FORECASTING OF FUTURE TUBERCULOSIS CASES USING AUTOREGRESSIVE MODELS

Abstract

This paper will provide a time-series analysis of Tuberculosis (TB) cases in Pakistan between 2002 and 2018. The main aim was to capture the trend of TB cases and predict the number of cases in the future using the autoregressive models. To capture the short-term, medium-term, and longer-term temporal dependencies, three models were developed; AR(1), AR(2) and AR(3). The lag variables were developed to make the past observations a predictor in the regression models. To estimate the model parameters, Ordinary Least Squares (OLS) was employed, and all the coefficients were estimated with the help of confidence intervals. To evaluate the performance of the model, good of fit measures such as, , Adjusted  and RMSE were computed. The AR(1) model was highly linear and related to the cases of the year before whereas AR(2) and AR(3) were more complex. The numerical findings showed that all the models were fitting the historical data very well with the values of the  above 0.95. Recursively calculated predictions of TB cases in 2019-2023 were obtained through the AR(1) model which gives 95 percent confidence intervals of the predictions. The findings show that there is a steady increase of TB cases during the forecast period. Comparison of AR models indicates that higher order models can be used to explain small fluctuations but not necessarily increase forecasting accuracy in the long run. The forecasts and methodology are useful to the health policy makers in Pakistan. This paper demonstrates the role of statistical modeling in disease dynamics and intervention planning strategy. New information can be used to revise the proposed models to make predictions more precise and assist in making evidence-based decisions. In general, this piece of work is relevant to predictive epidemiology and informs TB control and prevention strategies.