Comparison of Linear Trend Analysis and Double Exponential Smoothing Methods for Predicting Chronic Disease Drug Needs

Abstract

One of the factors influencing the standard of care given to patients and society is the process of identifying medical needs and accurately quantifying them. Therefore, finding new methods or mechanisms to evaluate the need for drugs has become essential. Due to faulty prediction methods, the Ministry of Health in Iraq was unable to achieve a precise match between actual consumption and estimated demand due to the employment of old methods. Therefore, the research aims to 1) select a suitable method from the two forecasting techniques, trend exponential analysis (TEA) and double exponential smoothing (DES), and 2) predict the needs for drugs for the next five years based on historical data. Accuracy measures (MAPE, MAD, and MSD) have been utilized as model selection criteria to best characterize the trend of estimating the demands of chronic drugs from 2017 to 2019. The results demonstrate a decrease in MAPE values in all medications compared with MAD and MSD. In addition, the MAPE of a DES technique for (Lovastatin = 3, Insulin = 12, Valproic acid = 4.36, Carbidopa = 1.74) is greater than the MAPE of a TEA (Lovastatin = 2, Insulin = 11, Valproic acid = 3.92, Carbidopa = 1.72). Depending on the value of MAPE, TEA was recognized as the optimal prediction model and can best fit for predicting the demand for chronic drugs in future needs. Annual forecasts using a trend exponential analysis for the quantities required for each drug from 2020 to 2024 show a significant increase in demand for all pharmaceuticals. Whereas insulin tablets are the smallest, with 559,367 packs in 2024, valproic acid is the largest, with 13,387,773 packs, indicating a shortage of competing drugs or an anticipated increase in patient numbers. The proposed method might be useful for inventory prediction at industrial sites as well .