arXiv:2611.21891v1 Announce Type: new Abstract: Automatic assignment of ICD-10 codes from short clinical diagnosis text is a practical medical text classification and normalization task. The task is difficult because diagnosis phrases are short and noisy, the ICD-10 label space is hierarchical, and the class distribution is strongly imbalanced. This paper presents a call-efficient ICD-10 code prediction pipeline that combines transformer embeddings, tree-based supervised candidate generation, candidate-level reranking, and a gated LLM assistance. The evaluation uses a leakage-safe validation protocol: model and threshold selection are based on out-of-fold cross-validation predictions, while final reporting is performed on a locked test set with no shared source-file groups. The supervised conditional parent-code candidate generator reaches 74.32% Top-1 accuracy on the locked test set and 93.05% Top-10 candidate coverage. A direct zero-shot GPT-5.6-Sol baseline reaches 81.01% Top-1 accuracy, but requires an invocation of this large model for every diagnosis. The final deployable system, Pre-LLM fusion, reaches 81.24% OOF-CV accuracy and 83.19% locked-test accuracy while calling the LLM for only 31.97% of cases. This improves the locked-test error rate from 25.68% for the supervised parent-code baseline to 16.81%, corresponding to a 34.5% relative error reduction. The results show that the most effective role of the LLM is controlled reranking on uncertain examples rather than unrestricted coding or full-dataset invocation.
Call-Efficient ICD-10 Code Prediction via Supervised Candidate Generation and Selective LLM Invocation
A call-efficient ICD-10 code prediction pipeline that combines transformer embeddings, tree-based supervised candidate generation, candidate-level reranking and gated LLM assistance reached 83.19% Top-1 accuracy on a locked test set while invoking the LLM for only 31.97% of cases, according to the arXiv paper 2611.21891v1. The supervised conditional parent-code candidate generator alone reached 74.32% Top-1 accuracy and 93.05% Top-10 candidate coverage, while a direct zero-shot GPT-5.6-Sol baseline reached 81.01% Top-1 accuracy but required a large-model call for every diagnosis. The Pre-LLM fusion system cut the locked-test error rate from 25.68% to 16.81%, a 34.5% relative reduction, indicating the LLM's most effective role is controlled reranking on uncertain examples rather than unrestricted coding.
Route only uncertain diagnoses to the LLM for reranking instead of calling it on every case; the gated pipeline matched full-invocation accuracy while invoking the model for 31.97% of cases.
more in playbook: Controlling LLM costRun your AI side-project on zahid.host
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.