A Bayesian Detection Model for Chronic Disease Surveillance: Application to COPD Hospitalisation Data

Abstract

Background/Aim: In chronic disease surveillance it is crucial to be able to detect any temporal changes that might occur in particular areas, as this might be indicative of an emerged localised risk factor. In this work, we propose an extension of BaySTDetect, a Bayesian hierarchical model introduced by Li et al.(2012), which is able to detect areas with an unusual temporal trend, and a simulation study is carried out to assess the performance of the model. The method is illustrated by application to chronic obstructive pulmonary disease (COPD) hospitalisation data in England. Methods: We extend the BaySTDetect method to a more general framework which can provide information on both the area and the time point of the unusual observation, and can be appropriate for longer time periods. In addition, we modify the model so that it is more flexible to detect different patterns of unusual behaviour. The performance of the proposed model is investigated through a simulation study. Finally, the model is applied to a set of hospitalisation data on COPD in England at the Clinical Commissioning Group (CCG) level between April 2010 and March 2011. Results: Simulation results showed that the model performs well under three different time scenarios, giving much lower false positive proportions than BaySTDetect and adequate sensitivity and specificity values. Under the proposed model, thirty three areas were found to have unusual COPD hospitalisation trends in at least one month during the time period considered. Conclusions: We proposed a flexible approach to perform disease surveillance on small area data. The case study on COPD data in England showed that a number of areas were detected as unusual in terms of time pattern over the months April 2010 and March 2011. Further investigation is needed to explain this behaviour.