue cases compared to probable and suspected ones and in the lack of significant difference between hospitalization times of confirmed DF and DHF cases. DF can slowly progress into DHF and not having a reliable diagnostic test requires doctors to keep the patient under observation for longer. Thus, the presence or absence of the DENV in a suspected case of DF enables doctors to make more accurate treatment decisions. RT-PCR dengue diagnosis does have its limitations because there is a short window of time in which the blood sample has to be collected (up to 7 days post infection), it requires expensive equipment and trained personnel to operate and results are not immediate. As reflected in the PATH analysis, dengue diagnosis affects hospitalization time while both are affected by demographic factors. The model explains little of the variance in the data probably which could be due to other variables that influence both outcomes not included in the model.

The conversion into bivariate categories and Yule’s transformation of the OR enabled us to generate a framework for categorical variables that could be greatly enhanced to identify risk factors that could predict severe dengue by the inclusion of more factors.

Currently, the WHO is trying to implement standardized definitions of dengue and severe dengue rather than classifying cases as DF and DHF (WHO 2009). Many developing countries, including Suriname, have failed to adopt these new definitions because those are more cumbersome and require more clinical testing to confirm the diagnosis (Kalayanarooj, 2011, Hadinegoro, 2012). The adoption, however, would benefit research as it would be easier to conduct

127 systematic reviews and metadata analysis from different countries. The DF and

DHF definitions are broad and subjective therefore viral detection is necessary to conclusively determine what factors affect the progression into severe dengue.

Aim 2.1 Integrate clinical and demographic information of all dengue cases reported in Suriname since 2001 into a GIS platform

Spatial visualization of dengue-related factors, such as cases, vector distribution, and land topography has contributed to dynamic models of dengue transmission.

In Suriname, an extensive geodatabase file compiled by the Land and Forestry

Planning (Ruimtelijke Ordening, Grond and Bosbeheer – RGB) includes district, resort, streets, rivers, canals, settlements, forests and topology shapefiles. This file did not contain population information. Therefore, to create the density maps new layers with population census data were created. The biggest challenge was incorporating individual dengue case information because the street code in the street shapefile did not match the street code assigned by the BOG which follows the ABS coding. This demonstrates a lack of communication and collaboration among the different ministries. We cross-referenced both the shapefile and the

ABS street codes to the street name by resort creating a shapefile that the BOG can use to map dengue cases, as well as other diseases, by referencing the ABS code. To furthermore improve these maps, it is necessary to include a more defined scale by incorporating premises (schools, parks, and hospitals) and units

(houses) polygons. In this manner, an identified dengue case can be visualized at the house rather than street level.

128

We used the improved street shapefile to input and visualize 10 years of dengue cases. At a glance, there is a concentration of cases in Nickerie, Wanica and

Paramaribo. These three districts are also the most populated therefore the density maps clarify whether incidence of disease is higher too. Thus, these maps show where dengue is a problem and time-lapse maps can help visualize a change in case concentration and incidence overtime. Inputting dengue reports to the BOG in a timely manner could be used to identify where the infected person lives and recognize whether it is a single case, a cluster or the beginning of an epidemic. Dengue transmissions do not always happen at home even though dengue risk does increase in households around index cases (Mammen et al., 2008, Anders et al., 2015). Thus, early recognition of dengue distribution can aid the BOG in understanding, managing and preventing the spread of dengue in a resource efficient manner.

Aim 2.2 Identify spatial and temporal and historic clusters of DF and DHF from

2001 to 2012

We identified two high epidemiological rate clusters (HRC) in Paramaribo and two low ones (LRC) in Paramaribo and Wanica from 2002 to 2012. Interestingly, some of the resorts with the highest population density (such as Flora and

Latour) had the lowest concentration of dengue cases over the 11 year span.

The resorts within the HRC and the LRC also differ in the percent of the population that has access to piped water and electricity both being lower in the

LRC. Conventionally, inadequate water supply and concentrated populations has been associated with more dengue cases (Cummings et al., 2004; Braga et al.,

129

2010; Schmidt et al., 2011), however, the retrospective spatial analysis indicates a statistically significant number of dengue cases in the LRC. Therefore, prompting the need to conduct field surveys to compare vector and environmental conditions between the clusters in order to identify factors that could affect the discrepancy in dengue cases. This is the first time dengue epidemiological data was spatially displayed, analyzed and used to inform environmental and entomological surveys in Suriname.

Aim 3.1 Characterize dengue-related entomological and environmental data in previously identified high transmission risk areas

Weather conditions, such as precipitation, temperature and humidity influence dengue transmission because it limits or augments Ae. aegypti potential breeding sites (Hopp & Foley, 2003, Altizer et al., 2006). Dengue outbreaks typically follow the rainy season and in Suriname a spike in cases have been reported after the

LRS and the SRS. Our weather and dengue cases correlation did not yield statistically significant results and no association was found between the four different seasons. However, previous dengue epidemics peaked in August

(towards the end of the LRS) and in January-February (at the end of the SRS and the beginning of the SDS) therefore we chose to conduct our surveys in

November before the start if the SRS and in February towards the end of the

SRS.

