初级保健连续性与急性病症可预防住院:基于机器学习的记录关联研究
该研究链接澳大利亚新南威尔士州45岁及以上人群的45 and Up Study数据与Medicare索赔数据,使用双机器学习模型分析11年(2007-2017)链接数据,控制年龄、多病共存、文化多样性、性别、教育水平、心理状态、身体限制、吸烟状况、社会经济十分位等患者特征。在54,376名参与者中,27,634人(50.8%)在11年研究期间至少经历1次急性可预防住院(PPH)事件。研究发现,与急性非PPH相比,即使连续照护(continuity of care)轻微改善也可降低急性PPH发生率;当连续照护从第45百分位(0.274)增至第50百分位(0.301)时,4个模型下急性PPH相对非急性PPH的概率降低范围为9.8%(95%CI,1.1%-17.8%)至23.5%(95%CI,14.1%-32.4%)。结论认为初级层面连续照护在减少急性PPH中起关键作用,以人为中心或整合型照护政策应包含促进连续照护的举措并支持全科医生提高连续照护。
要点
- 连续照护轻微改善即可降低急性可预防住院概率,4个模型降幅9.8%-23.5%。
- 研究纳入54,376名45岁及以上澳大利亚人,50.8%至少经历1次急性PPH。
- 政策应促进以人为中心和整合型照护以支持全科医生提高连续照护。
核心结论
- 初级保健连续照护在减少急性可预防住院中起关键作用。
- 以人为中心或整合型照护政策应包含促进连续照护的举措并支持全科医生提高连续照护。
研究设计与样本
- 研究链接澳大利亚新南威尔士州45岁及以上人群的45 and Up Study数据与Medicare索赔数据。
- 使用11年(2007-2017)链接数据,采用双机器学习模型控制年龄、多病共存、文化多样性、性别、教育水平、心理状态、身体限制、吸烟状况、社会经济十分位等患者特征。
- 使用4种算法(LASSO、随机森林、极端梯度提升、人工神经网络)进行估计以确保稳健性。
主要发现
- 54,376名参与者中,27,634人(50.8%)在11年研究期间至少经历1次急性PPH事件。
- 与急性非PPH相比,连续照护轻微改善即可降低急性PPH发生率。
- 连续照护从第45百分位(0.274)增至第50百分位(0.301)时,4个模型下急性PPH相对非急性PPH的概率降低范围为9.8%(95%CI,1.1%-17.8%)至23.5%(95%CI,14.1%-32.4%)。
关键数值与推荐
- 连续照护从第45百分位(0.274)增至第50百分位(0.301)时,急性PPH相对非急性PPH的概率降低范围为9.8%(95%CI,1.1%-17.8%)至23.5%(95%CI,14.1%-32.4%)。
临床怎么用
- 全科医生应关注并努力提高连续照护,以降低患者急性可预防住院风险。
- 政策制定者可将以人为中心和整合型照护政策纳入促进连续照护的举措。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
PURPOSE: Reducing potentially preventable hospitalization (PPH), also known as ambulatory care-senstive conditions, is a global concern. This study linked data from Sax Institute's 45 and Up Study on individuals aged 45 years and older from New South Wales, Australia, with Australian Medicare claims data to establish a causal relationship between continuity of care and acute PPH using a double machine learning model. METHODS: We utilized 11 years of linked data (2007-2017) to analyze the impact of continuity of care on acute PPH, controlling for key patient characteristics (ie, age, multimorbidity status, cultural diversity, sex, education level, psychological status, physical limitation, smoking status, socioeconomic deciles). Estimation was done using a double machine learning technique with 4 algorithms (ie, least absolute shrinkage and selection operator, random forest, extreme gradient boosting, artificial neural network) to ensure robustness. RESULTS: Among 54,376 participants, 27,634 individuals (50.8%) experienced at least 1 acute PPH episode during the 11-year study period. Our findings indicate that even a slight improvement in continuity of care can reduce the incidence of acute PPH compared with non-acute PPH. For example, the reduction in the probability of acute PPH compared with non-acute PPH ranges from 9.8% (95% CI, 1.1%-17.8%) to 23.5% (95% CI, 14.1%-32.4%) across 4 models when continuity of care increases from the 45th percentile (0.274) to the 50th percentile (0.301). CONCLUSION: Continuity of care at the primary level plays a key role in reducing acute PPH. Policies focused on person-centered or integrated care should include initiatives to promote continuity of care and support general practitioners in improving continuity of care.The authors of this article have provided Hindi and Vietnamese translations of the abstract.
引用
- Nguyen NMP, Borah BJ, Barr M, et al. Continuity of Primary Care and Preventable Hospitalization for Acute Conditions: A Machine Learning-Based Record Linkage Study. Ann Fam Med. 2025. PMID: 41285609. doi:10.1370/afm.240569
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。