Mosquito Surveillance and West Nile Prediction: Lessons Learned

The article "West Nile virus surveillance in mosquitoes in New York State, 2000-2004" (Lukacik, et al. Am. Mosq. Control Assoc. 22: 264-271) summarizes the mosquito surveillance program in New York State (NYS) from 2000-2004 based on sampling ca. 1.1 million mosquitoes that were tested for West Nile virus (WNV). By anyone's standards the testing of 35,280 mosquito pools for WNV using QPCR was a prodigious effort. What important lessons were learned that could benefit the Florida surveillance programs?

I recently heard of a White House staff member, Stuart Baker, who is the White House Director for Lessons Learned. I kid you not - he is Director of Lessons Learned with an annual salary of $106,641. Perhaps Mr. Baker is the person to evaluate different mosquito surveillance efforts for lessons learned.

The report in JAMCA points out several findings about the NYS WNV mosquito surveillance program. During the five years of the study, NYS detected 814 WNV positive mosquito pools representing 10 mosquito species. Culex pipiens/restuans pools account for ca. 91% of the WNV positives, while the Culex represented ca. 48% of the numbers tested. What did all the work show? Were the mosquito pools and sampling useful in predicting WNV outbreaks in New York? I submit this is the most important issue for Florida since we are well past the need to "detect the presence" of WNV.

The annual numbers of human cases in NYS ranged between 10 and 83 annually. Most of the mosquito pools were from what the authors define as the epicenter for WNV in NYS where it first appeared in i999. This also happens to contain New York's 12 most populous counties. If one takes into account the population size of NYS (ca. 20 million people), and the population sizes of the few focal areas, the incidence of West Nile in NYS is low and certainly below many western states and outbreaks in the midwest U.S., Pinellas County, and Coconut Grove, Florida, had much higher incidences of WNV. So did mosquito testing provide information to predict the risk to humans in NYS?

Elsewhere Roxanne Connelly has discussed the goals of mosquito surveillance (see Rutledge C. R. 2003. Mosquito Pooling: Getting a return on your investment. BuzzWords, 3(3): 8-91 and many BuzzWords coiumns have discussed the importance and utility of a well run sentinel chicken surveillance program for WNV surveillance as is done in many parts of Florida [Tabachnick and Day. 2001. BuzzWords l(6): 8-9; Tabachnick. 2004. Buzzwords 4(1): 7-53; Tabachnick. 2004. Buzzwords 4(2): 10-12; Tabachnick et al. 2005. Buzzwords 5(2):6]

The information reported for NYS is illustrative of what can be learned about predicting human risk for WNV from such data. For example, the authors report that between 2OO0 arid 2004, 82.8% of counties had a human case of WNV during a year it conducted virus surveillance! Is there a lesson here? Another reported finding is that of the 100 WNV human cases in these counties, 92% (n=92) occurred in a county conducting virus surveillance, with 77% (n=77) reporting a positive WNV pool 1 month before disease onset. I guess this means that of the 92 WNV positive humans, 77 were preceded by a positive mosquito pool, which also means that 15 human cases were not preceded by a mosquito pool despite the testing. BuzzWords readers should by now also be well aware that there are many, many areas of the U.S., including NYS, where the first WNV positive mosquito pool was not followed by a West Nile human case. Are these data useful to differentiate human risk? Can these data be used to predict a "big event?"

Some other findings are provided including: gravid traps were better at obtaining WVNV positive pools than CDC light traps, even when baited with CO2, supporting previousiy published reports: and C. pipiens/restuans are the primary WNV vectors in NYS supporting previously published repolts. The report also fulfills one of the goals of the NYS system which was to provide "necessary surveillance-based evidence of WNV presence since 2000." Readers need to continue to judge the utility and usefulness of such a prodigious effort to assess human risk. The authors of the report admit that the low numbers of human cases in NYS make it very difficult to use these data to predict human cases. However they believe that the program needs to become even timelier in reporting this information, and that these data can be used to identify geographic areas of increased risk, particularly when used with other data from avian, mammal and passive human surveillance. The authors recommend that improvements in infrastructure in NYS to handle larger numbers of mosquitoes is needed.

What can Florida learn from this report? Are there some lessons here for Florida?

