目的 利用层次聚类将保乳手术患者分组，结合评价指标平均使用率确定各组治疗质量高低，分析治疗质量的影响因素，为提高保乳手术患者治疗质量提供依据。方法 选择乳腺癌保乳手术治疗质量评价指标，收集保乳手术患者病历数据；利用Jaccard相似系数对患者进行聚类分组，结合分母权重法计算平均使用率确定高低质量组；应用多因素logistic回归分析保乳手术治疗质量的影响因素。结果 收集保乳手术患者279人。层次聚类将患者分为2组：组1患者108人，占比38.7%，评价指标平均使用率68.4%，为低质量组；组2患者171人，占比61.3%，评价指标平均使用率89.1%，为高质量组。2组间指标使用率差异较大的指标是保乳手术后进行放射治疗、病理报告记录手术切缘情况、术前细胞学或组织学诊断（P<0.001），高低质量组的差值分别为86.46%、30.09%和20.74%。多因素logistic回归分析显示：居住在城市（OR=2.461，95%CI:1.259~4.812）、高收入（OR=2.580, 95%CI:1.520~4.380）、有合并症（OR=1.990, 95%CI:1.104~3.587）、肿瘤病理分期为Ⅱ期（OR=2.511, 95%CI:1.474~4.278）、在专科医院治疗（OR =8.407，95%CI:4.738~14.917）的患者往往有更高的治疗质量（P<0.05）。结论 层次聚类结合评价指标平均使用率可以很好地区分患者治疗质量的高低；提高保乳手术后进行放射治疗、病理报告记录手术切缘情况、术前细胞学或组织学诊断的使用率可能会更好地改善治疗质量。医生应在临床诊疗中更关注居住在农村、低收入、无合并症、肿瘤病理分期较高患者的治疗质量。

Objective To cluster the patients undergoing breast conserving surgery by hierarchical clustering, determine the treatment quality of each group combined with the average utilization rate of evaluating indicators, analyze the influence factors for the treatment quality, and provide theoretical guidance for improving the treatment quality for breast conserving surgery patients. Methods We selected quality evaluation indicators of breast conserving surgery, and collected medical records of breast conserving surgery patients; the patients were clustered by using the Jaccard similarity coefficient, and the average utilization rates were calculated to determine the level of the treatment quality based on denominator-based weight method; we also analyzed the influence factors of the treatment quality using multiple logistic regression. Result Two hundred and seventy-nine patients were selected and divided into two groups based on hierarchical clustering: group 1 was the low-quality group with 108 patients, accounting for 38.7%, and the average utilization rate of evaluation indicators was 68.4%; group 2 was the high-quality group with 171 patients, accounting for 61.3%, and the average utilization rate of evaluation indicators was 89.1%. The indicators with significant difference in the utilization rate between the two groups were radiotherapy after breast-conserving surgery, pathological report recording surgical margins, and preoperative cytological or histological diagnosis (P<0.001), and the difference was 86.46%, 30.09% and 20.74%. Multivariate logistic regression analysis showed that: living in a city (OR=2.461, 95%CI: 1.259-4.812), high income (OR=2.580, 95%CI: 1.520-4.380), complications (OR=1.990, 95%CI: 1.104-3.587), pathological stage II (OR=2.511, 95%CI: 1.474-4.278), and patients treated in specialized hospitals (OR=8.407, 95%CI: 4.738-14.917) tended to have higher quality of treatment (P<0.05). Conclusion Hierarchical clustering combined with average utilization rate of evaluation indicators could distinguish patients' treatment quality well. The more use of radiotherapy after breast-conserving surgery, pathological report to record surgical margins and preoperative cytological or histological diagnosis may improve the quality of care better. Doctors should pay more attention to the quality of care for patients living in rural areas, with low income, without complications and in higher pathological stages of tumor.