Guidelines for Arbovirus Surveillance Programs in the United States
Click here for CONTENTS Page
For contact details regarding publication see Title Page

1. INTRODUCTION

Purpose of The Guidelines
General Considerations
Seasonal Dynamics
Patch Dynamics and Landscape Ecology
Meteorologic Data Monitoring
Vertebrate Host Surveillance
Domestic chickens
Free-ranging wild birds
Equines
Other domestic and wild mammals
Mosquito Surveillance
Human Case Surveillance
Laboratory Methods to Support Surveillance by Local and State Health Units
Human serum
Bird and wild mammal sera
Virus identification

Purpose of The Guidelines

Approaches to arbovirus surveillance in the United States vary from state to state (see Appendix I), and surveillance data are rarely comparable. Standardized data collected in a standardized fashion can document regional patterns in the spatial and temporal dynamics of disease activity. That information can be used to predict and help prevent major epidemics.

Our purpose is to provide guidelines for standardization of surveillance for mosquito-borne viral encephalitis. We emphasize predictive, proactive, and efficient methods whenever possible. Following a general discussion of the philosophy of surveillance and the range of available surveillance tools we present, in Chapter 2 - Surveillance Recommendations, recommended surveillance methods for each of the common encephalitides found in the U.S. In Chapters 3-6, we provide brief reviews of the biology and behavior of the vectors and vertebrate hosts of the major encephalitides. In the reviews we discuss only those biological and behavioral characteristics that are important to the surveillance effort. We also have tried to identify important research questions and areas where data are lacking. Finally, several appendices provide supplementary information on case definitions, techniques and equipment for mosquito surveys, and vertebrate surveillance methods. Rather than giving highly specific directions for each method, we refer readers to the original references for details. In addition, many state mosquito control associations or health departments publish guidelines for surveillance and control of mosquito-borne disease.^8,182,204

Surveillance is the organized monitoring of levels of virus activity, vector populations, infections in vertebrate hosts, human cases, weather, and other factors to detect or predict changes in the transmission dynamics of arboviruses. A sound surveillance program requires a thorough understanding of the biology, ecology and interactions of the vertebrate and mosquito hosts. The transmission of arboviruses depends on these interactions. The data needed to estimate the risk of transmission to humans are rarely available within a single agency. It is extremely important that the various data-collecting agencies actively communicate and exchange information.

The impact of prevention or control measures on the course of a potential epidemic is diminished by even the smallest delays. Biologic and ecologic factors influence the temporal pattern and intensity of arbovirus cycles. Optimal environmental conditions allow rapid increase of vectors and virus amplification in vertebrate hosts. It is urgent, therefore, that a well-organized surveillance program be in place well in advance of the virus transmission season. Virus isolation and identification techniques are rapid and new sampling methods can quickly define the vector situation. Still, these procedures require considerable time and effort.

Enzootic virus transmission may occur only at a low intensity among certain vertebrate host and mosquito species within specific habitats in rural or suburban environments. Thus, transmission may remain undetected by most monitoring programs. However, when low host immunity and an abundance of vertebrate hosts and mosquitoes are synchronized with favorable weather conditions, transmission may increase in intensity and expand in distribution, producing an epizootic. If epizootics begin early in the transmission season and if epizootic foci expand into urban centers that possess adequate host and vector populations, the risk of human involvement increases.¹⁷⁸

The prevention and control of arbovirus diseases depend upon identifying and monitoring vertebrate host and vector species involved in spring amplification and on monitoring the sequence of events and forces that lead to epizootics or epidemics. Enzootic vertebrate hosts and vectors also may be involved in epizootic or epidemic transmission. In Memphis, Tennessee, for example, many of the bird species that were involved in enzootic maintenance also participated in epizootic amplification of St. Louis encephalitis (SLE) virus .(McLean, R.G. Unpublished data. a)

A proactive surveillance system designed to provide early warning of epidemic activity should collect data on several variables rather than relying on a single predictor. Control measures should be started when a particular predictor exceeds the action threshold (usually determined from historical data and experience). For example, if early season climatologic data are compatible with epidemic activity, state and local agencies should make contingency plans. Such plans include contracting in advance for aerial ultra-low volume (ULV) insecticide application later in the season when, or if, needed. Ideally, the planning process involves other agencies and interest groups at the earliest possible time. This is the time to begin early-season control activities such as mapping larval habitats, source reduction and educating the public. Some or all of the following factors can increase the predictive ability of arbovirus surveillance programs: season, landscape ecology, meteorologic data, vertebrate hosts, vectors, and human case data.

Seasonal Dynamics

The power of a predictor is the likelihood that, if an outbreak is predicted, it will actually occur. There is a negative relationship between predictive power or accuracy and lead time between predictor and event. Predictions normally become more accurate as the season progresses, but provide less reaction time to carry out control measures to prevent human cases. By the time human cases are confirmed (a very accurate predictor), the epidemic may be waning of its own accord and control measures may have little impact.

