A Pilot Report on Extracting Symptom Onset Date and Time from Clinical Notes in Patients Presenting with Chest Pain

Area of Science:

• Medical Informatics
• Natural Language Processing
• Cardiology

Background:

• Acute coronary syndrome (ACS) is a rapidly evolving heart condition.
• Prognosis in no-ST-segment elevation ACS (NSTE-ACS) can be informed by symptom onset timing.
• Accurate extraction of symptom onset from clinical notes is challenging.

Purpose of the Study:

• To pilot-test natural language processing (NLP) pipelines for extracting patient-reported symptom onset date and time (DateTime).
• To evaluate the performance of 'parsedatetime' and 'regular expression' (regex) NLP tools on clinical notes.

Main Methods:

• Two NLP pipelines, parsedatetime and regex, were tested on 71 clinical notes (History and Physical, Emergency Department Screening, Triage Notes).
• Extracted DateTime information for chest pain symptoms was compared against manually annotated true onset times.
• Performance was assessed based on the number of successful DateTime extractions.

Main Results:

• Parsedatetime successfully extracted DateTime information from 36 out of 71 notes (50.7%).
• The regex pipeline yielded zero matched DateTime outputs.
• Parsedatetime demonstrated superior performance compared to regex, though results were suboptimal.

Conclusions:

• Off-the-shelf NLP pipelines like parsedatetime show potential but require significant customization for accurate DateTime extraction from EHRs.
• Further research into automated methods for large-scale DateTime extraction from clinical notes is necessary for improving NSTE-ACS patient management.