Time Series Analysis of Maternal Mortality in Ghana. A Case Study of the Korle-Bu Teaching Hospital, Accra (2001 - 2013)

Abstract

This study examined the pattern of maternal mortality ratios as well as a spectral analysis of maternal mortality at the Korle - Bu Teaching Hospital in Accra from 2001 to 2013. It also fitted a stochastic model to forecast maternal mortality ratios for four quarters. Analyses were based on data available at the Bio-Statistics Department of the Obstetrics & Gynaecology directorate of the Korle - Bu Teaching Hospital in Accra for the period 2001 – 2013. The R-Consol statistical analysis software as well as the Statistical Analysis Software (SAS) was used in analysing the data. It was observed that the average Maternal MortalityRatio (MMR) in Korle-Bu Teaching Hospital over the Thirteen (13) year period was 835 per 100,000 live births, which is over twice that of the national’s MMR of 380 per 100,000 live births. The time series plot of the monthly maternal mortality ratio data clearly shows volatility clustering in the data with constant mean and stable variance and hence considered stationary. An ARMA model was selected as the appropriate model for predicting future maternal mortality ratios for the hospital. The model satisfied all conditions of a good ARMA model and was used to predict Maternal Mortality Ratios (MMRs) for the next four quarters. The power spectrum of the series was obtained using the periodogram plot of the series and suggests smoothing. We conclude statistically that the maternal mortality ratio data has a platykurtic distribution, an ARMA model is adequate for forecasting Maternal Mortality Ratios (MMRs) and the power spectrum of maternal deaths has smoothing.