Different measures or predictors for epidemic transmission are effective at different times of the year. The earliest useful predictors are 95,295 climatologic factors that influence size of the early mosquito population. These include fall, winter, and spring temperatures, rainfall, snowpack, runoff, and flooding, depending on the virus(es), vector(s), and region of the country.

Mid-season predictors usually consist of population estimates of vectors, and vertebrate hosts (especially young of the year), and evidence of early virus transmission in the natural cycle. The likelihood of an outbreak is estimated by comparing current vector and vertebrate host population densities and age structures with long-term averages. Late-season predictors consist of evidence of virus spill-over to sentinel bird/chicken flocks, epidemic/epizootic vectors, and domestic animals. The likelihood of transmission to humans or domestic animals becomes more accurate as virus begins to circulate in vector and vertebrate host populations.

Patch Dynamics and Landscape Ecology

Localities vary in geography, weather, plant cover, soil type, host and vector distribution, host immune status, etc. Likewise, conditions at a given locality change with time. This spatial and temporal variation (called patch dynamics 227) makes it difficult to use a single criterion as a predictive measure over wide geographic areas 224 or even in one area over several years. Therefore, agencies will need to collect data in a range of different habitats over long periods (5 or more years) to improve the predictive capability of surveillance systems. Once long-term baseline data are available, it is more informative to express vector or host abundance indices as deviations (+ S.D. or S.E.) from the seasonally-adjusted (monthly, weekly) long-term mean index (e.g., as is done for stock market performance or volatility).

Meteorologic Data Monitoring

The great variety of local ecologic factors that influence transmission complicates the use of meteorologic data to predict epidemic arbovirus activity. Different vertebrate hosts and mosquito vector species respond to meteorologic changes in different ways, depending on geographic location and other factors.

In correlating meteorologic data with human disease incidence, problems arise from the focality of weather patterns, and the availability and appropriate choice of local weather data. For example, in correlating temperature and rainfall patterns with a statewide outbreak, which combination of weather stations does one choose as the data source? That is, at what scale should we examine the system? A second concern is the wide variations of temperature, precipitation and other indices that occur on a daily, monthly or annual basis. For a given station, the range in these observations may be extreme and the confidence intervals on the mean extremely broad. Deviations from the norm must, therefore, also be extreme to lie outside the normal limits. Combinations of less extreme deviations may be effective predictors. By comparing current measurements with long-term (e.g., 20-year averages) data, it is much easier to detect significant changes in these factors. 

Certain wind patterns can carry agriculturally important insects to new, distant locations.^139,181,261 Recently, interest has focused on the possibility that infected vectors species also are distributed in this manner. Trajectory analysis was used to match the geographic location of equine and human encephalitis cases with the convergence of southerly-moving warm fronts and northward- moving cold fronts.^256,257 Without large-scale mark-release-recapture studies, however, it is impossible to separate hypotheses based on wind-borne dispersal from hypotheses based on Hopkins' bioclimatic law. The bioclimatic law predicts seasonal retardation of biologic activity with increasing latitude and altitude.¹³⁴

Wild vertebrates are hosts for at least 63 registered arboviruses in North America and hundreds more throughout the world. Moreover, new viruses are 3 discovered continually. In the U.S., however, only four mosquito-borne arboviruses--St. Louis encephalitis (SLE), eastern equine encephalomyelitis (EEE), western equine encephalomyelitis (WEE), and La Crosse encephalitis (LAC)--have had a significant impact on human health. 

There are local and regional differences in vector and vertebrate host species, arbovirus strains, climate, habitats and urban development within the United States. Therefore, no single sentinel host species or specific surveillance technique is effective in all areas. For example, in west Texas, the number of WEE cases in humans was more highly correlated with virus isolation rates from house sparrows than with vector population densities or environmental conditions.^120,133 In California, the statewide surveillance program does not sample wild birds. Studies in that state found WEE virus isolations from Cx. tarsalis, seroconversions in sentinel chickens, and the incidence of WEE in humans all were positively associated with Cx. tarsalis abundance in light traps as indices rose to moderate levels. However, the relation became negative as light trap indices continued to rise.^224,237 Virus isolations from Cx. tarsalis generally preceded seroconversion in chickens.²³⁷ Each local health agency should conduct initial surveys to get information on the relative abundance, potential reproductive activity, and infection rates in vertebrate host species.^125,179,234 This background information is used to design a surveillance system to fit local capabilities and needs.

