Automated identification and predictive tools to help identify high-risk heart failure patients: pilot evaluation

Area of Science:

• Cardiology
• Medical Informatics
• Health Services Research

Background:

• Hospitalized heart failure (HF) management requires accurate patient identification and risk stratification.
• Manual review of patient data for HF identification is time-consuming and may miss critical cases.
• Predictive modeling can aid in assessing readmission and mortality risks for HF patients.

Purpose of the Study:

• To develop and evaluate an automated system for identifying hospitalized heart failure (HF) patients.
• To create a predictive risk report for HF patients, estimating 30-day readmission and mortality.
• To assess the impact of this automated system on patient care processes and outcomes.

Main Methods:

• Utilized natural language processing (NLP) to analyze dictated free-text reports for HF identification.
• Developed a predictive score to assess 30-day hospital readmission and mortality risk.
• Integrated NLP-identified HF patients and predictive scores into a comprehensive report.
• Evaluated the system in a hospital setting with high-risk HF patients.

Main Results:

• NLP integration significantly improved HF identification sensitivity (82.6% to 95.3%) and specificity (82.7% to 97.5%).
• The positive predictive value of the identification model reached 97.45%.
• Clinician review time for potential HF admissions decreased from 40 to 10 minutes.
• A significant reduction in 30-day mortality and an increase in home care discharges were observed.

Conclusions:

• Automated identification and risk reporting for HF patients using clinical decision support is effective.
• The system enhances HF patient identification and risk assessment accuracy.
• Implementation led to significant improvements in patient outcomes, including reduced mortality and increased home-based care.