This item has the following downloads:
1 DENGUE EPIDEMIOLOGY, PREVENTION, AND CONTROL IN KEY WEST, FLORIDA By ALI MESSENGER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Ali Messenger
3 To those who made all of this possible
4 ACKNOWLEDGMENTS A special thanks goes out to all of my committee members for their insight s, encouragement, and for always motivating me to set the bar high for future students. I thank my committee chair, Bernard Okech, for his time spent discussing and help improving various facets of my project and his confidence in my ability to complete th is of vector borne diseases and for teaching me the fundamental skills needed for a successful research career. Additional thanks go to Cynthia Lord for her construc tive criticism, advice on analysis, considerations for publications, and time and diligence in proofreading, without it, this dissertation would not be precise, clear, and comprehensive. I thank John Lednicky for his philosophical guidance and for always b eing supportive. I thank Gary Clark for his availability to assist with editing, his eternal optimism and encouraging words. I thank Andrea Leal and the rest of the Florida Keys Mosquito Control District for their countless hours of collaboration; without this assistance, all components within this dissertation would have been impossible. Through this journey, I have no doubt that the love and support from my friends and family is what helped me to complete this degree. I thank my sisters for their love, en thusiasm and unwavering support. My love for them is unconditional, our times together growing up have helped to make me the person I am today and their support through my academic years means more than words can describe. I am grateful to my parents for g iving me the opportunity to continue on in my academic career. I thank these two people for teaching me that hard work, tenacity, and perseverance are some of the most important virtues necessary for the journey to success. There is no doubt in my mind tha t without their help and without the strong sense of family that they have
5 helped to build, I would not have been able to complete this major task. Last but not least, I thank my dear sweet husband for his unconditional love, support and patience through t his extremely difficult accomplishment. Through times of struggle and success, your loyalty and confidence in me motivated me more than you will ever know
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 BACKGROUND AND INTRODUCTION TO RESEARCH ................................ ...... 13 Dengue Virus: The Most Important Arbovirus ................................ ......................... 13 History of Dengue Virus in the United States ................................ .......................... 16 Dengue Scenario in Key West: Focus of Research ................................ ................ 18 2009 2010 Outbreaks and Response ................................ ............................... 18 Hypotheses ................................ ................................ ................................ ...... 21 Origin of the virus ................................ ................................ ....................... 21 Transmission ................................ ................................ .............................. 24 Possible Consequences of Ongoi ng Transmission ................................ .......... 27 Potential Implications of Dissertation Research ................................ ............... 30 2 INTRODUCTION OF DENGUE VIRUS TO KEY WEST, FL BY VIREMIC VISITORS ................................ ................................ ................................ ............... 33 Introduction ................................ ................................ ................................ ............. 33 Methods ................................ ................................ ................................ .................. 33 Data Sources and Assumpti ons ................................ ................................ ....... 33 Viremic Air line Passenger Days Calculation s ................................ ................... 39 Viremic Cruise Ship Visitor Calculations ................................ ........................... 40 Results ................................ ................................ ................................ .................... 43 Discussion ................................ ................................ ................................ .............. 45 3 SEROPREVALENCE OF DENGUE VIRUS IN KEY WEST, FL RESIDENTS ........ 57 Introduction ................................ ................................ ................................ ............. 57 Methods ................................ ................................ ................................ .................. 57 Results and Discussion ................................ ................................ ........................... 61 4 MODELING ANALYSIS OF DENGUE VECTOR CONTROL STRATEGIES IN KEY WEST ................................ ................................ ................................ ............. 73 Introduction ................................ ................................ ................................ ............. 73 Model Background ................................ ................................ ................................ .. 73
7 Residual Spray Background ................................ ................................ ................... 79 Data Sources for CIMSiM Parameterization ................................ ........................... 80 Meteorology ................................ ................................ ................................ ...... 80 Pupal/demographic Survey of Old Town ................................ .................... 81 CIMSiM Parameterization for Key West, F lorida ................................ ..................... 83 DENSiM Parameterization for Key West, Florida ................................ .................... 84 Age specific Seroprevalence ................................ ................................ ............ 84 Virus Introductions ................................ ................................ ............................ 85 Simulation Runs ................................ ................................ ................................ ...... 88 Simulations Part 1: Baseline Control (No Residual Adulticide Applications) ..... 88 Simulations Part 2: Baseline Control and Residual Adulticide Applications ..... 89 Simulations Part 3: Sensitivity A nalysis of Proportion of Adults Resting Outdoors ................................ ................................ ................................ ....... 91 Simulations Part 4: Simulated Elimination of Transmission .............................. 92 Results ................................ ................................ ................................ .................... 93 Pupal/demographic Survey ................................ ................................ .............. 93 Simulations Part 1: Baseline Control (No Residual Adulticide Applications) ..... 94 Simulations Part 2: Baseline Control and Residual Adulticide Applications ..... 95 Simulations Part 3: Sensitivity Analysis of Proportion of Adults Resting Outd oors ................................ ................................ ................................ ....... 96 Simulations Part 4: Simulated Elimination of Transmission .............................. 96 Discussion ................................ ................................ ................................ .............. 97 5 GENERAL CONCLUSION ................................ ................................ .................... 118 Synthesis with Regard to Dengue Control ................................ ............................ 118 Limitations ................................ ................................ ................................ ............. 120 Viral Introductions ................................ ................................ ........................... 120 Seroprevalence Survey ................................ ................................ .................. 121 Simulation Studies ................................ ................................ .......................... 122 Areas of Future Research ................................ ................................ ..................... 125 Dengue Threat to the United States ................................ ............................... 125 Need for Im proved Surveillance ................................ ................................ ..... 126 Research to Guide Policy Changes ................................ ................................ 127 Conclusion ................................ ................................ ................................ ............ 129 APPENDIX: SEROPREVALENCE SURVEY QUESTIONNAIRE ............................... 130 LIST OF REFERENCES ................................ ................................ ............................. 135 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 144
8 LIST OF TABLES Table page 2 1 Highest DENV incidence rates (per 100,000) for dengue endemic countries where airline passengers departed for Key West, Florida as repor ted by PAHO, 2007 2011. ................................ ................................ ............................. 52 2 2 Highest country contributors for arriving airline passengers in Key West, FL, 2007 2011. ................................ ................................ ................................ ......... 53 2 3 H ighest country contributors to visitors arriving in Key West, FL via the cruise ship industry, 2007 2011. ................................ ................................ ................... 54 2 4 Estimated viremic airline industry passenger days in Key West, FL from 2007 201 1. ................................ ................................ ................................ ......... 55 2 5 Estimated viremic cruise ship visitor days in Key West, FL from 2007 2011. ..... 56 3 1 Demographic information of Key West, FL residents comparing participants in the 2009 and 2012 serosurveys and the associated 2010 census report. ...... 69 3 2 Logistic regression analysis results from serosurvey questionnaire data. .......... 71 4 1 Pupal productivity and dimensions by container type obtained during the June 2012 pupal demographic survey. ................................ ............................. 102 4 2 Estimate d reduction of simulated DENV infections in Key West, FL given residual adulticide treatments beginning in the 2009 outbreak. ........................ 115 4 3 Levels of residual adulticides required in 2008 and 2009 t o eliminate transmission for each assumed percentage of adult Ae. aegypti resting outdoors. ................................ ................................ ................................ .......... 117
9 LIST OF FIGURES Figure page 1 1 Weekly time series of reported dengue cases in Key West, Florida, 2009 2010. ................................ ................................ ................................ .................. 32 2 1 Calculation for determining viremic person days in Key West, FL due to airline passengers arriving on flights originat ing in a dengue endemic country .. 51 2 2 Calculation for determining viremic person days due to cruise ship passengers and crew arriving in Key West, FL after visiting each of three precedi ng dengue e ndemic ports of call ................................ ............................ 51 2 3 Y early fluctuations of raw airline passenger counts (total for year) and viremic airline passenger days in Key West, FL, 2007 2011. ................................ ......... 55 2 4 Y early fluctuations of raw cruise ship visitor counts (total for year) and viremic cruise ship visitor days in Key West, FL, 2007 2011. ............................. 56 3 1 Location of seroprevalence study enrollment and reported responded addresses in Key West, FL, March 2012. ................................ ........................... 70 3 2 Distribution of reported home and work addresses of dengue IgG and IgM positive serosur vey respondents in Key West, FL, March 2012 ......................... 72 4 1 Survey area and households examined during the pupal/demographic survey in Old Town, Key West, FL, and June 2012. ................................ .................... 101 4 2 Initial and final population proportions spanning the pre and post outbreak pe riod in Key West, FL, 2008 201 1 ................................ ................................ .. 104 4 3 Daily fluctuations in various weather parameters in Key West, FL, 2008 2011. ................................ ................................ ................................ ................ 105 4 4 Daily densities of Aedes aegypti females and pupae per hectare in Key West, FL reflecting FKMCD control efforts in Key West, FL, 2008 2011 as simulated by CIMSiM. ................................ ................................ ....................... 106 4 5 Daily densities of Ae. aegypti females overlapped with reported weekly dengue cases reflecting FKMCD control efforts in Key West, FL, 2008 2011 as sim ulated by CIMSiM. ................................ ................................ .................. 107 4 6 Daily estimates of female Ae. aegypti per person reflecting FKMCD control efforts, overlapped with weekly reported dengue cases in Key West, FL, 2008 2011 as simulate d by CIMSiM. ................................ ................................ 108
10 4 7 Correspondence of Ae. aegypti females observed in the field with BG Sentinel Traps to predicted daily densities of Ae. aegypti females per hectare from 2009 2011 as simula ted by CIMSIM. ................................ ....................... 109 4 8 Comparison of the correspondence of Ae. aegypti females observed in BG Sentinel Traps to predicted daily densities of Ae. aegypti females per hectare from 2009 2011 as s imulated by CIMSIM to reported daily rainfall. ................. 110 4 9 Predicted infected individuals (latent, symptomatic, and asymptomatic dengue cases) reflecting FKMCD control efforts, overlapped with rep orted weekly dengue cases in Key West, FL, 2008 2011 as simulated by DENSiM. 111 4 10 Daily densities of Ae. aegypti females and pupae in Key West, FL assuming residual adulticide treatments w ere begun in the 2009 outbreak as simulated by CIMSiM. ................................ ................................ ................................ ....... 112 4 11 Daily estimates of female Ae. aegypti per person in Key West, FL assuming residual adulticide treatments were begun in the 2009 outbreak as simulated by CIMSiM. ................................ ................................ ................................ ....... 113 4 12 Predicted reduction in infected individuals (latent, symptomatic, and asymptomatic dengue cases) in Key West, FL assuming residual adulticide treatmen ts were begun July 31, 2009 as simulated by DENSiM ...................... 114 4 13 Sensitivity of residual adulticide to control DENV transmission assuming varying proportions of Ae. aegypti adults resting outdoors as simulated by DENSiM.. ................................ ................................ ................................ .......... 115 4 14 Estimated reduction in predicted of dengue cases in Key West, FL assuming residual adulticide treatments eliminated local tran smission as simulated by DENSiM ................................ ................................ ................................ ........... 116
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DENGUE EPIDEMIOLOGY, PREVENTIO N AND CONTROL IN KEY WEST, FLORIDA By Ali Messenger August 2013 Chair: Bernard A. Okech Major: Public Health Dengue fever is a major global issue. It is the most important mosquito borne viral disease in the world. T he most recent autochthonous transmi ssion of dengue virus (DENV) in the continental United States occurred in Key West, Florida in 2009 and 2010 The work present ed herein aimed to document the epidemiologic scenario in Key West by estimating the potentially sizeable traveler introductions o f DENV to Key West, investig ating the vector that spread it throughout the local population, establishing a post outbreak estimate of the proportion of residents positive for DENV antibodies and therefore at risk for sequential infections and investigatin g the potential benefit of adding residual adulticide to vector control strategies in Key West by conducting simulation studies Results showed that viral introductions by visitors from dengue endemic areas were a significant potential source of DENV to lo cal Aedes aegypti females. Results showed that it is feasible for airline passengers and cruise ship visitors to bring DENV to Key West with seven to 16 viremic airline passenger days and 0.23, 0.89, and 0.55 viremic cruise ship visitor days for the immedi ately prior port, second prior port, and third prior port of call, respectively In March 2012 DENV antibody seroprevalence was
12 an estimated 3.5 percent for recent infections within the previous 90 days (IgM) and 7% for antibodies reflecting infection out side the previous three months (IgG) Inte gration of visitor introduction seropositivity, and vector abundance data into the stochastic container inhabiting mosquito simulation model (CIMSiM) and the dengue simulation model (DENSiM) revealed that current conditions allow for transmission of DENV to occur among local residents and that improved vector control such as the addition of a residual adulticide, could be helpful in the battle toward a long term goal of eliminat ing DENV transmission in Key West, F L
13 CHAPTER 1 BACKGROUND AND INTRODUCTION TO RESEARCH Dengue Virus: The Most I mportant A rbovirus Dengue fever is the most important mosquito borne disease in the world, with 40% of the global population at risk of infection (from one or more serotypes) w ith a projected 5.2 billion individuals to be at risk by 2085. 1 2 Currently dengue virus (DENV) is endemic in over 100 countries spanning the Americas, Caribbean Basin Asia, and Africa, with 50 100 million cases of dengue fever, 250 ,000 cas es of severe infection including dengue hemorrhagic fever (DHF) and dengue shock syndrome (DSS), and 25,000 deaths reported each year. 1 3 4 DENV belongs to the genus Flavivirus family Flaviviridae 5 6 It is a positive strand RNA virus that is composed of four serotypes ( DENV 1, DENV 2, DENV 3, and DENV 4 ) 5 6 Each serotype is unique and can be further subdivided into genotypes or clades. DENV 2 has the highest number of clades (n=6) followed by DENV 1 and DENV 3 (n=4 5 each). DENV 4 is the least genetically diverse serotype with three genot ypes. 6 Because each of the serotypes is unique, an individual may be infected up to four times within their lifetime. 5 Symptoms commonly reported in individuals with dengue fever (a primary infection) include sudden onset of a high grade fever, severe he adache, fatigue, and severe myalgia and arthralgia. 3 The latter two symptoms are the reason for the d Asymptomatic infections are common in individuals exposed to their first serotype; compared to symptomatic cases their frequency can range from 2:1 to significantly higher rates of nearly 40:1. 3 7 DENV 1 and DENV 3 may cause increased disease severity during primary infection while
14 DENV 2 and DENV 4 have been associ ated with more severe illness upon secondary infection. 6 Immunity to only the infective serotype will result after a single infection. 5 Infections with another serotype, or sequential infections, may result in more severe symptoms including DHF, DSS, and death. 5 Individuals with DHF often experien ce increased vascular permeability (plasma leakage), thrombocytopenia, and hemorrhagic manifestations. 6 DSS is marked by hypovolemic shock as a result of fluid leakage into interstitial spaces and without appro priate treatment may result in death. 6 Once a human is infec ted by a female mosquito, an individual typically begins experiencing symptoms in four to seven days, though this intrinsic incubation period can rang e from three to fourteen days. 5 8 Infected persons may begin their viremic period as early as 16 18 hours before th e onset of symptoms. 7 Symptomatic viremia usually persists for four to five days, but may extend as long as 12 days after infection. 7 Humans acquire DENV infections via the bite of an infective mosquito. 5 A mosquito becomes infected after taking a blood meal from a viremic person. Once the blood meal is ingested a nd a few days have passed, the virus is found in the posterior region of the midgut of the mosquito. 9 The virus then progresses throughout the body and can be found in the proventriculus, fat body, ovariole sheath, hemocytes, and nervous system. 9 DENV is later observed in the brain, thoracic and abdominal ganglia before it appears in the salivary glands. 9 Once in the salivary glands of the vector, DENV can be passed on to a susceptible host. The duration of the extrinsic incubation period depends on the size of the virus inoculum, viral titer in the infected host, and quantity of blood consumed by the female mosquito. 8
15 Globally, the most important vector for DENV is Aedes aegypti (L inneas ), a highly domesticated tropical mosquito. 5 These insects prefer to oviposit in artificial containers present in and around residential properties includ ing a variety of containers such as buckets, vases, and various trash items. 5 Th is mosquito is exceptionally anthropophilic and its feeding behavior makes it the ideal vector for dengue viruses Ae des aegypti females are nervous feeders and frequently interrupt their bl ood meals at the slightest movement, only to resume their feeding process moments later on the same or alternate host (up to two or three attempts per blood meal) 5 8 This behavior increases the likelihood for the mosquito to easily acquire and spread the virus. 5 The survival of the vectors is largely dependent upon weather conditions. Daily, seasonal, and yearly variations in atmospheric moisture, temperature, and rainfall all affect the survival in every life stage of the insect. 7 10 For instance, temperature influences bo th the gonotrophic cycle and adult survival W hen the temperature is warm adult survival is increased and the length of the gonotrophic cycle is shortened 7 10 The shorter the gonotrophic cycle within the female mosquito, the more frequently she will seek a blood meal. Additionally, temperature reduces the amount of time required to disseminate the infection in the vec tor. 8 Thus, the combination of increased adult survival, shorter gonotrophic period and reduced extrinsic incubation rate increases the probability of DENV transmission from vectors to hosts Relative humidity influences the survival of eggs prior to submersion in wet containers as well as adult mosquitoes such that moist conditions improve survival of desiccation by the atmosphere For instance when saturation deficit is high, vector surviv al at the adult and egg life stages are reduced Precipitation is another driver of
16 mosquito propagation as it is a controlling factor in the number and availability of oviposition sites. 7 10 However, rainfall is less influential when containers available for breeding are filled manually (e.g., rain is not the only source of clean water for oviposition sites). 8 Globally, transmission of dengue fever is often seasonal, occurring mostly in the rainy season as a result of increased oviposition sites, and thus higher mosquito densities but can also be a result of increased survival due to atmospheric mo isture and temperature in areas where manually filled containers predominate 5 History of Dengue Virus in the United States Historically, dengue fever was endemic in the United States throughout the 1800s until around 1945. 11 12 Dengue epidemics affecting as many as tens of thousands of people occurred every few years. 12 Following World War II, the Pan American Health Organization (PAHO) implemented a program to combat yellow feve r by controlling Ae. aegypti 12 13 Prior to this program, vector control focused on reduction of breeding and oviposition sites, fumigation of adult mosquitoes, and installation of adequate sanitation systems. 12 In addition to these strategies, incorporatin g the organochlorine insecticide dichlorodiphenyltrichloroethane (DDT) led to the successful suppression of DENV in the United States through the 1970s. 12 13 After a rapid increase in urban population densities, reduced support for vector control programs, and the ban of DDT, autochthonous cases were re ported once again. 12 Locally acquired cases returned in 1980 at the Texas Mexico border. 11 Additional autochthonous cases were reported at this location in 1986, 1995, and 2005 as well. 11 14 Outside of this location, and prior to the Key West outbreaks, the only other local transmission of DENV in the United States arose in Hawaii in 2001. 11 15 Despite the lack of endemicity of DENV in the United States since 1945, numerous imported cases of
17 dengue fever occur each year. For instance, outside of the few occurrences of loc al transmission, the majority of dengue cases in the United States in the past 60 years have been imported, travel associated infections. 3 12 Dengue incidence and the number of individuals visiting the American and Asian tropics are on the rise; most imported dengue infections are due to travelers returning from these areas. 5 16 G iven that DENV is not endemic to the United States, outside of episodic transmission events along the Te xas Mexico border, physicians may not include the disease in their differential diagnoses for febrile illnesses. 5 17 18 The epidemiology of the virus also contributes to reduced reporting of cases with the high rate of inapparent infections. 6 Thus, cases observed by physicians us ually represent only a small portion of total cases as only about 10% of symptomatic cases are reported. 5 6 19 Due to the lack of physician awareness of DENV infections and high proportion of asymptomatic infections passive surveillance from healthcare providers may not always be reliable thus more active surveillan ce such as serosurveys are needed to determine which individuals are infected and at risk of sequential infections in the future Currently therapy f or DENV is entirely supportive for symptoms, as no antiviral treatments exist, yet several vaccines are un der development. 3 Because of the current lack of treatments and vaccines, the only method to combat t ransmission and reduced disease spread is via surveillance of human disease and vector control 5 Surveillance in Florida relies largely on the Florida Department of Health which compiles weekly arbovirus reports from passively reported cases by health care professionals. 20 Current vector control strategies for DENV include elimination or treatment of aquatic habitats
18 where immature mosquitoes develop, and/or chemical control via application of insecticides to kill adult mosqui toes 5 H owever these may be costly and ineff ective in various local itie s Without effective control strategies, DENV may continue to spread to susceptible areas across the United States and the rest of the globe An example of this is in Key West, Florida. Dengue Scenario in Key West: Focus of Resea rch Key West is an island roughly four by two miles in size with a variety of ethnic groups resid ing across the entire area. The island can be described as having three major regions: Old Town (Western region), Mid Town (middle region) and New Town (Easte rn region). It maintains a population of roughly 24,000 people and receives nearly two million visitors each year. 21 It is expected that Key West is especially at risk of DENV introd uctions due to the high number of tourists visiting the island annually. 20 Given the warm climate and relaxed atmosphere, residents employ a casual lifestyle (e.g., wearing light layers of clothing, spendi ng the majority of biting hours outdoors, etc.). 21 22 This in addition to the uniquely high counts of Ae. aegypti and abundant breeding locations makes the island especially at risk for imported infections leading to locally acquired infections in the resident population. 21 22 2009 2010 Outbreaks and Resp onse Beginning in early July through mid October 2009, 27 cases of locally acquired dengue fever were reported in residents or visitors to Key West, Florida without travel histories to dengue endemic countries 23 After a hiatus of nearly 6 months, local ly acquired cases were again reported in April 2010, accumulat ing a total of 63 autochthonous infections by the end of November (Figure 1 1). 23 E xcluding cases along the Texas Mexico border, t hese outbreaks marked the first autochthonous cases
19 detected in the continental United States i n over 70 years, and the first locally acquired cases in the state of Florida since 1934. 24 25 As in most dengue endemi c regions across the globe, Ae. aegypti was the responsible vector for DENV infections in Key West residents, laying their eggs and developing in a multitude of ar tificial containers in Key West. These include but are not limited to empty flowerpots, bro meliads, and cisterns, such as those present in, around, and under residences. Vector control strategies and educational efforts were intensified at the beginning of the 2009 outbreak and continue to date, however they may be considered questionable in im ped ing additional local transmission Control strategies conducted in Key West include source reduction and education of the public by Florida Keys Mosquito Control District (FKMCD) domestic inspectors, larval control via aerial application of ultra low vo lume Bacillus thuringiensis israel ensis ( Bti ) and residual treatments of individual con tainer s and adult control by tr uck mounted and handheld ultra low volume applications of adulticide. 26 A potential inadequacy of vector control is evident in additional cases being reported in 2010 in addition to results of an evaluation of percent control of adult Ae. aegypti mosquito counts during Bti aerial applications. To analyze the percent cont rol of adult Ae. aegypti counts (as reported in FKMCD BG Sentinel trap data) by Bti sprays, counts of mosquitoes seven days pre and post aerial Bti applications were evaluated in collaboration with FKMCD 22 Larval control strategies varied in their percent con trol with regards to entomological indices, with only 17 percent control being achieved when considering housing indices, 30 percent for container indices and 15 percent control overall during Bti treatments. This potential inadequacy of control methods is problematic given the expense of current strategies
20 and the political pushback by the community and the Monroe County Board of Commissioners. In an attempt to increase support and heighten community awareness to the control of dengue transmission in Key West, the Monroe County Department of Health, in conjunction with FKMCD, initiated an educational program, titled Action to Break the Cycle of Dengue (ABCD) in 2011. 14 Educational efforts included involving residents in vector control on their own properties, town hall meetings, door to door visits by FKM CD field inspectors, press releases, editorials, television programs, and poster contests, among others. 14 It has not yet been determined if the educational campaign by the Department of Health had a direct impact upon the lack of cases being reported in 2011 and 2012, however it is evident that the Key W est community had access to a high level of resources to improve awareness. Increased surveillance also occurred once locally acquired cases began being reported. In the beginning stages of the 2009 outbreak (after three reported cases) a seroprevalence s urvey was conducted near index cases in Key West by the Florida Department of Health 27 However, after this survey, n o other seroprevalence studies were conducted in Key West. Because no serosurvey was conducted after the outbreak subsided, a post epidemic estimate of the population at r isk of sequential infections did not exist. It is now well known that sequential DENV infections have significantly more severe symptoms and can lead to DHF, DSS, and/or death. 28 Thus, a good public health policy would be to determine the populations at risk of sequential infection ; local policy makers should weigh carefully the cost of effective dengue control methods against the economic consequences of severe dengue.
