The early aberration reporting system (EARS) is a widely used surveillance tool that applies aberration detection algorithms to surveillance data and flags anomalies to help with the timely detection of disease outbreaks. ESR has adapted New Zealand's EARS from the original U.S. Centers for Disease Control and Prevention (CDC) version.
Aberration detection algorithms are used by the EARS system to flag events for follow-up. The methods can be applied to different data sets including surveillance data (e.g. notifiable diseases, sentinel influenza, chemical poisoning, vaccine safety) and syndromic data (emergency department data, absenteeism (workplace and school) data and over the counter drug sales).
The EARS program is applied to New Zealand EpiSurv notifiable disease data, and is updated weekly. Access is currently restricted to staff working in the New Zealand health sector as approved by the Ministry of Health. Email firstname.lastname@example.org to request access.
The analysis uses the notification date (data reported) in allocating week numbers to allow comparison between the data output.
An aberration is defined as a change in the distribution or frequency of health events when compared to historical data.
- This may or may not be an outbreak
- This may or may not be of public health interest
Principles and Practice of Public Health Surveillance (2002)
Not all flags will be of public health interest and the system relies on people who understand the data and system well to provide context to the data, e.g. the
aberration may be caused by a change in the laboratory testing method for an organism.
Flags can also be an isolated anomaly and as users become more familiar with their data, they are able to determine which flagging behaviour warrants an investigation.
Good communication between public health professionals is essential in determining whether the aberration requires public health response.
The EARS system has been redeveloped in R and now available in a dashboard. Red triangles are used to indicate anomalies.
EARS methods implemented for notifiable diseases
There are two types of methods implemented in the New Zealand EARS system, a spatial cluster & outlier method and a window method. The spatial method can be run on a single week of data and the window method can be run on six weeks of data.
Spatial cluster & outlier method
The spatial cluster & outlier method applies Local Moran's Index to identify spatial anomalies.
The disease rates in each DHB for a given week are clustered and compared with weights applied based on distance between their centre points.
Anselin, L.,1995, Local Indicators of Spatial Association-LISA. Geographical Analysis, 27, 93-115.(external link)
The window method model used is called 'rki'. The rki model compares the previous 6 weeks of values within a given DHB. An upper limit (threshold) is calculated using the non-outbreak case count from the previous 6 weeks and an alarm is raised if the actual value is bigger than the threshold) .If the mean of the previous 6 weeks is greater than 20 :
- An aberration is flagged if the current value is greater than two standard deviations from the mean of the previous 6 weeks.
If the mean of the previous 6 weeks is less than 20:
- An aberration is flagged if the current value is greater than the 95% confidence interval of the Poisson Distribution where mu is the mean of the previous 6 weeks, plus one.