Embed Wisely: An Ensemble approach to predict ICD Coding

Pavithra Rajendran, Alexandros Zenonos, Joshua Spear, Rebecca Pope

Abstract:

International Classification of Diseases (ICD) are the de facto codes
used globally for clinical coding. These codes enable healthcare providers to
claim reimbursement and facilitate efficient storage and retrieval of diagnostic

information. The problem of automatically assigning ICD codes has been previ-
ously approached as a multilabel classification problem, using neural models and

unstructured data. We utilise an approach for efficiently combining multiple sets

of pretrained word embeddings to enhance the performance on ICD code pre-
diction. Using post-processing and meta-embeddings techniques, we exploit the

geometric properties of word embeddings and combine different sets of word em-
beddings into a common dimensional space. We empirically show that infusing

information from biomedical articles, whilst preserving the local neighbourhood

of the embedding, improves the current state-of-the-art deep learning architec-
tures. Furthermore, we demonstrate the efficacy of this approach for a multimodal

setting, using unstructured and structured information.