Since the 2012 dengue epidemic, the BOG conducted Ae. aegypti breeding site surveys in Albina, Marienburg and Commewijne (Hiwat, 2012a & Hiwat, 2012b,

130

Hiwat et al., 2013) that guided the different container categories and design of the Ae. aegypti Larva Surveillance Surveys. This survey was modified to accommodate pupal surveillance as pupae are a better measure of dengue transmission in a community (Focks et al., 2000, Focks & Alexander, 2006).

Unfortunately, adult Ae. aegypti collection was not possible during the execution of the surveys due to a lack of resources. Nevertheless, pupal indices revealed seasonal and cluster differences associated with the transmission of dengue.

The surveys indicated that there is a strong association between the house and yard condition in HRC but not in the LRC. There is also an observed association between detection of positive Ae. aegypti breeding sites and not having access to piped water. In general, the houses and yard in the HRC were in better condition compared to LRC and field observations noted better infrastructure

(paved roads, no flooding) in the HRC. Therefore, it was unexpected to find more

Ae. aegypti pupae in the HRC during the SRS. We observed a behavioral change in both clusters between seasons. Before the start of the SRS, people in

LRC were more likely to use drums to collect water due to either a lack of or unreliable access to piped water. Once the rains began, fewer drums were needed to collect enough water for their daily use and these drums would often overflow after a heavy night of raining. Thus, fewer and shorter potential oviposition containers were available to Ae. aegypti. The more affluent neighborhoods in the HRC had kept gardens in which opportunistic breeding sites, such as plant plates and pots, were tended to by homeowner before the rains started. However, during heavy rain periods people are less willing or able

131 to keep up with their gardens increasing the number of Ae. aegypti breeding sites. The pupal indices were not a good measure of dengue transmission for the purpose of our study but they did allow us to compare the clusters during both seasons and infer on other factors that influence the number of Ae. aegypti breeding sites.

Pupae were more likely to be found in high volume, large circumferential opening plastic containers. Ae. aegypti breeding site preference is influenced by container availability. It is interesting to see a concentration of pupae dispersion during the

SRS in the HRC, while the LRC remains the same, when the heavy rains create more opportunistic water-holding containers. This decrease in the HRC in the dispersion index could be due to mosquito breeding site preferences. In the LRC, however, the dispersion index remained constant while the number of available containers decreased. We observed that many of the containers left to collect rain would overflow creating short-term breeding sites in which the eggs, larva and pupae had a decrease in survival. Similar variations between dry and wet seasons have been recorded for in other countries where the number of containers and the type of containers available vary between seasons (Quintero et al. 2014). During the surveys, we did not inspect the inside or neither second floors of houses nor the roof gutters limiting our conclusions to only outside containers. More detailed container studies are also necessary to understand

Ae. aegypti breeding site preferences in Suriname.

Furthermore, the socioeconomic status (SES) and ethnic composition between both clusters differed. The LRC is predominantly Maroon and Creole while the

132

HRC is composed of mostly Javanese and Hindustani people. As mentioned before, people of African descent are less likely to develop severe dengue compared to Asians or Caucasians. Thus, mild dengue could be a reason why these densely populated neighborhoods have fewer cases of reported dengue.

We were not able to obtain SES information at a resort level but percent electricity, percent piped-water and unpaved road or garbage collection services allow us to conclude that the SES is much lower in the LRC. The low SES and poor infrastructure in the LRC potentially influences health-seeking behaviors contributing to the discrepancies found between the clusters. More research is needed to determine which factors affect the infected person’s willingness and ability to access care. Such behavioral information is necessary to complement clinical and environmental dengue research presented in this thesis.

Aim 3.2. Design effective, low-cost dengue intervention activities based on local evidence-based information to reduce the risk of dengue transmission in

Suriname

Dengue is a dynamic disease. The transmission of the DENV is affected by host behavior and immunology, the virulence of the virus, the adaptive capacity and preferences of Ae. aegpti and environmental conditions ranging from sanitation practices to climatological phenomena like El Nino. Therefore, addressing one aspect of dengue transmission can help reduce the risk of transmission but is not an effective way to combat this disease. The Pan American efforts in the middle of the 20th century were aimed at eradicating Ae. aegypti. The success of this campaign was reflected in reduction of Yellow Fever cases and contributed to the

133 delayed introduction of DENV. Currently, there are no multinational, government sponsored efforts to eradicate Ae. aeypti and the focus has switched to control and reduction of the vector rather than elimination.

In Suriname, effective and timely mosquito control has to be complemented with real-time reporting of dengue cases. The current efforts of the BOG to incorporate GIS and mosquito surveillance into their routine household and private and public non-residential inspections are the first step to proactively address Ae. aegypti breeding sites within the population. These efforts can be strengthen by a continual, random collection of larvae and pupae samples to be tested for DENV infection status. There is a lagged relationship between dengue- infected mosquitos and dengue cases emerging in an affected human population. Recognition of infected mosquitos can prompt vector-control activities before dengue clusters arise.

