Introduction ICD编码作为WHO制定的疾病分类金标准，其手动编码过程耗时且易受主观性影响。 例如，美国每年因编码错误导致的成本高达250亿美元。 深度学习在药物推荐、放射报告生成等EHR任务中的成功，为自动化编码提供了可能。

针对电子病历ICD编码存在的语义模糊、结构特征缺失等问题，研究人员提出融合选区树 (Constituency Tree)和Tree-LSTM的TRIC模型，通过bioBERT预训练增强关键特征匹配，在MIMIC-III数据集上实现MiF 0.586、P@8 0.758的突破性性能，为医疗大数据标准化提供创新解决方案。

Conclusions: ICD-9 to ICD-10 translations do not always align, producing discontinuity over time. This may compromise ICD-based measurements and risk-adjustment. To address the challenge, we propose a ...

Congress should recognize the enormous impact the ICD-10 would have on the medical profession and stop or delay its implementation.

Injury codes in ICD-10 require a seventh character to indicate the number of the encounter that the patient is being seen for that injury (eg, initial encounter vs subsequent or follow-up encounter).

Having an ICD-10 code to document the proper diagnosis ensures that patients are categorized and treated accurately and helps us know what to expect for any given patient.

ICD-9 coding Use of specific diagnoses improves the quality of claims and is just as important as CPT coding.

Doctors are getting a two-year reprieve before they'll have to use the coding system known as ICD-10.The Centers for Medicare and Medicaid Services, which maintains the U.S. codes for billing and ...