Using intervention time series analyses to assess the effects of imperfectly identifiable natural events: a general method and example

This paper describes how to apply ITSA to analyse the impact of unplanned events on time series when the timing of the event is not accurately known, and so the problems of ITSA methods are magnified by uncertainty in the point of onset of the unplanned intervention.The methods are illustrated using the example of the Australian Heroin Shortage of 2001, which provided an opportunity to study the health and social consequences of an abrupt change in heroin availability in an environment of widespread harm reduction measures.Application of these methods enables valuable insights about the consequences of unplanned and poorly identified interventions while minimising the risk of spurious results.There are occasions when an unexpected event produces major changes in a process that provide an opportunity to investigate its nature and dynamics. These unplanned interventions can provide unique insights into environmental, sociological or physical phenomena that would be otherwise impossible to obtain, such as, for example, in the distribution of radioactivity after the Chernobyl disaster and its effect on the human food chain and economic processes [1,2]. Such natural experiments provide important research opportunities for public health, enabling investigations of phenomena which would not otherwise be practical or possible [3]. In such circumstances the method often involves the analysis of the impact of the event on the best available regularly collected surveillance data on outcomes of interest using intervention time series analysis (ITSA).ITSA has become a standard statistical method for assessing the impact of an intervention (usually a planned policy change) on a time series of relevant outcome indicators. In this paper we describe how ITSA can be adapted to model the impact of an unexpected event whose point of onset cannot be accurately identified. We illustrate the method in the context in which it was developed – modeling the health and social consequences of a