| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
 |
ABSTRACT |
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
A practical alternative to hard and fast numerical thresholds may be a conceptual model to forecast the risk of equine and human cases based on anomalies among environmental factors, mosquito abundance, and enzootic virus activity. The Vector-Borne Disease Section of the California Department of Health Services recently has developed such a model based on a variety of surveillance indicators that interact to produce seasons with low, normal, or epidemic risk of human infection.3 The model incorporates readily available environmental data along with surveillance data collected by local agencies to provide a straightforward way for personnel at these agencies to assess virus risk within their own area. However, before this model and response paradigm can be incorporated into local mosquito control and health planning and intervention programs, it seemed prudent to evaluate plan functionality using retrospective conditional simulations during years with no, enzootic, and epidemic levels of virus activity.
California has had a long interest in the epidemiology, surveillance, and control of two endemic mosquito-borne encephalitis viruses of public and veterinary health interest: western equine encephalomyelitis virus (WEE, family Togaviridae) and St. Louis encephalitis virus (SLE, family Flaviviridae).4 Both viruses are maintained in an enzootic cycle involving various bird species and mosquitoes in the genus Culex, particularly the primary vector, Culex tarsalis Coquillett.5 Humans and horses develop insufficient WEE viremias to infect mosquitoes and therefore are tangential, or dead-end, hosts that play no role in virus amplification. Similarly, humans are also dead-end hosts for SLE, a virus closely related genetically to WN.6 Since the formal inception of a state-managed surveillance program in 1969, the activity of the mosquito-borne encephalitides in California has been monitored intensively by vector control agencies, public health personnel, and University of California researchers. Statewide response plans and surveillance guidelines have been drafted to provide guidelines for effective and consistent responses by appropriate organizations during periods of increased risk for virus activity.7,8 These documents are similar to those prepared by the Centers for Disease Control and Prevention.9 However, these documents do not provide quantitative or semi-quantitative forecasts of risk or targets for intervention. Recently, the California Mosquito-Borne Virus Surveillance and Response Plan3 has been revised in preparation for the arrival of WN or other emerging arboviruses. In the current plan, environmental assessments are combined with various measurements of enzootic and epidemic virus activity to determine an overall risk for the occurrence of equine or human clinical cases and to provide escalating response recommendations to mosquito control agencies for intervention. Because the state’s surveillance program for mosquito-borne viruses has focused on WEE and SLE, our study evaluated the utility of the response plan using conditional simulations with historical data for years with varying levels of virus activity. These simulations provided an indication of whether the plan would have prescribed appropriate response levels as the seasonal cascade of surveillance measurements unfolded during years with no, enzootic, or epidemic levels of virus activity.
|
MATERIALS AND METHODS |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
and Figure 1). The largest recorded epidemic of mosquito-borne disease in California occurred during 1952, with 375 laboratory-confirmed cases of WEE and 45 cases of SLE; 100 and 16 cases occurred in Kern County, respectively.10,11 This year also allowed evaluation of the response plan for years and districts in which the surveillance system is skeletal because in California during 1952, sentinel chicken flocks, New Jersey light traps, and dry ice–baited traps were not used for surveillance. The remaining study periods and locations were selected from recent years that presumably were more comparable to the current situation in California for which the response plan was developed. Both WEE and SLE have been active in Kern County during the past 20 years. During 1989, an epidemic resulted in 15 human cases of SLE in Kern County,12,13 whereas during 1983,14 1996, and 1998,15 WEE enzootic activity occurred without the detection of human cases. These years were evaluated, and two others in which virus activity did not occur (1989 for WEE virus and 1995 for both viruses) were chosen as negative controls. In the Sacramento Valley, extensive enzootic WEE activity occurred during 1993,16 but human cases were not detected. Data from four counties (two vector control districts) (Figure 1) in this region were evaluated to compare risk assessments for neighboring mosquito control districts with similar levels of virus activity based on separate environmental and surveillance datasets. Los Angeles County (Figure 1) also was selected for evaluation of the risk model, because it is an entirely urban area that experienced an epidemic of SLE during 1984 involving 16 human cases.17
|
|
). All available risk factors are assigned a quintile ranking value between 1 and 5, and these values are averaged to calculate overall risk, which also is on a scale from 1 to 5. Overall risk for WEE or SLE activity is categorized into normal season (1.0–2.5), emergency planning conditions (>2.5–4.0), or epidemic conditions (>4.0–5.0). Because some measures of vector abundance and virus activity are available only for part of the year, and degree of urbanization in areas where virus activity has occurred can, by definition, only be included when virus activity has occurred, the number of factors included in the calculation of overall risk varies temporally depending on the number of factors being measured. For our study, overall risk for each half-month was calculated beginning with the second half-month of the calendar year and was based on data from the previous half-month to simulate information that would have been available to mosquito control districts if the Plan had been used in real-time to provide an estimate of risk.
