Segmented Harmonic Loss: Handling Class-Imbalanced Multi-Label Clinical Data for Medical Coding with Large Language Models

TL;DR

This paper introduces Segmented Harmonic Loss, a novel loss function designed to improve multi-label medical coding with large language models by addressing class imbalance and noise, leading to significant performance improvements.

Contribution

We propose Segmented Harmonic Loss and a segmentation algorithm to handle class imbalance and noise in medical coding datasets for LLMs, achieving notable accuracy gains.

Findings

Achieved over 10% F1 score improvement on noisy, long-tailed datasets.

Demonstrated effectiveness of the loss function on MIMIC III and IV datasets.

Showed LLMs outperform state-of-the-art methods with the proposed approach.

Abstract

The precipitous rise and adoption of Large Language Models (LLMs) have shattered expectations with the fastest adoption rate of any consumer-facing technology in history. Healthcare, a field that traditionally uses NLP techniques, was bound to be affected by this meteoric rise. In this paper, we gauge the extent of the impact by evaluating the performance of LLMs for the task of medical coding on real-life noisy data. We conducted several experiments on MIMIC III and IV datasets with encoder-based LLMs, such as BERT. Furthermore, we developed Segmented Harmonic Loss, a new loss function to address the extreme class imbalance that we found to prevail in most medical data in a multi-label scenario by segmenting and decoupling co-occurring classes of the dataset with a new segmentation algorithm. We also devised a technique based on embedding similarity to tackle noisy data. Our experimental results show that when trained with the proposed loss, the LLMs achieve significant performance gains even on noisy long-tailed datasets, outperforming the F1 score of the state-of-the-art by over ten percentage points.