21 Hypotheses Origin of the v irus Despite the s carcity of autochthonous cases prior to the outbreaks in tensified vector control strategies education of the public heightened surveillance for symptomatic individuals and numerous epidemiological studies stimulated by the outbreak, the manner in which DENV arrived in Key West has not been discovered Viral sequencing of mosquito pools and genetic sequencing of virus isolated from resident blood samples concluded that the infecting virus was most similar to two Nicaragua DENV 1 strains, however the exa ct origin of the virus remains unknown. 29 30 Several hypotheses may be plausible for the origin of the virus in Key West. First, circulation of DENV could have occurred for several years prior to its detection in 2009. 29 30 This is possible given the lack of physician awareness to DENV and the likelihood of primary infections being asymptomatic, however given that DENV has not been reported in the state of Florida since 1934, this is highly unlikely. 5 25 Further given the population of Key West (ca. 24,000) is too low by perhaps an order of magnitude for DENV to circulate endemically for a l ong period of time it is likely that the outbreaks were sparked in 2009 due to DENV introduction by travelers or infected mosquitoes with subsequent transfer to the resident population. 10 21 Mosquito introductions are far from a rare occurrence, as various mosquito species have been readily introduced across the globe including in the United States, Europe, Guam, th e Galpagos Islands, Australia and New Zealand, among others. 31 32 These vector introductions have been linked specifically to aircraft and ship transports in the travel and trade industries. 31 33 It is believed that Ae. aegypti originally arrived in the United States aboard slave ships and with arboviruses being introduced and
22 reintroduced by viremic passengers, led to outbreaks of yellow fever throughout the Atlantic Coast. 31 In tropical and mild temperate regions, the vector established itself, spreading deeper into the heart of the United States; arrivals of Ae. aegypti aboard later ships likely helped to propagate the species in areas where the vector could not overwinter. 31 M osquitoes are m ore likely to establish and propagate themselves in regions where climate is similar to its own country of endemicity, independent of the distance between the original and susceptible regions. 34 For Key West, it was expected due to the close proximity to many Caribbean countries, that the climate woul d be similar between these regions With these frequent vector introductions, it is apparent that by connecting climatically similar region s, ones with and without dengue endemicity, that the likelihood of exotic organisms est ablishing themselves increases 32 34 It has already been documented that in France in 1994, between eight and 20 malaria vectors were imported per airline flight, summing to between 2 000 and 5 000 mosquito introductions across a three week period 35 Additionally, it has been suspected that West Nile virus has been introduced i nto Hawaii and the Galpagos Islands by mosquito disease vectors 33 36 37 Despite the likelihood of mosquitoes traveling across the globe by way of air and sea and establishing themselves in the local environment it has been stated that infected mosquitoes traveling to susceptible areas via aircraft is unlikely when compared to the capabilities of human pa ssengers to deliver the disease as viremic humans are the most likely source of DENV introductions globally. 8 32 Given this, it was expected that an infected traveler or returning resident brought DENV to Key West.
23 Human migration has been a route for spreading infections for a large portion of our own history and it is lik ely that our travel will continue to shape the emergence, frequency, and distribution of infectious diseases across the globe. 38 Annually, o ver 100 million individuals from developed nations travel to tropical region s ; nearly half of these travelers obtain a health problem abroad and carry infections with them back to their country of origin 39 Infections spread by global transport (trade, air and sea travel) include but are not limited to measles, cholera, plague, influenza, and dengue fever 39 40 The United States is not excluded from the list of areas experiencing such disease introductions. Nationally in 2010, 648 cases of imported dengue were reported, 21% (n=133) occurred in the state of Florida alone. 41 In 2 011 (n=70) and 2012 (n=137), Florida comprised 25% of national imports (n=283, n=545 respectively). 63 64 Introductions of this sort can lead to local cases, a s is becoming more frequent and is expected to have occurred in Key West during the 2009 and 2010 outbreaks. Traveler introductions leading to subsequent local transmission has occurred in Taiwan as well as Hawaii. 15 42 Autochthonous transmission of dengue fever in Hawaii was sparked by tourists returning from French Polynesia in 2001 and 2002. 15 In Taiwan, numerous summer introductions lead to local outbreaks that are terminated by the onset of winter. 42 It is expected that the epidemiologic scenario of DENV in Key West resembles that of Taiwan. DENV transmission in Taiwan has been described as uniqu e; studies on the molecular epidemiology of the virus suggest that local outbreaks that occur each year a re a major result of continual introductions of multiple serotypes from nearby Southeast Asian countries through close com mercial links and air travel 42 43 The number of visitors or residents returning from foreign travel in Taiwan is nearly three
24 million annually; in Key West the number slightly exceeds two million, although most are from non endemic locations. 21 44 Additionally, t hrough heightened international trade and increased travel to and from the island the number o f annually imported cases has increased dramatically in Taiwan. 43 An increase in the international travel and trade has improved the movement of all four DENV serotyp es and has increased the probability of secondary infection s with other serotypes 2 Given DENV can be acquired in an endemic country and transported from the originating endemic area to a new region by a traveler and the close proximity of these locations it is reasonable to believe that the high incidence of dengue in the Caribbean an d Latin America during the epidemic years in Key West could be correlated to the magnitude of cases observed in 2009 2010. For example, in 2010, Puerto Rico experienced its largest dengue outbreak with over 21,000 reported cases. 45 Because of the low likelihood of endemicity and the large probability of visitor introductions being the source of DENV in 2009 and 2010, the magnitude of viral introductions by this population was estimated in the first component of this dissertation. Transmission Current hypotheses regarding the nature of DENV transmission in 2009 and 2010 in Key West include one or more elements of vertical transmission, the establishment of a unique endemic state and/or multiple virus introductions via travelers coming from dengue endemic countries. S everal lines of evidence make it unlikely that DENV persisted endemically in Key West from 2009 to 2010 the first being the hiatus of reported cases spanning six winter months (Figure 1 1). 23 It is a fundamental aspect o f dengue transmission that exposure to an infected Ae. aegypti
25 to reason that if the number of infected mosquitoes were small, so would be the risk of infection. Despite the fact that adult Ae. aegypti are present in Key West throughout the winter months, it is unlikely that virus overwinters there. 22 Given that the survival of adult mosquitoes decreases exponentially over time, the probability of daily survival varies, however by assu ming a probability of normal daily survival of 0.89 per day it is unlikely that an adult mosquito would be able to survive for six months. 8 46 Taking the integral of this 0.89 probability, the average lifetime of an adult Ae. aegypti is estimated at 8.6 days, falling well within the average survival rate of one to two weeks. 6 8 Given this average survival period and that the probability of an infec ted Ae. aegypti female surviving six month is 7.76x10 10 it is unlikely that DENV could overwinter in the adult Ae. aegypti population. 46 Despite it being unlikely th at a population of infectious adult mosquitoes surviving six winter months between cases in 2009 and 2010, it is still possible that virus was harbored in the vector. A second possibility for DENV to overwinter is via the passage of the virus to the next g eneration of vectors through survival of an infected egg or by venereal (sexual) transmission of an infected male mosquito (infected vertically itself) to a susceptible female. Vertical transmission of DENV from an infected Ae. aegypti female to her eggs can occur by one of two mechanisms. Eggs can be infected within the ovary of the infected female (transovarially), or mature eggs can be infected during fertilization at the time of oviposition in the genital chamber of the female. 6 9 Vertical transmission of all four DENV serotypes have been documented in the lab oratory and in the wild (in eggs and larvae reared to the adult stage without infective blood meals and in adult males). 35 Studies suggest that DENV transmission in successive generations of mosquitoes
26 occurs in laboratory settings a t rates ranging from five percent to 25%, but no more than 20% in nature. 37 However the efficiency of this mechanism in nature remains unknown. 47 Additionally, the impact of vertical transmission on persistence of infections is unclear. Modeling studies suggest vertical transmission has the abi lity to enhance persistence, but in the absence of human cases only has a modest impact on dengue. 48 These modeling studies were based on larger infection rates than those reported in the field (e.g. from laboratory estimates). If the efficiency of vertical transmission in laboratory settings (1 4%) is indicative of what occurs in the wild, the influence of vertical transmission on dengu e persistence is reduced from modest to weak. 48 Without vertical transmission enhancing the persistence of DENV along with the unlikel y event of an infected adult Ae. aegypti surviving the winter, it is most probable that the virus died out in 2009 and that visitors reintroduced it prior to infections in 2010. Moreover, if the epidemiology in Key West is indeed similar to the DENV scenar io in Taiwan, where the virus is repeatedly introduced by travelers but is not endemic, it is not likely that transovarial transmission occurs at rates that would enhance persistence of DENV. In Taiwan, transovarial transmission has been examined as a pote ntial mechanism for endemicity. By examining pools of adult Ae. aegypti in Taiwan, it was found that transovarial transmission of all four DENV serotypes was very limited or absent in the local vector population 43 No pools of adult males (n=5783 pools, 49,759 sampled males) were found positive for DENV and only 0.2% of pools (n= 7628 pools 43,133 sampled females) of adult females were positive. 43 This, along with the fact that the only positive mosquitoes found in the four year study were female and during an outbreak scenario discount s it as possible mechanism permitting
27 endemicit y. 43 This may also be the case in Key West Although not specifically testing for transovarial transmission the only viral testing of adult mosquitoes in the literatur e regarding Key West was conducted throughout the year in 2010 across Monroe County; DENV was detected in only three adult mosquito pools within Key West out of 1,178 pools collected 29 Additionally, in conjunction with increased mosquito control ef forts (lower adult abundance) it is possible that a rise in herd immunity in 2009 led to the virus dying out on its own due to unavailability of susceptible hosts, thus DENV was not transmitted during the winter between 2009 and 2010 (Figure 1 1 ). Moreover DENV maintenance in nonhuman hosts is not likely to occur in Key West given that only in Southeast Asia and Africa do free ranging nonhuman primates circulate DENV sylvatically and these primates are not present in Key West. 49 It is possible that if exotic pets, such as primates, are kept in Key West, they could propagate a sylvatic cycle, but the impact of this on the endemicity of DENV on the island cannot be determined since the abundance of these exotic pets in the area has not been documented. Possible Consequences of Ongoing Transmission Economic burdens on communities affected by sequential severe dengue outbreaks may involve loss of life and/or productivity of indi viduals within the workforce, increased medical expenditures, and loss of tourism due to negative publicity. 11 Because nearly half of privately employed Key West residents work in tourism, the area where Ae. aegypti would be most likely to pick up DENV from visitors and later transmit the vi rus to residents is in places that tourists commonly frequent (e.g, open air bars, restaurants, and boutiques ) 21 In 2004, tourism itself brought about $715 million into Key West. 50 With tourism as a main source of income for the city of Key West, the loss of
28 tourism funds due to negative publicity about dengue infections, lack of control of the disease, or any other cause could be catastrophic to the local economy. Considering the health effects caused by DHF and DSS and their potential to affect the local economy, the second component of this dissertation aimed to estimate the proportion of Key West residents at risk for sequential infections as well as to identify risk factors of DENV by conducting a post epidemic seroprevalence survey with corresponding risk factor and spatial analyses. Given the recent recess ions in the United States, many regions are facing the threat of reduced funding for vector control programs, which could significantly diminish capabilities to respond to future outbreaks and to interrupt transmission, lead to the inadequate detection of newly introduced exotic vector borne pathogens and diseases, and increase costs due to delayed resolution of ongoing epidemics. 51 52 Dealing with DENV in Key West alone, (e.g., on a single key ), has resulted in a disproportionate amount of control resources being spent in one two by four mile area, perhaps to the detriment of control measures along other keys and the rest of Monroe County, making it a political issue. Mosquito control in Monroe County is directed by a commission of five elected represen tatives who reside in the five major population centers of the county; those commissioners not representing Ke of funds may best be addressed by eliminating most of the dengue specific expenditures and sugge since 2010 FKMCD inspectors do not agree with this as they have encountered residents w ho have reported dengue like symptoms 22 Due to the fact the citizens of Monroe County have impose d a tax up on themselves to fund mosquito control and that drift based insecticide
29 aerosols have been deemed largely ineffective for controlling Ae. aegypti in ad dition to being expensive, strategies have shifted towards so urce reduction and larvicides. 22 Despite investigation of other vector control strategies, implementation to determine their success is difficult due to these challenges, thus to evaluate new and pot entially more efficacious strategies requires the use of stochastic models to evaluate alternative control strategies the final component of this dissertation. The container inhabiting mosquito simulation model (CIMSiM) and dengue simulation model (DENSiM ), have been utilized across several regions including the Caribbean Basin Central and South America, and Southeast Asia to study the epidemiology and control of dengue outbreaks. 46 53 While the models have been used to investigate many dengue related phenomena such as El Nio and climate change, these models were developed primarily to evaluate mixed control strategies. 10 46 53 When these models were first introduced, a novel result was th e pupal/demographic survey that for the first time provided transmission thresholds and target endpoints for control efforts; these have since been adopted for international use by the World Health Organization (WHO) The site specific parameterization of CIMSiM and DENSiM allows for evaluation regarding various vector control strategies tailored to conditions unique to Key West. This fitting of the models to Key West allows for analysis of new control strategies prior to their implementation, something use ful in an area where dengue is a highly political topic and debates on funding for control programs are ongoing. These stochastic models incorporate data reported in Chapter 2 and Chapter 3 Models were parameterized using weather data from the National Oc eanic and Atmospheric Administration (NOAA), the National Weather Service (NWS), introduction of DENV by
30 visitors as estimated in the first component, seropositivity values documented in the seroprevalence survey in the second component, a pupal demographi c survey conducted in June 2012, and percent control of current FKMCD efforts and various other control strategies. Potential Implications of Dissertation Research This dissertation serves to explore and clarify the epidemiologic scenario in Key West by pr oviding preliminary results and insights into how the virus arrived how it was spread, and how to eliminate it with the addition of a residual adulticide to vector control strategies Given the flexibility in parameterizing this work for other locations, all of the studies included herein could likely be repeated especially if a selected area experiences high tourism flow, high Ae. aegypti abundance, imported DENV infections, and autochthonous cases. Estimates of visitor introductions into other areas coul d be adapted based on the visitor counts an area receives from dengue endemic countries and year specific incidence rates. Seroprevalence surveys could be repeated using a low cost and less time intensive study design that employs grocery store enrollment and rapid diagnostic tests. Simulation studies are increasingly flexible and can be parameterized to nearly any location given the two previous studies, serotyping of virus, and a pupal/demographic survey are conducted appropriately. The dengue scenario in Key West should improve if research is conducted with the goal of guiding future policy decisions. Policy decisions on vector control are crucial for dengue prevention and control given the absence of treatment or vaccines for infections. If appropriate p olicy decisions regarding vector control are introduced and
31 enforced in enough areas, it is possible that an improvement could be seen in the global burden of dengue.