|
) from internet sources,18–20 whereas historical mosquito abundance and virus activity data were assembled from the archives of mosquito and vector control districts (MVCDs), annual statewide surveillance summaries published in the Proceedings of the Annual Meetings of the Mosquito and Vector Control Association of California,21–28 and from research publications29–31 (Table 1
).
|
). Biweekly method. Total precipitation, total runoff, and mean temperatures were determined for each half-month of each year from 1951 to 2000 for each station, and the 50 values (one for each year) for each of the 24 half-months were grouped into quintiles, so that an individual half-month could be assigned a risk value based on its quintile ranking. For this method, highest risk values were assigned to the periods with the greatest precipitation, runoff, and temperature levels. Thresholds for risk value assignments were calculated separately for each study area, half-month, and environmental condition variable.
The environmental conditions component of the risk model consisted of water and temperature factors. From November through May, precipitation and runoff quintile rankings were averaged as the water component, which then was averaged with temperature quintile rankings to determine the risk due to environmental conditions. Water-related risk results from runoff alone from June through October because precipitation is negligible during this period for most of the state, and as a result, even a very small amount of rainfall can result in a quintile ranking that overestimates the resulting risk. The area covered by the Greater Los Angeles Vector Control District (VCD) is almost entirely urban, and runoff from snowmelt does not measurably affect the availability of mosquito breeding sites, so risk related to water in this area was based entirely on precipitation.
Cumulative method. Mean temperatures were calculated for each half-month from 1971 to 2000, a period equal in length to that used by the National Oceanic and Atmospheric Administration for calculation of climatologic normals.32 Based on these temperature data, degree-day accrual since January 1 of each calendar year was calculated for each half-month of the 30-year period using the double sine method with a lower threshold of 7.3°C.18 This lower threshold temperature was the average of previous highest and lowest estimates (4.6–10.0°C) of the zero developmental temperature for Cx. tarsalis.33 Alternatively, minimum replication thresholds for the arboviruses could have been used, but our study sought to evaluate a general model for endemic arboviruses within California and WEE and SLE viruses have very different minimum growth thresholds of approximately 10°C and 15°C, respectively.34 For the same period, cumulative precipitation and runoff also were calculated for each half-month of each water year, which begins on October 1 and ends on September 30 of the following year. Individual cumulative precipitation, runoff, and degree-day values for each half-month were calculated as a percentage of their respective 30-year means. Runoff levels for the Coachella Valley were calculated as a percentage of the 15-year mean to minimize the effects of a marked downward trend observed through 1986. The percentage values for the second half of June were used to calculate risk assessment thresholds based on quintiles for the 30-year period. These thresholds for the percentage values were applied to all half-months through June, and the risk assessment for the second half of June was applied to all half-months thereafter. The second half of June was chosen as the terminal date for assessing environmental risk because by this time, runoff is usually at or beyond its annual peak, and rainfall during summer in most areas of California is negligible and would not dramatically change cumulative water levels after June. Also, the period for which runoff, rainfall, and degree-days have been accumulated encompasses the important late winter-spring virus amplification period and therefore establishes risk conditions for the remainder of the year. As in the biweekly method, risk due to rainfall and runoff were averaged as the water component of environmental conditions, and the risk level due to environmental conditions was equal to the average of water and temperature components. Greatest risk for WEE was assigned to periods with the highest temperature, runoff, and precipitation. In contrast, risk for SLE based on environmental factors was assessed in two ways: 1) with high risk assigned to high runoff and precipitation levels (as for WEE), and 2) with risk for water factors inverted so that highest risk occurred during lowest water levels. Method 2 was used because of the frequent association observed between hot, dry conditions and SLE epidemics.35 Temperature assessments were equivalent for all methods; i.e., anomalously warm temperature was ranked as high risk.
