Modeling births at a tertiary health-care facility in Ghana: Box-Jenkins time series approach

Abstract

Background & Aim: Changes in the trend of births among women have been studied worldwide with indications of peaks and troughs over a specified period. Periodic variations in the number of births among women are unknown at the Korle-Bu Teaching Hospital (KBTH). This study sought to model and predicts monthly number of births at the Department of Obstetrics and Gynaecology (O&G), KBTH. Methods & Materials: Box-Jenkins time series model approach was applied to an 11-year data from the Department of (O&G), KBTH on the number of births from January, 2004 to December, 2014. Box-Jenkins approach was put forward as autoregressive integrated moving average (ARIMA) model. Several possible models were formulated, and the best model, which has the smallest Akaike information criterion corrected (AICc) was selected. The best model was then used for future predictions on the expected monthly number of births for the year 2015. Analysis was performed in R statistical software (version 3.0.3). Results: Seasonal ARIMA (2,1,1) × (1,0,1)12 was selected as the best model because it had the smallest AICc. Furthermore, the forecasted values showed that the expected number of births were lowest in January (750 births) and highest in May (970 births) for the year 2015. Conclusion: Seasonal ARIMA (2,1,1) × (1,0,1)12 was identified as the model that best describes monthly expected births and its use to forecast the expected number of births at the KBTH in Ghana will facilitate formulation of health policies and planning for safe maternal delivery and prudent use of hospital obstetric services and facilities.

Keywords

• Forecasting
• Seasons
• Birth
• Models