Some general guidelines can be useful when an arbovirus surveillance program is in the planning stage. A separate publication gives detailed techniques for collecting and handling vertebrates and processing specimens for arbovirus studies.²⁷⁹ That publication includes information on permits required for trapping wild animals. The characteristics that define good vertebrate hosts for arbovirus surveillance include the following: 

1. Susceptibility to the monitored virus at rates that reflect virus activity in the surveillance area,

2. High titer and long duration of antibody response,

3. Low morbidity and mortality (except in those species where high mortality is easy to detect),

4. Locally abundant population,

5. Locally mobile to increase exposure to and dissemination of virus,

6. Frequent exposure to vector species (could overcome lack of mobility),

7. Attractive to and tolerant of vector feeding,

8. Easily captured by conventional methods,

9. Ease in handling and obtaining blood specimens,

10. Age determination possible, at least young of year, or the regular multiple captures of tagged animals permits detection of seroconversions,

11. Relatively long-lived for multiple sampling of same animal.

Probably no vertebrate species is universally suitable for arbovirus surveillance programs. Local abundance, distribution, exposure to vector mosquitoes, virulence of virus strains, and the competence of local vector species may vary regionally. For example, the house sparrow is a good sentinel for SLE virus in midwestern urban settings ^165,178 and for WEE and SLE viruses in rural west Texas.^120,133 It is inadequate as a sentinel for SLE in Florida and California^176,180, for WEE in rural areas in the northern plains states¹⁷⁹ or for EEE in southwestern Michigan.¹⁷⁷ Other species (e.g., the house finch in California ²³⁴) can be used in those areas. Conduct an initial survey to determine the most abundant local bird species exposed to the virus, the species that are easiest to sample, and the best sampling locations.^125,180,179

Arbovirus surveillance programs throughout the United States use a variety of species of birds and mammals. Many other species have been sampled only once as part of a survey to discover which arboviruses were present or which species were tangentially infected. Exposure is increased in long-lived species (wild ungulates) or in those with high mobility or particular feeding habits (carnivores). These latter species may be useful in determining the presence, distribution, and annual prevalence of a virus. Serosurveys of wild ungulates have provided valuable information in several states (see Appendix III for examples).

SLE and WEE virus infections in birds strongly correlate with reported human cases caused by these viruses in the same area.^{120,165,241,288} Some programs regularly sample passerine birds (e.g., house sparrows) or chickens every year during the transmission season to detect annual and seasonal changes in arbovirus activity. To provide more complete coverage of the surveillance area,^133,178 passerine and other free-ranging wild birds can be monitored in areas not covered by sentinel chickens. Some surveillance programs use free-ranging birds exclusively, some use only house sparrows, and others use a variety of wild bird species. The scope of such avian monitoring programs depends on the specific purposes and level of responsibility of the health department. Arbovirus surveillance programs may cover only metropolitan centers, may be regional programs covering parts of states, or they may be statewide.

Captive sentinel animals are used to establish the presence of arboviruses and to monitor temporal and spatial changes in virus activity in an area. Sentinels are sometimes used to attract mosquitoes for virus isolation. The use of sentinel animals allows flexibility. The primary advantage of using captive sentinels is that the time and place of exposure are known. The use of sentinels also assures uniformity in selection of location, habitat, number, breed, age and source of the animals, and sampling schedule. Seroconversion and field infection rates are reliably determined when the foregoing factors are controlled. The disadvantages of sentinel animals include the expense of buying animals, building shelters or cages and maintaining the animals in the field. Also, the lack of mobility of sentinel animals affects their exposure to mosquitoes, and limits the geographic area represented. The following paragraphs discuss the common species used as sentinels.

Domestic chickens:

Probably the most widely used sentinel animal for WEE and SLE surveillance is the domestic chicken. Chickens are attractive hosts for Culex mosquito vectors. They are susceptible to and can tolerate arbovirus infections, and they produce readily identifiable antibodies. Older birds are unlikely to contribute to local virus amplification because they usually develop only low titered viremia. Chickens are hardy and are easily handled and bled. They are inexpensively maintained on farms or in urban-suburban locations by residents or health officials. Eggs laid by the birds may provide an added incentive and help to defray any costs of maintaining the birds.

Six- to eight-week-old chickens are obtained in the spring. Each monitoring site is stocked with 10-30 pretested, non-immune, individually-banded birds. Dispersing smaller groups of birds throughout the area at risk yields a more representative estimate of arbovirus activity. It is important to base the choice of locations for sentinel chickens on historical records of virus activity, vector resting sites or flight corridors, and the likelihood of virus transmission rather than on convenience. The chickens are kept in standard sentinel sheds or similar structures.^231,279

Sentinel chickens are bled from the wing vein, the jugular vein, or from the heart biweekly or monthly throughout the transmission season. Seroconversions may occur 2-3 weeks before the detection of equine or human cases of WEE and weeks before human cases of SLE. If the intent of surveillance is to monitor season-long transmission, birds that seroconvert to positive are replaced by non-immune birds, preferably of the same age. In areas of low intensity of virus activity or where the only objective is to detect initial transmission, replacement is unnecessary since most individuals are still susceptible. All birds are still useful if more than one arbovirus is present in the surveillance area.

Sentinel chickens are used extensively for arbovirus surveillance.^130,156 Currently, a few states like Delaware, Florida, California and Utah use sentinel chicken flocks scattered throughout the areas of greatest risk for EEE, SLE, or WEE infection. Sentinel chickens were not useful for monitoring EEE virus activity in New Jersey.⁶³

Free-ranging wild birds:

Wild birds, principally passerine species, are the primary vertebrate hosts of SLE, EEE, and WEE viruses and serve as the principal hosts for mosquito infection. Virus activity and antibody seroprevalence for these viruses in local bird populations usually correlate well with the risk of human infection. Accurate monitoring of virus and antibody prevalence in wild birds should provide early warning of increased transmission that may constitute a risk to the equine and human populations.

Wild birds are monitored by repeated sampling of local populations to test for antibody or virus. Free-ranging adult and immature birds are captured in ground-level mist nets set at locations appropriate for the desired species. The Australian crow trap ¹⁸¹ also provides an effective method for collecting birds. Captured birds are bled, banded, and released for possible later recapture to check for seroconversions. Recapture data also gives useful insights on movement, survival, and other population characteristics of the birds. Successful use of this technique requires an intensive sampling effort because of low recapture rates. Since antibodies may persist for 2 or more years, the results from carefully identified juvenile birds may provide the most useful index of current virus activity.²⁶⁹ This technique is costly. It requires highly trained personnel as well as state and federal collecting permits.

Detection of viremia in nestling birds during the summer transmission season has been successfully used in WEE and SLE surveillance.^{120,125,133,179} Nestling birds are more susceptible to certain arboviruses than adults. They may produce viremia of longer duration and higher titer, providing a valuable early season indicator of transmission intensity.¹³² Additional information on location, reproductive stage, cycling of broods, and local abundance can be obtained from a survey of nesting activity.^179,190

House sparrow nestlings are a sensitive indicator of recent transmission, and are particularly useful in locations where they are the predominant avian species. They live in peridomestic settings, and are attractive to and frequently bitten by Culex mosquito vectors. The adults' gregarious behavior leads to nests being clustered at specific locations, so nestlings can be sampled easily. Virus isolations from house sparrow nestlings occurred early in the transmission season and correlated well with later human cases of WEE and SLE in Texas.^120,125,133 Nestling birds of other species such as pigeons, house finches, barn swallows, and mourning doves also may be valuable indicator hosts when abundant. These species could supplement or replace house sparrows as sentinels.

Equines: 

Surveillance for equine cases in areas with susceptible horse populations may provide the most practical and sensitive tool for the recognition of a potential public health problem caused by EEE and WEE viruses. This is especially true in areas that lack the resources to monitor virus activity in birds and mosquitoes. As a result of their field exposure, horses are subject to high vector attack rates. Equine surveillance can be active or passive. Reports by local veterinarians of equine encephalomyelitis give warning of increased arbovirus activity in an area.³⁷ This can alert public health officials to investigate the situation. Active surveillance requires regularly contacting large animal veterinarians, encouraging them to report clinically suspect equine cases, and to submit blood and autopsy samples for laboratory confirmation. Record sheets, containing a case history and vaccination history, must acompany samples for laboratory testing if the results are to be useful. Some limitations in using equines are their vaccination status, movement into and out of the surveillance area, and lack of prompt reporting of morbidity by attending veterinarians.

Other domestic and wild mammals:

Wild mammalian hosts are used as sentinels for California serogroup viruses. New Zealand white rabbits stationed in wire cages in wooded areas in eastern Canada confirmed local transmission of snowshoe hare (SSH) virus.¹⁷⁴ Domestic rabbits, eastern chipmunks, and red foxes have been used as sentinels in the north-central states to monitor LAC virus transmission.^109,305 Domestic rabbits¹⁴⁴ and cotton rats were used to detect transmission of Keystone (KEY) virus in the southeastern United States.²⁸² Cotton rats also were used in overwintering studies of SLE virus in the southeast and might be useful in a surveillance program.¹⁷⁶ State-wide surveillance for Everglades virus (EVE) activity in Florida used raccoons.²⁹

Appendix III describes several local and state surveillance systems that use vertebrates. It also lists species of birds and mammals that have been used in arbovirus surveillance programs throughout the U.S.

Mosquito surveillance should have two basic activities, 1) identifying and mapping larval habitats and 2) monitoring adult activity.^35,48 Both activities provide useful information in a proactive arbovirus surveillance system. Mapping and monitoring larval habitats gives early estimates of future adult densities and, under some conditions, provides the information necessary to eliminate mosquitoes at the source. Monitoring species, density, age structure, and virus infection rates in adults provides critical early, predictive data for the surveillance system.

Adult sampling stations usually should be located well away from larval habitats to reduce the number of males and young (nulliparous) females. Alternatively, the program can use gravid traps if they attract the species of interest. A high proportion of males in a collection usually indicates a nearby larval habitat. Data from both larval and adult collections are plotted to show mosquito density as a function of time for each station. Use these data to schedule control efforts and to evaluate control efficacy. Population changes are clearer when abundance is plotted on a logarithmic scale.²⁵

Well-prepared and maintained larval habitat maps to provide long-term baseline data. Maps are updated throughout the season to show the location of mosquito breeding sites and locations with high adult densities. Several automated data collection systems, using hand-held microcomputers, ease data collection and speed up the response to newly discovered larval habitats (b. Street, L.J. 1986. Larval data collection program for the HP-71B. Unpublished programs. Chatham Co. Mosquito Control Commission, Savannah, GA.). State and local agencies also can use computer-based geographic information systems (GIS) for a variety of planning and decision-making tasks.⁷ City, county, and state planning commissions frequently operate GIS programs and have extensive databases. GIS systems can greatly speed and simplify the process of mapping larval habitats, location of known virus foci, urban centers at risk, planning emergency response activities, etc. When several users share the cost of obtaining the data, GIS can be a highly cost-effective means of mapping and planning.