32 Figure 1 1. Weekly time series of reported dengue cases in Key West, Florida 2009 2 010.
33 CHAPTER 2 INTRODUCTION OF DENGUE VIRUS TO KEY WEST, FL BY VIREMIC VISITORS Introduction Approximately 90 Key West, FL residents without a history of travel to dengue endemic areas were reported to have locally acquired D ENV infections in 2009 and 2 010 20 23 Because of their lack of travel, the fact that the virus was likely not endemic in Key West prior to 2009 and 2010 ( discussed in Chapter 1), and the numerous reports of travelers sparking outbreaks in Hawaii and other countries, it is suspected that visitors or returning residents from dengue endemic countries brought the virus to the island. Local Ae. aegypti sub sequently acquired the virus from visitors or returning residents and local transmission ensued. To evaluate the reasonableness of visitors and/or returning residents introducing DENV to Key West, the number of people days mosquitoes could have acquired th e virus from viremic airline passengers and cruise ship visitors each year (2007 2011) was estimated using reported country and year specific DENV incidence rates for dengue endemic countries and airline passenger and cruise ship visitor counts. Methods Da ta Sources and Assumptions Several data sources were used in the estimation of viremic visitor days in Key West, FL for the years 2007 2011. Countries in the western hemisphere were selected for inclusion in the study based on which countries the CDC had r eported as endemic for DENV 4 Due to resource constraints and the cost s associated with airline passenger and cruise ship visitor data, the evaluation of introductions on a global scale was not possible. Thus, as a proxy of overall DENV introductions to Key West, airline and cruise
34 ship calculations were performed for dengue endemic countries located south of the United States and north of Argentina 4 For these countries, DENV incidence rates were obtained from PAHO 54 58 In most endemic regions, DENV transmission often occurs during the rainy season however for the purposes of this study, it was assumed that incidence rates did not vary seasonally during a particular year 59 I t was expected that months with high incidence rates, e.g., wet seasons and warmer climates were balanced by drought months with low levels of mosquito breeding and reduced DENV transmission. Thus, the incidence rat es obtained by PAHO were representative of actual symptomatic cases per 100,000 per year 54 58 Monthly airline passenger counts from each of the dengue endemic countries in the western hemisphere were obtained from OAG Aviation (OAG Worldwide, Downers Grove, IL). Several assumptions were required to estimate viremic person days in Key West from airline passenger counts and incidence data. T he first assumption wa s that returning US residents and visitors from dengue endemic countries could be treated as one airline passenger group. Airline passenger data obtained from OAG Aviation were reported as lump sums of monthly airline passengers from a n airport in a dengue endemic area with a connecting flight in Miami FL and an end destination of Key West. Reports did not include a travel history for airline passengers (e.g., if passengers were on a return flight or starting a new trip), so it was impossible to separate t he number of returning US citizens and visitors from dengue endemic countries entering Key West. Further, because of this lack of travel history separation of passengers making airline connections prior to Miami from those originating at the airport of re cord was not possible. Therefore, i t was assumed that returning residents and visitors had been
35 present in the dengue endemic countries where they were exposed to infectious mosquitoes prior to travel to Key West The second set of assumptions for airline passengers was related to risk of DENV infection within the dengue endemic countries. Because the only information regarding airline passengers were counts without any reports on travel history, it was not possible to determine where returning US residents stayed w hile in the endemic country and/or if their risk of exposure to DENV differed from a citizen that may live in conditions more favorable for dengue transmission. Additionally, because DENV prevalence observed in travelers ha s been reported to refle ct variations in DENV prevalence in residents of dengue endemic countries it is not highly unlikely that travelers and residents could have similar risks of exposure to DENV in a dengue endemic country 39 Th is allowed the assumption that returning US residents and citizens from endemic regions had equal risk of infection while in the dengue endemic country Because n o estimates existed for the probability of travelers acquiring infections while in these dengu e endemic areas, it was assumed that DENV incidence rates could be used as a proxy for the risk of infection Lastly, it was assumed that once airline passengers were infected in the dengue endemic country, they progressed through their entire incubation period prior to traveling to Key West. It was assumed that the location of the passenger prior to infection and the length of the incubation period was insignificant; the key consideration was that they were infected and progressed through the incubation p eriod in the dengue endemic country, such that they would be viremic upon entering Key West. This assumption was made to maximize the estimates of viremic person days due to airline
36 passengers, as the goal was to determine the feasibility of this route of introduction. It was further assumed that airline pass engers progressed through a range of viremic days in Key West to reflect both individual variations in viremic periods and the varying duration of stay between returning residents and visitors On aver age, a visiting airline passenger stays in Key West for 7.6 days while a resident would return home to normal daily activities. 60 Given this average length of stay for a visiting airline pa ssenger and an average viremic period of four to five days ( maximum 12 days ) the notion that a visitor was viremic for a minimum of four days in Key West is not unlikely. Additionally, because returning residents could carry out their entire viremic perio d on the island once they return home, it is possible that their respective number of viremic days in Key West could be longer than four days, thus the ten day estimate. W ith this assumption of a range of values of viremic days on the island, a n estimate o f viremic airline passenger days could be calculated for Key West. Given the impossibility of separation of the two sets of travelers in the OAG Aviation data, the fact that their risk of infection, and duration of viremia were assumed to be the same, retu rning US residents and visitors from dengue endemic countries were treated as one group throughout airline calculations and will from now on be referred to as one group: airline passengers. Cruise ship visitor data was acquired from Customs and Border Prot ection (CBP) using a Freedom of Information Act request. Information regarding any ships that docked in Key West after visiting a non US port of call was obtained from CBP for 2007 2011. Cruise ship visitor data from 2007 2010 listed up to three dengue end emic ports of call visited by cruise ship passengers and crew prior to Key West and the date the ship was docked in Key West; because information on the date a ship was docked in
37 each of the ports was lacking, 2007 to 2010 data were referred to as port lis ts The ports of call were abbreviated as follows: for the third prior port of call, POC3, the second prior, POC2, and the port immediately prior to docking in Key West, POC1. The CBP data on 2011 cruise ships reported the date in each of the ports visited prior to Key West as well as the date docked in Key West. Because the 2011 reports had dates ships were docked in each port and from these dates a timeline between each port of call and Key West could be created, 2011 data were referred to as itineraries D urations between each port of call and Key West were necessary to determine which ships could be carrying viremic individuals upon entry to Key West. Because 2007 2010 CBP reports lacked docking dates for preceding dengue endemic ports of call, 2007 2010 port lists were transformed into itineraries, in a process that is further discussed in the viremic cruise ship visitor calculations section below With all of the cruise ship visitor information in itinerary form, calculations to estimate the number of v iremic cruise ship visitor days in Key West could continue. Given the differences in the information reported on cruise ship visitors and airline passengers, and their differing timelines of infection and potential viremia in Key West, new sets of assumpti ons were required for individuals traveling by sea. For cruise ship visitors, neither estimates of probability of becoming infected in a port of call nor the specific DENV incidence rates for each port of call existed. Therefore, it was assumed that the es timate of probability of a citizen in a dengue endemic country being infected and a cruise ship visitor becoming infected in a port of call were the same. Thus, the DENV incidence rates for the dengue endemic countries used in the airline passenger calcula tions were also used for cruise ship visitor calculations. Next, based
38 on a conversation with a member of the Key West Port Authority, it was assumed all passengers and half the crew as reported by CBP on each ship visited each port. 61 Further, it was assumed that cruise sh ip visitors explore the port of call for approximately eight hours, during the biting times of Ae. aegypti This small window of time spent in port likely complicates this risk of exposure to DENV, however no information on risk of disease in exotic ports of call exists to refute this as sufficient time to acquire DENV at these locations. If a cruise ship visitor acquired a DENV infection in a dengue endemic port of call, in order for that visitor to be viremic in Key West, they must have passed through an incubation period in addition to having not progressed through their viremic period prior to docking on the island To account for this time frame between infection in a dengue endemic port of call and being viremic when docking in Key West, cruise ship it ineraries were screened against inclusion/exclusion criteria. For the purposes of this study, three sets of itineraries (one for each port of call) were screened corresponding to the duration of travel from each port of call to Key West. POC3 itineraries d ocumented travel from POC3 to POC2, POC2 to POC1, and POC1 to Key West. For POC2, itineraries documented travel from POC2 to POC1 and POC1 to Key West. Lastly, POC1 itineraries documented travel from POC1 to Key West. If durations of travel within itinerar ies for POC3 to Key West, POC2 to Key West, and/or POC1 to Key West were longer than a minimum incubation period and did not exceed the maximum viremic period ( e.g., greater than three days and less than 15 days) the port was included in analysis 7 62 If the duration of travel from a port of call to Key W est was less than three days or greater than 15 days (minimum incubation period plus maximum
39 viremic period) it was excluded from analysis, given cruise ship visitors that could have been infected in that port of call could not be viremic upon entering Key West. After screening against these inclusion/exclusion criteria, the cruise ship passengers and crew that visited the included ports of call were multiplied against the estimate of the probability that a visitor becomes infected in that port of call. Ass uming a minimum incubation period for DENV in cruise ship visitors allowed for the maximum number of cruise ship itineraries to be included, thereby increasing the estimate of viremic person days due to cruise ship visitors in Key West. The exact process o f inclusion or exclusion of port of call itineraries will be discussed in the cruise ship calculation section in greater detail. After gathering incidence rate (PAHO) and traveler data (OAG Aviation, CBP) and considering the above assumptions, calculations were performed to estimate the number of viremic person days (viremic airline passenger days and viremic cruise ship visitor days ) in Key West: the number of person days that an Ae. aegypti mosquito could acquire DENV from a viremic cruise ship visitor or airline passenger while in Key West during their stay Viremic Airline Passenger Days Calculations Airline calculations were performed as shown in Figure 2 1 The first calculation of the analysis of airline passengers was to translate incidence rates f rom a per 100,000 population level to the number of symptomatic cases (dividing the incidence by 100,000). Next, the total number of incident infections ( symptomatic and asymptomatic ) per year was calculated Given a range of values exist for the proportio n of asymptomatic infections, it was assumed that symptomatic cases only represented 20% of total infections (e.g., an inapparent infection rate of 80%) in a dengue endemic
40 country each year. 63 64 Thus symptomatic incident case rates were multiplied by 5 to account for asymptomatic infections (divided by 0.2; 1.0 0.8). This es timate of total infections was then divided by 365 days per year (366 for 2008) to arrive at an estimate of the probability that a resident is infected in the dengue endemic country. It was assumed that this estimate of probability of a resident being infe cted is the same estimate of probability for an airline passenger being infected in the dengue endemic country prior to when they are viremic as they board their flight to Key West. T wo values (4 and 10 days) for the number of viremic days spent in Key Wes t were considered to reflect the variations in viremic periods not only on an individual level, but also in the length of time a visitor or returning resident is viremic in Key West (visitors stay on average 7.6 days, residents stay longer). 60 Multiplying the estimate of the probability an airline passenger is infected when leaving the dengue endemic country by the number of days a passenger is viremic in Key West yields the viremic airline passenger days due to one airline passenger. The multiplication of the total monthly airline passengers by the estimate of viremic airline passenger days due to one airline passenger resulted in the total viremic airline passenger days in Key West of one d engue endemic country. The contribution of each dengue endemic country to viremic airline passenger days was summed by year (2007 2011) to arrive at an annual estimate of the number of viremic airline passenger days a mosquito could acquire DENV from an in fected airline passenger in Key West. Viremic Cruise Ship Visitor Calculations Cruise ship visitor calculations were performed as reported in the flow chart shown in Figure 2 2. An estimate of the probability that a cruise ship visitor is infected while in port was calculated in the same manner as an estimate of the probability that
41 an airline passenger was infected in a dengue endemic country. This makes the same assumption that the daily probability of a resident being infected can be used as an estimate of the probability of a cruise ship visitor becoming infected while in port. Before calculations could progress further, the transformation of 2007 2010 port lists into itineraries was required. To do this, 2011 itineraries and cruise ship brochures were e xamined for port pairs present in the 2007 were the two ports visited in POC3 and POC2, POC2 and POC1, or POC1 and Key West. Next, 2011 port data and cruise ship brochures were examined for inconsistencies in tim e frames between ports. Cruise ship brochures and 2011 data were inspected for variations individually, and then were compared to each other. These comparisons helped to confirm that time frames were consistent within the 2011 data and brochures prior to c omparisons to port pairs in the 2007 2010 data. Port pairs present in both sets of CBP data, or in cruise ship brochures, were assumed to have the same travel time between ports. After comparing 2007 2010 port pairs with 2011 itineraries and cruise ship br ochures, only 24 (6%) of all port lists (n=370) had at least one port pair without matches. For these port pairs without matches, duration (in hours) between each port of call (POC3 to POC2, POC2 to POC1, and POC1 to Key West ) were estimated by dividing th e nautical distance (mile) by an assumed average speed of a cruise ship of 23 knots (nautical miles/hour). Each 2011 itinerary and cruise ship brochure reported date arriving and departing each port of call. Thus, time spent exploring the port was included in these data. To account for time spent in port in each of the port pairs without matches, it was assumed that cruise ships follow a particular pattern of traveling overnight, then allowing
42 passengers and crew to explore ports of call the next day. The m ajority of itineraries in cruise ship brochures as well as the 2011 cruise ship itineraries followed this pattern. Given this assumption, if a ship arrived too late (assumed to be after 12pm) to allow for an eight hour stay in port, it was assumed they wou ld stay overnight or delay their arrival to the port such that the passengers and crew would be let off for their excursions the following morning. With this allotment of additional time, the duration between ports and Key West were calculated and the tran sformation of 2007 2010 port lists into itineraries were considered complete. With sets of itineraries for POC3, POC2, and POC1 calculated for 2007 2011, screening of cruise ships for inclusion based on the minimum incubation period and maximum viremic per iod was then performed. Each set of itineraries for each port of call (POC3: POC3 to POC2, POC2 to POC1, POC1 to Key West; POC2: POC2 to POC1, POC1 to Key West; POC1: POC1 to Key West) was screened against inclusion/exclusion criteria (Figure 2 2). Ports o f call were included if the total trip duration (as documented in its respective set of itineraries) from that port of call to Key West was greater than the minimum incubation period and did not exceed this minimum incubation period plus the maximum viremi c period (e.g., duration greater than three days and less than 15 days). Any one port was excluded if the total trip duration from the port to Key West (as documented in its set of itineraries) was less than the three day incubation period or more than the incubation period plus the maximum viremic period (15 days) as these individuals would not be viremic upon entering Key West. Excluded ports of call (those whose set of itineraries showed trip durations where visitors would not be viremic in Key West) wer e dropped from the analysis. Screening was performed for each port of call (and its respective itineraries)
43 individually as it was assumed no double infections occur in subsequent ports. Additionally, it was assumed that if a passenger was not infected in POC3, they had the opportunity to be infected in POC2 and if not in POC2, then POC1. The estimates of probability of becoming infected in each port of call varied, reflecting incidence rates of dengue endemic countries. Included itinerary sets progressed t o the multiplication of cruise ship passenger and crew counts by the estimate of probability of infections while in port. Multiplying the number of cruise ship visitors of included itinerary sets by the probability of a cruise ship visitor becoming infecte d in a port of call yielded the number of cruise ship visitors viremic in Key West. Because cruise ship visitor were assumed to visit ports of call for an average of eight hours, to convert the number of viremic cruise ship visitors to viremic cruise ship visitor days, they were multiplied by the proportion of a day spent in Key West (8 hours/ 24 hours). The resulting viremic cruise ship visitor days were then summed by year for each port of call to estimate their yearly contributions to the number of perso n days mosquitoes could obtain DENV from a viremic visitor in Key West. Results Upon initial examination, endemic countries in the Caribbean Basin represented the highest DENV incidence rates in the Western Hemisphere for all five years, except for Brazil, which carried the highest incidence in 2008 (Table 2 1). Incidence rates decreased initially after 2007, but subsequently increased considerably through 2010. Incidence rates were lower in 2011 than the previous two years, however were still higher than t hose observed in 2007 and 2008. In addition to having the highest incidence rates, the Caribbean also carried the highest passenger loads, specifically in
44 The Bahamas, Jamaica, and the Dominican Republic (Table 2 2). Visitors from the cruise ship industry did not differ from this pattern as the majority of cruise ship inhabitants visiting Key West had spent time in The Bahamas (Table 2 3). Mexico and Belize were also common previous ports of call prior to Key West. Between 2007 and 2011, airline passengers were estimated to be viremic in Key West for between approximately seven and 16 person days (Table 2 4). Initially in 2007, estimates of viremic airline passenger days were small, however the estimates increased over the five year period. Even for an estim ated four days being spent viremic in Key West in 2010 and 2011, there were over two person days where mosquitoes could have acquired virus from an infected airline passenger. For the longer estimated viremic period being spent in Key West (10 days), the e stimates for 2010 and 2011 are nearly five days and over six days, respectively. Viremic airline passenger days were positively, yet poorly correlated with counts of the total numbers of passengers for both the four day estimate of viremia (0.62, R 2 = 0.39 ) and the ten day estimate of viremia (0.59, R 2 = 0.35). Deviations from the pattern of yearly raw passenger counts are apparent in 2010 and 2011 (Figure 2 3). Viremic cruise ship visitor days were smaller than those estimated for airline passengers with 0 .23, 0.89, and 0.55 viremic visitor days for included POC1, POC2, and POC3 itineraries respectively across the five year study period (Table 2 5). If the contributions from each port of call to viremic cruise ship visitor days are totaled (POC1+POC2+POC3), it appears that the years with the highest estimates of introductions were 2009 (0.25 viremic visitor days), 2010 (0.94 viremic visitor days) and 2011 (0.28 viremic visitor days). Over the five year period, it was estimated that there
45 were nearly two pers on days mosquitoes could pick up virus from viremic cruise ship visitors in Key West. These viremic cruise ship visitor days were also poorly correlated with visitor counts. POC1 had the highest correlation of the three ports (0.59, R 2 = 0.35) followed by POC3 ( 0.40, R 2 = 0.16) then by POC2 ( 0.26, R 2 = 0.07). It seems that viremic cruise ship passenger days reflect the estimate of the probability of becoming infected in a port of call more than raw passenger counts, as person day estimates do not follow t he same patterns as visitor counts provided by CBP (Figure 2 4). Yearly introduction estimates are roughly consistent with reported cases from 2007 2010 with the exception of 2011 for both travel industries Furthermore, estimates of viremic person days du e to airline passengers and cruise ship visitors for 2007 ( airline: 0.79 1.98; cruise ship three port sum: 0.15) and 2008 (airline: 0.27 0.68; cruise ships sum: 0.06) were lower than those for 2009 ( airline: 0.87 2.17; cruise ships sum: 0.25 ), 2010 ( airlin e: 2.31 4.79; cruise ships sum: 0.94) and 2011 ( airline: 2.49 6.23; cruise ships sum: 0.28 ), which was estimated to have the highest introduction frequency by airline passengers, and the second highest for cruise ship visitors (Tables 2 4 and 2 5) These estimates suggest that DENV introductions by viremic airline passengers and cruise ship visitors in Key West do occur. Although these viremic person day estimates for the western hemisphere are small in size, results suggest it is reasonable for the airlin e and cruise ship industries to bring viremic individuals to Key West. Discussion Travelers have been documented as couriers and transmitters of several infectious diseases as they form an important bridge between diverse populations and geographical regio ns. 65 Visitors have brought pathogens to naive regions causing devastating outbreaks of infectious diseases including but not limited to measles,
46 smallpox, tuberculosis, malaria, yellow fever, and sleeping sickness. 6 5 Historically, the majority of outbreaks have been related to air travel, however as cruise ship vacations become increasingly popular, so does the risk of passengers carrying a disease to another country. 65 The majority of attentio n with regard to diseases on cruise ships is borne disease transmission has yet to be documented; future research should be geared towards exploring this phenomenon. W and susceptible areas becoming shorter and shorter, due to increased travel and trade technologies, and duration of travel between these areas well within incubation periods of infectious diseases, the risk of a pat hogen being carried from one region to another and establishing itself in susceptible areas increases. 65 Imported dengue cases in the United States have been on the rise for quite some time. From 1993 to 1994, samples from 148 United S tates residents with symptoms consistent with DENV infection were submitted to the CDC, of these DENV infection was confirmed in 46 residents who had recently returned from vacation in dengue endemic countries. 12 Compare this with 2006 2008 where 529 probabl e cases (136 confirmed) were submitted to the CDC Dengue Branch and to 2012, when there were 545 positive test results in residents returning to the United States. 8 66 Imported DENV totals in the state of Florida have been variable in the past couple of years ranging from 133 to 70 to 137 in 2010, 2011, and 2012 respectively likely reflecting fluctuations in travelers an d incidence rates in dengue endemic countries. 8 9 41 Such increases in imported cases in returning residents nationwide support the probability of dengue introductions due to airline and cruise ship passengers. The results of this analysis
47 support that DENV could have bee n imported into Key West, acquired and transmitted to local residents by the Ae. aegypti population, resulting in locally acquired cases in 2009 and 2010.The most likely sources of DENV in travelers are in areas with the highest passenger counts and/or lar gest dengue incidence rates. As expected, a high contribution of passengers was observed from the Caribbean. Large outbreaks were also observed in these areas within the study period. Imported dengue from these areas is not unusual, as studies have shown t hat these regions are common tourism destinations where tourists have picked up dengue in the past. 6 12 13 66 67 In 2009, Belize experience d a considerable outbreak in a large ecotour ism area, supporting a possible linkage between travelers and acquisition of DENV in that area. 68 In 2010, Puerto Rico encountered the largest outbreak of dengue with over 21,000 individuals reported as cases 45 Lastly, in 2011, The Bahamas endured a n outbreak nearly seven times the total of their previous two outbreaks combined. 69 The combination of the high passenger counts in these regions in addition to their seemingly concurre nt outbreaks could have resulted in a high influx of infected individuals into Key West. In combination with the higher number of viremic person days estimated here, it is possible that these individuals could have triggered the autochthonous cases seen in 2009 and 2010. Given the reasonability that autochthonous cases were initiated by traveler introductions in Key West due travel from and in the western hemisphere alone, it is likely that imported cases translating into local cases could be occurring in o ther areas where vectors are abundant and behavior of locals allows for appropriate contacts between mosquitoes and humans. The increase in traveler associated infections
48 combined with the increased attraction to tropical tourism destinations and the incid ence rates in dengue endemic locations, could be problematic for susceptible areas, such as Key West, in the years to come. 16 Estimates of viremic person days suggest that DENV introductions in susceptible areas with large amounts of visitors could be significant and interventions, such as educational campaigns or screening and quarantining travelers, are a solution that could be attempted with the goal of reducing transfer of the virus from viremic visitors into the local populatio n. Interventions could be appropriate for regions where large traveler counts exist and imported dengue translates into autochthonous cases in local residents. how to avoi d mosquitoes with repellents, protective clothing or bed nets, and which time of day to avoid spending outdoors. 3 70 Another intervention option is traveler screening and quarantine. One area that has already implemented visitor screenings is Taiwan, which receives about three million visitors annually. 44 By implementing infrared cameras to detect fevers in incoming visitors, 45% of imported dengue cases were identified, however no reduction in locally acquired cases was reported. 42 Further research should be conducted to determine the success of interventions impacting imported DENV; results should be used to guide policy decisions. The number of potential introductions in 2011 should not be ignored. The outbreaks in 2009 and 2010 appeared to coincide with outbreaks observed in commonly visited tourist destinations and with their respective passenger counts. It is unknown why cases were not reported in 2011 as estimates predicted otherwise, but it could be related to multiple factors First, improved vector control practices may have led to lower
49 probabilities of exposure for resi dents. 71 Alternatively, the virus could have disappeared in the winter months between 2010 and 2011 and no reintroductions occurred in so far as 2011 introduction estimates were incorrect. 71 If 2011 estimates were correct, another option is possible. It is possible that DENV simply did not translate into the resident population or that a less virulent strain could have been transmitted to residents in Key West, FL. Compared with 2010, in 2011, PAHO r eported less dengue cases, within these cases a smaller proportion of severe dengue, and fewer deaths in the western hemisphere, which could be suggestive of less virulent strains circulating in 2011 than 2010. 72 A less virulent strain could result in lesser symptoms leading residents to lack motivation for diagnostic testing and medical attention, and lending further support to the lack of reported cases. Another potential reality could be that rising h erd immunity from the 2009 and 2010 infections reduced the receptivity of the resident population to DENV 1 infection. 71 Whatever the cause, exploration into the reasons behind this lack of reporting should be conducted to determine if any cases presented in 2011 and if additional individuals are now at risk for sequential infections. Estimates show DENV introductions into Key West by viremic visitors could have initiated and driven the outbreaks in 2009 and 2010. Serotype data is needed from each of the endemic country outbreaks to confirm the potential linkages between their circulating strains and thos e observed in 2009 and 2010 in Key West (DENV 1). 27 29 30 Further, t he sheer number of visitors does not solely drive infections in Key West, but the incidence of dengue in endemic countries may also play a role. Because DENV is on the rise in many popular tourism locations, actions may be needed to protect local
50 areas fro m introductions. Future research should be guided towards estimating introductions from Africa and Asia, as dengue remains a substantial problem in these areas and DENV imports could be significant
51 Figure 2 1. Calculation for determining viremic pers on days in Key West, FL due to airline passengers arriving on flights originating in a dengue endemic country. Figure 2 2. Calculation for determining viremic person days due to cruise ship passengers and crew arriving in Key West, FL after visiting each of three preceding dengue endemic ports of call. Note: This was repeated for each port of call individually. T he number of visitors in the calculation is assumed to be all passengers and half of the crew aboard each cruise ship.
