No Association between Foehn and Daily Number of Emergency Calls or Calls to Medical Call Centers in Bavaria (Germany) – A Longitudinal Data Analysis from 2006 to 2009

Introduction

Foehn is a warm, dry, downslope wind descending the lee side of the Alps as a result of synoptic-scale, cross-barrier flow over the mountain range. The south foehn blows from northern Italy, where the air is warm, to the north of the Alps where the air is cooler. Usually foehn occurs only in the southern administrative districts of Bavaria (about up to Munich) while in the northern districts of Bavaria foehn occurs very rarely. A number of studies documented an association between foehn winds and different diseases. Some other studies could not verify such an association. The aim of our study was to find out whether foehn has an influence on the daily number of all emergency calls or calls to medical call centers in southern Bavaria during the period from January 2006 to December 2009. The reasons for the different kind of calls were not considered.

Data and Methods

To analyze the influence of foehn on the daily number of emergency calls and calls to medical call centers between January 2006 and December 2009 we used the data of the emergency calls to the twenty six rescue coordination centers in Bavaria as well as the data pool of all calls to medical call centers of the Association of Statutory Health Insurance Physicians in Bavaria. Information about foehn was provided by the Meteorological Observatory in Hohenpeißenberg (Germany Weather Service). The German Weather Service offered information on the type of foehn (foehn gaps, foehn in the upper atmosphere, foehn at ground level) and its intensity (light, moderate, unknown).

Expecting a large influence of several anthropogenic factors, different administrative variables like day of the week, public and school holidays, bridge days, begin/end of quarter and season were included in the analyses.

At first, we divided fifty six of the ninety six administrative districts of Bavaria into a northern and a southern part. For these regions we calculated the daily number of emergency calls and calls to medical call centers.

After a descriptive analysis, generalized linear models (GLMs), general additive models (GAM) and general additive mixed models (GAMM) were used to model the number of emergency calls and the number of calls to medical call centers. To account for possible over- or underdispersion Quasi-Poisson models were used. To consider the longitudinal structure of the data and the resulting problem of autocorrelation specific nonlinear time trends were integrated into the GAMs and GAMMs as well as auto regressive models of order 1 (AR1 processes) into the GAMMs. A p-value of lower than 0.05 was determined as statistically significant.

To verify the influence of foehn on the daily number of calls to medical and emergency call centers in Bavaria, Germany, we analyzed three different subdivided data sets:

• By sub set A, which includes only days with foehn, we compared the number of calls in the northern and southern regions of Bavaria.

• Sub set B includes only the southern region of Bavaria and was taken for a comparison between days with and without foehn.

• By our third approach, based on sub set C with data only from the southern region of Bavaria, we compared the differences between days with and without foehn at ground level (strongest kind of foehn).

The analyses were performed in R (version 2.8.0).

Results

Altogether, in northern and southern Bavaria 6.34 million calls (62.16 %) out of 10.21 million calls in the period under consideration were analyzed. Of those, 2.81 million (44.25 %) were emergency calls and 3.45 (55.75 %) were calls to medical call centers. For both types of calls there was a significant influence of the day of the week, public and school holidays, bridge days as well as the season independent of the model used (GLM, GAM, GAMM).

For no subdivided data set (A, B, C) the modelling of an AR1 process led to a better adaptation to the structure of the data and the remaining partial autocorrelation could not be reduced. For all subdivided data sets (A, B, C) the GAM had the lowest partial autocorrelation, was significantly better than the GLM and had a lower adjusted R² than the GAMM. Therefore, the presented results are based on the GAM.

For both data sets it could be shown that there is a statistically significant effect of the geographical region (sub set A) which is not attributed to foehn, because both days with foehn and without foehn had a statistically significant influence on the daily number of calls to emergency call centers and medical call centers. It could be assumed that other effects than foehn are responsible for these results.

The analyses of days with and without foehn in southern Bavaria (sub set B) and days with foehn at ground level in southern Bavaria (sub set C) did neither show a statistically significant result for emergency calls nor for calls to medical call centers (p-value for both data sets and both sub sets greater than 0.8).

Conclusion

The results indicate that foehn did not have an influence on the daily number of calls to medical or emergency call centers in general in Bavaria, Germany, in 2006 to 2009. These results are not in accordance to other studies which showed a correlation between foehn and different medical data sets. It might be that we would also get a relation if we specified the reasons for each call. The main disadvantage of both data set is, that we only have a first guess of the real diagnosis and not a final diagnosis of a physician. Therefore, further studies with more detailed information are necessary.