NegHunter

Negation Detection in Medical Documents Using Syntactical Methods

Team 

Stefan Gindl
Katharina Kaiser, Vienna University of Technology, Institute of Software Technology and Interactive System, Vienna, Austria
Silvia Miksch, Vienna University of Technology, Institute of Software Technology and Interactive System, Vienna, Austria
 

Project 

Motivation:
Medical information is often stored in a narrative way, which makes the automated processing a difficult and time-consuming task. Persons responsible for the authoring of medical documents do not take care of a further processing with automated systems. So, information stored in medical writings is not directly usable for the processing with computers. Due to this, efforts have been made to transfer these narrative documents in a format easier processable with computers.

This matter of fact also applies to clinical practice guidelines (CPGs). As many medical documents, CPGs are written in a narrative speech as well, without regards to a computer-assisted processing. For the implementation of CPGs in medical facilities an automated processing is therefore desirable. An important fact is that a lot of information in CPGs is provided in a negated form, expressing that certain circumstances in patients or treatments are not available, existing or applicable. Although negated, this information is nevertheless very useful, since it can express the absence of certain conditions or diseases in patients. Moreover, negations can describe which treatment options should not be taken into account for a given patient, helping a practising physician or nurse in his/her decision process for the assortment of a proper treatment. Thus, a proper Negation Detection in CPGs is an important task for the automated processing of this type of medical documents. It helps to accelerate the decision making process and can support medical staff in their care for patients.

We developed algorithms capable of Negation Detection in CPGs. We use syntactical methods provided by the English language to achieve a precise detection of occuring negations. According to our results we are convinced that the involvement of syntactical methods can improve Negation Detection, not only in medical writings but also in arbitrary narrative texts.


 

Methods:
We analyzed four clinical guidelines regarding their negations and categorized the found negations in five classes:

  1. Adverbial Negation:

    It is triggered by adverbs "not" and "never". The negated concepts appear in combination with a verb. Via the tense of the verb we decide, whether the sentence is written in active or passive voice. For example, the combination of "be-verb + adverb phrase + verb phrase" represents the simple present tense and the simple past tense in passive voice (e.g., "is not performed", "was not performed"). The simple present tense and the simple past in active voice are combined by "verb phrase + adverb phrase + verb phrase" (e.g., "do not recommend", "did not recommend"). With this information, we interpret the three preceding or succeeding noun phrases as negated. We use the number of three noun phrases as we receive the best results with it.
     

    Guideline developers do not recommend chemotherapy. (active voice)
     
    Chemotherapy is not recommended. (passive voice)
     
  2. Intra-Phrase Triggered Negation:

    These are negations in which the trigger is included in the noun phrase. "No" and "without" act as triggers.
     

    Evidence obtained from at least one well-designed study without randomization.
     
  3. Prepositional Negation:

    Triggers are followed by prepositional phrases, often introduced by the prepositions "of" or "from". The rule for this negation type is "(noun phrase or adjective phrase) + prepositional phrase".
     

    Patients with good performance status, ... , and the absence of systemic disease.
     
  4. Adjective Negation:

    It uses adjectives as negation triggers. We have identified, for example, the term "ineffective" and can interpret the first noun phrase before the trigger as the relevant phrase. The pattern of this negation is "noun phrase + [not noun phrases] + be verb + adjective phrase". The noun phrase is the negated phrase and it can be spatially separated from the trigger by several phrases, which are not noun phrases themselves.
     

    Recommendation indicates at least fair evidence that the service is ineffective or that harm outweighs benefit.
     
  5. Verb Negation:

    Some verbs are also negation triggers themselves. We identified the verbs "deny", "decline", and "lack". According to the verb's suffix we determine if it expresses active or passive voice.
     

    Information on final patient outcomes was also lacking.
     

 

Evaluation & Results:
We manually generated a gold standard from 17 clinical guidelines. These 17 guideline documents were also processed by a prototypical implementation of our algorithms. Its output was then compared to the gold standard. We received the following results:
 

Negation classRecallPrecision
Adverbial Negation78.81 %60.32 %
Intra-Phrase Triggered Negation   91.48 %97.94 %
Prepositional Negation94.74 %78.26 %
Adjective Negation83.33 %100 %
Verb Negation88.89 %55.17 %
83.92 %   69.36 %

 

Using NegHunter can support an automated structuring of the information in order to, for instance, decide which therapies or drug regimens are best applied in patients with certain diseases and which are not recommended. This helps to sort out treatment options and supports the medical personnel as well as patients in their decision-making.

Additionally, NegHunter's negation classification allows users to augment the trigger set by themselves. Therefore, new triggers need to be assigned a negation class. NegHunter applies its rule base to these new triggers. This makes NegHunter portable to be applied on other document types as well as extensible and maintainable.
 

Publications  S. Gindl. Negation Detection in Medical Documents Using Syntactical Methods Master's Thesis, Vienna University of Technology, Institute of Software Technology and Interactive Systems, Vienna, March 2008.
 
S. Gindl, K. Kaiser, and S. Miksch. Syntactical negation detection in clinical practice guidelines In Proc. of the Conference on Medical Informatics Europe (MIE'08), Göteborg, Sweden, 187-192, IOS Press, 2008.
 
S. Gindl. Negation Detection in Automated Medical Applications Vienna University of Technology, Institute of Software Technology and Interactive Systems, Vienna, Technical Report, Asgaard-TR-2006-1, p. 28, 2006.
 
Downloads  Master Thesis Poster (PDF, 356KB)
 
Related Work  W. W. Chapman, W. Bridewell, P. Hanbury, G. F. Cooper, and B. G. Buchanan. A simple algorithm for identifying negated findings and diseases in discharge summaries Journal of Biomedical Informatics, 34(5):301-310, 2001.
 
P. G. Mutalik, A. Deshpande, and P. M. Nadkarni. Use of general-purpose negation detection to augment concept indexing of medical documents: A quantitative study using the UMLS. J Am Med Inform Assoc, 8(6):598-609, 2001.
 
Y. Huang and H. J. Lowe. A novel hybrid approach to automated negation detection in clinical radiology reports J Am Med Inform Assoc, 14:304-311, 2007.
 
Related Projects 

EviX - Facilitating Evidence-based Decision Support Using Information Extraction and Clinical Guidelines
 

Funding 

Parts of this project were supported by "Fonds zur Förderung der wissenschaftlichen Forschung FWF" (Austrian Science Fund), grants P15467-INF and L290-N04.