52 Table 2 1. Highest DENV incidence rates (per 100,000) for dengue endemic countries where airline passengers departed for Key West, Flor ida as reported by PAHO, 2007 2011. Countries by Year IR s 2007 Martinique 1,316.58 Costa Rica 815.04 Guadeloupe 757.77 2008 Brazil 425.58 Honduras 288.08 French Guiana 270.59 2009 St. Bartolome 9,594.76 French Guiana 5,501.34 St. Mar tin 4,757.64 2010 Guadeloupe 10,257.05 Martinique 9,345.05 St. Martin 6,864.67 2011 Aruba 2,910.58 Bahamas 1,948.58 Curacao 658.53
53 Table 2 2. Highest country contributors for arriving airline passenger s in Key West, FL, 2007 2011. Cou ntries by Year No. Passengers 2007 Bahamas 2,848 Dominican Republic 1,178 Jamaica 1,097 Honduras 1,007 Nicaragua 818 2008 Bahamas 1,930 Dominican Republic 790 Jamaica 701 Aruba 653 Honduras 579 2009 Jamaica 1,193 Bahamas 1,104 Do minican Republic 972 Colombia 737 Aruba 728 2010 Bahamas 1,267 Dominican Republic 1,188 Jamaica 1,141 Colombia 920 Netherlands Antilles 874 2011 Bahamas 998 Jamaica 960 Dominican Republic 943 Haiti 905 Honduras 742
54 Table 2 3. H ighest country contributors to visitor s arriving in Key West, FL via the cruise ship industry 2007 2011 Countries by Year No. Visitors 2007 Bahamas 323,427 Mexico 59,395 Cayman Islands 36,587 Belize 10,335 Guatemala 9,944 2008 Bah amas 307,476 Mexico 77,837 Belize 29,990 Cayman Islands 29,283 Honduras 6,763 2009 Bahamas 293,362 Belize 46,985 Cayman Islands 37,239 Honduras 31,459 Mexico 20,685 2010 Bahamas 246,147 Mexico 55,038 Belize 38,768 Honduras 24,309 Guatemala 17,155 2011 Bahamas 193,114 Mexico 36,317 Belize 33,082 Guatemala 15,470 Honduras 14,602 Note: Visitors include passengers and half of total crew in Key West
55 Table 2 4. Estimated viremic airli ne industry passenger days in Key West, FL from 2007 2011. Year Cases Airline Visitors Airline Passenger Days 4 day Viremia 10 day Viremia 2007 0 11,788 0.79 1.98 2008 0 7,970 0.27 0.68 2009 27 9,376 0.87 2.17 2010 63 11,746 2.31 4.79 2011 0 1 0,789 2.49 6.23 Totals 90 51,669 6.73 15.85 Figure 2 3 Y early fluctuations of raw airline passenger counts (total for year) and viremic airline passenger days in Key West, FL, 2007 2011
56 Table 2 5. Estimated viremic cruise ship visitor days in K ey West, FL from 2007 2011. Year Visitor Counts POC1 POC2 POC3 Total 2007 208,453 0.05 0.05 0.04 0.15 2008 223,989 0.01 0.02 0.03 0.06 2009 206,058 0.03 0.12 0.10 0.25 2010 175,835 0.13 0.49 0.32 0.94 2011 124,630 0.22 0.06 0.28 Total 938,963 0.23 0.89 0.55 1.67 Note: Visitors include passengers and half of total crew in Key West Figure 2 4 Y early fluctuations of raw cruise ship visitor counts (total for year) and viremic cruise ship visitor days in Key West, FL, 2007 2011.
57 CHAPTER 3 SEROPREVALENCE OF DENGUE VIRUS IN KEY WEST, FL RESIDENTS Introduction In September 2009, after only three locally acquired cases were reported, a seroprevalence survey was conducted by the Florida Department of Health. 27 This serosurvey employed a clustered door to door sampling techni que beginning at the household of the index cases and spanned across the Old and Mid Town sections of Key West, Florida, a two kilometer square area. 27 No other serological studies had been conducted in Key West to date. Because only the Old Town area was sampled from and the serosurvey was conducted prior to the larger 2010 outbreak, and even before the height of the 2009 outbreak, it was expected that a post epidemic estimate of the overall Key West population at risk for sequential infections was absent. Severe symptoms such as DHF DS S and death are associated with sequential infections. 28 Besides leading to poor health outcomes, these severe symptoms can also lead to high medical costs and could have a significant negative impact on local t ourism. 73 74 Due to such the proportion of Key West residents at risk for sequential infections was estimated b y conducting a post epidemic seroprevalence survey Ri sk factor and spatial analyses were also conducted to identify further risk factors of DENV in Key West as well as to potentially guide v ector control efforts Methods This serosurvey was conducted in March 2012. All participants provided written informed consent under a protocol approved by the Western Institutional Review Board. Because of time and resource constraints, enrollment was con ducted in parking lots of
58 major regional grocery store chains (Publix Super Markets, Albertsons, Winn Dixie) using rapid diagnostic test kits in Key West, Florida. By sampling in grocery store parking lots, enrollment of participants may have been biased t o individuals that conduct the grocery shopping, but by sampling at a variety of grocery store chains with large spectrum of clients, it was assumed that a wider range of socioeconomic classes, age groups, and spatial distribution of individuals could be r eached than in traditional door to door surveys. 75 A power calculation was performed (assuming a predictive power level of 80%, a 95% confide nce interval, and a 5% error rate) to determine sample size necessary for the study. 76 It was estimated that a sample of 393 Key West residents would give sufficient predictive power to the risk factor analyses of this serosurvey however it was decided to raise this sample to 500 for increase d predictive power and 27 All individuals that frequented grocery stores during enrollment periods were invited to participate in the study given they were over the age of 18 years and live d or work ed on the island for at least seven months per year. These exclusion criteria were formulated to ensure that only year round residents were enrolled by excluding the sizable seasonal elderly popula tion that travel and reside in s outhern Florida during winter months. 77 This transient population may also be at risk for DEN V infection in Key West, however because they do not live of this survey. Additional work with these individuals could be performed in future works to determine their ris k for sequential infections. Again because of time and resource constraints, instead of conducting one of the gold standards in antibody detection, the enzyme linked immunosorbent assay (ELISA),
59 it was decided that rapid test kits would be sufficient. Deng ue Virus IgG/IgM Whole Blood/Serum/Plasma RapiCard TM InstaTests (Cortez Diagnostics, Inc., Calabasas, CA) the kits used in this study, are immunochromatographic tests for the simultaneous detection of recent (IgM) and previous (IgG) dengue antibodies. 78 Independent evaluations of similar rapid tests have shown that these assays are useful tools for screening individuals for current and recent DENV infections. 79 82 These tests have been used with increasing frequency in hospitals and in screening individuals for active dengue infections. 80 81 83 84 The accuracy of these rapid tests has been evaluated against other diagnostic te sts, establishing their reliability and validity; because of their diagnostic capabilities, low cost and time requirements, kits of this nature are be coming increasingly utilized worldwide For example, when testing whole blood samples collected via finger stick, the PanBio Dengue Duo Cassette has a sensitivity of 90% and specificity of 89%. 85 For accuracy, Cortez Diagnostics compared their own RapiCard TM InstaTests against these PanBio Dengue Duo Cassettes by testing 60 positive and 40 negative patient sera. 78 Agreement between the two rapid test kits was 100%. 78 Given the accuracy in cassettes to document DENV antibodies worldwide, and the agreeme nt with one of the most utilized rapid tests, it was expected that double testing with the RapiCard TM InstaTest was sufficient for confirming DENV antibodies. However, because these test kits are not approved as diagnostic tools for DENV in the United Stat es and the analysis of a test sample by these kits should not be used as a sole criterion for diagnosis, follow up with the Florida Department of Health was recommended by enrollment personnel to positive individuals to be retested for confirmed diagnosis and to be reported as confirmed cases. 85
60 The manufacturer suggests that samples of fresh serum or plasma are preferable to whole blood and thus is likely more specific in diagnostic testing. 78 However, becau se using only a few drops of whole blood from a finger stick could be used to adequately screen for antibodies and this would be less invasive for participants, this sample type was used. 79 82 Tests by first bringing the test kits and buffer to room temperature 78 Test kit components were kept in a cooler and the shade during enrollment to avoid overheating. A few whole blood drops were added into the well of the cartridge along with the buffering agent and retained. After approximately 20 minutes, anti body results could be observed. If only the control line of the test kit was illuminated, the individual had a negative result. 78 An invalid test occurred if no control line appeared and the te st was repeated. 78 A positive result was denoted by a complete line across the control line and IgM or IgG lines of the cartridge. 78 IgM antibodies are indicative of a primary infection within the previous 90 days while IgG antibodies represent a primary infection occurring prior to this three month interval. 5 Positive individuals were retested to confirm an antibody response. In addition to undergoing testing for DENV antibo dies, participants completed an enrollment questionnaire, in which they reported their home and work addresses, demographic information, travel history, as well as various indicators of mosquito exposure (Appendix) With the data acquired from these questi onnaires, statistical analys es were conducted to show which risk factors may have influenced the likelihood of previous infection for residents participating in the study Spatial analyses were performed to evaluate if any spatial patterns were present for previous infections (e.g., clustering of home or work addresses) and to potentially indicate common sites of
61 exposure in previously infected residents. Summary statistics, univariate and multivariate logistic regression were conducted in Stata 12.0 (Stata Corp, College Station, TX). Spatial analyses (e.g ., hot spot and cluster analyses ) were conducted in ArcGIS 10.0 (Esri, Redlands, CA) using reported home and work addresses to document the distribution and potential clustering of DENV antibodies across the island. Results and Discussion Of the 175 individuals enrolled in the survey, only two were excluded due to one being a nonresident and the other not completing the questionnaire and antibody test The current study design enrolled a subset of Key West re sidents expected to be reasonably representative of the general population with respect to age, gender, and ethnicity when compared to the 2010 US Census and the previous (2009) serosurvey The median age of enrolled participants was 52 years (range 18 84) ; the majority of the participants were between the ages of 50 and 60 years old (Table 3 1). It was expected that the oldest age classes were over represented in the current (2012) serosurvey because all individuals less than the age of 18 years were exclu ded However, the previous (2009) serosurvey found a similar median age of 53 (range 15 95), even when including children. Additionally m ales (52.4%) had higher participation than females (47.6%) These gender proportions underestimate the male population when compared percentages (58% male, 42% female), which were also three percent off of the census values. 27 86 W hites (79%) were the most frequently enrolled ethnicity and was consistent with the 2009 survey (78% white), showing another over representation by both enrollment strategies when compared to the 2010 US Census. 27 86 The current study sample being as representative of the ethnicity, gender and age of the resident
62 population as the previous (2009) serosurvey lend s support to the validity of the study design (Table 3 1). were well distributed spatially across the entire island, further suggesting a successful recruitment of diverse individuals representative of the overall resident pop ulation in Key West (Figure 3 1). The previous serosurvey in 2009 found three percent (n=6) of 240 Old Town residents IgM positive indicating a recent DENV infection (within three months) and six percent (n=16) IgG positive indicating an infection outside the three month interval 27 Rap id diagnostic testing from the current (2012) study revealed 12 (6.9%) enroll ed Key West residents to be IgG positive. Unexpectedly, six (3.5 %) tested participants were IgM positive. None of the participants were positive with both a recent and previous in fection. With seroprevalence rates slightly higher than the previous study, it is possible that Key West experienced an undetected outbreak in early 2012 potentially similar in scope to those that occurred from July to September 2009. It is also possible that continued transmission occurred when considering introductions of DENV from visitors as es timated in Chapter 2. I t is not unreasonable to assume that DENV transmission terminated in 2009 and later re started in 2010 once sparked by multiple introduct ions. Infections also could have arisen in 2011, given it had the highest estimates of visitor introductions in Chapter 2, leading to undetected cases. Further, because of the presence of IgM antibodies in March 2012, it is likely that an outbreak occurred possibly after travelers reintroduced the virus once again. With the evidence of IgM and IgG positive individuals, and the ability of travelers to bring virus to the island, it is
63 apparent that individuals in Key West are indeed at risk for sequential d engue infections. Univariate analysis uncovered several significant variables for each type of sero positivity (Table 3 2). For IgM positive individuals, central air conditioning use in the home was protective against recent infection (OR 0.11; CI [ 0.01, 0 .89 ] ). Although not significant, none of the IgM positive individuals reported being employed in an occupation that would re quire a high level of education. This could suggest that infections might have be en occurring in individuals of low socioeconomic st atus or those not reached by educational campaigns by mosquito control inspectors (education during door to door visits) or the health department (ABCD program), however without information on income or residents explicitly reporting a highest level of edu cation achieved, this cannot be proven 87 The greatest risk factors for infection in the prior three months were sleeping outdoors once per week (OR 14.70; CI [ 1.14, 190.27 ] ) and the self reported presence of birdbaths (OR 1 0.17; CI [ 1.85, 56.02 ] ) or bromeliads (OR 10.98; CI [ 1.24, 96.95 ] ) on reside ntial property. None of the IgM positive individuals reported leaving the state of Florida in the previous 12 months, suggesting that their exposure to DENV occurred locally and no t in an endemic region. Despite not being significant after logistic regression, all IgM positive individuals recalled not using repellent on t hemselves and being bitten at home With high odds ratios that were significant for items found on a residential property, the recollection of being bitten at home and not leaving the state of Florida, it appears that exposure to DENV in the preceding 90 days was related to residential life in Key West This lends further support to the likelihood of undetected local transmission of DENV in early 2012.
64 Work addresses for 2009 and 2010 cases were not obtained during interviews by the Monroe County Department of Health back when autochthonous transmission was ongoing. Thus no comparisons could be made with respect to l ikelihood of exposure sites (e.g., home vs. work) for the 2009 and 2010 outbreaks. Because of this, no comparisons could be drawn with respect to work addresses between the previous 2009 and 2010 cases and the IgM and IgG pos i tive participants in this stud y. Thus, current (2012) serosurvey IgM and IgG positive home addresses could be included. Spatially, IgM positive participants resided in the New Town (eastern) portion of Ke y West home addresses in the 2009 and 2010 outbreaks as reported by FKMCD (Figure 3 2). 26 Concurrently, for non retired IgM positive respondents work addresses were distributed in New and Old Town (western ) porti ons of the island (Figure 3 2). Unfortunately, hot spot and cluster analyses could not be performed due to insufficient sample sizes of positive ind ividuals. In comparison to those with evidence of DENV infections within the previous 90 days, risk factors for IgG positive subjects appeared less related to residential property factors and more related to the wo rk environment. None of the IgG positive i ndividuals reported sleeping outside. Risk factor analysis for these participants suggested that having a work environment cool ed by central air conditioning was protective against infection (OR 0.08; CI [ 0.01, 0.54 ] ) U sing open wi n dows to cool the work e nvironment (OR 5.44; CI [ 1.11, 26.54 ] ) increased the likelihood of IgG positivity (Table 3 2).
65 Experiencing a rash in the previous 12 months (OR 6.20; CI [ 1.82, 21.13 ] ) also increased the odds of IgG positivity. Each individual with IgG seropositivity not ed travel to an endemic area in the previous two years, suggesting that individuals infected outside of the preceding three months could have acquired infection during travel. T his travel history impedes the ability to claim that IgG positive individuals a re a direct result of the previous 2009 and 2010 outbreaks in Key West. However, this fact supports the notion that returning residents reintroduced DENV by bringing it back with them to the island. Thus, it is possible that if these IgG positive individua ls were not infected as a result of the 2009 and 2010 outbreaks, they could have contributed to the previous outbreaks reported in 2010, undetected in 2011, and possibly in early 2012. Despite this unknown time and location of possible previous DENV infect ion, these individuals remain at risk for sequential infections, DHF, and DSS if a new serotype of DENV virus were to gain a foothold in Key West. W hen spatially mapping home addresses of IgG positive individuals, it appears that previously infected reside nts trended towards clustering in the Old and Mid Town areas of Key West. This is consistent with what was observed in 2009 and 2010, however because of travel history, it cannot be explicitly stated that these individuals were infected at that time. W ork addresses are well dispersed across the entire island, showing very little clustering (Figure 3 2). This spatial pattern could be supportive of exposure to DENV in the preceding 2009 and 2010 outbreaks since infections were limited mostly to the Old and Mi d Town regions of the island but the lack of work
66 addresses for 2009 and 2010 cases in addition to travel histories to dengue endemic areas inhibit this as a definitive conclusion 23 27 Because of the relatively small sample size and previous DENV infection being a relativ ely rare outcome, IgM and IgG positive individuals were pooled into a single group and univariate an alysis was repeated for this Overall, similar to IgG positive individuals, this group was found to be less likely to be infected when using central air conditioning in the work environment (OR 0.16; CI [ 0.31, 0.86 ] ), but was at incr eased likelihood of infection when work areas were cooled by open windows (OR 4.26; CI [ 1.31, 13.89 ] ) or sleeping outdoors once per week (OR 18.27; CI [ 1.56, 213.58 ] ). All seropositive individuals recalled being bitten by a mosquito in the last 12 months, however, only two of 18 positive participants reported applying re pellent before going outdoors. After p ooling the positive individuals a trend in this behavior was discovered ; using repellent trended toward being protective (OR 0.23; CI [ 0.99, 8.76 ] ). Th ny p also trended toward supporting the notion of differing exposures from the two individual groups. Lending support towards differing exposures was the spatial distribution of the two groups as IgM positive individuals could have acq uired DENV infections locally based on their residential risk factors and lack of foreign travel, while IgG positive participants could have been exposed to virus either in the 2009 and 2010 outbreaks or in a dengue endemic country in the previous 24 month s. Regrettably, this relationship cannot be clearly explained given current sample sizes. Analysis of information related to additional seropositive individuals is necessary before establishing definitive correlations.
67 After adjusting for age and gender i n multivariate analyses, nearly all of the significant variables found in univariate analyses remained significant (Table 3 2). Thus, the majority of results of the current (2012) serosurvey are in accord with the previous ( 2009 ) serosurvey with air condit ioning being protective and open window cooling, recalling mosquito bites, and bird baths on the property increasing the likelihood of infection. 27 In addition to the above agreeing risk factors, the current study also found sleeping outdoors once per week, evidence of rash within the pr eceding 12 mo nths, and working outside to increase risk of infection The overall dengue scenario in Key West in early 2012 appears to be well distributed spatially across the entire island. Individuals ire Key West area, representative of all three regions, with respect to home address, however when broken down to IgM and IgG positive individuals, differing spatial patterns can be seen. These patterns could be reflective of the potentially differing exp osures for the different antibody responses (i.e. IgM: locally acquired cases vs. IgG: infected either locally or in endemic areas) however further research is needed to clarify this relationship The current study has developed valuable insights that wer e not observed in the 2009 seroprevalence study prior to the height of the 2009 and 2010 outbreaks. First, transmission in Key West appeared to be ongoing in the early months of 2012 and given the amount of visitors that could have brought or reintroduced DENV to the island, any previously infected residents, including returning travelers, could be at risk for severe sequential infections. Also, in contrast to the 2009 serosurvey, IgM positive individuals mostly resided in the New Town (eastern) portion of Key West, instead of the majority of cases being found in the Old Town (western) areas as previously
68 described. 23 27 Cur rent seropositivity could be indicative of an undetected outbreak in early 2012 however further research is required to determine if the circulating DENV was the same serotype and strain as the 2009 and 201 0 outbreaks. If the same serotype were found in K ey West, it could be implied that DENV is progressing toward becoming endemic, however, until this is confirmed, it is expected that travelers continue to bring the virus to the island multiple times each year. A new serotype being imported into Key West c ould cause the threat of sequentials to become an immediate reality. If a new serotype were found, it is possible that severe dengue illness could arise in the resident population, as all previously infected individuals are at risk for sequential infection s (currently in 2013, approximately 10% of Key West residents). To more accurately determine the number of individuals at risk of DHF and DSS repeating this serosurvey with an adequate sample size, the inclusion of gold standard diagnostics for confirmatio n of antibody responses, and serotyping of positive individuals should be performed Sequential infections in Key West could be catastrophic to the local health and economy and vector control with aims to eliminate dengue form the island should continue to prevent sub sequent infections.