Except when transovarial transmission is a major part of the enzootic cycle (as with LAC virus), the maintenance and transmission of arboviruses is strongly dependent upon adult female survival rates 86,100. It is more likely that older females have fed, acquired virus, and lived long enough to become infective. Surveillance programs often assume that older females are present at some more-or-less constant proportion in the total population (i.e., a stable age-distribution) and, therefore, that the total trap count has a direct relation to arbovirus transmission activity.^185,224 Frequently this is not a valid assumption. For example, as larval populations increase, competition for resources also increases. The availability of nutrients in some larval habitats can vary during a single season, further compounding the effects of competition.^101,259 Adults that emerge from highly competitive situations are smaller and less robust. The reduced adult survival rate leads to proportionately fewer old adults in the population.^1,163 Adult longevity, therefore, is dependent on larval population density. Thus, there is likely to be a stronger correlation between abundance of old vectors and arbovirus transmission rates than between total vectors and transmission.^88,235

Good estimates of changes in the density of parous females, not just of the total vector population, can improve the predictive capability mosquito surveillance. In New Jersey's EEE surveillance program, percent parity in Ae. sollicitans is determined by ovarian dissections.⁶⁴ To selectively sample older components of the vector population, susrveillance programs should use female-retaining gravid traps (see Appendix II) instead of light traps whenever such traps are appropriate for the species being sampled.

Human Case Surveillance

The primary purpose of a surveillance system is to provide information to direct prevention and control activities. The surveillance system has no value if the data collected are not used to implement control measures in a timely fashion. Arbovirus surveillance requires input from many different agencies. Coordination and sharing of data between those agencies are essential for the surveillance system to function properly. State and local public health officials need to be contacted immediately if evidence is found of increased arbovirus activity in a mosquito, avian, or equine population. Similarly, vector control officials should be contacted when a suspected human case of arboviral encephalitis occurs so additional environmental monitoring and appropriate control strategies can be planned.

At the national level, the Division of Vector-borne Infectious Diseases (DVBID), Centers for Disease Control and Prevention (CDC), collects information from the states on cases of arboviral encephalitis. Although state and federal laws do not require physicians or hospitals to report human cases, there has been good cooperation between local, state and federal agencies in reporting cases of arboviral encephalitis.

Standardized report forms and electronic reporting systems are used by state epidemiologists to notify CDC of most reportable illnesses. Forms with demographic, clinical, and epidemiologic information are used to determine whether patients meet the surveillance case definition. Case definitions for the common arboviral illnesses found in the United States are published periodically (see Appendix I).⁵² Although the routine reporting of human cases of encephalitis was discontinued in 1983, many states still report cases and other relevant data, on an informal basis, using the forms shown in Appendix I. Since 1983, DVBID has informally collected information on human arbovirus cases by telephone from state and local agencies. This surveillance system is useful for immediately identifying possible outbreaks of arboviral disease. However, it is very time-consuming, and detailed epidemiologic data on cases of arboviral illness are seldom available. CDC is currently revising human surveillance procedures for arboviral encephalitides to include reporting cases electronically using a standardized report format based on the forms shown in Appendix I.

Arboviral illnesses are widely under-reported in the United States.²⁸⁵ These illnesses have varied clinical presentations that cannot be clinically distinguished from other forms of viral encephalitis, and serologic testing is therefore critical for diagnosis. Because there is no specific therapy for these illnesses, local physicians are often reluctant to obtain samples for serologic tests. Moreover, they must be regularly reminded of the public health importance of arboviral disease outbreaks and encouraged to report suspected cases to state and local health departments rapidly so that investigations and control can be initiated if necessary.

Because several arboviral illnesses have a high inapparent-to-apparent infection ratio, the prevalence of arbovirus antibodies can be high in some populations. A diagnosis of arboviral encephalitis requires that the patient have signs and symptoms compatible with neuroinvasive disease. For reporting purposes, clinical data should be obtained to ensure that the patient meets the criteria for the surveillance case-definition (see Appendix I).⁵² From patients with such signs and symptoms, physicians should obtain both acute phase (1-7 days post-onset) and convalescent phase (>14 days post-onset) serum and cerebrospinal fluid specimens.

When a case of suspected human arboviral encephalitis is reported, the individual's site of exposure and the risk of additional human cases should be assessed. The patient's age, sex, race, and place of residence should be recorded. To determine sites of possible exposure and risk factors for illness, data can be collected on:

a) recent travel to areas with known viral activity in mosquito populations,

b) peridomestic, neighborhood, occupational, or recreational exposure,

c) conditions that promote peridomestic mosquito breeding (e.g., empty tires and containers), and

d) conditions that increase contact with vectors (e.g., gardening, lack of air conditioning).

