Title: ‘Extending Scottish Exception Reporting Systems Spatially and Temporally’ Abstract: Exception reporting systems allow medical conditions and micro-organisms to be automatically monitored for unusually high levels. Typically, statistical models are used to predict expected levels. Where observed levels exceed the predicted ones by some pre-determined amount, an ‘exception’ is reported to give warning. We focus on developing suitable models for use in the predictive component of such systems. Two particular systems are extended spatially to monitor counts at the regional health board level. The first of these uses call data from the 24-hour medical helpline NHS24, to monitor particular medical syndromes. The second uses counts of positive lab identifications of micro-organisms collected by Health Protection Scotland (HPS). Regional incidences tend to have very small counts, and for these, we use negative binomial Generalized Linear Models (GLMs). However, GLMs assume that observations are independent, which is rarely, if ever, the case in the systems we consider. Two approaches are investigated for dealing with serial correlation and capturing local trend, both of which improve the models. We also investigate links between the health boards and investigate if these links can be used to further improve the models. A new system is produced for monitoring daily all-cause mortality in Scotland, using data collated by the General Register Office. Fitting models to this data is challenging because of the sharp peaks present in the annual seasonality; to address this, we use Generalized Additive Modelling. There is also a marked delay in the reporting of deaths, which must be dealt with if the system is to detect unusually high levels of mortality in a timely fashion. We present a straight forward ‘correction’ to do this. Combining these elements, a mortality surveillance system is produced, which has been used by HPS to monitor mortality during the swine flu pandemic (2009).