69 Table 3 1. Demographic information of Key West, FL residents comparing participants in the 2009 and 2012 serosurvey s and the associated 2010 census report 2012 Study 2009 Study Census Reports Median Age (Age Range) 52 (18 84) 53 (15 95) 41.5 (all ages) Age Groups Percent Percent Percent 18 35 21.21% 27.6% a 35 50 22.42% 24.00% 50 60 29.70% 16.00% 60+ 26.67% 20.20% Gender Male 52.38% 58% 55.40% Female 47.62% 42% 44.60% Race Whi te 79.04% 78% 66.10% Black 7.78% 9.00% Hispanic 9.58% 21.20% Other/ Multiple 3.59% 3.70% a age group spans 15 35
70 Figure 3 1. Location of seroprevalence study enrollment and reported responded addresses in Key West, FL March 2012.
71 Table 3 2. L ogistic regression analysis results from serosurvey questionnaire data Univariate Multivariate Variable Odds Ratio p value 95% Confidence Interval Odds Ratio p value 95% Confidence Interval IgM Home Air Conditioning 0.11 0.038 (0.01, 0 .89) 0.09 0.038 (0.01, 0.87) Sleep Outdoors Once per Week 14.70 0.040 (1.14, 190.27) 55.72 0.020 (1.90, 1637.09) Bird Bath on Property 10.17 0.008 (1.85, 56.02) 9.31 0.020 (1.43, 60.74) Bromeliads on Property 10.98 0.031 (1.24, 96.95) 9.97 0.041 (1.10, 90.68) Miss Work >10 days due to Fever 6.50 0.045 (1.05, 40.43) 5.96 0.102 (0.70, 50.74) IgG Work Air Conditioning 0.08 0.010 (0.01, 0.54) 0.09 0.016 (0.01, 0.64) Window Cooling at Work 5.44 0.036 (1.11, 26.54) 4.97 0.057 (0.95, 25. 90) Work Outside (absence of screens) 6.50 0.057 (0.94, 44.80) 6.41 0.062 (0.91, 45.26) Rash in Last 12 Months 6.20 0.004 (1.82, 21.13) 5.82 0.006 (1.67, 20.30) Any Work Air Conditioning 0.16 0.032 (0.31, 0.86) 0.17 0.045 ( 0.03, 0. 96) Window Cooling at Work 4.26 0.016 (1.31, 13.89) 3.79 0.036 (1.09, 13.18) Sleep Outdoors Once per Week 18.27 0.021 (1.56, 213.58) 44.75 0.009 (2.55, 784.22) Using Repellent on Self 0.23 0.058 (0.05, 1.05) 0.24 0.066 (0.05, 1.10) Rash in Last 12 Mont hs 2.95 0.051 (0.99, 8.76) 2.84 0.067 (0.93, 8.65)
72 Figure 3 2. Distribution of reported home and work addresses of dengue IgG and IgM positive serosurvey respondents in Key West, FL March 2012 Note: Work and home addresses not plotted if blank i n questionnaire. Work addresses also not plotted if individual was retired.
73 CHAPTER 4 MODELING ANALYSIS OF DENG U E VECTOR CONTROL STRATEGIES IN KEY WEST Introduction Today, in many parts of the world, the control and prevention of dengue continues to be pr imarily an issue of vector control. With this in mind, a pair of simulation models was previously developed with the goal of establishing and evaluating new integrated vector control strategies. 10 46 53 The stochastic simulation models used here describe the daily dynamics of transmission of DENV in the urban environment and have been orient ed in such a manner to permit them to be site specific such that the models have been used and continue to be used for the evaluation of dengue control programs in South and Central America, the Pacific islands, and South East Asia (unpub lished data). These models have been used to estimate transmission thresholds ( Ae. aegypti pupae per person) for targeted source control and have been adopted by WHO. 1 6 78 Other applications have included studies on the impacts of weather anomalies associated with the El Nio/Southern Oscillation (ENSO) and climate change. 88 89 Model Background The entomology model used in this analysis is the container inhabiting mos quito simulation model (CIMSiM). 10 53 CIMSiM is a weather driven dynamic life table model of container inhabiting mosquitoes such as Ae. a egypti 10 53 Specifically, CIMSiM estimates on a daily basis the number of each cohort passing to the next life stage as a function of a multitude of variables and relationships. 10 CIMSiM integrates daily local weather such that it is indicative of adult and immature survival. To do this, the entomological model takes into account the microclimate that influences each life stage.
74 Mic roclimates are a key determinant of survival and development for all life stages of mosquitoes, thus CIMSiM incorporates the extensive database of weather information. The duration of embryonation, the pupal stage and gonotrophic cycles are based solely o n temperature in the entomological model. 10 53 Temperature dependent development is considered specifically in an enzy me kinetics model within CIMSiM. 10 53 The premise behind this enzyme kinetics approach is that when temperatures are a t high or low extremes, the enzyme controlling developmental rates of eggs, larvae, pupae and the duration of the gonotrophic cycle is reversibly denatured. 10 53 It is modeled as a reversible process in that when temperatures return to tolerable levels, development proceeds at normal rates. Because there were very few data on egg survival and development when CIMSiM was created, estimates were put in as defaults that could be changed at a later date. Embryonation is purely based on temperature, increasing when temperatures rise. Egg survival is density independent and nominal daily survival is assumed to be 0.99 based on field estimates without predation. 10 53 Survival reflecting egg desiccation is 1.00 if the container is holding water. 10 53 For dry containers, survival reflects saturation deficit and sun exposure to account for desiccation. 10 53 The predation of eggs is thought to be temperature dependent, thus foraging values vary according to temperature. 10 53 From estimates derived from the literature when CIMSiM was created, it was assumed that 19.7% of eggs hatch spontaneous ly without being submerged with water, while 59.6% of the remaining eggs hatch each day they remain flooded. 10 53 In the model, it is assumed that eggs submerged when embryonation has concluded hatch immediately. 10 53
75 The modeling of l arval development in CIMSiM is complex, integrating temperature dependent relationships with resource competition. For immature stages, the microclimate is calculated using a series of equations and biological relationships to provide estimates of water g ains and losses, and wat er temperature. 10 53 For rain filled containers, water gains and losses are calculated based on rainfall and evaporative losses. 10 53 In CIMSiM the amount of rainfall considered to fill a container is the produ ct of a watershed ratio and the daily reported rainfall. 10 53 Evaporative loss is a function of the combination of su n exposure and relative humidity. 10 53 For covered containers evaporative losses are modeled as reduced by some fact or between zero and one. 10 53 For manually filled containers, the models assume that water levels fluctuate daily and that water gains and los s es are equivalent. 10 53 Water temperature is an important factor influencing the developmen t rate and survival of all immature stages. Minimum and maximum water temperatures are calculated within each container as a function of minimum and maximum air temperatures, sun exposure, and container size. 10 53 For containers with the ability to hold greater than five liters of water, a moving average of daily values of minimum and maximum water temperature is used in the model as an attempt to simulate reduced temperature fluctuations within the large container caused by its own thermal inertia. 10 53 The enzyme kinetics model determines only a portion of larval development ba sed on these water temperatures; r esource competition within the immature environment also plays a role. Biologically, density dependent mortality of Ae. aegypti larvae is often regulated by indirect intraspecific competition for food. 86 87 CIMSiM models daily arrivals and
76 losses of food in each container using three inputs: 1) the initial amount of food, 2) the amount added to, or created photosynthetically, in the container each month, and 3) the daily proportion of food that decomposes or is lost (not consumed). 10 53 The model integrates food, larval density, and temperature into an estimate of larval development using a series of differential equations. By allowing the user to adjust larval food larval survival such that predicted values of pupae from simulations are consistent with values observed in the field. 10 48 62 Pupation occurs once larvae have reached a minimum weight (estimated by assimilation of food into larval biomass) and completion of a minimum thermal experience as determined by the enzyme kinetics model. In CIMSiM, pupal development is assumed to be dependent solely on temperature; males and females emerge at equal ratios (50% each). Not on ly does CIMSiM provide DENSiM with estimates of emergence of mosquitoes it also gives inputs of gonotrophic development and adult survival. To calculate adult survival and gonotrophic development, CIMSiM first uses estimates of dry weights from larval cal culations to establish the wet weight of an adult female. It is assumed that larval rearing conditions influence both size and energy reserves of adults. 10 53 Thus, the frequencies in blood meals per gonotrophic cycle vary as a function of adult size. If female adults are well fed, as a result of larval rearing conditions, they are assumed to only need one blood meal prio r to oviposition. 10 53 The necessity of females to take replete feeds before oviposition is also considered in CIMSiM. For Ae. aegypti females that need more than one blood meal to complete their
77 first cycle, it is assumed that replete meals occur on the second day, and for consecutive cycles, replete feeds are sought on the first day. 10 53 The daily number of contacts between hosts and blood seeking females (per one hectare area) is a product of the number of mosquitoes seeking a reple te feed and the average number of feeding attempts per replete feed (n=3). 10 53 The rates of oviposition in differing containers (small vs. large) are considered in the model such that CIMSiM distributes oviposition between container types as a product of container density and size. 1 0 53 Nominal adult survival is adjusted by survival factors that reflect extremes in air temperatures and moisture. 10 53 To parameterize CIMSiM requires daily information on weather and the completion of a pupal/demographic survey because the model takes into account both heterogeneity of larval habitat and microclimates 10 Larval estimates are not used to parameterize the model due to the fact that pupal estimates have been shown to more accurately predict adult abundance in addition to dengue transmission thresholds. 47 90 Pupal productivity documented in the survey by container type is scaled to reflect relative abundance of adults in the s tudy area. 10 Once CIMSiM, which underwent extensive validation by comparing model output to field/laboratory observations, estimates adult survival, gonotrophic development, weight of adult females, and emergence of mosqui toes per hectare, DENSiM uses these inputs to define the adult biting population. 10 46 53 In DENSiM survival and emergence estimates drive the size of the Ae. aegypti population while values regarding weight and gonotrophic development impact the female biting frequency. 42 DENSi M takes into account that daily temperature, in addition
78 to viral titer, influences the extrinsic incubation period within the mosquito. 42 Using the user supplied information on the human age distribution and age specific birth and death rates, DENSiM creates a population with the desired age distribution. 42 Once the initial number of humans per hectare is specified, DENSiM determines the size of the simulation are a based on the total number of humans and their density. 42 Taking into account the abundance of human and mosquito populations, DENSiM incorporates additional sources that influence contact rates between the v ector and host. 42 Replete feeds and the influence of temperature on the gonotrophic cycle, interrupted feeding attempt, and alternate hosts all have been documented to affect DENV transmission, hence they are considered by the model. 42 Virus is introduced into DENSiM one of two ways, via an infected human or infectious mosquito. Following the introduction and inoculation of DENV into the population, humans are mod eled to progress through incubation and viremic periods as specified in the DENSiM database. 42 To estimate human to mosquito transmission, each day DENSiM computes the average number of contacts (considered to be at least a portion of a replete feed) by different mosquitoes per person then divides the total number of contacts by the number of people in the simulation area. 42 To assess the probability of a mosquito consuming enough virus to become infectious, the viral titer within the blood meal is taken into account, with the default value of a probability of 0.55 for a virus titer of 10 6 mosquito infectious dose 50 (MID 50 ) per milliliter for a titer of 10 3 a prob ability of 0.30 is the default. 42 The number of Ae. aegypti receiving sufficient virus to become infectious is a product of the infection probability of a serotype, the number of hosts infected with that serot ype, and the contact rate. 42 To calculate mosquito to
79 human transmission, DENSiM considers the number of mosquitoe s with a disseminated infection 42 This propor tion of infectious mosquitoes is a function of the adult survival and the temperature and titer dependent extrinsic incubation rate, both of which are considered in DENSiM. 42 DENSiM then multiplies the number of infectious females by the number of different individuals to be bitten by these females, and the probability of mosquito contact (initial probing through replete feed), which results in the transfer of enough DENV to spark an infection, if the host is not protected by antibodies. 42 Residual Spray Background T he ability of residual sprays to control dengue has been known for quite some time. 91 93 Residual sprays have been reported to control dengue vectors in areas that use them for malaria control. 93 95 In Key West, FKMCD inspecto rs often do not have access or permission to perform vector control inside residences. Thus, the only locations residual adulticides could be applied would be outdoors. However, residual adulticide applications are not frequently utilized for controlling A e. aegypti since female mosquitoes commonly rest indoors and on cloth products. 96 Given that the majority of residual adulticide applications occur indoors, the entomological model was previously set up to reflect indoor use. 10 46 53 CIMSi M takes into account the proportion of indoor surfaces being treated, the percentage of buildings being treated and the proportion o f Ae. aegypti adults resting outdoors 10 46 53 Because indoor use is h ighly unlikely or impossible in Key West, the model simulations regarding residual adulticide applications were conducted with the goal of estimating a proportion of mosquitoes that would need to come into contact with insecticides (regardless of adult res ting locations) to retrospectively reduce or eliminate d engue transmission in 2009 and 2010 in Key West, and potentially in 2011 given an
80 outbreak occurred. Ideally, places in Key West with the highest risk for DENV transmission (those with all components necessary for infection) such as places where viremic visitors, vectors and residents overlap should be treated with residual adulticides. However without proper field estimates, it is unknown how feasible and practical a residual adulticide regimen in t hese locations would be. In discussions with FKMCD on alternate methods to control transmission, residual sprays in areas where local residents and visitors overlap spatially was considered viable in Key West but simulation studies were deemed essential be fore implementation could take place. 22 The simulations herein represent a residential area, however the concept of control of DENV transmission can be useful in guiding policy decisions regarding a residual spray in the overlap area. Current vector control me thods in Key West are insufficient to prevent local transmission of DENV to residents; given the current political view of the disproportionate spending in Monroe County on dengue control, the long term sustainability of current efforts are in jeopardy. 22 The dengue models were parameterized with data specific to Key West, Florida with the goal of evaluating spatially targeted residual adulticide sprays. Hopefully this effort may provide sufficient evidence to encourage research and promote further evaluation o f this method. Data Sources for CIMSiM Parameterization M eteorology Daily weather data for Key West including minimum, average and maximum temperatures, dewpoint, and precipitation were obtained for the years 2000 2012 from the National Weather Service (NW S) of the National Oceanic and Atmospheric Administratio n (NOAA) From this information, saturation deficit was calculated using
81 vapor pressures from dew points and temperatures in a process defined previously. 97 Relative humidity was calculated using a ratio of vapor pre ssures. All weather data were directly input into the model as an XML file type to populate the water database used by CIMSiM and DENSiM. Pupal/demographic Survey of Old Town As a part of the modeled environment for larval habitat, CIMSiM requires estimates of the average standing crop of Ae. aegypti pupae by type of container and various container information including shape, sun exposure, watershed area, and drawdown fr equency. 10 For this, a pupa l/ demographic survey was conducted in 2012 in a 45 ha area of Old Town, Key West (Figure 4 1) 93 96 98 The area selected for surveillance was chosen by overlapping several GIS shape files. These included a list of hot spots of mosquitoes, overall Ae. aegypti co unts, and previously confirmed cases all provided by FKMCD. 22 Houses within this overlapping area (n=702) were pulled from the parcel file and listed by address. Each address was assigned a random number then put into numeric order; every seventh house was sel ected for the survey. Permission for entry to each residence was obtained by an FKMCD inspector. If entry was denied, neighboring houses were surveyed; if the house number was odd, the neighboring house on the right was selected for surveillance, even hous e numbers to neighbor on the left. At each residence visited, every outdoor water holding container was examined for pupae. All pupae were removed from containers and stored in collection jars labeled with the container description and home address until i dentification could be performed. Pupae were identified in the FKMCD lab space using a taxonomic key. Information on each water holding container found outside of the home was collected and included the
82 method of water fill (e.g., rain, manually filled, or both), shape and dimensions, and sun exposure (0.0 [completely shaded] to 1.0 [full sun]). Container information was collected in a field notebook; data entry occurred at a later date. All standing crops were recorded during fieldwork in Key West with th e exception of gutters and cisterns. Gutters represent an unknown larval habitat in Key West in terms of productivity. Because of the potential for gutters to be a significant producer of Ae. aegypti in Key West the total density of gutters in 100 houses was recorded during the pupal/demographic survey, however logistical constraints made it impossible to determine if the containers contained larvae or pupae, or if they were dry, or clogged and retaining water. Since FKMCD did not have an estimate for gutt er productivity in Key West, a literature search was performed to see measured productivities in other locations in this type of container. 22 Unfortunately, the literature review yielded very few estimates of the productivity of roof edge gutters and forced an assumption that Key West gutters resembled the productivity of Brazilian gutters. 99 Specifically, the observed number of gutters per hectare in the pupal/demographic survey wer e scaled using a proportion of wet/clogged gutters in the Brazilian study; the food fitter was used to adjust daily food into this class of container so that the simulated and the Brazilian estimates were within a few percent of each other. 99 By inserting gutters into the model, inevitably any unknown larval habitat, even perhaps those responsible for cryptic breeding, was account ed for. Another potentially important class or type of container known to exist in Key West is the cistern; prior to the construction of the water line from south Florida, cisterns were built beside or under residences for the purpose of water storage dur ing drought
83 seasons. FKMCD knows of the existence of 134 cisterns in the 45 ha survey area and attempts to treat these; however they do not know the efficacy of these treatments or the productivity of cisterns. The decision was made to not attempt to model this class of containers in as much as their number, size, water holding status, and production of Ae. aegypti pupae was unknown 22 CIMSiM Parameterization for Key West, Florida After the pupal/demographic survey, containers were grouped by volume, fill metho d, and shape to determine the representative container types to be used in the models (n=34). Average standing crop (total Ae. aegypti pupae divided by number of containers in each particular type per hectare) as well as the average container dimensions an d sun exposures were calculated by type of container using Microsoft Excel spreadsheets. These container types were then created in CIMSiM using the user interface menus. Information for each container type recorded in the field included shape and size, mo nthly turnover, sun exposure, if the container was covered or uncovered (if covered, an estimate for reduced evaporation), and frequency and mode of water filling. Container density was calculated using total area size, the number of overall houses, and th e 100 houses examined for pupae. Left at the default value was water drawdown frequency. Once all container information was input into the model, the food fitter function was run in CIMSiM. Because s urvival during the larval stage is a function of intraspe cific competition for food resources and temperature, the food fitter is used to match the number of pupae observed in the field to simulated estimates of pupal abundance. 10 48 62 To do this, the dates, weather inputs, and productivities of containers during the 2012 itting menus and run six times.
84 In these runs, the final population of eggs, larvae, pupae, and adults of run one are used for initial values and the average standing crop for runs two through six are calculated and compared to observed data. In the food f itter function of CIMSiM the initial default for daily food additions (0.1 mg/day) was modified such that the average standing crop of Ae. aegypti pupae in the simulation population for each type of container in the model was within 0.2% of the observed st anding crop as estimated by the pupal/demographic survey. In addition to the similarity in average standing crop per hectare, the standing crop of Ae. aegypti pupae in each of the simulated types of containers was within a few percent of those observed in the pupal/demographic survey. After the food fitter process was completed, CIMSiM was run for the years 2007 to 2011 to establish entomological inputs for the transmission model. DENSiM Parameterization for Key West, Florida Age specific Seroprevalence Th e human age structure (age specific birth and death rates) was estimated using demographic data for Monroe County and Key West, Florida obtained from Florida Charts and the 2010 US Census. 86 100 The initial population size of the study area (45 ha; n=1700) and host density (38 residents per ha) was determined using parcel public GIS files from FKMCD. 22 Age class proportions were derived from raw number counts and were hand entered into DENSiM. 86 100 Initial and final age classes for the study period (2008 2011) can be seen in Figure 4 2. Age sp ecific birth and death rates were converted from a denominator of a 100,000 population to per a 1,000 population and were entered into the transmission model.