It is necessary to obtain near real-time information of dengue cases reported in point-of-care settings to better understand patterns of dengue transmission. As exemplified by Bowman et al, few studies have collected concurrent real-time dengue case and vector data (Bowman et al., 2014). A lag in dengue case reporting, which includes the time a symptomatic person takes to seek care and the speed with which health care practitioners take to report to the BOG, can decrease the impact and success of vector control practices. Mosquito foci tend to be temporary and even though there is an association between an index dengue case and household clusters (Anders et al., 2015) not all mosquitos bite at home. Therefore, the incorporation of private and public non-residential vector

134 surveillance practices during the BOG inspections is a step in the right direction in reducing the risk of dengue transmission.

To strengthen surveillance efforts, dengue case information should also be incorporated into a real-time GIS platform to complement the vector data. An efficient system of case reporting is not currently in place in Suriname where dengue case reports in paper form are mailed to the BOG. A systematic, standardized, digital platform requires governmental and financial support and political will. Smaller improvements can be made by expanding the use of GIS into the epidemiology branch of the BOG rather than limiting it to the environmental health department. Visualization of mosquito breeding sites and dengue cases (even if there is a time difference) would allow us to explore the relationship between the mosquito, the virus, the host and the environment to shift dengue control measures to predictive and preventive in Suriname.

Because dengue is understudied in Suriname, no intervention would be complete without an educational component. As indicated by the surveys, different populations even within the same city have different vector-control needs and one simple unified message is not enough to create awareness in Suriname. As a next step, dengue perception surveys to understand what the population knows and how they deal with dengue and Ae. aegypti is essential to design effective educational campaigns. Suriname is composed of different ethnicities that have different cultural practices and beliefs influencing how they view and treat dengue. Therefore, a comprehensive and representative sampling of the

135 population should illustrate the best ways to inform and educate people on dengue prevention practices.

9. Conclusions and Recommendations

Dengue has become hyperendemic and a significant public health problem in

Suriname. The retrospective analysis of the BOG and SVZ databases show that since 2001 the frequency, magnitude and severity of dengue have increased as the disease became endemic in the population. Exogenous factors, such as disease awareness and better diagnostic practices, probably account for some of the increase in dengue reporting. Therefore, it is necessary to standardize clinical surveillance and diagnostic practices. During inter-epidemic periods, dengue cases are underreported due to the generic symptoms of mild-dengue and the reliance on rapid blood tests. During an epidemic, over reporting is a result of hyper-awareness among health care practitioners and the general public and an over utilization of resources that limits diagnostic and treatment capacities. A shift towards virus-detection diagnostic practices, such as the AZP laboratory started in 2011 with the introduction of RT-PCR technology, will enhance surveillance in

Suriname if this practice becomes ubiquitous and provides timely information of the DENV serotypes in circulation. Furthermore, a laboratory-confirmed diagnosis reduces the probability of false positives, enhancing research validity and reliability.

Certain demographic characteristics, such as age and ethnicity, increase the risk of severe dengue in Suriname. The predictive model developed in this

136 dissertation is limited in its predictive power. To enhance the model, more specific dengue case information is necessary. The associations between age and ethnicity and DHF we observed have been reported in the literature as risk factors for severe dengue. However, dengue has not been endemic in Suriname as long as in Southeast Asian countries and we suspect that as dengue becomes more established, the disease dynamics within the Surinamese population will change. It is necessary to continue to develop models that identify vulnerable populations to put in place public health practices that mitigate dengue risk factors and strengthen surveillance based on local research.

Dengue is an urban problem in Suriname. Most of the Surinamese population concentrates in the two major cities, Paramaribo and Nickerie, and along the coastal region. An increased degree of urbanization in recent years has created unplanned expansions of these cities which are not prepared to deal with the density. Therefore, it was expected to see a concentration of DF cases in the two major cities and along the coast line. DHF frequency and incidence was also higher in areas with high population concentration but was notably lacking in

Nickerie. A significantly lower number of reported DHF cases in Nickerie is likely due to a discrepancy in recognition or reporting rather than in the treatment of

DF. It is unlikely that there is an unrecognized protective factor against severe dengue within the population. The ethnic composition is similar to that of

Paramaribo and is predominantly Hindustani and Javanese. Thus, further dengue research in Nickerie is necessary to elucidate the situation.

137

There are historic clusters of dengue in Suriname, particularly in Paramaribo.

Taking into account population density, high rate clusters within Paramaribo presented in the North while a low rate cluster was recognized in the South. A high concentration of dengue in the North was expected since cases are usually first reported in the northern neighborhoods before spreading to the rest of the district and neighboring districts. It was unexpected to find a low rate cluster in the South which has the highest population density and a less developed infrastructure compared to North Paramaribo. However, the ethnic composition of the population also differs between North and South. In the Northern resorts people of Southeast Asian descent predominate while Maroon and Creoles compose the majority of people in the South. Environmental and entomological factors can also affect the transmission of dengue therefore these clusters were of significant value to our field research. Epidemiological maps have public health practice implications particularly in developing countries were health departments are often underfunded and have limited resources to conduct large scale research.

The is research represents the first time dengue epidemiological data are used to guide environmental surveys in Suriname. The surveys indicated that there were environmental and entomological differences between the two clusters that can be explained by the entomological indices. However, the pupal indices are not a good measure of dengue transmission because areas with higher number of dengue cases should have higher documented indices. Having real-time case information for the two clusters would have enabled us to correlate the indices to

138 number of dengue cases. The entomological information also indicates that weather seasons have an influence on breeding site availability and influences the presence of Ae. aegypti in each cluster differently. This has implications for vector-control activities because a unified strategy might not have the same impact on different areas within Paramaribo. More research and the incorporation of vector surveillance into routine environmental health inspections will reveal spatial pattern of mosquito foci usable in the development of targeted control practices in Suriname.

