Forecasting Diagnostic Imaging Utilisation Rate for Effective Healthcare Delivery

Abstract

The variation of diagnostic imaging utilisation rate (DIUR) per 10,000 patients for 2008 to 2017 was studied using time series analysis. The DIUR monthly series is not stationary and it showed a slightly depreciated downward trend from January 2008 to December 2017. A first order difference was done to obtain stationarity. ARIMA model techniques were employed to identify a suitable model that will explain the variation in the series. ACF, PACF and their plots were determined. Akaike information criterion and Bayesian information criterion were used for identification and selection of the most suitable model. The suitable model identified is ARIMA(0, 0, 1) or MA(1) with no seasonal variation. This model was used to predict for two years and the forecast values show a constant trend with lower and upper limits of 4.62 and 86.17 respectively. We recommend this to healthcare facility managers as a planning guide for sustainable healthcare development.