“…Previous approaches to the winter surge in emergency demand in adult intensive care have involved using disease surveillance (Hiller et al, 2013;Moriña et al, 2011;Nguyen et al, 2016), and/or weather and seasonal information (Batal et al, 2001;Boyle et al, 2012;Diehl et al, 1981;Jones et al, 2002;Marcilio et al, 2013;Shiue et al, 2016), and/or previous demand (Abraham et al, 2009;Jones, 2007;Jones et al, 2008;Proudlove et al, 2003), using a range of techniques including 4 regression, stochastic Markov models and time series analysis methods such as Autoregressive Integrated Moving Average (ARIMA) models. In general, while emergency demand was universally found to be strongly seasonal and autoregressive it was also extremely variable, with its stochastic nature making accurate forecasts beyond seasonal or monthly means difficult.…”