BID Brown Bag Seminar: Analysing weekly attendances to major Accident and Emergency departments in England with panel- and time-series methods
|Speaker:||Katharina Hauck, Imperial College London|
|Date: ||Tuesday 7 March 2017|
|Time: ||12.30 - 13.30|
|Location: ||Forum Seminar Room 3|
There is conflicting evidence on the determinants of attendances to Accident & Emergency (A&E) departments. Such evidence could help A&E providers devise strategies to cope with increasing or fluctuating attendances, or devise interventions that moderate demand for emergency services. The objective of this study is to shed light on the determinants of weekly attendances and their geographical and seasonal patterns. It uses both panel and time-series methods and data on providers and their catchment areas merged from a variety of sources for 135 English A&E providers over 156 weeks between April 2012 and March 2015. The forecasting ability of the two modelling approaches are compared. Both approaches show that attendances are crucially influenced by unobservable characteristics of the providers or the area they are located in, so-called common trends. Only the share of the working population, weather and socioeconomic deprivation are statistically significant in the panel models. Reorganizations between major A&E departments and minor injury units within one provider lead to structural breaks in attendances. Modelling of unobservable factors with Common Trend models show that weekly attendances are not a random walk, and there are patterns related to season and location of providers in coastal versus urban areas. These might be related to temporary population movements. We conclude that policy interventions need to recognize unique differences between providers. It is unlikely that there is a policy intervention that works for all. The use of time series analysis to model attendances and fine tune funding allocations across trusts and seasons should be explored further.