A surveillance system based on a short memory scheme

A surveillance system is proposed to detect an increase in the mean of a Poisson distribution of cases of a disease. This system, called short memory (SM), is based on conditional binomial tests which are performed sequentially at fixed time intervals. The probability of rejection at each test defines the run length distribution which has a geometric tail. A standard SM scheme outperforms other SM schemes. The CUSUM outperforms the SM schemes when the baseline mean is specified correctly. This type of misspecification does not affect the SM scheme.