Studying Associations Between Heart Failure Self-Management and Rehospitalizations Using Natural Language Processing

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Natural Language Processing

Background:

• Effective self-management is crucial for heart failure (HF) patients to prevent hospital readmissions.
• Identifying patients with ineffective self-management from clinical notes is challenging but important for targeted interventions.

Purpose of the Study:

• To develop and validate a natural language processing (NLP) algorithm for automated identification of ineffective HF self-management from discharge summaries.
• To analyze the association between identified ineffective self-management and 30-day hospital readmissions.

Main Methods:

• Development of an NLP algorithm to extract self-management status (diet, physical activity, medication, appointments) from narrative discharge summaries.
• Validation of the NLP system using a test set of 300 notes annotated by human reviewers.
• Analysis of 8,901 HF patient records to determine the prevalence of ineffective self-management and its association with readmissions using adjusted regression models.

Main Results:

• The NLP system achieved high accuracy in identifying ineffective self-management (F-measure = 86.3%, precision = 95%, recall = 79.2%).
• 14.4% of 8,901 HF patients had documented ineffective self-management.
• Ineffective self-management, particularly skill-related deficits and non-specific deficits, was significantly associated with increased 30-day hospital readmissions (ORs ranging from 1.3 to 1.5).

Conclusions:

• Automated identification of ineffective heart failure self-management from electronic discharge summaries using NLP is feasible.
• Ineffective self-management is a significant, identifiable risk factor for preventable 30-day hospital readmissions in HF patients.
• NLP tools can support clinical decision-making by flagging at-risk patients for proactive interventions.