Even if the immediate danger for other human illnesses seems remote, these data should be sought to provide a basis for future control measures. This list is not meant to be exhaustive, and the epidemiologic data collected should be tailored to each arboviral illness under consideration.

When an outbreak is suspected or anticipated, increased surveillance for human cases should be considered. Special surveillance measures that might be initiated include undertaking active surveillance for encephalitis or meningoencephalitis admissions to local hospitals and enhancing the testing of undiagnosed encephalitis patients. Contacting local physicians and infection control nurses about the need for arbovirus testing and reporting of all suspected cases will increase the sensitivity of the surveillance system to detect cases of arboviral encephalitis. This can be accomplished through direct mailings, participating in local hospital meetings and grand rounds, and giving lectures/seminars to local medical groups. Special studies to detect unrecognized cases, such as routine testing of all cerebrospinal fluid samples drawn during the transmission season, should also be considered. Private diagnostic laboratories also should be included in the list of contacts.

Increased or early arbovirus activity in animal populations may herald an upcoming outbreak of arboviral illness in humans. Five risk categories for arbovirus outbreaks have been defined and appropriate responses established (Table 1). Data collected in vector control investigations may be useful in determining a qualitative probability of an epidemic as well as a stepwise response to this threat. In addition, knowing the type of infected vector, the predominant type of arbovirus, and the location of viral activity may help state and local health departments provide a more focused public health message to groups at high risk for infection. It is critical, therefore, that vector control/surveillance specialists work closely with health department officials to ensure that data can be analyzed and used to direct an appropriate response as early as possible.

Locally relevant predictors of arboviral disease in humans may be obtained if human surveillance data can be correlated with sentinel surveillance data.²²⁴ Parameters of arbovirus activity in defined geographic areas, such as census tracts or mosquito abatement districts, may be collected routinely and consistently over a period of several years by vector control personnel. These data then can be correlated with human arbovirus infections occurring within the same areas during the same time period. With this information, sensitivity, specificity, and positive predictive value calculations can be made to predict subsequent cases of human disease. Such models may be useful in predicting the eventual occurrence of a human outbreak and instituting control measures prior to the appearance of human illness.

Evidence of increased or early arbovirus activity in animal populations may herald an outbreak of arboviral illness in humans. Data collected in vector control investigations can be useful to health departments that monitor human populations for the occurrence of cases. Knowing the vector species, the virus, and the location of viral activity should help health departments to provide a more focused public health message to groups at high risk for infection.

Natural disasters and encephalitis outbreaks:

Natural disasters such as floods and hurricanes can create a potential for epidemics of vector-borne disease. When a response to these disasters or emergencies is beyond the capability of state or local governments, the president may determine that a disaster or emergency exists. A presidential disaster declaration makes state and local agencies eligible for reimbursement of disaster-related expenses. The Federal Emergency Management Agency (FEMA), which oversees all federal disaster activities, calls upon CDC to evaluate the risk of vector-borne disease. Reimbursement for vector control depends on the presence of a clear risk of vector-borne disease that can be related to the emergency or disaster.

In order for CDC to rapidly and accurately evaluate the risk of vector-borne disease, it is important for state and local health and vector control agencies to have readily accessible as much data as possible. Historical data should be available for comparison with current data, to show how the disaster is related to any increase in vector or virus activity. The types of information that are needed to estimate the risk of an epidemic are the following:

a) Mosquito population indices (Are vector species present? How do light trap indices compare with previous years and with this year prior to the current disaster?)

b) Virus infection rates in mosquitoes (What is the minimum infection rate (MIR) this year? How does it compare with MIRs in epidemic years? Is virus activity localized or is it widespread?)

c) Evidence of increased virus transmission in vertebrate amplifying hosts (What temporal and spatial patterns are seen and how do they compare with the norm for this locality?)

d) Evidence of disease in equines (WEE/EEE)

e) Rainfall and temperature data (Is there any evidence to show an association between past outbreaks/epidemics and specific weather patterns?)

f) Time of year (Is it relatively early in the virus transmission season for this locality?)

g) Risk to the human population (Is virus activity near populated areas? Is vector movement between areas of virus activity and populated areas?)

If all of the foregoing information is readily available, a rapid risk assessment can be made using the categories in Table 1. If insufficient information is available, it is necessary to collect at least part of the data before a decision can be made. This frequently delays efforts by state or local agencies to implement the appropriate response. The delay may, in turn, result in increased virus and vector activity and human or equine encephalitis cases. State and local agencies should consider the components of Table 1 and points a) through g) above in designing surveillance programs.

Table 1.1. Definitions and stepwise response for risk categories for mosquito-borne arboviral disease outbreaks in the United States. Risk categories are tentative and approximate. Local and regional characteristics may alter the risk level at which specific actions must be taken.

In addition to federal disaster assistance provided through FEMA, some states have established their own funding procedures for vector- borne disease emergencies. Similar requirements for supporting data may be required for access to state emergency funding.