Mosquito abundance and virus activity.
Risk for biotic factors was based on the categorizations summarized in Table 2. Risk categories assigned for mosquito abundance and infection rates, chicken seroconversion rates, equine cases, and human cases were generated subjectively, relying on expert opinion and published studies,36–39 rather than calculated thresholds. In the current study, risk values for mosquito abundance were based on a comparison of the number of Cx. tarsalis females per trap-night for the half-month in question to the average for the same half-month during the previous 10 years. If the period of available mosquito abundance data encompassed fewer than 10 years prior to the study period, the maximum number of available years was used. The risk from mosquito virus infection was based on the minimum infection rate per 1,000 Cx. tarsalis females for the previous half month. Extent and intensity of virus transmission to sentinel chickens was measured by the number of flocks with at least one seroconversion and the number of seroconversions per positive flock. For human and equine cases, the risk level escalated when cases were detected within the state and was highest when cases occurred within the local district or region, indicating local tangential transmission. The risk component for proximity of virus activity to populated areas was included during half-months when virus activity had been detected and was assessed based on the most densely populated area in which virus activity had been detected.
To determine whether horse and human cases should represent one or two risk components, overall risk for WEE activity was calculated with the two factors considered separately and averaged as a single risk factor. These two methods, in combination with the biweekly and cumulative methods of environmental risk assessment, resulted in a total of four risk assessment methods for WEE and three for SLE.
|
RESULTS |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
). However, when equine and human cases were combined into a single risk factor, risk for both biweekly and cumulative methods reached emergency planning levels during the second half of April, but immediately subsided to normal season conditions afterward.
|
). The cumulative method with horse and human cases combined indicated emergency planning conditions for each half-month from January through the end of April when surveillance for enzootic virus activity was initiated. Risk then decreased into the normal range for the remainder of the year. When horse and human cases were combined using the biweekly method, only two half-months reached the emergency planning range. For SLE during the same year (Figure 3A), risk levels were equal to those for WEE with horse and human cases combined when risk for environmental conditions was calculated in the same way for both viruses, but when risk level assignments for water factors were inverted for SLE, overall risk only reached emergency planning during late April.
|
and Table 1).14 Risk for the biweekly method with horse and human cases combined reached emergency planning conditions twice during winter and early spring, forecasting WEE activity that would occur later in the year. After July, risk for all methods remained within the emergency planning range for most half-months during the remainder of the enzootic monitoring period.
Only one half-month would have reached emergency planning levels for WEE during 1996 in Kern County, based on the biweekly method with horse and human cases as separate risk factors, despite enzootic virus activity from the second half of June through September (Figure 2D). For the cumulative method with horse and human cases considered separately, only three half-months reached emergency planning levels, including two during the period of enzootic activity. Risk for both methods with combined horse and human cases reached emergency planning conditions three times during the enzootic period; emergency planning conditions also were reached during the winter and spring before virus activity began. For the cumulative method with horse and human cases combined, emergency planning conditions persisted from the beginning of the year through the end of May, but fell into the normal range just before enzootic WEE activity began.
