Institute for Health Services Research and Clinical Epidemiology, Philipps-Universität Marburg, Karl-von-Frisch-Strasse 4, 35043, Marburg, Germany.
BMC Med Res Methodol. 2024 Sep 19;24(1):212. doi: 10.1186/s12874-024-02317-z.
In longitudinal health services research, hospital identification using an ID code, often supplemented with several additional variables, lacks clarity regarding representativeness and variable influence. This study presents an operational method for hospital identity delimitation and a novel longitudinal identification approach, demonstrated using a case study.
The conceptualisation considers hospitals as evolving entities, identifying "similar enough" pairs across two time points using an automated similarity matrix. This method comprises key variable selection, similarity scoring, and tolerance threshold definition, tailored to data source characteristics and clinical relevance. This linking method is tested by applying the identification of minimum caseload requirements-related German hospitals, utilizing German Hospital Quality Reports (GHQR) 2016-2020.
The method achieved a success rate (min: 97.9% - max: 100%, mean: 99.9%) surpassing traditional hospital ID-code linkage (min: 91.5% - max: 98.8%, mean: 96.6%), with a remarkable 99% reduction in manual work through automation.
This method, rooted in a comprehensive understanding of hospital identities, offers an operational, automated, and customisable process serving diverse clinical topics. This approach has the advantage of simultaneously considering multiple variables and systematically observing temporal changes in hospitals. It also enhances the precision and efficiency of longitudinal hospital identification in health services research.
在纵向卫生服务研究中，使用 ID 码识别医院，通常辅以几个额外的变量，在代表性和变量影响方面缺乏明确性。本研究提出了一种用于医院身份界定的操作方法和一种新的纵向识别方法，并用案例研究进行了演示。
概念化将医院视为不断发展的实体，使用自动相似度矩阵在两个时间点识别“足够相似”的对。这种方法包括关键变量选择、相似度评分和容差阈值定义，根据数据源特征和临床相关性进行定制。通过应用与最小病例量要求相关的德国医院识别方法来测试这种链接方法，利用德国医院质量报告（GHQR）2016-2020 年的数据。
该方法的成功率（最小：97.9% - 最大：100%，平均：99.9%）超过了传统的医院 ID 码链接（最小：91.5% - 最大：98.8%，平均：96.6%），通过自动化大大减少了 99%的手动工作。
这种方法基于对医院身份的全面理解，提供了一种可操作、自动化和可定制的过程，适用于各种临床主题。这种方法具有同时考虑多个变量和系统观察医院时间变化的优势。它还提高了卫生服务研究中纵向医院识别的精度和效率。