The choice of laboratory diagnostic tests depends on the needs, approach, and surveillance philosophy of a given health agency. The most commonly used methods include direct and indirect fluorescent antibody (DFA and IFA) tests, hemagglutination- inhibition (HI), complement-fixation (CF), neutralization (N), and IgM and IgG enzyme-linked immunosorbent assay (ELISA) for detection of antibody.^38,39,40,41 Antigen-capture ELISA⁹⁴ is used for direct detection of antigen in mosquito pools, and in human and animal tissues. Various cell cultures⁴² or baby mice are used for virus isolation. The most common methods used to identify virus isolates are DFA, IFA, CF, N, or ELISA. Although it is not yet available for routineuse, the polymerase chain reaction (PCR) shows promise as a rapid and specific arbovirus detection method.¹⁵⁷

Specimen collection:

Specimens may consist of whole blood, serum, cerebrospinal fluid, or tissue samples. These should be processed immediately or placed on dry ice (-70°C) or other suitable deep-freezing agent if virus isolation is to be attempted. Although this may not be critical for antigen detection, shipment and storage of specimens at low temperatures prevents further degradation of proteins. Serum specimens to be tested only for antibody can be shipped at ambient temperatures for brief periods, provided they are collected aseptically and kept free of contaminating microorganisms. If transit time to the laboratory is longer than several days, refrigeration or the addition of antibiotics is necessary to prevent deterioration of the specimen.

Human serum:

One or more of many methods are used for detecting antibody in human serum (see above). Laboratory confirmation of clinical diagnosis depends on direct detection of antigen, virus isolation, or serologic tests. However, the likelihood of SLE, EEE, WEE, LAC, or other arboviral encephalitides being isolated from blood or spinal fluid taken during the acute stage of illness usually not great. Often the viremic stage has is passed before the individual becomes ill. This is not the case with a few viruses for which humans are the principal viremic host in the transmission cycle (dengue fever and yellow fever). These latter viruses may be consistently isolated during the first 5 or 6 days after onset of symptoms.¹¹³ SLE virus may be isolated more often from, or antigen detected by immunofluorescence in, brain collected post-mortem.

Antibody generally is not detectable until the end of the viremic phase. Detectable IgM antibodies usually appear soon after onset of illness and usually persist for only a few months. Their presence can serve as an indicator of recent infection. Detectable IgG antibody appears shortly after IgM and contains antibodies by neutralization, HI, and CF. IgG antibody produced after infections with arboviruses persists for months, years, or even for the life of the individual. Therefore, the presence of IgG antibody does not necessarily denote an active or recent arbovirus infection. The fetus or neonate produces IgM, but not IgG in response to infection in utero or shortly after birth. The large size of the IgM molecule prevents it from crossing the placenta. Thus, the presence of IgG in the fetus or neonate indicates passive transfer of IgG across the placenta.

Measurement of IgM antibody in cerebrospinal fluid is extremely useful for serodiagnosis. Because IgM antibodies do not cross the blood-brain barrier, finding IgM antibodies in cerebrospinal fluid implies intrathecal antibody synthesis in response to central nervous system infection. Moreover, the titer of IgM antibody in cerebrospinal fluid may be a prognostic indicator in certain encephalitides. However, IgM antibodies to some viruses have been detected for long periods, and a minority of patients may have prolonged IgM antibody responses. This limits somewhat the value of these assays as a measure of very recent infection. IgM antibodies seem relatively type-specific for arboviral encephalitides, but complex- and serogroup-reactivity also are observed.

HI antibody is broadly reactive among viruses of a serogroup, making this a useful test for preliminary screening. CF antibody is more complex-specific, short-lived, later to appear, and of lower titer than HI antibody. Finding antibody to a particular virus by CF usually indicates the individual was recently infected with that or a closely-related virus. Certain individuals infected with arboviruses never produce CF antibody, or produce it too late to be of diagnostic value. Nevertheless, the presence of CF antibody in a patient can be used as presumptive evidence of recent infection. As with HI and NT tests, a fourfold rise in titer between paired acute-and convalescent-phase serum samples is confirmatory of infection with that or a closely related virus. CF tests now are considered relatively insensitive for antibody detection and, unfortunately, are no longer widely used. Because birds do not produce CF antibodies, the CF test is not useful for determining antibody in this group of animals.

The HI, CF, and IgM antibody capture (MAC) ELISA tests are not virus-specific. The MAC ELISA is at present, and for the foreseeable future, the test of choice for making provisional serodiagnoses with single serum specimens or with cerebrospinal fluid. It is of great value even when paired acute- and convalescent-phase serum samples are available. The MAC ELISA is comparatively easy to perform, and can be used to test large numbers of serum samples. Furthermore, the presence of IgM antibody usually signifies recent infection, the sine qua non of surveillance.

Bird and wild mammal sera:

Specimens usually are tested for antibody to detect changes in population immunity. This provides evidence for virus amplification in a population. As with human serum, antibody is determined by one or more of the following tests: IFA, HI, IgM and IgG ELISA, and N. N tests are the most sensitive and specific, but are costly and complex to perform. IFA, HI, and IgM ELISA tests often are used to screen serum, with N tests used for confirmation of positive and negative specimens.

