Natural Language Processing and Machine Learning Classification Model for Injury Mechanism in Trauma.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Machine Learning

Background:

• Trauma research databases require significant resources to create and maintain.
• Natural language processing (NLP) offers a potential solution for extracting structured data from unstructured clinical text.
• Automating data extraction can streamline the development of trauma research databases.

Purpose of the Study:

• To develop and evaluate a machine learning algorithm using NLP for classifying traumatic injury mechanisms.
• To determine the accuracy and efficiency of the NLP algorithm in distinguishing between penetrating and non-penetrating injuries.
• To assess the potential of NLP as a tool to supplement the work of trauma registry abstractors.

Main Methods:

• Clinical notes from trauma patients were collected and split into training, tuning, and testing sets.
• A bag-of-words model with term frequency-inverse document frequency weighting was used, followed by singular value decomposition for dimensionality reduction.
• Random forest, support vector machine, and logistic regression models were trained and evaluated using area under the receiver operating characteristic curve and accuracy metrics.

Main Results:

• The NLP models achieved high performance, with area under the receiver operating characteristic curve values of 0.99-1.00 and accuracy of 0.98 across models.
• The algorithm was trained on 4944 unique terms, identifying 30 distinct injury topics.
• Training and classification were highly efficient, with run times of 234 seconds and 0.15 seconds, respectively.

Conclusions:

• A straightforward NLP machine learning algorithm demonstrated excellent accuracy and discrimination for classifying injury mechanisms in trauma patients.
• NLP presents a scalable and accurate method for automating clinical note review in trauma research.
• This technology can effectively supplement human abstractors in building and maintaining trauma databases.