Natural Language Processing for Enhancing Anamnesis Documentation in Typhoid Fever Cases

Abstract

The implementation of Natural Language Processing (NLP) is crucial for enhancing the quality of medical records. This study aimed to develop an NLP model to improve the accuracy of documenting disease anamnesis for typhoid fever. The problem addressed by this research is the difficulty in analyzing and classifying patient complaints recorded in electronic medical records, which can affect the accuracy of diagnosis and treatment. The urgency of this study lies in ensuring that documented medical information is used accurately to support diagnosis and patient management. A quantitative approach was used, focusing on electronic medical records of patients who underwent anti-salmonella IgM tests in 2023, involving 424 individuals. The study assessed the performance of three models: Support Vector Machines (SVM), Naive Bayes Bernoulli, and Logistic Regression. The SVM model achieved the highest accuracy at 81.4%, compared to 76.7% for Naive Bayes Bernoulli and 79.1% for Logistic Regression. Additionally, four topic models were identified, highlighting common complaint words and their impacts. The most frequently occurring symptoms in the anamnesis of typhoid fever were "defecation," "nausea," "vomiting," "fever," "diarrhea," "heartburn," "weakness," "loss of appetite," "abdominal pain," "cough," and "cold." This study demonstrates that the SVM model provides superior accuracy in analyzing medical records compared to other models.