Reply: Seasonal Variation in Emergency General Surgery
We thank Eisele et al for their interest and thoughtful letter regarding out article. In our study, we adjusted the data in a way that accounts for both annual population changes in the United States and also for general hospital admissions in NIS database, both of which had sinusoidal patterns. However, we had to report the raw data difference (22.1%) between the lowest (February) and highest (August) months to establish the fact that this difference is not due to overadjustment, that is, the mirror effect of decrease in the nonemergency general surgery (EGS) cases in August.1 As Eisele et al correctly mentioned, the number of days in February, which is 28 in each year except for 2004 and 2008, does have an effect on the raw data; however, our further analysis adjusts the data for both monthly admissions for all reasons in NIS database, and monthly US population which eliminates the effect of February on our final analysis and conclusion. Since the number of EGS admissions in February is adjusted for total NIS admissions, the proportion of EGS admission is decreased in February by 17.2% and the fewer days in month does not necessarily affect the analysis. Furthermore, the EGS admissions were adjusted for the estimated total US population in each month in a secondary analysis; thus the fewer days in February does not affect the secondary analysis either (a reduction by 21.5%). By taking into account the adjustments, there is no reason to believe that proportionally more EGS patients could have been admitted in the additional 2.5 days in February in consecutive years to flatten the difference. We believe the average daily admission rate is not required in this setting since the proportion of admissions in emergency general surgery is analyzed mainly and used for conclusions, and not the raw numbers. As a result, there is no particular reason to assess the average daily admission rate. Also, given the database limitations, it is impossible to perform such analysis as only month of admission is provided in this national database.2 In addition to the above mentioned analysis, we further analyzed the data based on US census regions, age, sex, race, and performance of a major operation. Seasonality exists in all subanalyses performed.
We read with interest the mentioned paper by Schuld et al, which concludes rupture events are more dependent on patient-related factors than to external circumstances such as meteorological, seasonal, or other reasons; however, a 4-day variance in barometric pressure is found to have a significant relationship with rupture rate of abdominal aortic aneurysm.3 To further improve the conclusions of this paper, we encourage the authors to perform rate adjustments for general hospital admission rate, population fluctuation, or geographical, racial, or age differences, as the discovered relationship could be merely a correlation effect due to an unmeasured covariate.