The Relationship of WN Dead Birds and Human Cases
The significance of dead wild birds to detect the presence of WN virus and in predicting the risk for human cases has been discussed in previous Buzzwords columns (i.e., Oct. 2000, Jan. 2001, Nov. 2001). A recent article "Guptill et al. 2003. Early-Season Avian Deaths from West Nile Virus as Warnings of Human Infection, Emerg. Infect. Dis. 9: 483-484" can be found at http://www.cdc.gov/ncidod/eid/vol9no4/02-0421.htm, deserves discussion.
The authors report that 1719 U.S. counties reported WN-positive dead birds during 2002. Of these, 632 counties reported at least one WN infected bird before August 4. The use of August 4 is explained in the article. 45% of these later reported a human WN case. In comparison, only 19% of the 1,162 counties not reporting a WN-positive bird by August 4 had a WN human case. The two groups add up to 1794, an error in the paper. However, counties with a WN dead bird before August 4 were more likely to report a human WN case. They write that work is needed to assess sampling differences due to effort, i.e., the effort is greater in more populous areas and so is the chance of a human case. However, they do not address this critical point in their paper. The authors hypothesize that where an avian epizootic occurs early in the transmission season, subsequent WN in humans is more likely because the early epizootic may indicate viral activity that has sufficient time to escalate to high levels before the end of the transmission season.
We recognize that a WN positive wild bird may provide early evidence for the presence of WN virus. However, will this event predict human cases? How accurate is the prediction? Let's evaluate Guptill's data for 2002.
First, consider the imprecision of the risks. 55% of counties with an early indicator bird had no human cases. 20% of counties with no early indicator had a human case. Second, a dead WNpositive bird does not distinguish a random human case from epidemic transmission. Likewise, it does not increase our ability to target epidemic intervention or vector control in a specific region. We need better predictors to achieve efficient, effective, and proper vector control during medical emergencies.
Does the predictive value of a WN-positive bird found before August 4 make sense? Although plausible, remember the old song, "It Ain't Necessarily So." Do the two groups really represent different levels of transmission before August 4? Perhaps, but how can this date be of value for the myriad of ecologies and habitats throughout North America? An equally likely hypothesis is there are differences in sampling effort, and ease of observation, e.g. a dead bird in an urban shopping center stands out better then a dead bird in a coastal marsh. The mean human population in counties with a WN bird before August 4 is 30% higher then those with no reports, counties with WN animals before August 4 have ca. 2X's the human population of those with no WN animals. All of the counties with WN human cases were in the top 60% of counties ranked by population size. One cannot discount human population size, effort, ease, and chance.
Look at Florida's data in 2002 at the FL Dept. of Health site at: http://www9.myflorida.com/disease_ctrl/epi/htopics/arbo/data/2002/12-16-02.pdf Florida had 10 counties reporting WN-positive wild birds prior to July 26, 2002 with 4 later having a human case--a 40% risk. But wait--of the 23 counties with later WN positive wild birds there were 5 with human cases--a 22% risk. Surprisingly similar to the previous percentages, but in Florida there is no difference (X2=1.05, P>0.05). The prediction is not improved with other indicators, i. e., dead birds, chickens and/or horses. Of 24 counties with a WN indicator prior to July 26, 9 later had human cases--a 38% risk. Of 30 Florida counties with a WN positive animal after July 26, 5 had human cases - a 20% risk. Again no difference (X2=0.47, P>0.05) in Florida. Do your best and play with the numbers and the cutoff dates. We can all play the numbers game and develop some nifty stories. The risk of a human case in a county with a WN positive animal reported by or before a date is April 5--29%, May 5--22%, June 14--18%, July 19--33%, Aug. 9--33%, Aug. 30--30%, Sept 30--29%. The WN-positive animal detects the presence of virus prior to the detection of human cases, but it is not predictive of the case. The fallacy is trying to make connections to predict sporadic human cases, which is very difficult. So far Florida has escaped a major WN virus epidemic. This will likely change, and our forecasting abilities will be tested.
Can the predictive precision be improved by taking sampling effort and ease into account? I predict that this will only weaken the predictability of the 2002 human cases. It will be very difficult because sampling effort and ease will differ in dynamic ways over time and space. What information does a dead WN-positive wild bird found prior to August 4 really provide and is it a basis for initiating vector control actions? For now it is clear that such a finding before August 4 is not a good thing. A sentinel chicken seroconversion in April is not a good thing. A WN-positive horse in April is not a good thing. The presence of WN virus is not a good thing--and there will be sporadic isolated human cases. How much action is necessary and is it cost effective? My friend Alan Curtis quipped "and if the WN dead bird falls in the forest and no one sees it, is it really a WN transmission event?" Florida has low-level transmission now, sporadically in a few counties. Expanded sentinel chicken surveillance will provide more precise information for public health and mosquito control. We need to effectively monitor virus transmission to sentinels, viral amplification in wild birds, and mosquito infection rates. We need to be able to quickly evaluate changes in mosquito transmission frequencies in time and space. We can obtain the data to provide accurate risk assessment for the "big event." We can do better risk assessment from the combination of surveillance tools based upon a sentinel surveillance program that is capable of adapting and changing to meet local conditions and needs. This will not be easy. Dead bird reporting has limited use but is relatively inexpensive to obtain. Perhaps this is the reason for the continued efforts to refine and promote this tool. It ain't necessarily so.
Walter J. Tabachnick, Director
Florida Medical Entomology Laboratory,
University of Florida/IFAS
- Back to Buzz Words Archive