Real-time outbreak and disease surveillance

Real-time outbreak and disease surveillance system (RODS) is a syndromic surveillance system developed by the University of Pittsburgh, Department of Biomedical Informatics.[1] It is "prototype developed at the University of Pittsburgh where real-time clinical data from emergency departments within a geographic region can be integrated to provide an instantaneous picture of symptom patterns and early detection of epidemic events."[2]

RODS uses a combination of various monitoring tools.[3]

1. The first tool is a moving average with a 120-day sliding phase-I-window.
2. The second tool is a nonstandard combination of CUSUM and EWMA, where an EWMA is used to predict next-day counts, and a CuSum monitors the residuals from these predictions.
3. The third monitoring tool in RODS is a recursive least squares (RLS) algorithm, which fits an autoregressive model to the counts and updates estimates continuously by minimizing prediction error. A Shewhart I-chart is then applied to the residuals, using a threshold of 4 standard deviations.
4. The fourth tool in RODS implements a wavelet approach, which decomposes the time series using Haar wavelets, and uses the lowest resolution to remove long-term trends from the raw series. The residuals are then monitored using an ordinary Shewhart I-chart with a threshold of 4 standard deviations.

References

1. RODS Laboratory Website
2. Public Health-Related Activities at the US HHS government website. Accessed December 2, 2010.
3. Shmueli, G., Burkom, H.S., "Statistical Challenges Facing Early Outbreak Detection in Biosurveillance", Technometrics (Special Issue on Anomaly Detection), vol. 52, issue 1, pp. 39–51, 2010