Evaluation of natural language processing from emergency department computerized medical records for intra-hospital syndromic surveillance
2011

Evaluating a System for Detecting Epidemic Risks in Hospitals

Sample size: 100 publication 10 minutes Evidence: moderate

Author Information

Author(s): Gerbier Solweig, Yarovaya Olga, Gicquel Quentin, Millet Anne-Laure, Smaldore Véronique, Pagliaroli Véronique, Darmoni Stefan, Metzger Marie-Hélène

Primary Institution: Hospices Civils de Lyon

Hypothesis

Can a natural language processing system effectively extract and encode medical concepts from emergency department records for syndromic surveillance?

Conclusion

The study shows that an automated method for extracting and encoding medical concepts from emergency department reports is feasible and useful for identifying potentially infectious patients.

Supporting Evidence

  • The system achieved an overall recall of 85.8% for processing medical concepts.
  • Precision was measured at 79.1%, indicating a good level of accuracy in identifying relevant concepts.
  • The study evaluated 1,674 medical concepts from 100 emergency department records.

Takeaway

This study created a computer program that helps hospitals find patients who might spread infections by reading their medical notes.

Methodology

The study used a natural language processing application called UrgIndex to analyze narrative reports from emergency department medical records.

Potential Biases

Potential for false positives due to misinterpretation of medical terms and abbreviations.

Limitations

The system may struggle with unrecognized terms and does not contextualize medical concepts based on their occurrence timeline.

Participant Demographics

Adult patients admitted to the emergency department of a university hospital.

Statistical Information

P-Value

0.0001

Confidence Interval

95% CI: 84.1-87.3

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1186/1472-6947-11-50

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication