Background and Goal of Study: Based on historic data it is often difficult to determine a trend. Seasonal variations are implicitly well known, but difficult to model. Therefore seasonal variations can obscure a trend or show a trend although there is absolutely no trend.
Time series analysis are well established but until now not often used to supply decision finding.
Materials and Methods: Historical data of a Swiss general hospital were used. The hospital is a referral and teaching hospital. The number of surgical procedures in the central operating theatre was counted over more than six years. To correct the effect of public holidays the data was cleared manually
The second diagram shows the average of each month and the trend within the month.
Results and Discussion: Between 1.1.2000 and 31.10.2008 88'009 interventions were done in one of the eight central operating theatres.
The data ware aggregated per month and separated by the loess algorithm (STL: A Seasonal-Trend Decomposition Procedure Based on Loess).
The first diagram shows the time series, the seasonal decomposition, the trend and the error.
Conclusion(s): The decomposition by Loess algorithm allows us to detect trends which are occurred by seasonal variations. This information can be an advantage for strategic decisions. Operative decisions are supported by seasonal levels for each month and trends within. So the required resources can be determined by month and adjusted for these effects.