Human behavior is also an important component of the vector – host transmission dynamic. During the SRS, we expected an increase in Ae. aegypti breeding sites in the LRC where previously we had identified more available containers and observed poor infrastructure. However, in both the HRC and the

LRC people’s behavior towards heavy rains increased or decreased the number of available breeding sites, respectively. Behavior has important public health repercussions because there is a uniform educational campaign on how to decrease your personal risk of getting dengue that is unlikely to yield maximum results. To the contrary, educational campaigns have to take into consideration the different infrastructure problems, traditional and cultural beliefs and current knowledge and attitudes people have towards dengue. Suriname’s ethnic diversity is unique in South America creating opportunities to study to what extent demographic and environmental factors influence the risk of dengue transmission.

139

In conclusion, dengue prevention and control in Suriname has to happen at the intersection of science, practice and policy. As is depicted in the framework, efforts in practice to incorporate GIS into vector surveillance or the introduction of

RT-PCR technology fill in previous knowledge gaps regarding spatial distribution of Ae. aegypti or which serotype is circulating within the population. This local knowledge can be used in research to identify risk factors and the most efficient ways to mitigate. In turn, science can inform public health policy and dengue campaigns to translate research findings into applicable methods that improve the surveillance, diagnostic and treatment of dengue. Any change in or addition to the efforts to control and prevent dengue have to be implemented in a culturally competent and timely manner. New methods to survey, diagnose and treat dengue or survey and control Ae. aegypti have to be monitored when put in practice to assess improvement of the dengue incidence and severity in

Suriname.

Dengue in Suriname Framework.

140

10. Appendix

Supplement Figure 1. Transmission of disease

The triangle of disease for dengue fever includes the human host, the environment and the dengue virus. In addition, a vector, Aedes aegytpi is necessary to completed the transmission to humans. Additional external factors influence the rate of transmission.

141

Supplement Figure 2. Demographic and clinical factors that influence DHF

Based on Supplemental Figure 1, additional factors that influence the transmission of disease can influence the progression of DF into DHF.

142

Supplement Table 1. Larval and Pupae Indices

Index Description Levels of Risk Limitations

CI Container Safety Zone: - Does not Index # of containers w/ larvae or pupa CI ≤ 10%* account for # of containers examined X 100 number of containers/area§

HI House Index Low: -Does not indicate HI < 0.01% # # of houses w/ larvae or pupae Medium: containers/house§ # of houses inspected X 100 HI 0.1 – 5% High: HI > 5%**

BI Breteau Low: -Does not account Index BI < 5 for container # of containers w/ larvae or pupae Medium: productivity§ 100 houses inspected BI 5 – 50 High: BI > 50**

PI Pupae / Unable to -More laborious demographic sustain -Large containers # of pupae in a community survey transmission: hard to survey§§ # of people in a community PI < 0.26 – 1.05***

*Based on yellow fever studies by Connor & Monroe, 1923 **PAHO, 1994 *** Simulation threshold value dependent on temperature and population level of immunity. Focks et al., 1995 §Focks & Chadee, 1997 §§Focks & Alexander, 2006

143

Supplemental Figure 3. Street maps of neighborhoods in HRC and LRC

144

145

146

147

148

Supplemental Figure 5. Aedes Larvae/Pupae Surveillance Survey

149

Supplemental Figure 6. Water-holding containers surveillance tool

150

11. References

ABS. (2012). Statistical yearbook 2011. ( No. 288). Paramaribo: Algemeen

Bureau voor de Statistiek.

ABS. (2013). Statistical yearbook 2012. (No. 296). Paramaribo: Algemeen

Bureau voor de Statistiek.

AFP. (2012). Suriname hit with dengue epidemic, health ministry says. Retrieved

01/25, 2012, from http://medicalxpress.com/news/2012-01-suriname-dengue- epidemic-health-ministry.html

Allicock, O. M., Lemey, P., Tatem, A. J., Pybus, O. G., Bennett, S. N., Mueller, B.

A., Suchard, M. A., Foster, J. E., Rambaut, A., & Carrington, C.V. (2012).

Phylogeography and population dynamics of dengue viruses in the Americas.

Molecular Biology and Evolution, 29(6), 1533-1543.

Altizer, S., Dobson, A., Hosseini, P., Hudson, P., Pascual, M., & Rohani, P.

(2006). Seasonality and the dynamics of infectious diseases. Ecology Letters,

9(4), 467-484. doi: 10.1111/j.1461-0248.2005.00879.x

Anders, K.L., Nga le, H., Thuy, N.T., Ngoc, T.V., Tam, C.T., Tai, L.T.,…

Simmons, C.P. (2015). Households as foci for dengue transmission in highly urban Vietnam. PLoS Neglected Tropical Diseases, 9(2).