Enzootic WEE activity also occurred from late July, through early October 1998 in Kern County, but risk assessments with horse and human cases as separate factors indicated normal conditions throughout the year (Figure 2E). When horse and human cases were combined as a single fac tor, biweekly and cumulative methods would have been slightly more indicative of the enzootic conditions that actually occurred, with emergency planning conditions occurring during one and three winter half-months, respectively, and for the biweekly method, during one half-month of the enzootic period.
Risk levels estimated by all methods for Sacramento-Yolo and Sutter-Yuba MVCDs would have been within the emergency planning range for most half-months during which enzootic virus activity occurred during 1993 (Figures 2F and G), a year with very widespread and intense enzootic transmission.16 In Sutter and Yuba Counties, emergency planning conditions were indicated for all half-months following the onset of enzootic activity and were reached during one or more half-months preceding the onset of virus activity. In Sacramento and Yolo Counties, risk levels generally were similar to those in the Sutter-Yuba MVCD, and emergency planning conditions were indicated for most half-months during the enzootic activity period. Risk assessed with horse and human cases as separate factors did not reach the emergency planning range prior to the enzootic activity, but when horse and human cases were combined, emergency planning conditions for biweekly and cumulative methods were reached two or five half-months, respectively, prior to the onset of enzootic activity.
The cumulative risk assessment method with inverted rainfall and runoff risk provided the best indication of the enzootic SLE activity that occurred during the summer and early fall of 2000 in the Coachella Valley (Figure 3B). Using this method, emergency planning conditions were reached following each of the first five half-months during which focal enzootic virus activity occurred.40 This method also foreshadowed the summer virus activity during winter and early spring, with risk reaching the emergency planning range during four half-months from February through April. Risk based on other methods indicated normal conditions during the first four half-months of the enzootic activity period and reached emergency planning conditions only during the fifth half-month. None of the methods, including the cumulative method with inverted water risk, indicated emergency planning conditions following the last three half-months with enzootic activity.
During the following year in the Coachella Valley, enzootic SLE activity began one month earlier, and again, the cumulative method with inverted risk for water factors provided the best indication of risk during the period when virus activity was detected, reaching emergency planning conditions during three of the seven half-months following detection of SLE activity (Figure 3C). However, during 2001, none of the risk assessment methods provided an early-season indication of virus activity that would occur later in the year,40 and risk remained within the normal range for all methods until one month after the first detection of enzootic virus activity.
Epidemics.
Prior to the 1952 WEE epidemic in Kern County, early season emergency planning conditions would have foreshadowed the occurrence of human cases when risk was calculated with horse and human cases combined as a single factor (Figure 2H). However, the emergency planning period was brief (1–2 half-months, depending on the calculation method), and risk subsequently fell into the normal range until virus activity was detected in late May. Risk calculation methods were similar in their ability to represent virus activity levels during this epidemic, and risk based on each of the methods was above the emergency planning threshold for every half-month during the epidemic period. However, the estimates of risk during the epidemic were, in all cases, conservative, and epidemic conditions were indicated during only 2–4 of nine half-months during which human cases occurred.
Human cases of SLE also coincided with the large WEE epidemic during 1952 in Kern County.10 Risk calculation methods showed more variation in their ability to indicate appropriate risk levels for SLE than they had for WEE (Figure 3D). Risk based on the biweekly method and the cumulative method with higher risk assigned to wetter conditions reached emergency planning levels during four half-months prior to epidemic activity. Risk for the cumulative method with higher risk assigned to drier conditions remained within the normal range for the entire year before the epidemic, and epidemic risk levels were never reached, even when human cases were occurring. Risk based on the biweekly method provided the most accurate reflection of SLE epidemic activity, with five of six half-months showing appropriate epidemic risk levels.
