Quality Improvement Study Using a Machine Learning Mortality Risk Prediction Model Notification System on Advance...

Area of Science:

• Medical Informatics
• Healthcare Quality Improvement
• Clinical Decision Support

Background:

• Advance care planning (ACP) is crucial but underutilized in patient care.
• Machine learning (ML) offers potential for identifying patients needing ACP.
• Current methods for identifying patients for ACP are suboptimal.

Purpose of the Study:

• To evaluate the impact of ML-driven provider notifications on ACP documentation rates.
• To assess the effect of ML-identified patient alerts on ACP completion.
• To determine if ML-based interventions improve patient outcomes related to ACP.

Main Methods:

• A pre-post quality improvement (QI) study was conducted at a tertiary academic hospital.
• Adult general medicine patients identified at high risk of mortality by an ML model were included.
• The intervention involved email and pager notifications to providers for identified patients.

Main Results:

• Covariate-adjusted ACP documentation rose from 6.0% to 56.5% post-intervention (RR=9.42).
• Patients with ACP were more likely to have reduced code status (RR=2.69) and increased hospice referral (OR=2.16).
• A longer mean length of stay (LOS) was observed in patients with documented ACP (1.29 event time ratio).

Conclusions:

• Provider notifications utilizing ML models effectively increase ACP documentation in inpatient settings.
• ML-driven alerts can enhance clinician engagement with advance care planning.
• This approach shows promise for improving end-of-life care discussions and documentation.