Emergency Medical Text Processor


1. What is EMT-P?
Emergency Medical Text Processor (EMT-P) is a natural language processing (NLP) system for cleaning emergency department (ED) chief complaint (CC) data. EMT-P version 1 (v.1) was initially developed as a semi-automated prototype of Debbie Travers' doctoral dissertation research in 2002-03. In 2004-05, EMT-P version 2 (v.2) was automated and released as open source software.

2. Who uses EMT-P?
The system is intended for use by researchers, data managers and others who have a need to process ED chief complaint data for secondary uses such as clinical research, public health surveillance, syndromic surveillance and administrative tracking. All users must sign an EMT-P license and register as users via the EMT-P website.

3. What kind of technical knowledge is necessary to install EMT-P?
Although installing EMT-P is straightforward, some experience installing and using software will be helpful.

4. Can you use EMT-P on other types of data beside ED chief complaints?
EMT-P was developed through a sublanguage analysis of ED chief complaint data. While theoretically it could be used on other types of ED text, it has not been tested for other purposes. Text passages (e.g., clinical notes) differ from the brief, fielded input found in CC entries so EMT-P would likely need adaptation to perform well on text passages.

5. Can you use EMT-P with Linux/Unix?
EMT-P has been tested only in Windows environments. Users are encouraged to try EMT-P in other environments but need to have versions of Perl, MySQL and Java that run in Linux/Unix.

6. Is EMT-P available for processing CC data in languages other than English?
No, EMT-P was designed to process English terms only.

7. How do you use the output of EMT-P?
Use of the EMT-P output will depend on the user's needs. The output includes one or more cleaned CC segments for each original CC entry, a matching Unified Medical Language System (UMLS) preferred term (if the cleaned CC segment matched the UMLS) and a Concept Unique Identifier (CUI). Potential uses for the output include counting the UMLS concepts for the local site(s), listing the most frequent CC segments for the local site(s), and using the cleaned CC segments for syndromic surveillance.

8. What if I only want to use some EMT-P modules, e.g. unnec_abbrev?
By default, EMT-P uses all modules in the cleaning process. Excluding some of the modules easily is beyond the scope of the current version, but this can be made possible by modifying and recompiling RunEMTP.java.