151

Barrera, R., Amador, M., & MacKay, A.J. (2011). Population dynamics of Aedes aegypti and dengue as influenced by weather and human behavior in San Juan,

Puerto Rico.PLoS Neglected Tropical Diseases, 5(12), e1378

Benedict, M. Q., Levine, R. S., Hawley, W. A., & Lounibos, L. P. (2007). Spread of the tiger: global risk of invasion by the mosquito Aedes albopictus. Vector

Borne Zoonotic Diseases, 7(1), 76-85.

Blanton, R.E., Silva, L.K., Morato, V.G., Parrado, A.R., Dias, J.P., Melo, P.R., ...,

& Teixeira, M.G. Genetic ancestry and income are associated with dengue hemorrhagic fever in a highly admixed population. European Journal of Human

Genetics, 16(6), 762-765.

BOG. (2000). Draaiboek ter controle van dengue en dengue epidemieén in

Suriname. Paramaribo, Suriname.

Bowman, L.R., Runge-Ranzinger, S., & McCall, P.J. (2014). Assessing the relationship between vector indices and transmission: a systematic review of the evidence. PLoS Neglected Tropical Diseases, 8(5), e2848

Braga, C., Luna, C.F., Martelli, C.M., de Souza, W.V., Cordeiro, M.T., Alexander,

N., …, & Marques, E.T. (2010). Seroprevalence and risk for dengue infection in socio-economically distinct areas of Recife, Brazil. Acta Tropical, 113(3), 234-240

152

Bravo, J.R., Guzman, M.G., & Kouri, G.P. (1987). Why dengue hemorrhagic fever in Cuba? 1. Individual risk factors for dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS).

Burke, D. S., Nisalak, A., Johnson, D. E., & Scott, R. M. (1988). A prospective study of dengue infections in Bangkok. The American Journal of Tropical

Medicine and Hygiene, 38(1), 172-180. Transactions of the Royal Society of

Tropical Medicine & Hygiene, 81(5), 816-820

CAREC. (2012). Report on communicable diseases for weeks 1 - 8, 2012.

CAREC Surveillance Report, 32(1), 1.

Carrington CV, Foster JE, Pybus OG, Bennett SN, & Holmes, E. C. (2005).

Invasion and maintenance of dengue virus type 2 and type 4 in the Americas.

Journal of Virology, 79(29), 14680-14687.

Cachon-Duque, J.C., Adhikari, K., Avendano, E., Campo, O., Ramirez, R., Rojas,

W., Ruiz-Linares, A., Restrepo, B.N., & Bedoya, G.(2004). African genetic ancestry is associated with a protective effect on dengue severity in Colombian populations. Infection, Genetics & Evolution, 27, 89-95.

Chang, C.X., Tan, A.T., Or, M.Y., Toh, K.Y., Lim, P.Y., Chia, A.S.,…, &

Grotenbreg, G.M. (2013). Conditional ligands for Asian HLA variants facilitate the definition of CD8+ T-cell responses in acute and chronic viral diseases.

European Journal of Immunology, 43(4), 1109-1120.

153

Chiewslip, P., Scott, R.M., Bhamarapravati, N. (1981). Histocompatibility antigens and dengue hemorrhagic fever. American Journal of Tropical Medicine

& Hygiene, 30(5), 1100-1105

Connor, M. E., & Monreo, W. M. (1923). Stegomyia indices and their value in yellow fever control. American Journal of Tropical Medicine and Hygiene, sl-3(1),

9-19.

Coronie, N. (2007). Health in the Americas: Suriname. Health in the Americas

2007, 2, 640.

Cummings, D.A., Irizarry, R.A., Huang, N.E., Endy, T.P., Nisalak, A., Ungchusak,

K., & Burke, D.S. (2004). Travelling waves in the occurrence of dengue hemorrhagic fever in Thailands. Nature, 22(427), 344-347. de Haas, R. A., & Arron-Leeuwin, A. E. F. (1975). Arboviruses isolated from mosquitoes and man in Surinam. Tropical and Geographical Medicine, 27, 409-

412. de la C Sierra, B., Kouri, G., & Guzman, M. G. (2007). Race: A risk factor for dengue hemorrhagic fever. Archives of Virology, 152(3), 533-542. doi:

10.1007/s00705-006-0869-x F de Wet, N., Ye, W., Hales, S., Warrick, R., Woodward, A., & Weinstein, P.

(2001). Use of a computer model to identify potential hotspots for dengue fever in

New Zealand. The New Zealand Medical Journal, 114(1140), 420-422.

154

Díaz-Quijano, F., González-Rangel, A., Gómez-Capacho, A., Espíndola-Gómez,

R., Martínez -Vega, R., & Villar-Centeno, L. (2008). Pluviosidad como predictor de consulta por síndrome febril agudo en un área endémica de dengue. Revista

De Salud Pública, 10(2), 250-259.

Diaz-Quijano, F. A., & Waldman, E. A. (2012). Factors associated with dengue mortality in Latin America and the Caribbean, 1995-2009: An ecological study.

The American Journal of Tropical Medicine and Hygiene, 86(2), 328-334. doi:

10.4269/ajtmh.2012.11-0074; 10.4269/ajtmh.2012.11-0074

Duncombe, J., Clements, A., Hu, W., Weinstein, P., Ritchie, S., & Espino, F. E.

(2012). Geographical information systems for dengue surveillance. The American

Journal of Tropical Medicine and Hygiene, 86(5), 753-755. doi:

10.4269/ajtmh.2012.11-0650; 10.4269/ajtmh.2012.11-0650

Eisen, L., & Lozano-Fuentes, S. (2009). Use of mapping and spatial and space- time modeling approaches in operational control of Aedes aegypti and dengue.

PLoS Neglected Tropical Diseases, 3(4), e411. doi:

10.1371/journal.pntd.0000411; 10.1371/journal.pntd.0000411

The feasibility of eradicating aedes aegypti in the americas. (1997). Revista

Panamericana De Salud Pública, 1(1), 68-72

Ferreira, M.C. (2014). Geographical distribution of the association between El

Nino South Oscillation and dengue fever in the Americas: a continental analysis

155 using geographic information system-based techniques. Geospatial Health, 9(1),

141-151.

Focks, D. A., & Alexander, N. (2006). Multicountry study of Aedes aegypti pupal productivity survey methodology: Findings and recommendations. Geneva,

Switzerland: WHO/TDR.

Focks, D. A., Brenner, R. J., Hayes, J., & Daniels, E. (2000). Transmission thresholds for dengue in terms of Aedes aegypti pupae per person with discussion of their utility in source reduction efforts. The American Journal of

Tropical Medicine and Hygiene, 62(1), 11-18.

Focks, D. A., & Chadee, D. D. (1997). Pupal survey: An epidemiologically significant surveillance method for Aedes aegypti: An example using data from

Trinidad. The American Journal of Tropical Medicine and Hygiene, 56(2), 159-

167.

Focks, D. A., Daniels, E., Haile, D. G., & Keesling, J. E. (1995). A simulation model of the epidemiology of urban dengue fever: Literature analysis, model development, preliminary validation, and samples of simulation results. The

American Journal of Tropical Medicine and Hygiene, 53(5), 489-506.

Gagnon, A. S., Bush, A. B. G., & Smoyer-Tomic, K. E. (2001). Dengue epidemics and the el nino southern oscillation. Climate Research, 19, 35.

156

Githeko, A. K., Lindsay, S. W., Confalonieri, U. E., & Patz, J. A. (2000). Climate change and vector-borne diseases: A regional analysis. Bulletin of the World

Health Organization, 78(9), 1136-1147.

Guzman, M. G., Halstead, S. B., Artsob, H., Buchy, P., Farrar, J., Gubler, D. J., .

. . Peeling, R. W. (2010). Dengue: A continuing global threat. Nature

Reviews.Microbiology, 8(12 Suppl), S7-16. doi: 10.1038/nrmicro2460;

10.1038/nrmicro2460

Guzman, M. G., Kouri, G., Valdes, L., Bravo, J., Vazquez, S., & Halstead, S. B.

(2002a). Enhanced severity of secondary dengue-2 infections: Death rates in

1981 and 1997 cuban outbreaks. Revista Panamericana De Salud Publica,

11(4), 223-227.

Guzman, M. G., Kouri, G., Bravo, J., Valdes, L., Vazquez, S., & Halstead, S. B.

(2002b). Effect of age on outcome of secondary dengue 2 infections.

International Journal of Infectious Diseases : IJID : Official Publication of the

International Society for Infectious Diseases, 6(2), 118-124.

Guzman, M.G., Alvarez, M., & Halstead, S.B. (2013). Secondary infection as a risk factor for dengue hemorrhagic fever/dengue shock syndrome: an historical perspective and role of antibody-dependent enhancement of infection. Archives of Virology, 158(7), 1445-1459.

157

Hadinegoro, S. R. (2012). The revised WHO dengue case classification: does the system need to be modified? Pediatrics and International Child Health, 32(Suppl

1), 33-38.

Hales, S., de Wet, N., Maindonald, J., & Woodward, A. (2002). Potential effect of population and climate changes on global distribution of dengue fever: An empirical model. Lancet, 360(9336), 830-834. doi: 10.1016/S0140-

6736(02)09964-6

Halstead, S. B. (2006). Dengue in the Americas and southeast Asia: Do they differ? Revista Panamericana De Salud Publica, 20(6), 407-415.

Halstead, S. B., Lan, N. T., Myint, T. T., Shwe, T. N., Nisalak, A., Kalyanarooj, S.,

. . . Endy, T. P. (2002). Dengue hemorrhagic fever in infants: Research opportunities ignored. Emerging Infectious Diseases, 8(12), 1474-1479. doi:

10.3201/eid0812.020170

Halstead, S.B., & O’Rouke, E.J. (1977). Antibody-enhanced dengue virus infection in primate leukocytes. Nature, 265(5596), 739-741

Halstead, S. B., Rojanasuphot, S., & Sangkawibha, N. (1983). Original antigenic sin in dengue. The American Journal of Tropical Medicine and Hygiene, 32(1),

154-156.

Hammond, S.N., Balmaseda, A., Perez, L., Tellez, Y., Saborio, S.I., Mercado,

J.C., …, & Harris, E. (2005). Differences in dengue severity in infants, children,

158 and adults in a 3-year hospital-based study in Nicaragua. American Journal of

Tropical Medicine & Hygiene, 73(6), 1063-1070.

Hiwat, H. (2012a). Determination of larval indices (Aedes breeding sites) in

Commewijne as a follow-up of Bti introduction in the communities. AFD Proposal

C for Dengue Prevention. Paramaribo: Bureau of Public Health.

Hiwat, H. (2012b). Adult mosquito collections and determination of larval indices

(Aedes aegypti) in Albina. AFD Proposal B for Dengue Prevention. Paramaribo:

Bureau of Public Health.

Hiwat, H., Doerdjan, K., Kerpens, M., Samjhawan, A., Sardjoe, P., & Soekhoe, T.

(2013). Importance of domestic water coutnainers as Aedes aegypti breeding sites in Suriname, implications for dengue control. Academic Journal of

Suriname, 4, 403 – 407.

Hopp, M. J., & Foley, J. A. (2003). Worldwide fluctuations in dengue fever cases related to climate variability. Climate Research, 25, 85.

Huy, N.T., Van Giang, T., Thuy, D.H., Kikuchi, M., Hien, T.T., Zamora, J., &

Hirayama, K. (2013). Factors associated with dengue shock syndrome: a systematic review and meta-analysis. PLoS Neglected Tropical Diseases, 7(9), e2412

Juliano, S. A., & Lounibos, L. P. (2005). Ecology of invasive mosquitoes: effects on resident species and on human health. Ecology Letters, 8(5), 558-574.

159

Kalayanarooj, S. (2011). Clinical Manifestations and Management of

Dengue/DHF/DSS. Tropical Medicine & Health, 39(Suppl. 4), 83-87.

Kovats, R. S., Bouma, M. J., Hajat, S., Worrall, E., & Haines, A. (2003). El Niṅo and health. Lancet, 362(9394), 1481-1489. doi: 10.1016/S0140-6736(03)14695-8

Mammen, M.P., Pimgate, C., Koenraadt, C.J., Rothman, A.L., Aldstadt, J.,

Nisalak, A., ..., & Scott, T.W. (2008). Spatial and temporal clustering of dengue virus transmission in Thai villages. PLoS Medicine, 5(11), e205

Mir, D., Romero, H., Fagundes de Carvalho, L.M., & Bello, G. (2014).

Spatiotempoeral dynamics of DENV-2 Asian-American Genotype Lineage in the

Americas. PLoS, 9(6), e98519

Montoya, M., Gresh, L., Mercado, J.C., Williams, K.L., Vargas, M.J., Gutierrez,

G., Kuan, G., Gordon, A., Balmaseda, A., & Harris, E. (2013). Symptomatic versus inapparent outcome in repeat dengue virus infections is influenced by the time interval between infections and study year. PLoS Neglected Tropical

Diseases, 7(8), e2357

Morrison, A. C., Zielinski-Gutierrez, E., Scott, T. W., & Rosenberg, R. (2008).

Defining challenges and proposing solutions for control of the virus vector Aedes aegypti. PLoS Medicine, 5(3), e68. doi: 10.1371/journal.pmed.0050068;

10.1371/journal.pmed.0050068

160

MZ. (2012). Primary Health Care Suriname. Retrieved 02/02, 2015, from http://www.medischezending.sr/index.php?option=com_content&view=article&id=

68:diensten&catid=28&Itemid=257&lang=nl

Nisalak, A., Endy, T. P., Nimmannitya, S., Kalayanarooj, S., Thisayakorn, U.,

Scott, R. M., . . . Vaughn, D. W. (2003). Serotype-specific dengue virus circulation and dengue disease in Bangkok, Thailand from 1973 to 1999. The

American Journal of Tropical Medicine and Hygiene, 68(2), 191-202.

O’Meara, G. F., Evans Jr., L. F., Gettman, A. D., & Cuda, J.P. (1995). Spread of

Aedes albopictus and decline of Ae. Aegypti (Dipetera: Culicidae) in Florida.

Journal of Medical Entomology, 32(4), 554-562.

Quintero, J., Garcia-Betancourt, T., Cortes, S., Garcia, D., Alcala, L, …, &

Carrasquilla, G. (2014). Effectiveness and feasibility of long-lasting insecticide- treated curtains and water container covers for dengue vector control in

Colombia: a cluster randomized trial. Transactions of the Royal Society of

Tropical Medicine and Hygiene, 109(2), 116-125.

PAHO. (1994). Dengue and dengue hemorrhagic fever in the Americas:

Guidelines for prevention and control. PAHO Scientific Publication, 548

PAHO. (1997). Emergence of dengue in the Americas. Epidemiological Bulletin,

18(2).

161

PAHO. (2002). Health systems and services profile of Suriname. Organization and Management of Health Systems and Services, 1st ed.

