Weather, Climate, and Respiratory Health in the Shenandoah Valley, Virginia

Chronic cases of respiratory distress arise when individuals are exposed to environmental conditions that trigger an event. These environmental triggers can be from indoor or outdoor exposures, the latter of which include factors related to weather variations. Our goal is to examine a suite of weather and climate variables to determine what relationships, if any, exist between chronic respiratory distress and the atmospheric environment.

The response variable is total daily hospital admissions from 2001–2006 for individuals who reside in the Shenandoah Valley, Virginia. A listing of respiratory conditions likely to be related to external environmental conditions were identified by a respiratory disease specialist. The resulting time series exhibits significant daily and intra-annual variation, with higher admission rates in the winter and lower rates in the summer. As our goal is to determine departures from the normal, seasonal cycle, we used a mean-based LOWESS filter to remove the seasonality, and we standardized the data to remove a seasonal variance bias related to a poisson-like process.

The admissions data were related to surface weather observations and day-to-day changes in these variables at Martinsburg, West Virginia and Roanoke, Virginia, the Spatial Synoptic Climatology (SSC) daily air mass types, HYSPLIT 72-hour back-trajectories clustered into natural groupings using a two-stage hierarchical/non-hierarchical cluster analysis, ambient ozone concentrations (hourly and daily maximum) for a variety of stations in and around the Shenandoah Valley, and daily aeroallergen counts (tree, grass, and ragweed pollen, and mold spores) from two different stations in the general vicinity.

We identified significant relationships between several atmospheric pollutants and air mass types and back trajectories and the within-synoptic-type trajectory clusters that result in the highest pollutant concentrations in the Shenandoah Valley. In general, the worst pollution episodes occur under air masses associated with high pressure when back-trajectories extend over the Ohio River valley. Further, we find significant relationships between pollutant concentrations and standard meteorological variables that correspond to previous research. Preliminary results show that pollen concentrations are highly autocorrelated, and that weather data provides little or no additional predictive information on pollen counts.

Preliminarily, we find a small but statistically significant portion of the daily variance in hospital admissions can be accounted for by a combined time series model and weather/pollution factors. SSC air mass type proves to a be useful discriminator in some cases, but sample size considerations suggest that the results must be viewed with caution. Current and ongoing work will involve evaluating our model in a predictive mode to determine the efficacy of an respiratory alert system for the region.