Predicting Hospital Medicine Needs Based on a Multiple Linear Regression Model

Abstract

The study's issue is that hospitals are unable to accurately match the predicted demand and actual consumption for medications due to the use of traditional and ineffective forecasting techniques. The study's objective: 1)Finding a systematic approach based on modern scientific forecasting methods, based on multi-factormathematical models. 2) Selecting the optimal forecasting model among the methods utilized in this study topredict the future needs of hospitals for medicines in the coming years. 3) Calculating the discrepancy ratefor each drug annually to check the forecasting accuracy. Three statistical analyses were performed:correlation and regression, time series, and retrospective forecast between estimated demand and actualconsumption. Three accuracy indicators (MAPE, MAD, and MSD) were used as criteria for selecting theoptimal model that best describes the pattern of future demands for six pharmaceuticals and five factorsinfluencing drugs' consumption from 2015 to 2019. Three time series forecasting techniques (ALT, SES, andDES) were tested and compared with the MLR model to verify its forecasting accuracy. Time seriestechniques were compared to each other; DES was selected as the optimal technique among them. The resultsshowed that the mean MAPE for all medications by using MLR and DES models was 15.63 and 24.61,respectively. Therefore, we conclude that MLR is the optimal model for hospital inventory management andforecasting future needs since it has a lower relative error rate compared to DES. This indicates that theinfluence of independent factors on demand is stronger than the time factor. Therefore, MLR outperformedDES, which relies on the time factor. With the exception of Infliximab 100mg and Tocilizumab 20mg/ml,whose values exceed 25%, the average discrepancy rates between the estimated demand and actualconsumption of each medication over a 5-year period are statistically significant and within the acceptablebounds.