Randomized controlled trial of an automated problem list with improved sensitivity.

Area of Science:

• Clinical Informatics
• Natural Language Processing
• Electronic Health Records

Background:

• Electronic problem lists are crucial for patient care but often suffer from incompleteness and delays.
• Manual updating of problem lists is time-consuming and prone to errors.
• Automated methods are needed to improve the efficiency and accuracy of problem list management.

Purpose of the Study:

• To develop and evaluate a Natural Language Processing (NLP) system for automatically extracting medical problems from clinical text.
• To improve the completeness and timeliness of electronic problem lists.
• To propose extracted problems for inclusion in an electronic problem list management application.

Main Methods:

• A prospective randomized controlled trial of the Automatic Problem List (APL) system was conducted in intensive care and cardiovascular surgery units.
• 247 patients were enrolled, with 171 in the randomized controlled trial phase.
• The system's performance was evaluated based on sensitivity, specificity, predictive values, likelihood ratios, and timeliness of problem list updates.

Main Results:

• The APL system significantly increased problem list sensitivity in the intensive care unit from 9% to 41% (or 77% with unacknowledged suggestions).
• Timeliness of problem addition improved dramatically, reducing the delay from ~6 days to <2 days.
• No significant improvements were observed in the cardiovascular surgery unit.

Conclusions:

• Automated problem extraction using NLP can substantially enhance the quality of electronic problem lists, particularly in critical care settings.
• The APL system demonstrates potential for improving clinical data accuracy and patient care coordination.
• Further research may be needed to optimize the system for diverse clinical environments.