A two-stream deep model for automated ICD-9 code prediction in an intensive care unit.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Medicine

Background:

• Accurate patient medical coding is crucial for healthcare operations, including billing, record-keeping, and procedure analysis.
• Manual assignment of International Classification of Diseases, Ninth Revision (ICD-9) codes is labor-intensive due to the vast number of codes.

Purpose of the Study:

• To develop and evaluate an automated system for predicting ICD-9 codes.
• To leverage multi-modal data (numerical and text-based) for enhanced prediction accuracy.

Main Methods:

• A two-stream deep learning framework was designed to process multi-modal patient data.
• Text representation models and Gated Recurrent Unit (GRU) were employed for text and numerical data, respectively.
• The framework was applied to the MIMIC-III dataset for intensive care unit (ICU) code prediction.

Main Results:

• The proposed two-stream model demonstrated superior performance compared to existing state-of-the-art methods.
• The integration of multi-modal data significantly improved ICD-9 code prediction accuracy.

Conclusions:

• The developed deep learning framework offers an effective solution for automated ICD-9 code prediction.
• This approach can reduce the manual burden on medical specialists and improve the efficiency of healthcare data management.