Leading Autonomous Pharmacy Company
Predictive Model to Reduce the Number of Drug Stock-Outs
Since 1992, the customer has been inspired to create safer and more efficient ways to manage medications and supplies across all care settings. The company is revolutionizing the patient medication experience from hospital to home by empowering providers to keep each patient at the center of care. The company’s autonomous approach to medication and supply management leverages a differentiated platform for hardware and workflow software solutions, real-time predictive intelligence, and performance-driven partnerships to help drive operational, financial, and clinical success for customers.
The customer engaged with the SpringML team for a proof of concept to build a model using Google Cloud Platform that predicts and ultimately helps in reducing the number of stock-outs of a medication that occurs within each medicine dispensing cabinet in a hospital. Data from around 100 complete cabinets from one hospital system over the past five years was made available for analysis, model development, training, testing, and validation.
Google Cloud Implementation
The stockout proof-of-concept model is a forecast system to predict which drug from a cabinet will be out-of-stock in the next four hours. The model can be categorized as a forecasting system as it looks at the dispensing rhythm of a specific drug by analyzing its history (thus, the model requires a minimum of 50 data points per drug/comp/code, and this is based on our agreed 3-months minimum required historical data). The implemented model is a linear regression that can handle the requirement of predicting stock outs 4 hours in advance and triggerable at any given point in time.