Weather and Influenza and Pneumonia Mortality: Does the Impact of Cold, Dry Air Vary between Climates?

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 11:45 AM
228AB (Phoenix Convention Center - West and North Buildings)
Robert E. Davis, Univ. of Virginia, Charlottesville, VA; and E. M. Dougherty and R. Kolkmeyer

Biometeorological research over the past decade has posited a potential relationship between unusually dry and/or cold air and increased human mortality from influenza (and the pneumonia that often results). Theories that support this relationship include increased viral stability, host susceptibility, and human behavioral factors, among others. Most research indicates a lag of several weeks between the occurrence of anomalously cold, dry periods and the onset of high mortality. Less clear is how these relationships vary spatially between cities that have markedly different climates. In this research, we examine the strength of the dry air-influenza linkage between cities with differing climates.

Daily pneumonia and influenza (P&I) mortality counts were acquired for five large U.S. metropolitan areas (Chicago, IL; Detroit, MI; Philadelphia, PA; Houston, TX; Phoenix, AZ) as well as for Auckland, New Zealand. Data for the U.S. cities were acquired for 1987–2000 from the National Mortality, Morbidity, and Air Pollution Study archives. The Auckland data were acquired from 1979–2009 from New Zealand government sources. Hourly weather data were gathered from nearby weather stations and included temperature and dew point temperature observations. Mortality data were adjusted for possible biases associated with coding errors linked to transitions between coding protocols that occurred in 1998 by converting the mortality time series to z-scores.

Because of inherent variations in human responses to a (putative) environmental factor and the latency between infection and mortality, the data must be smoothed to accentuate the relationships between variables. After testing for a variety of smoothers and lags, we employed between 17 and 21-day leading moving-average smoothers, with the lag and smoother length allowed to vary by city. The weather data were similarly smoothed using a 3 to 5-day moving average filter. Mortality data were characterized by single day “events” and more prolonged “episodes” in which the z-score exceeded a threshold that varied by location based upon that city's mortality distribution. Through this method, the strong seasonality inherent in P&I mortality is removed and the resulting analysis emphasizes relationships to weather specifically. T-tests and linear regression were used to compare mortality rates to weather observations.

In Auckland, both low temperature and dew point temperatures were associated with increased mortality events and episodes at a statistically significant level (alpha<0.05). Similar results were found for both morning and afternoon temperature and dew point temperature. The optimum lag tended to vary temporally but results were fairly consistent using a 19-day latency period. No significant relationships were found between the severity of the cold and dry air period and total episode mortality. For U.S. cities, a significant relationship was found for all cities based on 17–21 day smoothers and lags of 9–12 days. Low maximum temperatures exhibited slightly strong relationships to P&I mortality than mean dew point temperatures, but both were statistically significant.

These results suggest a surprisingly consistent relationship between weather and P&I mortality in cities located in different climatic zones. Anomalously cold and dry air tends to precede peak mortality events by 2–3 weeks, but the specific temperature and humidity conditions that serve as the purported trigger vary significantly between locations. The lack of a consistent relationship between the intensity of the cold and/or dry air and mortality suggests that other factors are also involved in accounting for the substantial inter- and intra-annual P&I mortality variability that is observed. In ongoing research, we are currently exploring differences in influenza virus types and subtypes in relation to weather.