Sentinel Surveillance and Human Risk for West Nile Virus


wtfig1At the recent 2004 American Mosquito Control Association meeting in Savannah Georgia various methods to assess human risk of WN virus infection were discussed. Figure 1 is a CDC slide presented at the meeting suggesting that dead birds precede WN virus in sentinels, mosquitoes, veterinary cases and humans. Some might also misinterpret this figure to suggest that dead birds are the best surveillance tool for WN infection. However, note that the figure is not an actual graph of WN virus data. Is the figure a composite of information from many regions? What of the many areas where there were dead birds and other indicators of the presence of WN and no human cases? What are the numbers on the axes? Is this the temporal sequence found everywhere? Is the appearance of each indicator in this figure always correlated with the conceptual level of "disease activity?" On what basis was the figure compiled? Where are the data used to compose such a figure? At best this figure suggests that dead birds may be an early "detector" of the presence of the virus in a region but is the figure useful to assess the risk of human virus infection?.

wtfig2Why should we conduct arboviral surveillance? Why should we seek actual indicators that provide a real assessment of infection risk to humans? Figure 2 shows the 1999 New York City WN epidemic sequence. It is clear that mosquito control had little impact on WN transmission since the numbers of human cases were already declining by the time vector control intervention was implemented. Are we any better now? Consider Fort Collins Colorado in 2003. Figure 3 is from a presentation by Dr. Roger Nasci. You can see Dr. Nasci's presentation in full at the web address Note that the mosquito adulticiding began in week 34 when mosquito populations were already in decline. Nasci presents data suggesting control did impact mosquito populations. Did mosquito control reduce human cases in week 35? The cases are reported as week of onset. This means that these cases were infected several days prior and that the time of infection is at least a week earlier to the left. Adulticiding was applied late in the epidemic. Hopefully Florida Mosquito Control will do better. Our goal should be to conduct control in week 26-28 when vector abundance was increasing, and we hope to also have evidence of increasing transmission to sentinels. We must avoid implementing control and intervention strategies too late, e.g., week 34 in Ft. Collins, after most human cases are already infected. This is why WN virus surveillance is essential in Florida.

Various surveillance methods provide valuable data that must be assessed in the context of specific location, times and environmental conditions. Many Florida counties maintain sentinel chicken flocks that provide information to gauge human risk of WN infection. We need to evaluate this information carefully and improve the way sentinels are configured to improve the reliability of the information. Sentinel chicken surveillance information can be used to estimate the mosquito transmission frequency at a specific time and location. If there are 48 birds tested and 5 seroconversions, then one needs an estimate of the numbers of mosquitoes biting birds during the infection period. If we estimate 1000 mosquitoes per bird, then the mosquito transmission frequency is ca. 1 in 10,000 (5 per 50,000 bites). Obviously estimates of the number of biting mosquitoes are important. However, this information can estimate human risk in different counties with different numbers of people. The purpose is to provide a rough prediction of the magnitude of the expected problem.

wtfig3For example, we will use a mosquito transmission frequency of 1 infective bite per 1000 biting mosquitoes, and estimate that on average each person will receive ca. 10 mosquito bites during the period. This is a very broad based estimate on a population level. Certainly one could try to refine this with estimates of subpopulation numbers that differ in exposure, i.e., a group of 10,000 with 100 bites etc. However in the simplest case, Figure 4 shows estimates of the number of human WN cases expected in three counties assuming 20 in every 100 infected people develops a clinical WN infection. This sentinel information can gauge the risk, and we can evaluate expected differences in the numbers of cases under the same levels of risk. There will be more cases in population centers despite the fact that individual risk is no different. 250 cases in Indian River vs. 4500 in Miami Dade. Florida must address this issue of incidence of disease rather then the absolute number of cases. What level of incidence of risk will require a vector control response? What number of cases requires a particular response? If we consider WN disease (20% of human infections), Miami-Dade County will have 230 WN disease cases when the incidence is 1 in 10,000; Indian River County. Ft. Collins had a weekly incidence of ca. 4 in 10,000. Should these jurisdictions be under medical alert? What level should trigger a medical emergency? Suppose there is surveillance information that Miami-Dade and Indian River are at risk for 1 WN disease case per 1000 population. This means 2300 cases in Miami Dade, 130 in Indian River. Is this an emergency situation in both counties? Should outdoor activities be canceled?

wtfig4Why surveillance? Why alert the public? In the above illustrations a 90% reduction in mosquito populations reduces the number of biting females 10 fold. Now include personal protection, messages to reduce human exposure, reducing nighttime and morning outdoor activities, further reducing biting activity on the public 10 fold. A case incidence of 1-10 per 1000 would give 2300-23,000 WN cases in Miami-Dade that could be reduced 100 fold to 23-230 cases. Should night time activities be canceled, should public parks be closed in advance of any human cases with such available surveillance information?

We can impact the big event, but we have to make that impact early in the epidemic! We can use reliable information in the context of the Florida WN Response Plan with different tiers of action commensurate with the estimated danger. Our goal is to reduce the big event in areas that are at risk in advance of human cases. Recall week 26 in Fort Collins. Eventually, as more people become immune to WN virus we will need information on the population immune status to assess overall population risk. The more information we have on mosquito transmission frequencies the better our chances to use our Fig. 4 resources quickly, wisely, improve quality of life, avoid unnecessary negative effects on the economy and at the same time protect the public from WN virus.

Walter J. Tabachnick, Director,
Florida Medical Entomology Laboratory
University of Florida/IFAS