Forecasting healthcare cost in Australia using health insurance claims data

Private health insurance (PHI) companies have a growing claims dataset that could be used for cost forecasting. This study focused on addiction, mental health, obesity and musculoskeletal disorders disease groups, the costliest disease groups in Australia. PHI claims dataset has been employed to forecast cost using the Seasonal Autoregressive Integrated Moving Average (SARIMA) model. PHI company experts had been working along with the study to provide assistance in understanding data and relevant information from the PHI point of view. Ten years worth of data has been split into 90% training (2008 – 2016) and 10% testing (2017) datasets. SARIMA model parameters have been chosen while minimizing the error component. As per the outcomes, addiction, mental health, obesity, and musculoskeletal disorders cost forecasting have been done with an accuracy of 88.43%, 91.08%, 81.93% and 88.82%, respectively. As per the further improvements, exogenous variables not present in the used dataset could be tested with other forecasting models to improve forecasting accuracy.