Session ID: 
141

Machine Learning to Estimate Acute Care Length of Stay

11:30am - 12:30pm Wednesday, March 11
Orlando - Orange County Convention Center
W230A

Description

Analysis of charge data from Ascension’s hospitals indicates that substantial opportunity exists to standardize clinical practice and length of stay (LOS) for our admitted patients. This session will discuss how Ascension data scientists and clinical analysts created and implemented a machine learning model leveraging internal administrative billing data to develop an “expected” LOS for each patient, thus better identifying areas of clinical opportunity than previous methods using the Center for Medicare and Medicaid Services (CMS) geometric mean LOS (GMLOS) for each diagnosis related group (DRG). Our results show that machine learning performs significantly better than the CMS GMLOS in terms of producing an efficacious “expected” length of stay. These results are used to provide a more reliable and targeted course for intervention while evaluating facilities, care groups, or provider groups for excess LOS days.Ni?o/a

Learning Objectives

  • Assess the current industry and internal benchmarks for LOS management, including the CMS GMLOS for each DRG
  • Evaluate the quality and consistency of administrative data for capturing clinical events that impact LOS
  • Develop a machine learning model to highlight LOS opportunities that is trusted by clinicians and useful within an organization new to data science solutions

Speaker(s)

Data Scientist,
Ascension
Manager, Data Analytics,
Ascension

Continuing Education Credits

ABPM
1.00
AHIMA
1.00
CAHIMS
1.00
CME
1.00
CNE
1.00
CPHIMS
1.00

Audience

Chief Quality, Chief Clin Transformation Officer
Data Scientist
Healthcare Financial Professionals

Level

Advanced