A joint use of monitoring and forecasting methods to detect change points in daily hospitalizations

Surveillance of hospitalization trends is a crucial task in health care and life sciences, especially during periods of a health emergency, such as those occurred in the recent pandemic. A good surveillance system supports decisions towards an optimal allocation of human, technical and economic resources and allows the development of better and innovative health policies. Under this framework, change points analysis represents a useful statistical tool to monitor the temporal trends in hospitalization, able to detect changes in the evolution of a phenomenon by estimating the corresponding time location. In particular, change point algorithms, such as control charts, can discern substantive causes of directional variation, from other causes of variation, i.e. random fluctuations around a baseline trend. The prediction of a trend is analogously a challenging objective, as it anticipates a change and, thus, further supports better decision-making processes towards early and more effective solutions. Statistical literature provides a large set of proper methods able to fulfil this aim. The monitoring methods and the forecasting techniques of time series are combined into a unique set of statistical tools proposed to detect change points in the trend of daily hospitalizations and applied to the forecast of COVID hospitalization in the Emilia Romagna region.