Physician Characteristics Associated With Advance Care Planning After a Machine Learning-Based Nudge

Area of Science:

• Medical Informatics
• Clinical Decision Support
• Healthcare Delivery Research

Background:

• Prognostic machine learning models show promise in identifying patients requiring advance care planning (ACP).
• Physician comfort and prioritization of ACP conversations remain critical, even with predictive model assistance.
• Understanding physician-specific factors is essential for optimizing the integration of these tools into clinical practice.

Purpose of the Study:

• To explore the relationship between internal medicine physician characteristics and the likelihood of initiating advance care planning (ACP) conversations.
• To identify which physician traits, such as training background and practice patterns, influence ACP conversation rates after receiving a machine learning-generated mortality risk notification.
• To inform strategies for enhancing physician engagement with prognostic tools for advance care planning.

Main Methods:

• Secondary analysis of a cluster randomized trial data.
• Examination of internal medicine physician characteristics (training, practice patterns).
• Assessment of ACP conversation likelihood following notification by a mortality risk prediction machine learning model.

Main Results:

• Analysis focused on the association between physician attributes and ACP conversation initiation.
• Investigated how different physician backgrounds and practices correlate with acting on machine learning-generated risk alerts.
• Results aim to elucidate barriers or facilitators to ACP discussions driven by prognostic technology.

Conclusions:

• Physician characteristics significantly influence the adoption and prioritization of advance care planning (ACP) conversations prompted by machine learning tools.
• Identifying these factors can guide the development of more effective implementation strategies for prognostic models in clinical settings.
• Further research is needed to tailor interventions that support physicians in utilizing predictive analytics for timely ACP discussions.