Tailored Stream Differentiation: Customizing streams for each ICD code optimizes prediction accuracy.
•
Synergy of Text and Numerical Methods: Integrating text and numerical techniques enhances ICD code prediction.
•
Token Length Significance: Token length's significance is evident in ICD prediction using NLP algorithms.
Abstract
Assigning medical codes for patients is essential for healthcare organizations, not only for billing purposes but also for maintaining accurate records of patients' medical histories and analyzing the outputs of certain procedures. Due to the abundance of disease codes, it can be laborious and time-consuming for medical specialists to manually assign these codes to each procedure. To address this problem, we discuss the automatic prediction of ICD-9 codes, the most popular and widely accepted system of medical coding. We introduce a two-stream deep learning framework specifically designed to analyze multi-modal data. This framework is applied to the extensive and publicly available MIMIC-III dataset, enabling us to leverage both numerical and text-based data for improved ICD-9 code prediction.
Our system uses text representation models to understand the text-based medical records; the Gated Recurrent Unit (GRU) to model the numerical health records; and fuses these two streams to automatically predict the ICD-9 codes used in the intensive care unit. We discuss the preprocessing and classification methods and demonstrate that our proposed two-stream model outperforms other state-of-the-art studies in the literature.
6. Discussion
ICD code prediction is a complex task due to the vast number of classes, but it can greatly benefit doctors by automating the process. The advantages of automatic ICD code prediction include disease detection, providing suggestions to doctors when entering their own procedure's ICD codes, and giving reminders to healthcare professionals regarding which tests to administer based on tests given to similar patients. However, this task is challenging due to the extensive number of ICD classes, the vast array of tests conducted at hospitals with missing data where not all tests are available for all patients, and the possibility of multiple diseases within a single patient.
This study introduces a novel two-stream method for ICD-9 code prediction, but there are notable limitations: reliance on the specific MIMIC-III dataset may limit generalizability, the model's performance could vary in more complex medical scenarios, and the method demands substantial computational resources. While a significant advancement, these limitations emphasize the need for future research to enhance its applicability and accuracy in diverse healthcare settings.
7. Conclusion
In this study, we analyzed a substantial dataset of adults in an intensive care unit and proposed preprocessing methods, as well as natural language and machine learning models, to predict the ICD-9 codes of patients based on this data. The primary objective of this research is to enhance the performance of NLP methods using text-based data by incorporating numerical data in a two-stream network. To achieve this, we proposed both text and numerical-based methods and combined them to create a model that achieved the best results. For the two-stream models, a ratio of 0.75 with GRU and KEPT-based models yielded the best micro F1 score.
As a result, our study provides an effective and comprehensive approach for achieving improved results in the multi-class classification task of ICD code prediction. This research holds significant potential in assisting medical professionals with accurate ICD code predictions, ultimately leading to better patient care and outcomes; however, its potential should still be investigated across other ICU units and even other countries.