During 1984 in Los Angeles County, all three risk calculation methods for SLE were equally indicative of the epidemic periods when human cases occurred, with four of five epidemic half-months correctly reflected by overall risk levels in the epidemic range during the following half-month (Figure 3E). The best early forecasts of epidemic activity were given by the cumulative method with inverted water risk assessment. Using this method, nine of 14 half-months preceding the epidemic were in the emergency planning range. The other two calculation methods resulted in 4–5 of 14 half-months in the emergency planning range before the epidemic.
An SLE epidemic with 15 human cases occurred during late August, through early October 1989 in Kern County. Neither the biweekly method nor the cumulative method with greater risk assigned to wetter conditions reached epidemic conditions, even during the period when human cases were occurring (Figure 3F). However, the cumulative method with inverted water risk did adequately reflect epidemic activity following three of four half-months with human cases. Otherwise, all three methods were similar in their ability to provide an early warning of virus activity, reaching emergency planning conditions during 1–2 half-months prior to the first detection of SLE activity. During the first half-month of the epidemic, all three methods indicated emergency planning conditions as a result of enzootic SLE activity during the previous half-month.
|
DISCUSSION |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
For SLE during all study years, except for 1952 when weather conditions and surveillance methodology differed considerably from recent years, the inversion of risk for runoff and rainfall (i.e., drier conditions were assigned higher risk levels) led to overall risk levels that were indicative of future or concurrent virus activity. During years when epidemic or enzootic SLE activity did occur, overall risk was higher with inverted water risk, but during 1995 in Kern County when SLE activity did not occur, overall risk was lower with inverted water risk. In both cases, the inversion of the assessment of water risk yielded a better indication of actual virus activity levels. These results agreed strongly with the association of SLE activity with La Niña conditions that have been observed since the discovery of SLE as a human pathogen.35,41,42 These associations were related to the positive impact of limited rainfall on Cx. quinquefasciatus population dynamics; however, our assessment was based on Cx. tarsalis as the vector and therefore must be related to the temperature required for effective SLE amplification within host populations.34
Extensive consideration was given to whether horse and human cases should constitute one or two risk factors in the WEE model because both represent dead-end hosts infected tangentially from the basic Culex-bird cycle and are not part of virus amplification. Widespread immunization of horses in California against WEE also has reduced their sensitivity as an indicator of virus activity and limited the number of clinical cases recognized in recent years.43 Because human or equine cases of WEE never have been detected during the spring virus amplification period, inclusion of both factors during that time reduces overall risk levels and therefore limits the ability of the risk model to provide an early warning of pending virus activity (Figures 2C–H). Averaging horse and human cases as a single factor provides a partial remedy by reducing their respective impacts on overall risk, but in years when virus activity did not occur, combining these two factors resulted in early season emergency planning conditions that falsely forecasted pending virus activity (Figures 2A and B). During epidemic periods when both equine and human cases of WEE occurred, treating them as a single factor limited their upward influence on overall risk (Figure 2H), an effect which is not desirable if the goal is to indicate current epidemic activity. Therefore, treating equine and human cases separately has both advantages and disadvantages, and the decision on how to treat them will depend on whether the goal is to predict future WEE activity or to provide an accurate index of current activity.
