Cedars-Sinai Uses Machine Learning to Continuously Forecast Demand, Supply, and Outcomes

Michael Thompson, MS, Executive Director of Enterprise Data Intelligence, Cedars-Sinai. Over the past three years, Cedars-Sinai has merged together its analytics capabilities across finance, clinical informatics, and IT to form an enterprise data intelligence function, encompassing an enterprise data warehouse, data science team, end-user content delivery, and an advisory team to teach others how to use the data. The data analytics teams have designed an applied machine learning platform that has been used to forecast patient volumes, staffing and supply needs, patient outcomes, and financial modeling. Combining human-in-the-loop with reinforcement machine learning, the data science process is designed to adapt as the situation changes. During the COVID pandemic, the data science team at Cedars-Sinai used these methods to implement a machine learning model capable of predicting daily COVID patient volumes with an 85 to 95 percent accuracy. The models were used to meet daily bed, staff and PPE needs, by modeling the disease progression within the local community.