Predictive Modeling for West Nile Virus and Mosquito Surveillance in Lubbock, Texas.

West Nile virus (WNV) was first detected in North America during 1999, and has since spread throughout the contiguous USA. West Nile virus causes West Nile fever and the more severe West Nile neuroinvasive disease. As part of a WNV vector surveillance program, we collected mosquitoes in Lubbock, Texas, using CO2-baited encephalitic vector survey (EVS) traps. During 219 wk from 2009 through 2017, EVS traps were operated for 1,748 trap nights, resulting in more than 101,000 mosquitoes captured. Weekly, selected female mosquito specimens were pooled by species and trap site, and screened for WNV using reverse transcription-polymerase chain reaction assay. Mosquitoes positive for WNV were detected during 16.9% (37/219) of the weeks. Using this information, we constructed a statistical model to predict the probability of detecting an infection within a mosquito pool as a factor of weather variables. The final model indicated that detection of WNV in mosquitoes was negatively associated with the week of year squared and average wind from 3 wk prior to sampling, and was positively associated with week of year, average visibility, average humidity from 2 wk prior to sampling, and average dew point from 4 wk prior to sampling. The model developed in this study may aid public health and vector control programs in swift and effective decision making relative to city-wide mosquito control efforts by predicting when the chances of mosquitoes having WNV are at their greatest.

Estimating serum concentrations of dioxin-like compounds in the U.S. population effective 2005-2006 and 2007-2008: A multiple imputation and trending approach incorporating NHANES pooled sample data.

Dioxin-like compounds (DLCs) are monitored in the U.S. population using data collected with the National Health and Nutrition Examination Survey (NHANES). Until recently, participants' serum samples have been analyzed individually, and summary statistics defining reference ranges by age, gender, and race/ethnicity have served as the background by which other biomonitoring data can be evaluated. In the most recent NHANES DLC data, 2005-2006 and 2007-2008, participants' sera have been physically pooled prior to laboratory analysis, introducing major challenges to their utility as a reference population: variability among individuals and relations with covariates are lost, and individual design effects cannot be applied. Further, the substantial drop in limits of detection (LODs) in pooled sample biennials prevents reliable comparisons to individual data, and has complicated estimates of change over time. In this study, we address the drawbacks introduced by pooled samples by generating U.S. population reference ranges based on individual-level data adjusted to 2005-2006 and 2007-2008 levels. Using publicly available data, multiple imputation (MI) generated four NHANES biennials (2001-2008) of individual DLC data; we then trended the change over time in each DLC by demographic stratum. NHANES 2003-2004 individuals were adjusted by the trended change over time. Population estimates of toxic equivalency (TEQ) concentrations were calculated using traditional MI survey analysis methods and reference tables provided for 2005-2006 and 2007-2008 by age, race, and gender. Demographic differences in TEQ concentrations and trended change are reported, e.g. TEQ continues to drop in young adults aged 20-39, but distributions appear stable in older adults 60+; Mexican Americans have consistently lowest dioxins, furans, and PCBs, with non-Hispanic Blacks dropping to the same levels as non-Hispanic Whites in dioxins and PCBs and significantly below non-Hispanic Whites in furans by 2007-2008. Additionally, the ratio of 95th percentile to mean in DLC distributions was found to vary by age, between dioxins, furans, and PCBs, and across mean, making a simple ratio approach impractical for describing population concentrations using pooled samples. We discuss the practical implications of the pooled sample method, the performance of this trending solution in the context of other methods, and expected effects of distribution assumptions on variability and TEQ estimates, particularly in largely undetected congeners. These updated reference populations of individuals, along with information on trending, provide a common and valid basis for interpreting other individually sampled biomonitoring data.