ComplicaCode: Enhancing Disease Complication Detection in Electronic Health Records through ICD Path Generation
Xiaofan Zhou
TL;DR
This work tackles disease complication detection within EHR ICD coding by reframing coding as ICD path generation and introducing ComplicaCode. The framework uses a path generator G_θ and a path discriminator D_φ, augmented with a copy module to identify complicating diseases, and is trained via adversarial reinforcement learning with a rich EHR–ICD fusion mechanism. Empirical results on the MIMIC-III top-50 subset show that ComplicaCode achieves a high complication-detection rate (57.30%) and outperforms CNN baselines, while surpassing transformer models on complication detection; ablations confirm the critical role of the copy module. The approach demonstrates that integrating complication-aware generation and adversarial feedback yields meaningful gains in multi-label ICD coding, with practical impact for improved coding accuracy and downstream epidemiology, billing, and clinical research workflows.
Abstract
The target of Electronic Health Record (EHR) coding is to find the diagnostic codes according to the EHRs. In previous research, researchers have preferred to do multi-classification on the EHR coding task; most of them encode the EHR first and then process it to get the probability of each code based on the EHR representation. However, the question of complicating diseases is neglected among all these methods. In this paper, we propose a novel EHR coding framework, which is the first attempt at detecting complicating diseases, called ComplicaCode. This method refers to the idea of adversarial learning; a Path Generator and a Path Discriminator are designed to more efficiently finish the task of EHR coding. We propose a copy module to detect complicating diseases; by the proposed copy module and the adversarial learning strategy, we identify complicating diseases efficiently. Extensive experiments show that our method achieves a 57.30\% ratio of complicating diseases in predictions, and achieves the state-of-the-art performance among cnn-based baselines, it also surpasses transformer methods in the complication detection task, demonstrating the effectiveness of our proposed model. According to the ablation study, the proposed copy mechanism plays a crucial role in detecting complicating diseases.