PAHO. (2009). Health surveillance and disease prevention and control, communicable diseases, dengue. Retrieved 02/02, 2013, from http://amro.who.int/english/ad/dpc/cd/dengue-cases-2008.htm

PAHO. (2015). Number of reported cases of Chikungunya Fever in the Americas, by country or territory, 2013-2015 (to week noted). Retrieved 02/22, 2015, from http://www.paho.org/hq/index.php?option=com_docman&task=doc_download&It emid=&gid=29127&lang=en

Parker, D., & Holman, D. (2012). Event history analysis of dengue fever epidemic and inter-epidemic spells in Barbados, Brazil, and Thailand. International Journal of Infectious Diseases : IJID : Official Publication of the International Society for

Infectious Diseases, 16(11), e793-8. doi: 10.1016/j.ijid.2012.07.002;

10.1016/j.ijid.2012.07.002

Pinto Severo, O. (1955). Eradication of the Aedes aegypti mosquito from the

Americas. Yellow Fever, a Symposium in Commemoration of Carlos Juan Finlay,

6

RGD. (2015). Stichting Regionale Gezondheid Dienst. Retrieved on 02/02, 2015, from http://www.stgrgd.org/index.php?option=com_content&task=view&id=63&Itemid= 10

162

Rico-Hesse, R., Harrison, L. M., Salas, R. A., Tovar, D., Nisalak, A., Ramos, C., .

. . da Rosa, A. T. (1997). Origins of dengue type 2 viruses associated with increased pathogenicity in the Americas. Virology, 230(2), 244-251.

Rodenhuis-Zybert, I. A., Wilschut, J., & Smit, J. M. (2010). Dengue virus life cycle: Viral and host factors modulating infectivity. Cellular and Molecular Life

Sciences : CMLS, 67(16), 2773-2786. doi: 10.1007/s00018-010-0357-z;

10.1007/s00018-010-0357-z

San Martin, J. L., Brathwaite, O., Zambrano, B., Solorzano, J. O.,

Bouckenooghe, A., Dayan, G. H., & Guzman, M. G. (2010). The epidemiology of dengue in the Americas over the last three decades: A worrisome reality. The

American Journal of Tropical Medicine and Hygiene, 82(1), 128-135. doi:

10.4269/ajtmh.2010.09-0346; 10.4269/ajtmh.2010.09-0346

Sanchez, L., Vanlerberghe, V., Alfonso, L., Marquetti Mdel, C., Guzman, M. G.,

Bisset, J., & van der Stuyft, P. (2006). Aedes aegypti larval indices and risk for dengue epidemics. Emerging Infectious Diseases, 12(5), 800-806. doi:

10.3201/eid1205.050866

Schmidt et al 2011

Schreiber, K. V. (2001). An investigation of relationships between climate and dengue using a water budgeting technique. International Journal of

Biometeorology, 45(2), 81-89.

163

Scott, T. W., & Morrison, A. C. (2010). Longitudinal field studies will guide a paradigm shift in dengue prevention. In P. W. Atkinson (Ed.), Vector biology, ecology and control (pp. 139). Washington, D.C.: The National Academies Press.

Service, M. (2008). Medical entomology for students. Cambridge, UK: Cambridge

University Press.

Simmons, C. P., Farrar, J. J., Nguyen, v., & Wills, B. (2012). Dengue. The New

England Journal of Medicine, 366(15), 1423-1432. doi:

10.1056/NEJMra1110265; 10.1056/NEJMra1110265

Stabroek News. (2012). Dengue continues to worry Suriname. Retrieved 02/02,

2012, from http://www.stabroeknews.com/2012/archives/01/24/dengue- continues-to-worry-suriname/

The World Bank. (2013). Suriname. Retrieved 02/02, 2013, from http://data.worldbank.org/country/suriname

Thein, S., Aung, M. M., Shwe, T. N., Aye, M., Zaw, A., Aye, K., . . . Aaskov, J.

(1997). Risk factors in dengue shock syndrome. The American Journal of

Tropical Medicine and Hygiene, 56(5), 566-572.

UNDESA. (2012). World urbanization prospects, the 2011 revision. ( No. 2013).

New York: United Nations.

Vaughn, D. W., Green, S., Kalayanarooj, S., Innis, B. L., Nimmannitya, S.,

Suntayakorn, S., . . . Nisalak, A. (2000). Dengue viremia titer, antibody response

164 pattern, and virus serotype correlate with disease severity. The Journal of

Infectious Diseases, 181(1), 2-9. doi: 10.1086/315215

WHO. (1997). Dengue hemorrhagic fever: diagnosis, treatment, prevention and control (2nd Edition). Geneva: WHO Press.

WHO. (2009). Dengue guidelines for Diagnosis, treatment, prevention and control (New Edition ed.). Geneva: WHO Press.

WHO. (2012). Dengue and severe dengue. Retrieved 02/02, 2012, from http://www.who.int/mediacentre/factsheets/fs117/en/

WHO. (2013). DengueNet. Retrieved 02/02, 2013, from http://apps.who.int/globalatlas/default.asp

WHO. (2015). Dengue and severe dengue. Retrieved 02/02, 2015, from http://www.who.int/mediacentre/factsheets/fs117/en/

Wills, B. A., Nguyen, M. D., Ha, T. L., Dong, T. H., Tran, T. N., Le, T. T., . . .

Farrar, J. J. (2005). Comparison of three fluid solutions for resuscitation in dengue shock syndrome. The New England Journal of Medicine, 353(9), 877-

889. doi: 10.1056/NEJMoa044057

Wilson, M. E., & Chen, L. H. (2002). Dengue in the Americas. Dengue Bulletin,

26, 44-61.

165