目的: 分析创新抗肿瘤药物使用与肿瘤类疾病诊断相关分组（DRG）费用结构、分组效能及支付偏差的关系,并比较血液系统肿瘤与实体瘤的差异。方法: 采用单中心回顾性真实世界研究设计,纳入A医院2024年1月至2025年12月54个肿瘤相关DRG组55425例住院病例。采用变异系数（CV）和总体方差减少系数（RIV）评价分组效能;以年龄、住院天数和是否手术为协变量实施1∶1倾向性评分匹配（PSM）,并开展肿瘤类型分层分析;采用多元线性回归、单药费用关联、DRG结余归因和回归决策树分析费用及支付偏差。结果: 原始样本增加创新抗肿瘤药物使用标识后,病例数加权平均CV由0.7719降至0.7619,RIV由0.5549升至0.5647。PSM后纳入10732例,全量样本RIV提高0.0513,血液系统肿瘤和实体瘤分别提高0.0746和0.0518。FM5B、FR1B、FW19、GJ1B和FR21等病组药物使用标识的RIV超过0.5。控制年龄、住院天数和是否手术后,创新抗肿瘤药物使用与住院总费用高约80.1%相关（P<0.001）。奥希替尼、氟马替尼和阿美替尼等药物用量增加与DRG结余下降相关;回归决策树识别出重组人血管内皮抑制素等药物的关键用量阈值。结论: 创新抗肿瘤药物使用能够补充解释肿瘤相关DRG组内费用差异,其分层价值具有肿瘤类型和病组异质性。应结合RIV、CV、病例规模、费用差距和临床可解释性筛选重点病组,衔接病组细化、过渡性支付支持、特例单议和权重动态调整。

Objective: The paper examines the association of innovative antineoplastic drug use with cost structure, grouping performance, and payment deviations in tumor-related diagnosis-related groups (DRG), and compares hematologic malignancies with solid tumors. Methods: This single-center retrospective real-world study included 55425 inpatient cases from 54 tumor-related DRG groups at Hospital A from January 2024 to December 2025. The coefficient of variation (CV) and reduction in variance (RIV) were used to assess grouping performance. A 1∶1 propensity score matching (PSM) model based on age, length of stay, and surgical status was applied, followed by tumor-type subgroup analysis. Multivariable linear regression, single drug cost association analysis, DRG balance attribution, and a regression decision tree were performed to analyze cost and payment deviations. Results: Adding an innovative antineoplastic drug-use indicator reduced the weighted mean CV from 0.7719 to 0.7619 and increased the RIV from 0.5549 to 0.5647 in the original sample. After matching, 10732 cases were included. The RIV increased by 0.0513 overall, 0.0746 for hematologic malignancies, and 0.0518 for solid tumors. The drug-use indicator yielded an RIV greater than 0.5 in several DRG groups, including FM5B, FR1B, FW19, GJ1B, and FR21. After controlling for age, length of hospital stay, and surgery status, innovative antineoplastic drug use was associated with approximately 80.1% higher total hospitalization costs (P<0.001). Greater use of osimertinib, flumatinib, almonertinib, and several other agents was associated with lower DRG balances. The regression tree identified key dosage thresholds for recombinant human endostatin and other targeted agents. Conclusion: Innovative antineoplastic drug use provides additional explanatory value for within-group cost differences, but its grouping value varies across tumor types and DRG groups. Priority groups should be selected by jointly considering RIV, CV, sample size, cost gaps, and clinical interpretability, with refined grouping, transitional payment support, special-case review, and dynamic weight adjustment used in coordination.