85 Seroprevalence of anti dengue IgG was taken from the 2009 seroprevalence survey (minimum age samp led: 15 years) and the IgM/IgG survey of March 2012 (minimum age 18 years) 27 Given that only one serotype (DENV 1) was introduced to the island in 2009 and 2010 the lack of reported cases of DHF or DSS, it was assumed that DENV 1 was the only circulating serotype and continued to circu late in 2011. 27 29 30 Because viral titers were not available from e ither study, the default value in DENSiM (10 6 MID 50 ) was used. 46 Virus I ntroductions To input virus into DENSiM one specifies the number of humans in their first day of being viremic and/or infectious mosquitoes to be introduced. When modeling endemic locations such as Indonesia and simulating large areas with over 250,000 people, the introduction of only a few viremic humans introduced early in only the first year for a multi year simulation results in the endemic condition being established. 46 The v iremic airline passenger day estimates from Chapter 2 were multiplied by 10 before being added to viremic cruise ship visitor day estimates. This multiplication of viremic airline passenger days was warranted based on the limited scope of the introductions analysis (Chapter 2). First, the analysis of viremic travelers only took into acco unt airline passengers from the western hemisphere. Passengers from the eastern hemisphere could also have an impact on the dengue scenario in Key West. Second, there are multiple airports that individuals could depart from on a connecting flight to Key We st, but the study only considered Miami as a connecting airport to Key West. Lastly, the analysis did not take into account individuals who fly into a Florida airport, rent a car, and commute to Key West. Based on these three factors, even the maximum esti mate of viremic traveler days in Key West in Chapter 2 is likely only a small portion
86 of the total DENV introductions via travelers. Once the viremic airline passenger days were multiplied by 10, they were added to the average viremic cruise ship visitor d ays (from the three previous ports of call). Because DENSiM does not have the capacity to add the number of introductions as a series of people days, adjustment from viremic airline passenger and cruise ship visitor days to number of potentially viremic i ndividuals was necessary. The number of potentially viremic visitors due to both the airline and cruise ship industries was calculated in Chapter 2 prior to calculating the time frame they were viremic in Key West (multiplying by 4 or 10 for airline passen gers and by 8/24 hours for cruise ship visitors). To estimate the total number of viremic individuals introduced into Key West, the number of viremic airline passengers was added to the number of viremic cruise ship visitors. This sum was rounded to the ne arest whole number to estimate the total number of viremic individuals in Key West in each particular year (2007 2011). The number of viremic airline passengers were defined to be the same for both estimates of viremic days (4 and 10 days), therefore only one set of viremic individuals were input into the models. The resulting number of DENV 1 human introductions used for the simulation study was 2, 1, 2, 6, and 7 for the years 2007, 2008, 2009, 2010, and 2011, respectively. Unfortunately, DENSiM lacks the ability to randomly add introductions on a yearly basis. Given the lack of randomization capabilities, introductions were scheduled manually using a periodic schedule. For each year, viremic individuals were introduced during the spring and summer; the tim e between introductions varied depending on how many introductions occurred in the particular year. The exact number of viremic visitors
87 arriving in Key West at each introduction event was unknown because viremic passenger days were calculated on a yearly scale. Thus, virus could theoretically be introduced by clusters of more than one returning resident or visitor (i.e. in a family) or via individual introductions. DENSiM has the capacity for predicting epidemics over long periods with viral introductions of the same serotype, given their frequency of arrival does not vary on a yearly basis. Thus, another obstacle with DENSiM for this project was the fact that when running outbreak simulations, introductions varying in number and frequency of the same serot ype across the study period could not be observed in one set of runs. In other words, infections resulting from a given set of introductions (say in 2009, n=2) did not graphically carry over to another when adding a new set of introductions (say in 2010; n =6). Therefore, to graph the progression of infections from 2008 2011 and to and if the virus died out between each year, each set of yearly introductions were entered as differing serotypes. Serotype variations within the model were removed such that all serotypes acted in the same manner after viral introduction. From this limitation arose another. Because there were 5 years of introductions calculated in Chapter 2, o ne year needed to be dropped from the simulations due to DENV only having four serotypes. For 2007, no matter which month viremic persons were introduced, no infections resulted. Therefore, it was determined that excluding 2007 was the best option, leaving two years with reported local transmission (2009 and 2010) and two years where transmission was not reported (2008 and 2011) to be evaluated in simulations.
88 Introductions in 2008 did not produce infections no matter the month of entry, thus a single intro duction was added in the middle of the year (June 1). In an attempt to closely model the 2009 and 2010 outbreaks, introductions were initiated based on the estimation of a time frame from mosquito acquisition of DENV, transmission to residents, and develop ment of symptoms. For 2009, this translated to a periodic introduction of viremic visitors beginning July 1, occurring every 45 days and ending August 31. For 2010, introductions were set to begin March 10, repeat every 50 days and end November 30, for a t otal of six introduction events. Visitor introductions for 2011 were set to occur every 52 days, spread throughout the year. Simulation Runs Simulations P art 1 : Baseline Control (No Residual Adulticide Applications) After model parameterization, the first set of simulation runs made in this study took into account control efforts practiced by FKMCD from 2008 2011. These runs were considered the baseline control for all simulations conducted, as an estimate of zero control in Key West did not exist at the in itiation of this project. Operational vector control targeting Ae. aegypti in Old Town in Key West using aerially applied granular formulations of Bti began in 2011 and continued in 2012; 15 applications were made in 2011 between January and late September 22 Using pre and 7 day post treatment data from CO 2 baited BG Sentinel traps and house and container indices, 7 day efficacy for FKMCD treatments was estimated to be less than 16%. Because there is neither data on what particular types of containers received treatment nor container specific efficacy, it was not possible to apply it as a strategy in the model. Hence, it was assumed that the percent control of Bti applications were included in the pupal/demographic survey and when predicted pupae in the model w ere fitted to match the observed counts, this
89 percent control was also integrated into CIMSiM. Thus, the model fitting took into account the impact of Bti. Runs without residual adulticide additions were used to verify that the models as parameterized with field data corresponded with the number of cases reported in Key West during 2007 2011. In addition the output from these baseline control estimates, adult mosquito abundance predicted by the model was compared to existing BG Sentinel Trap data provided b y FKMCD to verify if predicted mosquito counts followed the same pattern as what was observed in Key West. These control runs were repeated between five and ten times with identical results. Because DENSiM becomes stochastic only when the number of virus p ositive mosquitoes is less than a specified number per simulation area (default value of 10) and because the simulation area of Old Town was 45 ha, repeated runs gave essentially identical results; it is expected that the models did not go stochastic. 46 However, it is still possible for the model to go stochastic but give similar results for multiple runs with the same parameters. 46 It is expected that this possibly occurred during simulations, however because the model went stochastic towards the end of runs, it is likely that its impact was minimal. Thus, the results reported herei n only represent one set of simulations due to the fact that estimates from replicate runs were not substantially different. Simulations Part 2: Baseline Control and Residual Adulticide Applications The second set of simulations took into account baseline control values in addition to residual adulticide applications. For the simulation work, the efficacy data of ICON 10CS (active ingredient: lambda cyhalothrin), a synthetic pyrethroid, was used. 101 103 Though local Ae. aegypti in Key West have been documented as resistant
90 to pyrethroids, any residual spray that behaves in a manner that is similar to what is depicted by the models could be a reasonable addition to vector control strategies by FKMCD. 22 Because Ae. aegypti is a highly domesticated mosquito and it often lives within the home, the models were previously set up such that residual adulticide treatments would occur inside the residence. As mentioned above, the key factor in the output is not where mosquitoes rest, but the proportion of adult mosquitoes coming into contact with insecticides. Therefore, the default value of DENSiM was used regarding the location of adult resting sites for initial residual appl ications (60% outdoors with daily movement between indoor and outdoor areas). 46 The addition of this control intervention was input manually under the label 10CS in the field, it was specified that 100% of adults contacting this material died during the first 7 days (maximum effect period); this efficacy was also input such that it decli ned linearly over the course of 180 days (residual effect period) 101 102 In modeling the residual applications for residential areas, the proportion of indoor surfaces to be treated was set at 75% while 55% of the buildings were treated every 120 days, assuming 60% of mosquitoes rested outdoors. For the remaining simulations, this 75% of indoor resting sites, 55% of buildings and 120 days was considered a baseline treatment regimen. Buildings wi thin the models are not specifically parameterized; the models are not developed for such spatial applications. Therefore, treatments of buildings were based on probabilities of treated resting sites, assuming complete mixing of Ae. aegypti and human popul ations and regular flow of mosquitoes indoors and
91 outdoors. This treatment schedule was used for simulations because it was felt that this would only take moderate funds and effort to achieve maximum protection against dengue transmission. Simulated treatm ents in 2009 were scheduled periodically and were simulated as initiated on July 31 under the assumption that two weeks after the index case (reported by the Florida Department of Health July 19) was a sufficient time period to acquire residual adulticides and begin their use. Multiple runs (5 to 10) were conducted for residual adulticide application simulations. These also remained nearly identical, such that if models went stochastic, they did not considerably affect the output. Residual applications were also scheduled in another set of simulations to reflect dengue control if treatments were initiated in 2008. For simulations of residual adulticide applications beginning in 2008, 75% of interior surfaces, 55% of buildings were treated periodically (every 120 days) beginning January 1, again assuming a 60% percentage of adults resting in outdoor sites. Simulations P art 3: Sensitivity Analysis of Proportion of Adults Resting Outdoors Because the percentage of adults resting indoors or outdoors in Key West is unknown and adult Ae. aegypti can rest either indoors or outdoors, a sensitivity analysis was conducted on the percentage of adult resting outdoors to see the effect of percentage of adults resting outdoors on the control of DENV transmission. Simulated treatment applications were set up such that maximum effect and residual effect periods of the pyrethroid adulticide remained at seven and 180 days, respectively. Additionally, the percentage of indoor surfaces treated (75%) and buildings treated (55%) re mained identical to the baseline plus residual adulticide application simulations above 101 102 The only changes in the sensitivity analysis simulations were to
92 vary the percentage of adults res ting outdoors (30%, 60% 90%) for each of the simulated treatment years (2008 initiated vs. 2009 initiated). Simulations P art 4: Simulated Elimination of Transmission The last set of simulations were to determine if elimination of the 2010 and 2011 outbreak s could be achieved if residual applications were initiated at the end of July in 2009 as well as all three outbreaks being avoided if residual adulticides were initiated in 2008 as a proactive vector control strategy. If transmission existed at the baseli ne treatment (treating 75% indoor surfaces, 55% buildings, every 120 days) for each proportion of Ae. aegypti adults resting outdoors (30%, 60%, 90%) percentages of indoor surfaces to be treated with adulticides were increased, while if transmission was ab sent at baseline values, the percentage of indoor surfaces to be treated was decreased until a threshold for transmission was reached. The first set of simulations for elimination of transmission occurred with the baseline treatment level (75% of indoor su rfaces being treated and 55% of the buildings treated every 120 days) beginning on July 31, 2009 and assuming 60% of mosquitoes rested outdoors. Parameters of percent of buildings being treated (55%), frequency between applications (120 days) and mosquitoe s resting outdoors (60%) were left at baseline values, while the percentage of indoor surfaces treated was increased at five percent intervals until transmission was not observed in output, then decreased by one percent to find the threshold value. Assumin g 30% of adults were resting outdoors, beginning at the 75% of indoor surfaces being treated, 55% of buildings, every 120 days starting in 2009 was the second set of transmission elimination simulations. Because there was no transmission at the baseline tr eatment schedule, indoor surfaces being treated was reduced at 5%
93 intervals until transmission was observed, then was increased at 1% intervals until it transmission ceased to be observed. The last set of simulations regarding the elimination of transmissi on given baseline treatment values (75% indoor surfaces treated in 55% of buildings every 120 days) in 2009 assumed a 90% proportion of adults resting outdoors. Given transmission was observed at the baseline, indoor surface treatment percentages were incr eased at intervals of 5% until they reached 100%. When transmission continued, the percentage of buildings being treated was increased until it reached 100%. The same set of parameters from the above three sets of simulations (30%, 60%, 90% assumed adults resting outdoors) were conducted identically for the runs simulating treatments beginning January 1, 2008. Baseline values were increased/decreased initially by intervals of 5% until transmission or an absence of transmission occurred and were increased/de creased by intervals of 1% until a threshold for transmission or maximum coverage was reached. Results Pupal/demographic Survey A total of 100 residential properties were sampled from an entire study area spanning 45 ha, containing a total of 1,306 proper ties that housed 1,700 residents in Old Town, Key West (Figure 4 1). CIMSiM models the entomology for a representative one hectare area; needed for parameterization are the numbers of containers per hectare. Therefore, container and pupal densities were co nverted to the area of the 100 residences surveyed to parameterize the model accordingly. The area associated with the 100 houses and their respective yards specifically was 3.4 ha A total of 1 332 containers were examined during the survey, of these 971 contained water Larvae
94 were present in 82 containers and 15 were positive for Ae. aegypti pupae (n=182) and Culex quinquefasciatus (n = 13) The average standing crop of Ae. aegypti pupae indicated that most types of water holding containers did not produ ce pupae (Table 4 1). The most productive container types documented by the survey were miscellaneous cylindrical containers (3.00 pupae per container), buckets (1.03 pupae per container), rectangular planters (2.50 pupae per container), birdbaths (0.50 pu pae per container), and yard trivets (0.48 pupae per container) all filled by rainfall and not manually. Simulations Part 1: Baseline Control (No Residual Adulticide Applications) CIMSiM output of total females followed seasonal weather patterns as expecte d (Figure 4 3 and 4 4 ). Interestingly, female mosquito densities were predicted to be low in the weeks coinciding with the 2009 2010 outbreaks while large numbers of mosquitoes were observed in weeks when no cases were reported (Figure s 4 4 and 4 5). Despi te low vector totals overall, female mosquitoes per person remained over one per person for the majority of the four year period reaching upwards of six f emales per person (Figure 4 6). After comparing BG Sentinel Trap data to CIMSiM predictions of female mosquitoes, the only year of predictions that corresponds reasonably well with counts observed in trap data was 2009 (Figure 4 7). Levels of observed female Ae. aegypti are substantially higher in months w h ere nearly no adult females w ere predicted by the models. When the mismatch of adult data in Key West and modeled predictions became apparent, inputs were re evaluated for errors, but none were found. It remains unknown why model predictions did not agree with field estimates, therefore, after considering this mismatch, simulations continued as planned with the caveat that additional work with the model would be necessary before making definitive claims
95 regarding the benefit of residual adulticide applications. It was however, noted that when reviewing mod el inputs that mosquito estimates in CIMSiM appeared to be more responsive to rainfall events than Ae. aegypti females as captured by BG traps (Figure 4 8). After the input of viral introductions, consistent with observed cases, no outbreaks were seen in t he models during 2007 and 2008 (Figure 4 9). Dividing the number of infected persons seen in this figure by the sum of the human incubation and average viremic periods (9 days) gives the estimated number of individuals infected. The estimated numbers o f la tent, symptomatic, and asymptomatic infections during 2008 2011 ranged between 1,700 and 1,900 per year (ca. 7 8% of the population of Key West). If 20% of these infections developed symptoms, the number of cases would range between 353 376, reported cases would be lower still. 64 The correspondence between reported cases and model estimates include that there were no infections predicted in 2007 and 2008, the seasonality of cases in 2009 (2010 has less of a fit), and very roughly, the sheer number of cases assuming an inapparent infection rate of 80%. 64 In striking contrast with reported cases is the model prediction of an outbreak in 2011; this will be discussed in the conclusion and discussion section (Chapter 5). Simulations Pa rt 2: Baseline Control and Residual Adulticide Applications Simulated residual adulticide treatments initiated in 2009 caused mosquito populations and female mosquitoes per person to decrease substantially (Figures 4 10 and 4 11 ). With residual adulticide treatments of 55% of building and 75% of resting sites infections were reduced by 0.1% in 2009, 99.5% in 2010, and 7.0% in 2011
96 (Figure 4 12, Table 4 2). When simulating baseline control as initiated in 2008, no outbreaks were observed in any of the subse quent years. Simulations Part 3: Sensitivity Analysis of Proportion of Adults Resting Outdoors The models appeared to be sensitive to variations in proportion of adults resting outdoors given the other baseline treatment values were left the same. Variatio ns in DENV transmission after simulated initiation of treatment on July 31, 2009 were observed at each level of percentage of Ae. aegypti resting outdoor s (30%, 60%, 90%) (Figure 4 13). Assuming the 2009 start date of residual applications, when 75% of ind oor surfaces were treated in 55% of houses every 120 days, assuming only 30% of Ae. aegypti adults rest outdoors, transmission was observed in 2009 and not in the following two years. For baseline treatment values assuming 60% of adults rest outdoors, data was identical to that seen in baseline control simulations, (e.g., lesser infected individuals in 2010 and 2011). Predicted transmission during baseline treatments assuming 90% of adults rested outdoors was higher than that of the other outdoor resting pr oportions simulated. Simulations Part 4: Simulated Elimination of Transmission Successful elimination of transmission in 2010 and 2011 was observed for treatment schedules simulated as beginning in both 2008 and 2009 (Table 4 3). When introducing the resid ual spray two week s after onset of reported cases in 2009, the 2010 and 2011 outbreaks could easily be reduced; with more effort they could be successfully eliminated. After simulating a 2009 initiation of residual adulticide applications, a complete elimi nation of DENV transmission can be seen in 2010 and 2011 with a residual adulticide being applied every 120 days assuming 60% of adult mosquitoes rest outdoors, treati ng 83% of resting sites of only 55% of buildings
97 Similarly, treatments of 49% of indoor surfaces in 55% of houses (every 120 days beginning in 2009) eliminated transmission in 2010 and 2011. Assuming 90% of Ae. aegypti adults rest outdoors, transmission could not be interrupted even when 100% of buildings and all interior surfaces were treat ed. The only peaks that were observed after residual adulticide applications were initiated and DENV transmission was eliminated were the multiple visit or introductions from 2008 2011, assuming the treatment schedules as dictated in Table 4 3 assuming 30% and 60% of mosquitoes resting outdoors (Figure 4 1 4 ). For simulated treatments beginning in 2008, treatment regimens less than that of the baseline (75% of indoor surfaces, 55% of buildings every 120 days) were able to eliminate transmission in all years. At 30% assumed adults resting outdoors, only 40% of indoor surfaces of 55% of houses were necessary to eliminate transmission. For 60%, treatment of 70% of interior surfaces eliminated infections. Similarly to 2009, elimination of outbreaks in 2009 and 201 0 could not be achieved assuming 90% of adults rest outdoors, however, a reduction in the amount of infected individuals was observed. Discussion Simulations resulting from estimated visitor introductions and the entomologic scenario in Key West suggest t hat outbreaks should have continued into 2011. This, however, could be disputed given the high numbers of mosquitoes predicted in 2011 are not in agreement with BG trap data from Key West. However, mosquito control field inspectors have suggested that ongo ing transmission in 2011 occurred based on residents reporting similar symptoms in themselves and friends/associates. It is not understood why the model output on Ae. aegypti females did not match what occurred
98 in the field, as recorded in BG trap data. Ho wever, it seems that CIMSiM more closely reacts to daily reported rainfall than Ae. aegypti in Key West. One possible explanation is that this disagreement could be related to cryptic breeding on the island in containers not observed during the pupal/demog raphic survey. The phenomenon of cryptic breeding was not taken into account in the pupal/demographic survey, most obviously because it was not explicitly sought out (due to logistical limitations). During the pupal/demographic survey, only outdoor contain ers were observed. Containers that were not documented during the pupal/demographic survey that could have an impact include manually filled containers, containers inside the home, and containers such as cisterns. Manually filled containers within the model are assumed to produce mosquitoes independent of rainfall conditions; water gains and losses are due to an individual filling or removing water from a container. It is possible t hat manually filled containers have more of an impact on mosquito counts in Key West due to the fact that model output reflects rainfall patterns more closely than the local population as estimated by BG traps. The productivity of these containers is unkno wn given the fact that all containers observed in the pupal/demographic survey were outdoors and filled by rain. Similarly, the productivity of containers within the home was not estimated, therefore their respective contributions to the mosquito populatio n in Key West is uncertain. Finally, the importance of cisterns in Key West is unclear; FKMCD does not have estimates of their productivity or their true abundance on the island. Repeating a pupal/ demographic survey and repeating simulations could change the disagreement between BG traps and model predictions. This repeated survey
99 should be significantly larger than the 100 houses used to parameterize the models here. Moreover, it should consider not only productivity in outdoor containers, but containers within residences and the productivity of gutters. To get a better grasp on cryptic breeding, the survey should also seek out potential sites of cryptic breeding (such as cisterns) and should aim to document their productivity. Furthermore, an in depth loo k into the hydrology of containers in Key West and its impact on adult mosquito abundance should be considered prior to model parameterization from this repeated pupal/demographic survey. Because of the magnitude of viral introductions in addition to trans mission despite FKMCD efforts in 2009 and 2010, and the presumable outbreak in 2011, the exploration of alternative vector control strategies is warranted; possibly even the use of a residual adulticide regimen. Since approximately 40% of residents private ly employed work in tourism, the potential sites where travelers, local residents, and Ae. aegypti overlap could profitably be examined for control implications; such sites would include open bars, restaurants, guesthouses, and boutiques catering to the to urist. 14 Residual sprays could be an effective treatment for these areas, since all components needed for infection, such as the source of virus, vector, and susceptible population are all potentially present. However, without the mosquitoes predicted by CIMSiM matching what was observed in the field, this cannot be considered a definitive solution to DENV transmission in Key West. Residual sprays have been used to successfully combat dengue vectors in various locations. 101 102 Results from CIMSiM and DENSiM simulations suggest that residual sprays could be prophylactic in managing dengue outbreaks in Key West
100 Mosquito counts predicted by the models were effectively controlled using a moderate amount of effort and funds, despite the mismatch with BG counts. Given the high cost and relatively low efficacy of current aerial larvicides and source control via inspectors, this work suggests that a research program designed to tailor a residual adulti cide regimen in Key West may be warranted and should be considered by FKMCD a nd their Board of Commissioners.
101 Figure 4 1. Survey area and households examined during the pupal/demographic survey in Old Town, Key West, FL, and June 2012.
