SHM Converge Scientific Abstract ePoster Gallery-Implementation of Machine Learning for High Mortality Risk Hospitalized Patients

Pdf Summary

This project aimed to improve goal-concordant care by increasing advanced care planning (ACP) documentation for hospitalized adults at high risk of dying within 30 days, while also reducing unnecessary ICU use. A key barrier to ACP had been the difficulty of identifying appropriate patients and limited clinician time.<br /><br />To address this, the team implemented a machine learning algorithm using electronic health record data to accurately predict 30-day mortality. When a patient was identified as high risk, hospitalists in both academic and community hospitals received an electronic medical record notification prompting them to initiate a goals-of-care conversation. Completing the ACP note remained voluntary.<br /><br />Preliminary results were encouraging. Across an average of 23,557 encounters per quarter, the proportion of inpatient admissions with at least one ACP note increased from 3.15% in the first quarter of 2022 to 6.8% in the third quarter of 2023. Among patients who received the intervention in the first three quarters of 2023, 8.65% were later admitted to the ICU, compared with 14.2% for the overall patient population at participating hospitals.<br /><br />Overall, the findings suggest that combining machine learning–based mortality prediction with provider notification can improve ACP documentation and may help reduce ICU utilization for patients nearing end of life. Further analysis is needed to evaluate system-wide adoption and the intervention’s impact on hospital departments.

Asset Subtitle

Michael Lin

Meta Tag

Author List Daniel Paget, Lucas Jorgensen, Matt Reuter, Michael Lin, Nathan Moore

Category Innovations

Concept Machine Learning Algorithm

Concept 30-Day Mortality

Concept Advance care planning

Concept Provider Notification

Concept Shared Electronic Medical Record Results Pool

Distinguished Non-Finalist

Presenter Organization Washington University School of Medicine St. Louis

Presenting Author Michael Lin

Track Technology in Hospital Medicine

Keywords

advanced care planning

ACP documentation

30-day mortality prediction

machine learning

electronic health record

goals-of-care conversation

hospitalized adults

ICU utilization

provider notification

end-of-life care

Machine Learning Algorithm

30-Day Mortality

Advance care planning

Provider Notification

Shared Electronic Medical Record Results Pool