Abstract：Objective To explore the application of long short term memory neural networks (LSTM) in the demand management of
medical consumables, to analyze and predict the usage demand of medical routine consumables in the future, and to realize the fine
management of the demand of medical consumables. Methods The LSTM model was used to analyze the amount of intravenous
indwelling needles received by the depot of a tertiary hospital in Wuxi from 2015 to 2021, and to predict the future quarterly or
even half-yearly receipt. Results Based on the comparison of the prediction results between 2019 and 2021 and the analysis of
each evaluation index of the model, it was found that the minimum average absolute percentage error was 2.27%, and the maximum
value was 4.54% in the six months of 2021. The average absolute percentage error of all the predictions was less than 5%, and the
prediction accuracy was limited when affected by COVID-19. Conclusion The LSTM neural network model can predict the demand
of medical consumables in hospitals more accurately, and can be used as reference data for the inventory base and procurement
strategy development of medical consumables.
杨燕，钱正瑛，庄希，金伟. 一种基于LSTM模型的医用耗材需求量预测方法[J]. 中国医疗设备, 2022, 37(6): 123-126.
YANG Yan, QIAN Zhengying, ZHUANG Xi, JIN Wei. LSTM Model-Based Method for Forecasting the Demand of Medical Consumables. China Medical Devices, 2022, 37(6): 123-126.
