روندهای فصلی و الگوهای قابل توجیه در داده‌های کشوری نظام مراقبت بیماری سرخک: رویکردهای شناسایی و حذف

Background & Objectives: Knowledge of the presence of seasonal trends and other explainable patterns in the prediagnostic data sources and removing such patterns before applying outbreak detection methods seem very important. This study aimed to detect and remove the explainable patterns such as seasonality, day-of-week (DOW) and holiday effects of the daily counts of suspected cases of measles in Iran.Methods: Data on daily counts of suspected cases of measles as a pre-diagnostic data source were obtained from Iranian national surveillance system between 21 March 2008 and 20 March 2011. We used lines plot, moving average chart, autocorrelation and partial autocorrelation functions for detecting explainable patterns. Moving average (MA) and Holt- Winters (HW) exponential smoothing method are used for removing explainable patterns. Results: Our findings indicate the presence of seasonality, DOW effect, holidays and weekend effects in the daily counts of suspected cases of measles. The good performance of HW exponential smoothing technique in removing seasonal patterns is evident. MA technique showed better performance regarding assumption violation on outbreak detection methods. Conclusion: Because of the presence of explainable patterns in the daily counts of suspected cases of measles, considering such patterns before applying outbreak detection algorithms is very important. Implementing both MA (7 days) techniques for its simplicity as a pre- processing method and HW method for its efficacy in removing seasonal patterns is recommended.