102 Table 4 1. Pupal productivity and dimensions by container type obtained during the June 2012 pupal demographic survey. Container Type Average No. Pupae Average Length (cm) Average Width (cm) Average Height (cm) Average Diameter (cm) Boat_Rain 0.00 487.68 243.84 182. 88 Bromeliad_Manual 0.00 8.05 13.68 Bucket_Accessory_Rain 0.00 3.23 19.63 Bucket_Manual 0.00 20.32 20.32 Bucket_Rain 1.03 29.21 28.10 Conical_ Fountain_Rain 0.00 10.16 45.72 Cyl_Birdbath_Manual 0.00 7.20 38.95 Cyl_Birdbath_ Rain 0.50 11.43 38.10 Cyl_Cup_Rain 0.00 14.45 8.23 Down_Pool_Rain 0.00 426.72 304.80 45.72 Fountain_Manual 0.00 7.62 7.62 15.24 Fountain_Rain 1.00 30.48 21.59 11.43 Gutter_Rain 2.00 91.44 15.24 2.54 Hot_Tub_Manual 0.00 152.40 152.40 9 1.44 Jacuzzi_Rain 0.00 121.92 152.40 Misc_Cylinder_Manual 0.00 41.06 57.15 Misc_Cylinder_Rain 3.00 12.65 35.65 Misc_Rect_Container_Manual 0.00 14.61 11.75 9.84 Misc_Rect_Container_Rain 0.15 25.55 19.01 13.24 Planter_Manual 0.00 19.69 33.53 Planter_Rain 0.33 15.91 23.16 Pond_Rain 0.00 142.24 101.60 30.48 Pool_Rain 0.00 335.28 426.72 182.88 Rect_Planter_Rain 2.50 16.51 24.13 13.65
103 Table 4 1. Continued. Container Type Average of No. Pupae Average Length (cm) Average Width (cm) Average Height (cm) Average Diameter (cm) Rect_Trivet_Rain 0.00 55.88 13.06 4.72 Recycle_Bin_Rain 0.00 42.33 37.74 33.38 Tire_Rain 0.00 8.42 57.95 5.64 Trash_Can_Rain 0.00 89.26 65.31 Treehole_Rain 0.00 15.24 10.16 2.54 Trivet_M anual 0.06 5.12 35.50 Trivet_Rain 0.48 3.62 22.65 Waterfall_Rain 0.00 45.72 30.48 15.24 Watering_Can_Manual 0.40 25.91 12.19 Watering_Can_Rain 0.00 23.71 16.93 Note: Gutter productivity estimated by the food fitter and productivities documented in the literature
104 Figure 4 2 Initial and final population proportions spanning the pre and post outbreak period in Key West, FL 2008 2011.
105 Figure 4 3 Daily fluctuations in various weather parameters in Key West, FL 2008 2011.
106 Figure 4 4. Daily densities of Aedes aegypti females and pupae per hectare in Key West, FL reflecting FKMCD control efforts in Key West, FL, 2008 2011 as simulated by CIMSiM
107 Figure 4 5. Daily densities of Ae. aegypti female s overlapped with reported weekly dengue cases reflecting FKMCD control efforts in Key West, FL 2008 2011 as simulated by CIMSiM
108 Figure 4 6. Daily estimates of female Ae. aegypti per person reflecting FKMCD control efforts, overlapped with weekly reported dengue cases in Key West, FL 2008 2011 as simulated by CIMSiM
109 Figure 4 7. Correspondence of Ae. aegypti females observed in the field with BG Sentinel Traps to pre dicted daily densities of Ae. aegypti females per hectare from 2009 2011 as simulated by CIMSIM.
110 Figure 4 8. Comparison of the correspondence of Ae. aegypti females observed in BG Sentinel Traps to predicted daily densities of Ae. aegypti females per hectare from 2009 2011 as simulated by CIMSIM to reported daily rainfall.
111 Figure 4 9 Predicted infected individuals (latent, symptomatic, and asymptomatic dengue cases) reflecting FKMCD control efforts, overlapped with reported weekly dengue cases in Key West, FL 2008 2011 as simulated by DENSiM
112 Figure 4 10 Daily densities of Ae. aegypti females and pupae in Key West FL assuming residual adulticide treatments were begun in the 2009 outbreak as simulated by CIMSiM
113 Figure 4 11 Daily esti mates of female Ae. aegypti per person in Key West, FL assuming residual adulticide treatments were begun in the 2009 outbreak as simulated by CIMSiM
114 Figure 4 1 2 Predicted reduction in infected individuals (latent, symptomatic, and asymptomatic deng ue cases) in Key West, FL assuming residual adulticide treatments were begun July 31, 2009 as simulated by DENSiM. Note: Assumes baseline treatment regimen of 75% of indoor sites and 55% of buildings treated with residual adulticides every 120 days.
115 Table 4 2. Estimated r eduction of simulated DENV infections in Key West, FL given residual adulticide treatments beginning in the 2009 outbreak. 2008 2009 2010 2011 Pre IRS totals 1 1725 1826 1888 Post IRS totals 1 1723 8 1756 Percent Reduction 0.0% 0.1% 99.5% 7.0% Note: Based on 75% coverage of resting sites and 55% of buildings being treated. Totals include asymptomatic cases. Figure 4 13. Sensitivity of residual adulticide to control DENV transmission assuming varying proportions of Ae. aegypti a dults resting outdoors as simulated by DENSiM. Note: Assumes baseline treatment regimen of 75% of indoor sites and 55% of buildings treated with residual adulticides every 120 days, only adjustment is proportion of resting sites outdoors.
116 Figure 4 14. Estimated reduction in predicted of dengue cases in Key West, FL assuming residual adulticide treatments eliminated local transmission as simulated by DENSiM. Note: This pattern was seen for treatments begin in 2008, occurred every 120 days in 55% of buil dings, treating both 40% of indoor surfaces (assuming 30% adults resting outdoors) and 70% of indoor surfaces (assuming 60% of adults resting outdoors). For treatments beginning in 2009, cases would be observed in 2009, individual spikes can be seen in 201 0 and 2011 if applications occurred every 120 days in 55% of buildings, treating both 49% of indoor surfaces (assuming 30% adults resting outdoors) and 83% of indoor surfaces (assuming 60% of adults resting outdoors). Repeated one infected individual spike s are representative of viral introductions (not visible in previous figures due to scale).
117 Table 4 3. Levels of residual adulticides required in 2008 and 2009 to eliminate transmission for each assumed percentage of adult Ae. aegypti resting outdoors. 2008 2009 Percent of Adult Ae. Aegypti Resting Outdoors Percent of Indoor Surfaces Percent of Buildings Percent of Indoor Surfaces Percent of Buildings 30% 0.40 0.55 0.49 0.55 60% 0.70 0.55 0.83 0.55 90% 1.00 1.00 1.00 1.00 Note: Treatment sche dule reflects 120 days and 55% of buildings. This table shows the percentages of indoor surfaces required to be treated to eliminate transmission assuming each percentage of adults resting outdoors. Elimination of transmission could not be achieved assumin g 90% of adults rest outdoors for either year, however DENV transmission was reduced given treatments began in 2008.
118 CHAPTER 5 GENERAL CONCLUSION Synthesis with Regard to Dengue C ontrol Based on the work conducted here it is possible that imported DENV f rom viremic traveler (s) visiting Key West, Florida resulted in local transmission beginning in 2009. It was the first time the disease was reported in the state in over 70 years. 24 Estimates of visitor introductions were seen to vary annually as a function of incidence rates in dengue endemic countries and the numbers of visitors from these locations; estimates of viremic passenger days per year were lower in 2007 and 2008 than estimates for 2009 and 2010, the two years with local transmission. A s the transmission process is stochastic, it is understood that there is not a one to one relationship between introductions and subsequent local transmission. Lifestyle, the numbers of potentially viremic visitors, and high populations of Ae. aegypti that exist in Key West likely allows this location to be receptive to DENV; the absence of this complex of factors makes other areas north of the Florida Keys less receptive. Autochthonous i nfections in Key West increased in 2010. The estimation of viremic pas senger days per year suggested that additional autochthonous transmission was likely in 2011, yet no cases of dengue were reported. Perhaps dengue was under reported in 2011 and 2012. Diagnostic tests for dengue specific antibodies were available at no cos t to residents by the Monroe County Department of Health in Key West during the 2009 and 2010 outbreaks but not in 2011 and subsequent years. 22 Additionally, it is possible that the public health campaign (ABCD) by the Monroe County Department of Health initia ted in 2011 in Key West lead to an increased awareness to dengue and hence potentially the knowledge that serologic testing did not
119 improve treatment options. 14 Despite the lack of official case reports, FKMCD field inspectors in Key West frequently mentioned meeting with residents reporting or having sym ptoms consistent with primary dengue infection in 2011 and 2012. 22 This is consistent with estimat es of traveler introductions for 2011 and the evidence provided by the 2012 serosurvey where local residents without a recent travel history to dengue endemic are as were IgM positive Additionally, the dengue models parameterized with local data indicated that the ratio of Ae. aegypti to humans in 2011 was sufficient for transmission to occur if there were introductions. However, this prediction of cases in 2011 by the models could be related to high mosquito counts that do not agree with BG trap data from Key West. Additional work within the models is needed before this can be put forth as a definitive conclusion. Historically, with the exception of a few years in Cuba and a few decades in Singapore, dengue control via vector control has been a failure. 94 101 The dengue models were parameterized for Key West to evaluate a residual insecticide based control strategy ta rgeting adult Ae. aegypti in residential areas The models suggest that current control efforts in Key West were likely to be insufficient as well in eliminating local transmission associated with introductions, but without further knowledge regarding the m ismatch of CIMSiM output and BG trap data, it is unknown if residual adulticides could be the best solution. Locations where local residents and visitors overlap spatially, e.g., open air restaurants and bars, shops and boutiques should be considered for s pecific residual use given the simulated success of mosquito reduction and retrospective elimination of transmission in modeled areas. The use of a residual adulticide was not expected to significantly reduce absolute adult densities, rather they
120 were only expected to slightly reduce adult daily survival resulting in the majority of female Ae. aegypti dying before viral dissemination occurs and they transition from being infected to being infectious. It was surprising that with only moderate effort and fund s, that mosquito populations could be substantially reduced. The successful simulation of elimination of transmission supports the notion that a residual adulticide could be successful in Key West given a residual adulticide that behaves in the manner simu lated in the model is used. As Ae. aegypti in Key West are resistant to pyrethroids, this residual adulticide would be rendered ineffective. While the pupal/demographic survey was limited to one hundred homes in the Old Town section of Key West, the models appeared to be sufficiently parameterized to capture the essential features of the history of local transmission in Key West for the years studied, 2008 2011. However, as noted several times previously, mosquitoes predicted by the models did not agree wit h what was observed in Key West BG traps from 2009 to 2011. Based on this, more work is needed in the models. However, residual adulticide applications could still be explored by FKMCD as a cost effective method for reducing the likelihood of local transmi ssion in the face of continued introductions. Limitations Viral Introductions This dissertation is not without limi tations. First, assumptions in C hapter 2 may have led to over estimation of viremic visitor days to Key West If estimates of viremic airline passenger and viremic visitor days were indeed overestimated, the likelihood of traveler introductions of DENV in Key West could be minimal, thus further exploration into other pathways that DENV arrived to the island should be considered. In contrast,
121 bec ause a irline passenger data for flights from the western hemisphere through Miami to Key West is a proxy of the total visitors traveling from dengue endemic countries it is possible that introductions to Key West could be higher than those reported herein More accurate estimates of introduction s by the airline industry could be deduced if airline passengers originating in the Eastern Hemisphere, such as Africa and Asia, as well as alternate connecting airports to Key West were considered. Limitations in c ruise ship data also could have had an impact on estimates of viremic cruise ship visitor days. First, assuming the minimum incubation period (3 days) for cruise ship passengers may have led to an overestimation of individuals that were viremic upon arrivi ng on the island. If a more accurate estimate of incubation periods of DENV in cruise ship passengers were to exist in the future, this analysis could be repeated to obtain a more accurate estimate of viremic cruise ship passengers in Key West. Next, port lists and itineraries from CBP documented a maximum of three prior ports of call prior to Key West. It is possible that ports outside this range (e.g., cruises that make more than 3 stops before docking in Key West) could have a greater possibility to brin g viremic cruise ship visitors because of the increased duration of travel to Key West. Tracking these cruise ships that have more than three stops prior to Key West, the number of passengers and crew on board, and the dates of docking in each dengue endem ic port as well as Key West could help give a more accurate depiction of viremic cruise ship passenger days in Key West. Seroprevalence Survey This serosurvey was also not without limitations. Rapid diagnostic kits did not distinguish between DENV serotyp es, thus follow up with seropositive individuals is needed to determine their DENV serotype exposures. Initially a plan was set in place for
122 the Monroe County Department of Health to collect and ser otype samples from IgM positive individuals, however it wa s not possible to complete this aspect of the project If serotyping were to occur on all positive participants, it is possible that a c learer relationship between IgG positive individuals and the 2009 and 2010 outbreak could be determined. Additionally, c rossreactivity with other flaviviruses has not been well documented for these kits, however because no other flaviviruses have been reported in Key West, it is expected that the risk of crossreactiv ity is low. If crossreactivity did occur, it is possible t hat IgM and IgG positive individuals were overestimated and could have been previously infected with another illness. This would mean that the risk of sequential infections in Key West would be lower than estimated (Chapter 3) and that the impact of DENV could be minimal. Another potential limitation is small sample size. During enrollment, the general public responded well to kits and questionnaires, but political reasons caused enrollment procedures to be terminated prematurely. Because of the small sample size, the power of the study to detect meaningful differences in the study population was reduced. This means that with this smaller sample size, it is possible that estimates for the ri sk for sequential infections could be different than the 10 percent estimated (Chapter 3). Regardless of this limitation, the trends uncovered in the serosurvey provide information that can still be helpful to develop vector control policies. Simulation S tudies The modeling analysis in the Chapter 4 was limited by 100 residences being used to parameteriz e of the model. Because this lack of container information and the variability in containers and pupal density in Key West, it is likely that the true scen ario
123 of Ae. aegypti productivity in Key West was not observed. This is evident by the disagreement in mosquito counts by CIMSiM and BG trap data 46 F or more accurate model projections, a more in depth pupal demographic survey will need to be conducted considering the factors mentioned in Chapter 4 Though 100 residences were sufficient in predicting retrospective transmissio n herein, increased accuracy of mosquito counts could be achieved if additional residences were surveyed in addition to documented productivity of gutters, cisterns, or other cryptic breeding sites in Key West. These two containers alone could have a large impact on the dengue scenario in Key West, thus further information regarding their productivity should be documented. Another limitation is the lack of age specific serotype data for Key West from the 2012 serosurvey To accurately represent the populat ion at risk of sequential infections, more information regarding seroprevalence of dengue serotype specific antibodies should be obtained across all age groups. As the models were parameterized based on two studies with small sample sizes, the true proport ion of individuals that are seropositive in Key West could be substantially different. A larger serosurvey should take place to determine seroprevalence across the island. Moreover, viral isolation and sequencing of all seropositive individuals should also be conducted to determine which strain(s) of virus have circulated in Key West. This would provide the model with more accurate antibody data and would make outbreak predictions more accurate given mosquito counts and infected visitor or infectious mosqui to introductions were correct. Simulation studies were further limited by the lack of randomization possibilities with the introduction of virus into DENSiM. Because of this, the models are likely biased and probably do not reflect hat actually occurred. M odels could be improved not only by
124 allowing random introductions to occur throughout the year, but also by allowing the same serotype to be introduced at varying frequency across several years. Lastly, introductions of DENV can occur via one of two mechan isms: humans and mosquitoes. 32 This study utilized the introduction of virus via visitors or returning resident to Key West, however mosquitoes have also been linked to vector borne disease outbreaks in susceptible regions. 32 When comparing the two, transporting mosquitoes to new regions via air travel is unlikely when compared to the capabilities of human passengers to deliver the disease thus this mechanism of introduction was not considered here. 32 If mosquitoes were the true source of virus, the models should be repeated once introduction frequencies and amounts have been estimated. Limitations on the residual adulticide applications include the lack of information on outdoor resting sites for adult Ae. aegypti Given the fact that residual applications rely strongly on where vectors rest, it is imperative to know what proportion of adult mosquitoes rest in which location for treatments to be effective. Additionally, because Ae. aegypti in Key West are resistant t o pyrethroids, exploration of a different residual spray should be conducted prior to implementation. Further research should be conducted to determine the proportion of resting adults in and around areas where tourists, vectors and residents overlap in Ke y West (e.g., open air restaurants, bars and boutiques) using backpack aspirators to determine where these resting sites occur and if there are sufficient mosquito counts to warrant these applications. This could be better understood given it was explored, perhaps in the improved pupal/demographic survey or during a separate study on the island.
125 Areas of Future Research Dengue Threat to the United States Imported dengue fever has been on the rise for several decades and shows no sign of slowing down in the immediate future. 6 12 13 66 67 Because of the increasing dengue incidence in popular tourism locations, the amount of dengue infected travelers increases, as does the possibility of outbreaks within the United States given the ve ctor is present in that area. 16 51 As our global connectedness incre ases and our climate changes, newly susceptible areas may become more favorable for both vectors and pathogens to establish themselves. 34 Additional research is necessary into the influence of the cruise ship, airline, and trade industries on disease introductions into vulnerable populations; specific ally the introduction of DENV into susceptible areas by both infectious mosquitoes and returning residents or visitors. This research should take into account any location that encounters trade items, returning residents or visitors from dengue endemic loc ations. When looking at each industry individually, special precautions should be taken to separate returning residents from visitors since these two groups of individuals are likely to have differing exposure rates during travel. The potential of airline passengers to be at differing points within their own viremic period (and viremia thereafter) upon entry to susceptible areas should be considered in future works estimating viremic traveler days. Moreover, for cruise ships, vectors and their respective tr ansmitted diseases on board should be examined; there is a significant lack of evidence in the literature on this topic. Another aspect of the cruise ship industry is with regard to DENV infection risk in ports of call. Differing risks within the ports of call and the rest of the
126 dengue endemic country should be explored as well as how behaviors of visitors change when visiting a port of call compared to the rest of the country. As the world becomes increasingly connected, it is expected that traveler intro ductions of DENV could continue to become a significant issue in the United States and other nations where virus, vectors, and susceptible people overlap. Given the probable magnitude of introductions, policies regarding vector control and disease surveill ance should be put in place to reduce the probability of DENV introductions and sequential infections. It has already been stated that the increase in travel and trade has led to a rise in the incidence of DHF and DSS by allowing various serotypes to circu late more freely, as this increases the likelihood of reinfection with a different strain. 2 The probability of travelers acquiring severe dengue infections could be reduced should research be conducted to guide policy decisions in the area of surveillance and education of travelers to dengue endemic areas. Need for Improved Surveillance The CDC currently has a national surveillance system for arboviral diseases in the United States, ArboNET. 19 It has been charged with organizing national epidemiology, incidenc e and geographic spread of all of the arthropod borne diseases and to provide timely information to public health officials, clinicians, researchers, and members of governmental parties. 19 Dengue cases are passively reported to ArboNET from state and metropolitan health departments, however this is often problematic due to the fact that physicians are often not educated in arboviral diseases, hence they do not include them in dif ferential diagnoses. 5 66 If a clinician is not familiar with DENV, it is likely that imported dengue cases go unnoticed and are not reported. 51 In an area where the population is susceptible, vectors are abundant, and behavior of residents and
127 mosquitoes allow for adequate contact, these imported cases could translate into local transmission. Acti ve surveillance of dengue cases could be made more affordable especially with further exploration of the use of rapid diagnostic kits, as these are often less expensive, invasive, and time consuming than other tests with comparable diagnostic capabilities. 79 84 Not only could active surveillance find more cases of imported or locally acquired dengue, but they can also help to guide vector control by providing mosquito co ntrol districts with areas to target for various control strategies. Proactive prophylactic vector control is not only highly effective in preventing dengue infections (in agreement with conducted simulation studies), it also costs less than delayed respon ses to DENV outbreaks. 52 By having a proactive instead of retroactive surveillance of cases, vector contr ol can be more effective in reducing or eliminating transmission and the economic and community impacts that follow dengue outbreaks. Research should be conducted to determine if surveillance of imported and local cases is able to evolve into more of an ac tive approach to save time and money during potential outbreak scenarios as well as if policy changes may be possible such that future populations may benefit. Research to Guide Policy Changes Without changes in policy, it is not likely that practices to c ontrol dengue will improve. Research should be conducted with the goal of guiding policy changes. Such research should focus on implementing interventions to help reduce traveler introductions in the United States and other areas where dengue is not yet en demic. It is evident that traveler introductions are not likely to resolve on their own, thus one intervention could be to reduce the risk of DENV infection in dengue endemic countries.
128 Educational campaigns could be utilized to show appropriate mosquito a voidance techniques, proper clothing to wear on their vacation, and which hours of the day to avoid spending outdoors. With this preparation, travelers to dengue endemic areas could reduce their contact with mosquitoes and virus during their visit. Because infectious mosquitoes are also a source of viral introduction, vectors have been documented to travel in airplanes from one area to another, and their ability to spread disease in these novel locations, the magnitude of infectious mosquitoes traveling the globe should be further explored. 35 Research could be directed to collect or model mosquito travel and vira l introductions via aircraft and cruise ships. This could be performed in conjunction with exploring the impact of mosquito introductions into DENV susceptible areas, the manner in which the vectors arrived in the area, and how to prevent vector introducti ons. 35 Residual sprays on aircraft can prevent importations of vectors and reduce the risk of disease transm ission. 35 To this end, it could be reasonable to explore the feasibility of mandatory disinsection of aircra fts, specifically from areas that are experiencing outbreaks of vector borne disease. Given disinsection could not occur in various locations, research to guide policy regarding improvement of vector control strategies should occur. Simulation studies coul d be of assistance in areas with a known mosquito population, population estimates, and DENV serology and seroprevalence. With simulations of vector control strategies prior to implementation, areas could avoid having to implement strategies before they kn ow a reasonable value of their success, thereby saving time and resources. Research into CIMSiM and DENSiM themselves should also be conducted
129 such that randomization of viral introductions is possible, spatial components may be considered, and a wider ran ge of areas may use them. Conclusion The hypothesis that dengue virus introductions led autochthonous transmission in Key West in 2009 and 2010 is supported by the work reported here. The estimate that 7% IgG seropositivity and about 4% IgM seropositivity of Key West residents should inform any deliberations on vector control as sequential infections and associated severe illness are a real possibility. Research to develop and document efficacy of a more effective vector control strategy is warranted. Model ing studies suggest that targeted residual adulticide applications may prove prophylactic and sustainable by virtue of being more cost effective.
