Simulation of Oncology Drugs Inventory Controls by Drug Inventory Department using Vendor Managed Inventory through Decision Trees Classification in National Cancer Institute of Thailand

Abstract

Background: Vendor Managed Inventory (VMI) is one of many methods used in inventory management. Its concept is to allow the vendor be able to access inventory data including product tracking in the customer’s inventory. It also lets the vendor manage customer’s inventory by filling products to customer’s inventory instead of using the usual purchase orders. Its goal is to reduce the potential of product reservation that might exceed the amount that is required. Today is the age of technology. Machine learning is a branch of artificial intelligence that is developed by learning patterns. It makes computer able to self-learn from data samples. It utilizes algorithm that creates model, predicts outcome, and helps decision making. It works out of order sequence unlike ordinary programs do, which is why it is widely used in several fields nowadays. It is a challenge to study and apply VMI system for hospitals. It will let the drug inventory department manage the medicine reservation to improve and optimize the process for the service department, and will make the medicine reservation process more efficient. Objective: To simulate cancer medicine reservation situation by drug inventory department using Vendor Managed Inventory (VMI) via decision trees classification comparing with the real situation in medicine subinventory in order to have enough quantity for services and reduce value of medicine reservation. Method: Since the characteristics of the demand for medicine usage were different, there could not be only one policy for medicine reservation. The study of oncology drug usage in National Cancer Institute has been using machine learning to analyze through decision trees classification, and simulate such situations and had come up with many beneficial results. Result: It reduced the average inventory value by 788,328.53 baht per day, which is 13.25%. It also reduces average days of stock from 18.04 days down to 15.84 days with 99.96% service rate. It even reduced the number of times for medicine reservation 72.73%. Conclusion: Vendor Managed Inventory (VMI) with an appropriate model from medicine demand usage classification using machine learning through information classification of a decision tree could reduce average inventory value, average days of stock, and quantity of medicine reservation by maintaining the same service rate. It had found the medicine reservation shortage 0.04%. The main reason for that was because the major changes in the characteristics of the medicine usage. In case of targeted therapy that its characteristic depended on the individual patient, using rules to classify the demand of medicine by number of patients and rate of medicine usage might not be appropriate. It might even results in medicine reservation inaccuracy from the actual patients.