Correcting for day of the week and public holiday effects: improving a national daily syndromic surveillance service for detecting public health threats

BMC Public Health

Research Article

Elizabeth Buckingham-Jeffery¹  Gillian E. Smith²  Roger Morbey²  Sally Harcourt²  Alex J. Elliot²  Thomas House³ 

[1] Centre for Complexity Science and Warwick Infectious Disease Epidemiology Research Centre, University of Warwick, Coventry, UK;School of Mathematics, University of Manchester, Manchester, UK;Real-time Syndromic Surveillance Team, National Infection Service, Public Health England, Birmingham, UK;School of Mathematics, University of Manchester, Manchester, UK;

关键词: Syndromic surveillance;    Day-of-the-week effect;    Smoothing;   

DOI  :  10.1186/s12889-017-4372-y

 received in 2016-08-17, accepted in 2017-05-07,  发布年份 2017

来源: Springer

【 摘 要 】

BackgroundAs service provision and patient behaviour varies by day, healthcare data used for public health surveillance can exhibit large day of the week effects. These regular effects are further complicated by the impact of public holidays. Real-time syndromic surveillance requires the daily analysis of a range of healthcare data sources, including family doctor consultations (called general practitioners, or GPs, in the UK). Failure to adjust for such reporting biases during analysis of syndromic GP surveillance data could lead to misinterpretations including false alarms or delays in the detection of outbreaks.The simplest smoothing method to remove a day of the week effect from daily time series data is a 7-day moving average. Public Health England developed the working day moving average in an attempt also to remove public holiday effects from daily GP data. However, neither of these methods adequately account for the combination of day of the week and public holiday effects.MethodsThe extended working day moving average was developed. This is a further data-driven method for adding a smooth trend curve to a time series graph of daily healthcare data, that aims to take both public holiday and day of the week effects into account. It is based on the assumption that the number of people seeking healthcare services is a combination of illness levels/severity and the ability or desire of patients to seek healthcare each day. The extended working day moving average was compared to the seven-day and working day moving averages through application to data from two syndromic indicators from the GP in-hours syndromic surveillance system managed by Public Health England.ResultsThe extended working day moving average successfully smoothed the syndromic healthcare data by taking into account the combined day of the week and public holiday effects. In comparison, the seven-day and working day moving averages were unable to account for all these effects, which led to misleading smoothing curves.ConclusionsThe results from this study make it possible to identify trends and unusual activity in syndromic surveillance data from GP services in real-time independently of the effects caused by day of the week and public holidays, thereby improving the public health action resulting from the analysis of these data.

【 授权许可】

CC BY   
© The Author(s). 2017

【 图 表 】

12902_2016_104_Article_IEq1.gif

12864_2017_3670_Article_IEq16.gif

12864_2017_3670_Article_IEq15.gif