Another reason for minimizing the influence of human and equine cases on overall risk is related to the poor sensitivity of the current passive case detection system, including case recognition, laboratory confirmation, and reporting. Case detection requires well-directed suspicion by the attending physician or veterinarian, collection of proper laboratory specimens, rapid testing, and reporting of cases to a centralized agency. Missed cases or delays in diagnoses would artificially reduce the risk level during enzootic or epidemic periods. For example, all of the SLE cases during the 1989 epidemic in Kern County were diagnosed retrospectively after the epidemic had waned,13 so risk based on the perceived lack of human cases during the epidemic would have underestimated the actual risk and mostly likely limited the emergency response. In the current study, cases were included by date of onset and not actual detection. Although case numbers must be included in the model to increase risk during periods when cases are detected, environmental and enzootic factors that are better harbingers of virus risk should receive greatest emphasis. In addition, infection rates in California human populations currently are low,44 perhaps because humans now have reduced contact with vector mosquitoes by choosing indoor activities during the early evening period when mosquitoes seek blood meals.45
An additional confounding factor in our retrospective evaluation of the model’s ability to predict human infection has been the possible success of intervention by mosquito control agencies. In each of our representative enzootic transmission years, local agencies intensified larval control and initiated adulticiding to eliminate infected adult mosquitoes. It may be that the lack of human cases detected during these years was the direct result of intervention and public education reducing the risk of human infection. In support of this possibility, serosurveys conducted in Sacramento after widespread and intensive enzootic transmission during 1993 indicated that less than 2% of the human population had been infected.44
Risk for each factor can be assessed only when that factor is monitored, and therefore the basis of overall risk during late fall, winter, and early spring is limited essentially to environmental conditions and equine and human cases, which are monitored year-round (Figure 4). Although some mosquito control districts monitor mosquito abundance year-round, particularly in the southern and coastal regions of the state, most mosquito trapping does not begin until spring. Shortly thereafter, sentinel chicken flocks are deployed. As each additional risk factor becomes available for inclusion, the relative impact of each individual factor on the overall risk is reduced (Figure 4). During a typical year, if early-season environmental conditions and mosquito abundance favor virus amplification, assessed risk levels frequently will reach emergency planning conditions because these factors constitute a major component of overall risk. However, because testing of chicken flocks and mosquito pools commonly begins during mid-spring when virus activity does not occur, the addition of these negative estimates to the overall risk calculation reduces the level to normal conditions. For this reason, overall risk frequently falls into the normal range immediately prior to enzootic or epidemic virus activity, even when early-season risk had been in the emergency planning range. Later, as enzootic virus surveillance factors begin to detect WEE or SLE activity, overall risk levels again increase.
|
Received September 17, 2002. Accepted for publication February 13, 2003.
Acknowledgments: We thank the management and staff of the Kern MVCD, the Sutter-Yuba MVCD, the Sacramento-Yolo MVCD, the Greater Los Angeles County VCD, and the Coachella Valley MVCD for their invaluable assistance in providing historical surveillance records. In particular, we are grateful for the help of Richard Takahashi, Debbie Lemenager, Kenneth Boyce, Rhonda Laffey, Deborah Dritz, Minoo Madon, Jacqui Spoehel, Branka Lothrop, and Arturo Gutierrez.
Financial support: This study was supported by grant NAO6GP0665 from the Office of Global Programs, National Oceanic and Atmospheric Administration.
Authors’ addresses: Christopher M. Barker, Arbovirus Field Station, 4705 Allen Road, Bakersfield, CA 93312, Telephone: 661-588-6957, E-mail: cmbarker{at}ucdavis.edu. William K. Reisen, Arbovirus Field Station, 4705 Allen Road, Bakersfield, CA 93312, Telephone/Fax: 661-589-0891, E-mail: arbo123{at}pacbell.net. Vicki L. Kramer, California Department of Health Services, Vector-Borne Disease Section, 601 North 7th Street, MS 486, PO Box 942732, Sacramento, CA 94234, Telephone: 916-324-3738, E-mail: vkramer{at}dhs.ca.gov
Reprint requests: William K. Reisen, Arbovirus Field Station, 4705 Allen Road, Bakersfield, CA 93312.
* An abbreviated description of this research was published previously in the Proceedings and Papers of the 70th Annual Conference of the Mosquito and Vector Control Association of California.1 
|
REFERENCES |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. L. GATTON, B. H. KAY, and P. A. RYAN ENVIRONMENTAL PREDICTORS OF ROSS RIVER VIRUS DISEASE OUTBREAKS IN QUEENSLAND, AUSTRALIA Am J Trop Med Hyg, June 1, 2005; 72(6): 792 - 799. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
