Forecasting the Incidence of Tuberculosis in Algeria

Abstract

Aims and objectives: Tuberculosis (TB) remains a health public issue in Algeria. In spite of the great efforts made for controlling the disease, the number of new reported cases exceeds the twenty thousand cases per year. One worthy of note epidemiological characteristics of TB is its seasonality. The seasonality of TB has been explored in some countries but no study has been conducted in Algeria. The aim of this study is to find out the seasonal pattern and the trend of TB incidence in Algeria. Methods: The monthly TB notification data from January 2008 to December 2017 extracted from the National Health Tuberculosis Register were examined to find out the seasonal pattern and trend of TB in Algeria. The seasonal autoregressive integrated moving average (SARIMA) model was used in analysing and predicting TB cases from 2008 to 2016. Cases registered in 2017 were used to assess the prediction accuracy of the selected models. The metrics mean square error (MSE) and mean absolute error (MAE) were applied to assess the better performance of prediction between the selected models. Results: From January 2008 to December 2017 there were 215,581 TB notified cases in Algeria with a yearly average 21,467 cases leading to an average annualized morbidity rate of 56,7 per 100,000 inhabitants. The monthly mean was 1796.5, the maximum number (2499) of reported cases was registered in April 2015 and the minimum (1202) in November 2009. The monthly TB data are normally distributed. The incidence of notified pulmonary TB cases decreased from 24 cases per 100,000 inhabitants in 2008 to 15 cases per 100,000 inhabitants in 2017; thus registering a drop of 37.6%, whereas the incidence of notified extrapulmonary TB cases increased from 27.5 cases per 100,000 inhabitants in 2008 to 35.8 cases per 100,000 inhabitants in 2017; thus registering an increase of 29.9%. The heatmap of TB incidence per quarter per province showed a significant temporal and geospatial variation. Time series analysis revealed seasonality with peaks in spring and summer and troughs in autumn and winter and showed that a (3,1,6)×(1,0,1)12 SARIMA model offered the best fit to the TB surveillance data for the period 2008-2016. This model was used to predict TB cases for the year 2017, and the fitted data showed considerable agreement with the actual data. Conclusions: Our findings are optimistic for forecasting TB by means of SARIMA models and afford useful information to public health policy makers in formulating targeted prevention and preparedness measures.