目的 通过建立急性心力衰竭（AHF）患者服药依从性预测模型,提高AHF患者的服药依从性和临床管理效果。方法 纳入2021年1月—2023年12月在广州市番禺区何贤纪念医院住院治疗的580例AHF患者,通过收集患者的一般人口学资料、疾病相关资料及出院后6个月的服药依从性数据,应用Logistic回归模型分析患者服药依从性的影响因素,并基于影响因素建立预测模型。结果 患者服药依从性总体良好（75%）。依从性良好组与依从性差组的年龄、独居情况、合并基础病、服药种类、疾病了解评分、治疗信心评分和自我控制信心评分比较差异有统计学意义（P<0.05）。Logistic 回归分析显示危险因素包括年龄≥60岁（OR=1.774）、独居（OR=1.871）、合并基础病≥2种（OR=1.719）和服药种类≥7种（OR=1.456）。而疾病了解评分（OR=0.923）、治疗信心评分（OR=0.946）和自我控制信心评分（OR=0.901）是保护因素（P<0.05）。基于上述因素建立的预测模型,通过ROC曲线验证,曲线下面积为0.815（95%CI：0.780~0.850）,提示所构建的模型具有良好的区分度。对该模型的校准度进行评价,P=0.528,提示该预测模型拟合度良好。此外,该预测模型的一致性指数为0.738,说明模型的预测性能良好。绘制的决策曲线中,曲线位于极端线之上,当阈概率取值在9%~59%时,对应的净获益率为0~27%,提示建立的模型具有优秀的临床有效性。结论 AHF患者的服药依从性受到多种因素的影响,包括年龄、居住状态、合并基础病种类及服药种类等。

Objective To establish a predictive model for medication compliance among acute heart failure（AHF）patients in order to enhance their therapeutic compliance and optimize clinical outcomes. Methods A total of 580 AHF inpatients at He Xian Memorial Hospital in Panyu District, Guangzhou between January 2021 and December 2023 were enrolled. Demographic information, disease-specific data,as well as post-discharge medication compliance records within six-month were collected by investigators. Utilizing logistic regression analysis revealed several influential determinants affecting medication compliance which formed the basis for constructing our predictive model. Results Generally,patient compliance was good（75%）. The comparison between the good compliance group and the poor compliance group showed that there were significant differences in age, living alone,combined with underlying diseases, types of medication, disease understanding score, treatment confidence score and self-control confidence score（P<0. 05）. Logistic regression analysis showed that independent risk indicators including individuals aged ≥60 years（odds ratio[OR]=1. 774）, those living alone（OR=1. 871）, presence of two or more underlying diseases（OR=1. 719）, along with consumption of seven or more medications daily（OR=1. 456）. Conversely,disease awareness score（OR=0. 923）, treatment confidence score（OR=0. 946）, and self-control confidence score（OR=0. 901）were identified as independent protective factors. Validation using receiver operating characteristic curves demonstrated robust predictive performance with an area under curve value of 0. 815（95%CI：0. 780-0. 850）, affirming its efficacy. The calibration of the model was evaluated, with a P-value of 0. 528, indicating good fit of the predictive model. Additionally, the concordance index（C-index）of the model was 0. 738, suggesting its excellent predictive performance. The decision curve analysis revealed that the curve was above the extreme lines, with a net benefit rate ranging from 0 to 27% when the threshold probability falls between. Conclusions The medication compliance of AHF patients is influenced by various factors, including age, living arrangement, the number of underlying diseases, and the number of medications taken. Targeted interventions such as enhancing patient education, simplifying treatment regimens, and improving social support can effectively improve the medication compliance of AHF patients. The predictive model established in this study provides a scientific basis for clinicians to develop more precise and effective individualized intervention measures,thereby improving the prognosis and quality of life.