Early season crow mortality as a sentinel for West Nile virus disease in humans, northeastern United States.

The 1999 New York epidemic of human West Nile virus (WN) encephalitis and meningitis was preceded by a crow die-off also caused by WN infection. As one component of the subsequently developed national surveillance system, crow mortality data were collected to detect WN activity before humans might become infected. However, predicting areas at risk for human WN disease likely requires assessment of multiple factors, including the intensity and timing of crow epizootics. To identify early season measures of WN activity in crows associated with subsequent WN disease in humans, county-level crow mortality data from seven northeastern states were analyzed. A predictive model was developed based on analysis of 2000 surveillance data and then assessed for 2001. To characterize the intensity of early season WN activity in crows, 15 variables were constructed from surveillance data of 52 counties that tested at least four crows during the early season (defined as June 17-July 28, 2000). County values for each variable were dichotomized at the 75th percentile into "high" and "low" activity. Multivariate analysis indicated that "high" early season activity of two variables-density of reported dead crow sightings (reported dead crows/area) and [(WN-infected crows/tested crows) x (human population)]--were associated with report of at least one human WN disease case (for each variable: adjusted odds ratio, 6.9; 95% confidence interval, 1.2-40.6). An assessment of this model using 2001 surveillance data from 61 counties yielded similar findings. With emphasis on early season WN activity, crow surveillance may allow timely targeting of interventions to protect the public health.