It is apparent and widely known by Florida's mosquito and public health professionals that sentinel chicken surveillance as done in Florida has not found much use in many U. S. states, which like NYS have invested heavily in WNV surveillance using mosquito pools. There are many reasons for this including the perception that sentinel chickens do not work in these states. There is no published scientific support for this belief, and indeed there is not even a scientifically testable reason why this should be so. Some are also still confused about the difference between detection and surveillance to assess risk. Despite this issue, let's evaluate the amount of effort being expended in NYS and Florida on surveillance. NYS tested 1.2 million mosquitoes and the results from this effort are those reported above.

How many sentinel chickens would be needed over 5 years to obtain WNV surveillance data comparable to NYS? The results from mosquito testing would not provide the same information as sentinel chicken testing because, after all, we all recognize that mosquito infection data is only infection, and of course does not necessarily translate into actual mosquito transmission. Indeed the NYS data has the further caveat that the mosquito pools were tested using QPCR so that even some of the pools may not have been detecting live or "infectious" virus. There is no doubt that a sentinel chicken seroconversion represents an actual transmission event. All one needs to know, as we have stressed before in several previous BuzzWords, is some estimate on the number of mosquitoes actually being drawn to the chickens. We know this can vary widely, and of course one does not place sentinel flocks routinely in areas where they will not or cannot be exposed to mosquitoes. One wonders how some of the sentinels were placed, and just what the tests using sentinel chickens in some other areas of the U. S. were trying to determine. However, for arguments let's say on average Florida's sentinel chickens receive 1000 bites per sampling period, whether this is every 1 week or 2 weeks. We can debate an actual number of mosquitoes per sentinel, and it will vary, but for purposes of discussion let's just say 1000 which to me is actually realistic. Then we should look at some year round programs and again let's consider they bleed the chickens either every week or every other week. Here are some estimates for 4 different mosquito control districts:

Period (ca.
Indian River MCD 48 48000 124800 249600
Lee County MCD 96 9600 249600 499200
Manatee County MCD 72 72000 187200 374400
Orange County MCD 84 8400 218400 4368000
TOTAL     7800000 15600000

The above shows that 4 Florida counties effectively sample more mosquitoes in 1 year than NYS did in 5 years. Even if one reduces the estimates of the mosquitoes by 10-fold, clearly these 4 counties exceed NYS's efforts. Even with a 100-fold change in the mosquitoes sampled, these 4 counties are comparable to all of NYS's efforts. In sum there is simply no comparison between the entire state of Florida's WNV surveillance programs and the considerable effort in NYS. Florida's Tampa laboratory would process ca 7800 serum samples annually for antibody for the listed counties if testing biweekly. Certainly Florida's plan is to let the chickens do most of the work!

The goal of Florida's surveillance program is not to detect the presence of WNV as apparently was one of the NYS goals. Florida is well past the need to detect WNV presence. Florida's goals are to get information to assess and differentiate human risk, identify high risk areas and predict a "big event" in advance. Florida recognizes that sporadic isolated human cases are a random chance event that is not entirely predictable in time or space. Florida's surveillance programs assess the transmission of many, many more mosquitoes. The sentinei chicken data assesses wide spread areas enhancing the ability to pick up focal outbreaks as occurred in Pinellas County in 2005. It provides direct assessment of transmission frequency, not "infection" rates. Certainly one might use mosquito infection surveillance to get focal data in those high risk areas identified by the sentinel chickens. Currently in Baton Rouge, Louisiana, where there is certainly high transmission, the mosquito pools are showing rates approaching 15/1000. These data are useful, but so would sentinel chicken surveillance be useful in Baton Rouge at less cost, provide wider coverage, and if done properly, sampling more mosquitoes.

Florida's choice here is clear. How much does one invest in the different types of WNV surveillance? What resources should be devoted to mosquito surveillance, sentinel chickens, wild avians, mammals, passive human? Is processing 35,000 mosquito pools using PCR a desirable useful goal as a widespread WNV sentinel surveillance program?

Lukacik, et al. (2006) have amply demonstrated the advantages and disadvantages of mosquito surveillance for WNV to assess human risk. Florida needs to stay focused on the goals of surveillance for WNV and use its available resources wisely. Rather than expand resources to increase widespread mosquito surveillance for WNV, Florida should evaluate the costs of a widespread, large WNV mosquito surveillance compared to Florida's sentinel chicken surveillance program. I suggest that Florida use its resources to expand the areas being covered and increase the timeliness of reporting sentinel chicken data.

Walter J. Tabachnick, Professor
Florida Medical Entomology Laboratory,
Department of Entomology and Nematology
University of Florida/IFAS