130 APPENDIX SEROPREVALENCE SURVEY QUESTIONNAIRE
135 LIST OF REFERENCES 1. WHO Global Strategy for Dengue Prevention and Control, 2012 2020. Geneva : World Health Organization, 2012 2. Knowlton K, Solomon G, Rotkin Ellman M. Mosquito Borne Dengue Fever Threat Sprea ding in the Americas. New York : Na tural Resources Defense Council, 2009. 3. Wilder Smith A, Schwartz E. Dengue in travelers. N Engl J Med 2005; 353 (9): 924 32. 4. Tomashek KM. Dengue Fever & Dengue Hemorrhagic Fever. In: Brunette GW, editor. CDC Health Information for Internationa l Travel 2010. 2010 ed. Atlanta : U.S. Department of Health and Human Services, Public Health Service 2011. 5. Gubler DJ. Dengue and dengue hemorrhagic fever. Clin Microbiol Rev 1998; 11 (3): 480 96. 6. Kyle JL, Harris E. Global spread and persistence of dengue. Ann u Rev Microbio l 2008; 62 : 71 92. 7. Halstead SB. Dengue. London, UK: Imperial College Press; 2008. 8. Halstead SB. Dengue virus mosquito interactions. Annu Rev Entomol 2008; 53 : 273 91. 9. Gubler DJ, Kuno G. Deng ue and dengue hemorrhagic fever New York, USA: CAB Inter national; 1997. 10. Focks DA, Haile DG, Daniels E, Mount GA. Dynamic life table model for Aedes aegypti (Diptera: Culicidae): analysis of the literature and model development. J Med Entomol 1993; 30 (6): 1003 17. 11. Gratz N. Vector and rodent borne diseas es in Europe and North America: distribution, pu blic health burden, and control Cambridge UK: University Press 2006. 12. Bouri N, Sell K T, Franco C, Adalja AA, Henderson D, Hynes NA. Return of Epidemic Dengue in the United States: Implications for th e Publ ic Health Practitioner. Public H ealth R ep 2012; 127 (3): 259 66. 13. Gubler DJ. Resurgent vector borne diseases as a global health problem. EI D 1998; 4 (3): 442. 14. Adalja AA, Sell TK, Bouri N, Franco C. Lessons learned during dengue outbreak s in the United States, 2001 2011. EID 2012; 18 (4): 608 14.
136 15. Effler PV, Pang L, Kitsutani P, et al. Dengue fever, Hawaii, 2001 2002. EID 2005; 11 (5): 742 9. 16. Farrar J, Fo cks D, Gubler D, et al. Towards a global dengue research agenda. Trop Med Int Health 2007; 12 (6): 695 9. 17. Ramos MM, Mohammed H, Zielinski Gutierrez E, et al. Epidemic dengue and dengue h emorrhagic fever at the Texas Mexico border: results of a household based seroepidemiologic survey, December 2005. Am J Trop Med Hyg 2008; 78 (3): 364 9. 18. Centers for Disease Control and Preventi on Dengue fever at the U.S. Mexico border, 1995 1996. MMWR Morb Mortal Wkly Rep 1996; 45 (39): 8 41 4. 19. WHO Strengthening implementation of the global strategy for dengue fever/dengue haemorrhagic fever prevention and control. Geneva: World Health Organization 1999. 20. Anil L, Stanek DR, Blackmore CG, Stark L, Mock V. Florida Arbovir us Surveillance Week 52: December 26 January 1, 2011: Department of Health Bureau of Environmental Public Health Medicine; DOH Bureau of Laboratories; 2011. 21. Key West Chamber of Commerce Key West and Monro e County Demographics. 2013 http://www.keywestchamber.org/PDF/demographics.pdf (accessed Feb 8, 2013 ) 22. Doyle M, Leal A. Dengue Fever and Entomological Scenario in Key West, Florida. Personal Communication 2011. 23. Centers for Disease Control and Prevention Locally acquired Dengue -Key West, Florida, 2009 2010. MMWR Morb Mortal Wkly Rep 2010; 59 (19): 577 81. 24. Centers for Disease Control and Prevention Imported Dengue Florida, 1997 1998. MMWR Morb Mortal Wkly Rep 1999; 48 (50): 1150 2. 25. Ehrenkranz NJ, Ventura AK, Cuadrado RR, Pond WL, Porter JE. Pandemic Dengue in Caribbean Countries and the Souther n United States Past, Present and Potential Problems. N Engl J Med 1971; 285 (26): 1460 9. 26. Florida Keys Mosquito Control District M osquito Control Methods. 2012 http://keysmosquito.org/programs domestic field ent services offshore trucks aerial/ (accessed Jun 4, 2013)
137 27. Radke EG, Gregory CJ, Kintziger KW, et al. Dengue outbreak in Key West, Florida, USA, 2009. EID 2012; 18 (1): 135 7. 28. Guzman MG, Halstead SB, Artsob H, et al. Dengue: a continuing global threat. Nat ure Rev Microbiol 2010; 8 (12 Suppl): S7 16. 29. Graham AS, Pruszynski CA, Hribar LJ, et al. Mosquito associated dengue virus, Key West, Florida, USA, 2010. EID 2011; 17 (11): 2074 5. 30. Munoz JJL, Santiago GA, Margolis HS, Stark L. Genetic Relatedness of Dengue Viruses in Key West, F lorida, USA, 2009 2010. EID 2013; 19 (4): 652 4. 31. Lounibos LP. Invasions by insect vectors of human disease. Annu Rev Entomol 2002; 47 : 233 66. 32. Tatem AJ, Hay SI, Rogers DJ. Global traffic and disease vector dispersal. Proc Natl Acad Sci U S A 2006; 1 03 (16): 6242 7. 33. Bataille A, Cunningham AA, Cedeno V, et al. Evidence for regular ongoing introductions of mosquito disease vectors into the Galapagos Islands. Proc Biol Sci 2009; 276 (1674): 3769 75. 34. Tatem AJ, Hay SI. Climatic similarity and biologi cal exchange in the worldwide airline transportation network. Proceedings. Biological Sciences 2007; 274 (1617): 1489 96. 35. Martins VEP, Alencar CH, Kamimura MT, et al. Occurrence of Natural Vertical Transmission of Dengue 2 and Dengue 3 Viruses in Aedes aegypti and Aedes albopictus in Fortaleza, Cear a Brazil. PloS O ne 2012 ; 7 (7): e41386. 36. Kilpatrick AM, Daszak P, Goodman SJ, et al. Predicting pathogen introduction: West Nile virus spread to Galaipagos. Conserv Biol 2006; 20 (4): 1224 31. 37. Joshi V, Mourya D, Sharma R. Persistence of dengue 3 virus through transovarial transmission passage in successive generations of Aedes aegypti mosquitoes. Am J Trop Med Hyg 2002; 67 (2): 158 61. 38. Wilson ME. Travel and the emergence of infectious diseases. EID 1995; 1 (2): 39. 39. Torresi J, Leder K. Defining infections in international travellers through the GeoSentinel surveillan ce network. Nat ure Rev Microbiol 2009; 7 (12): 895 901. 40. Tatem AJ, Rogers DJ, Hay SI. Global transport networks and infectious disease spread. Adv Parasitol 2006; 62 : 293 343.
138 41. United States Geological Survey Dengue Fever (imported) Human 2010. http: //diseasemaps.usgs.gov/2010/dep_us_human.html (accessed Jun 4, 2013) 42. Kuan MM, Lin T, Chuang JH, Wu HS. Epidemiological trends and the effect of airport fever screening on prevention of domestic dengue fever outbreaks in Taiwan, 1998 2007. Int J Infect Dis 2010; 14 (8): e693 7. 43. Chen CF, Shu PY, Teng HJ, et al. Screening of dengue virus in field caught Aedes aegypti and Aedes albopictus (Diptera: Culicidae) by one step SYBR green based reverse transcriptase polymerase chain reaction assay during 2004 2 007 in Southern Taiwan. Vector Borne Zoonotic Dis 2010; 10 (10): 1017 25. 44. Taiwan Tourism Bureau MOTC Republic of China 2012. Visitor Arrivals by Year (1956 ). 2012 April 3 2013 http://admin.taiwan.net.tw/statistics/year_en.aspx?no=15 (accessed Jun 4, 2013) 45. Centers for Disease Control and Prevention Dengue Branch 2010: Largest Dengue Outbreak in Puert o Rico History. San Juan : Centers for Disease Control and Prevention 2011. 46. 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. 47. Zeidler JD, Acosta PO Barr e to PP, Cordeiro Jd a S. Dengue virus in Aedes aegypti larvae and infestation dynamics in Roraima, Brazil. Rev Sa u de P u blica 2008; 42 (6): 986 91. 48. Adams B, Boots M. How important is vertical transmission in mosquitoes for the persistence of dengue? Insights from a mathematical model. Epidemics 2010; 2 (1): 1 10. 49. Weaver SC, Vasilakis N. Molecular evolution of dengue viruses: contribut ions of phylogenetics to understanding the history and epidemiology of the preeminent arboviral disease. I nfect Genet Evol 2009; 9 (4): 523 40. 50. Murray TJ. The Impacts of the Cruise Ship Industry on the Quality of Life in Key West. City of Key West Naval Properties Local Redeve lopment Authority, 2005. http://www.keywestcity.com/egov/docs/1156864276_445343.pdf (accessed Jun 4, 2013) 51. Franco C, Hynes NA, Bouri N, Henderson DA. The dengue threat to the United States. Biosecur Bioterror 2010; 8 (3): 273 6.
139 52. Vazquez Prokopec GM, Chaves LF, R itchie SA, Davis J, Kitron U. Unforeseen costs of cutting mosquito surveillance budgets. PLoS Negl Trop Dis 2010; 4 (10): e858. 53. Focks DA, Haile DG, Daniels E, Mount GA. Dynamic life table model for Aedes aegypti (diptera: Culicidae): simulation results and validation. J Med Entomol 1993; 30 (6): 1018 28. 54. Pan American Health Organization, WHO Number of Reported Cases of Dengue and Severe Dengue (DS) in the Americas by Country: Figures for 2010 (to week noted by each country). Washington D.C : World Health Organization 2010. 55. Pan American Health Organization, WHO Number of Reported Cases of Dengue and Severe Dengue (DS) in the Americas by Country: Figures for 2007 (to week noted by each country). Washington D.C : World Health Organization 2007. 56. Pan American Health Organization, WHO N umber of Reported Cases of Dengue and Severe Dengue (DS) in the Americas by Country: Figures for 2008 (to week noted by each country). Washington D.C .: World Health Organization, 2008. 57. Pan American Health Organization, WHO Number of Reported Cases of Dengue and Severe Dengue (DS) in the Americas by Country: Figures for 2009 (to week noted by each country). Washington D.C : World Health Organization 2009. 58. Pan American Health Organization, WHO Number of Reported Cases of Dengue and Severe Dengue (D S) in the Americas by Country: Figures for 2011 (to week noted by each count ry). Washington D.C. : World Health Organization 2011. 59. Schwartz E, Weld LH, Wilder Smith A, et al. Seasonality, annual trends, and characteristics of dengue among ill returned travelers, 1997 2006. EID 2008; 14 (7) : 1081 8. 60. Leeworthy VR, Loomis DK, Paterson SK. Visitor Profiles: Florida Keys/Key West 2007 08: US Department of Commerce, National Oceanic and Atmosph eric Administration, National Ocean Service, Office of National Marine Sanctuaries 2010. 61. Fitton J Key West Port Authority Average Crew into Key West. Personal Communication 2011.
140 62. Chan M, Johansson MA. The incubation periods of dengue viruses PloS O n e 2012; 7 (11): e50972. 63. Naze F, Le Roux K, Schuffenecker I, et al. Simultaneous detection and quantitation of Chikungunya, dengue and West Nile viruses by multiplex RT PCR assays and dengue virus typing using high resolution melting. J V irol M ethods 2009; 162 (1): 1 7. 64. Chen LH, Wilson ME. Dengue and chikungunya infections in travelers. Curr Opin Infect Dis 2010; 23 (5): 438 44. 65. Wilson ME. The traveller and emerging infections: sentinel, courier, transmitter. J Appl Microbiol 2003; 94 Suppl : 1S 11S. 66. Centers for Disease Control and Prevention Travel associated Dengue surveillance United States, 2006 2008. MMWR Morb Mortal Wkly Rep 2010; 59 (23): 715 9. 67. Rynor B. Dengue fever on the rise at tourist getaways. CMAJ 2010; 182 (4): E195 E6. 68. Tropical Medical Bureau Dengout Outbreak Confirmed in Western Belize. http://www.tmb.ie/destinations/news.asp?title=Dengue outbreak confirmed in western Belize&id=1739 62 (accessed Jun 4, 2013) 69. Bain SV. Dengue Fever: An Emerging Infectious Disease in The Bahamas. Int J Bahamian Studies 2011; 17 (2): 67 72. 70. Wichmann O, Jelinek T. Dengue in travelers: a review. J Travel Med 2004; 11 (3): 161 70. 71. Tabachnick WJ. Key West Must Continue the Fight Against Dengue. Buzz Words 2012. http://fmel.ifas.ufl.edu/buzz/wtfight.shtml (accessed Jun 4, 2013 72. Pan American Health Organization, WHO Epidemiological Alert: Dengue In the Americas: Pan American Health Organizatio n 2011. 73. Shepard DS, Coudeville L, Halasa YA, Zambrano B, Dayan GH. Economic impact of dengue illness in the Americas. Am J Trop Med Hyg 2011; 84 (2): 200 7. 74. Puwar TI Murtola T M, Vasan SS, Mavalankar DV Quantifying the Impact of Chikungunya and Dengue on Tourism Revenues. eSocialSciences ; 2010 Jul. Working Papers id:2687
141 75. Carpenter JM, Moore M. Consumer Demographics, Store Attributes, and Retail Format Choice in the US Grocery Market. I nt J Retail Distrib Manag 2006; 34 (6): 434 52. 76. Israel GD. Determining Sample Size. Gainesville, FL: Institute of Food and Agricultural Sciences, University of Florida 2009. p. 1 5. 77. Smith SK, House M. Snowbirds, sunbirds, and stayers: seas onal migration of elderly adults in Florida. Journal of Gerontology Psychological Sciences and Social Sciences 2006; 61 (5): S232 S9. 78. Beatty M, Boni MF, Brown S, et al. Assessing the pote ntial of a candidate dengue vaccine with mathematical modeling. PLoS Negl Trop Dis 2012; 6 (3): e1450. 79. Lam SK, Devine PL. Evaluation of capture ELISA and rapid immunochromatographic test for the determination of IgM and IgG antibodies produced during dengue infection. Clin Diagn Virol 1998; 10 (1): 75 81. 80. Kidwai AA, Jamal Q, Saher, Mehrunnisa, Farooqi FU, Saleem U. Serodiagnosis of dengue infection using rapid immunochromatography test in patients with probable dengue infection. J Pak Med Assoc 2010; 60 (11): 936 9. 81. Aikat A, Sarkar A, Tudu N, Moitra S, Pal PP, Sarkar TK. Diagnostic dilemma of dengue fever: how much reliability are there in rapidity. J Indian Med Assoc 2011; 109 (8): 543 5. 82. Blacksell SD, Jarman RG, Bailey MS, et al. Evaluation of six co mmercial point of care tests for diagnosis of acute dengue infections: the need for combining NS1 antigen and IgM/IgG antibody detection to achieve acceptable levels of accuracy. Clin Vaccine Immunol 2011; 18 (12): 2095 101. 83. Wang SM, Sekaran SD. Early d iagnosis of Dengue infection using a commercial Dengue Duo rapid test kit for the detection of NS1, IGM, and IGG. Am J Trop Med Hyg 2010; 83 (3): 690 5. 84. Goljan J, Myjak P, Nahorski W, Kubica Biernat B, Felczak Korzybska I, Kowalczyk D, et al. Dengue ant ibodies in Polish travellers returning from the tropics. Evaluation of serological tests. Int Marit Health 2010; 61 (1): 36 40. 85. PanBio. Dengue Duo Cassette. Australia: PanBio; 2008. http://www.alere.co.jp/products02/panbio/dengue_duo_cassette.pdf (acces sed Jun 4, 2013)
142 86. United States Department of Commerce Profile of General Population and Housing Characteristics: 2010. 2010 http://factfinder2.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid =DEC_10_DP_DPDP1 (accessed Jun 4, 2013) 87. Fl orida Keys Mosquito Control District Combating Dengue Fever. 2011 http://keysmosquito.org/combating dengue fever/ (accessed Jun 4, 2013) 88. Jetten TH, Focks DA. Potential changes in the distribution of dengue transmission under climate warming. Am J Trop Med Hyg 1997; 57 (3): 285 97. 89. Patz JA, Martens WJ, Focks DA, Jetten TH. Dengue fever epidemic potential as projected by general circulation models of global climate change. Environ Health Perspec t 1998; 106 (3): 147 53. 90. Romero Vivas CM, Falconar AK. Investigation of relationships between Aedes aegypti egg, larvae, pupae, and adult density indices where their main breeding sites were located indoors. J Am Mosq Control Assoc 2005; 21 (1): 15 21. 91. Esu E, Lenhart A, Smith L, Horstick O. Effectiveness of peridomestic space spraying with insecticide on dengue transmission; systematic review. Trop Med Int Health 2010; 15 (5): 619 31. 92. Kroeger A, Lenhart A, Ochoa M, et al. Effective control of dengue vectors with curtains and water container covers treated with insecticide in Mexico and Venezuela: cluster randomised trials. BMJ 2006; 332 (7552): 1247 52. 93. Camargo S. History of Aedes aegypti era dication in the Americas. Bull World Health Organ 1967; 36 (4): 602 3. 94. Focks D. A Review of Entomological Sampling Methods and Indicators for Dengue Vectors: World Health Organiza tion 2003. 95. Suleman M, Arshad M, Khan K. Yellow fever mosquito (Diptera:Culicidae) introduced into Landi Kotal, Pakistan, by tire importation. J Med Entomol 1996; 33 (4): 689 93. 96. Focks DA, Chadee DD. Pupal survey: an epidemiologically significant su rveillance method for Aedes aegypti: an example using data from Trinidad. Am J Trop Med Hyg 1997; 56 (2): 159 67. 97. Murray FW. On the computation of saturation vapor pressure. J Appl Meteorol 1967; 6 (1): 20 3 4.
143 98. Chadee DD, Huntley S, F ocks DA, Chen AA. Aedes aegypti in Jamaica, West Indies: container productivity profiles to inform control strategies. Trop Med Int Health 2009; 14 (2): 220 7. 99. Pilger D, Lenhart A, Manrique Saide P, Siqueira JB, da Rocha WT, Kroeger A. Is routine dengue vector surveillance in central Brazil abl e to accurately monitor the Aedes aegypti population? Results from a pupal productivity survey. Trop Med Int Health 2011; 16 (9): 1143 50. 100. Florida Department of Health Florida Charts. 2013 http://www.floridacharts.com/charts/default.aspx (accessed Jun 4, 2013) 101. Raghavendra K, Ghosh S, Eapen A, et al. Field evaluation of lambda cyhalothrin (ICON 10 CS) indoor residual spraying against Anopheles culicifacies in India. J Vector Dis 2011 ; 48 (1): 18 26. 102. Syngenta Company ICON 10CS Advanced Residual Treatment: Advanced microencapsulated insecticide for high performance indoor resi dual spraying. Schwarzwaldallee : Syngenta, Professional Products 2007. 103. WHO Manual for Indoor Residual Spraying: Application of Residual Sprays for Vector Control: World Health Organization Communicable Disease Contr ol, Prevention and Eradication WHO Pesticide Evaluation Scheme; 2002.
144 BIOGRAPHICAL SKETCH Ali Messenger was born in Tallahassee, Florida and raised in Gainesville, FL at the core of the Gator Nation. She was a second generation gator and has two sister s that decided to become second generation gators as well. After finishing her undergraduate courses in just three years at the University of Florida, she earned her Bachelor of Science in wildlife ecology and conservation, with honors. Given her interests in zoonotic and vector borne diseases, she decided to further her education by pursuing a Master in Public Health degree at the same institution. Throughout her coursework, she found that she had an extreme interest in not only infectious disease, but als o in teaching others. Her internship experience with Dana Focks sparked a yearning for further exploration into vector borne diseases, specifically dengue fever. After graduating with her MPH in 2011, she began her doctorate in public health at the University of Florida During her Ph D, Ali dedicated herself to dengue researc h in addition to various teaching experiences including several guest lectures and serving as a teaching assistant. After receiving her PhD in August 2013 she aimed to employ her collective research and teaching experiences to acquire an academic teaching position to facilitate education of individuals in the field of public health