Hospital Length of Stay Prediction System
Hospital Length of Stay Prediction System

OVERVIEW

  • Hawaiian Telcom is the primary provider of comprehensive communications and entertainment services, solutions, and products in Hawaii. With its headquarters based in Honolulu, Hawaiian Telcom is the prominent integrated communications provider for both residential and business customers in the region.
  • Hawaiian Telecom partnered with Microsoft – OCP and PeopleTech to develop machine learning use cases in the healthcare industry. One of the key projects involved predicting Hospital Length of Stay (LOS) to define the number of days from initial admission to discharge in any hospital facility.
  • The company was facing issues such as.
  • Inaccurate discharge planning: Hospitals faced challenges in accurately planning discharges, leading to issues with quality measures such as readmissions.
  • The need to predict the length of stay in hospitals to facilitate effective resource planning and patient care.

SOLUTION PROVIDED BY PEOPLE TECH GROUP

  • The machine learning model was deployed in a hybrid architecture, utilizing both Azure Cloud and Azure Stack.
  • Missing values in continuous laboratory measurements were imputed to ensure data completeness.
  • Continuous laboratory measurements were standardized to maintain consistency in the data.
  • The best model was selected using favorite bagging and boosting techniques, with gradient boosting outperforming Random Forests in accurately predicting the length of stay.

BENEFITS

  • The implementation of the machine learning solution increased workload efficiency from approximately 65% to 76%, enabling hospitals to better manage resources and patient care.
  • Accurate prediction of the length of stay allowed hospitals to plan discharges more effectively, leading to improved quality measures such as reduced readmissions.