Virus identification:

No single virus isolation system is adequate for all arboviruses. More sensitive isolation systems (inoculation of mosquitoes in vivo, inoculation of arthropod cells in vitro) are being increasingly employed.²⁵⁰ It is becoming apparent that there are many virus strains or viruses that have not been detected because of the bias incurred by use of traditional systems, such as suckling mice and vertebrate cell cultures.

Traditional methods for virus isolation are still used in many laboratories. Suckling mice have been used as laboratory hosts for amplifying virus in diagnostic specimens and from field-collected mosquitoes, ticks, and animal tissues. They are inoculated intracranially with clarified suspensions of specimens. Because suckling mice are available to nearly all laboratories, particularly those that isolate rabies virus, this system holds certain advantages over others. Nevertheless, mosquito cell cultures, particularly C6/36 (Aedes albopictus), AP-61 (Aedes pseudoscutellaris), TR-284 (Toxorhynchites amboinensis), and other cell lines are increasingly being used for virus isolation.^111,155

Arthropod cell culture systems have the advantage of ease of containment and reduction of aerosols. These cell lines are highly stable and mammalian cells. Cultures and mosquitoes may be taken to the field, inoculated with clinical specimens, and returned to the laboratory days or even weeks later, during which time virus amplification has occurred. For several viruses, mosquito cell cultures are more sensitive than mice or mammalian cell culture systems for virus isolation. However, they have the disadvantage in some cases of not producing cytopathic effects. Thus, they require secondary steps such as IFA to detect the presence of virus in the culture. Intrathoracic inoculation of Toxorhynchites and have optimal growth at lower temperatures than do male Aedes mosquitoes, which do not take blood meals but in which dengue and other viruses replicate, have also been used with sensitivity and safety.¹¹²

The classical procedure for the initial isolation and identification of an arbovirus begins with inoculation of suckling mice or a cell culture system in which cytopathic effects or plaques develop. The isolate is characterized by testing its ability to pass through a filter that excludes bacteria and its sensitivity to lipid solvents such as ether, chloroform, or sodium deoxycholate. It is often useful to determine the pathogenicity of the agent for, and titers in, various laboratory animals and cell cultures. A crude alkaline extract or partially purified (sucrose-acetone extracted) antigen is prepared for use in serologic tests. The antigen is tested for its ability to agglutinate the erythrocytes of male domestic geese (Anser cinereus) and to react in CF tests with homologous antibody preparations. The antigen is then tested by HI or CF with a battery of antibody preparations. The test will include antibodies to: a) viruses representing various serogroups, b) viruses suspected as the etiologic agent of the disease, and c) viruses known to be present in the area in which the specimen was collected or in which the patient contracted the illness.

The best method for identifying an arbovirus is one that is rapid, specific, and inexpensive. In some laboratories, electron microscopy can be used at an early step to provide an identification at the family level. This can greatly facilitate later characterization. The application of DFA or IFA tests using polyclonal or monoclonal antibodies can provide a rapid and simple means of virus identification. Because a complete battery of reagents is not yet available, this method is only used for the identification of certain viruses at present. Both DFA and IFA tests have been applied to direct detection of viral antigen in clinical specimens.

Once the isolate is characterized to the level of serogroup or antigenic complex by these less specific assays, N tests are performed with antisera against individual viruses to confirm the identification. If necessary, an antiserum is also prepared against the isolate and cross-tested against antigens of viruses in the serogroup to which it belongs. Most of the data regarding antigenic characterization of arboviruses have been generated using these tests. They remain the standards by which newly isolated viruses are to be judged. Newly developed reagents and procedures will add significantly to our diagnostic armamentarium and expand our ability to more fully characterize the epitopes and other antigenic moieties of viruses. For example, monoclonal antibodies are available with group-specificity against many arboviruses. In addition, antibodies have been characterized that show complex-reactive as well as type-specific and even strain-specific reactivities.

Virus is amplified in an in vitro system (C6/36, Vero, other cells), in baby mice inoculated intracranially or in mosquitoes inoculated intrathoracically. The virus is detected by DFA, IFA, antigen-capture ELISA, CF, or N tests. If facilities are available in the local or state health laboratory, definitive identification can be done with reagents obtained from CDC. Alternatively, unidentified or provisionally identified viruses can be submitted to CDC for further studies. Tests performed at CDC include those for biologic characterization (host susceptibility, titer, presence of hemagglutinin, presence of essential lipids, etc.) and IFA, CF, and N tests for definitive taxonomic placement.

Although this general approach has been used successfully for decades, various adaptations of the ELISA test are being applied to virus (antigen) detection and identification. Direct detection of viral nucleic acid using molecular probes (polymerase chain reaction, hybridization) is now being used to detect viruses directly. Furthermore, gene sequencing is used for molecular epidemiologic studies of viruses. Nevertheless, N tests are recommended for definitively identifying viruses that have been provisionally identified by HI, CF, IFA, and ELISA or detected directly.