Bridging the Version Gap: Multi-version Training Improves ICD Code Prediction, Especially for Rare Codes

TL;DR

This paper explores multi-version training for ICD code prediction, demonstrating that combining data from different ICD versions significantly improves accuracy, especially for rare codes.

Contribution

It introduces a multi-version training approach that enhances ICD code prediction accuracy across versions, particularly benefiting rare code classification.

Findings

Adding ICD-9 data improves rare code prediction by 27%.

Multi-version training enhances macro metrics for frequent codes.

Fewer model parameters are needed with multi-version training.

Abstract

Clinical coding maps clinical documentation to standardized medical codes, an essential yet time-consuming administrative task that could benefit from automation. Current models on ICD coding are typically optimized for codes from a specific ICD version. However, in reality, ICD systems evolve continuously, and different versions are adopted across time periods and regions. Moreover, ICD coding suffers from the long-tail problem, and rare code performance can be a bottleneck for developing implementable models. We examine whether it is viable to train version-independent models by combining data annotated in different ICD versions, which may help address these challenges. We add ICD-9 data to the training of a modified label-wise attention model for ICD-10 prediction, and find that despite the version mismatch, adding ICD-9 yields a 27% increase in micro F1 for 18K rare ICD codes compared to training on ICD-10 alone. On 8K frequent ICD-10 codes, the multi-version training also substantially improves macro metrics, with far fewer model parameters.