From the National Center for Human Factors in Healthcare, MedStar Institute for Innovation, MedStar Health.
J Patient Saf. 2021 Sep 1;17(6):e509-e514. doi: 10.1097/PTS.0000000000000400.
The increase in patient safety reporting systems has led to the challenge of effectively analyzing these data to identify and mitigate safety hazards. Patient safety analysts, who manage reports, may be ill-equipped to make sense of report data. We sought to understand the cognitive needs of patient safety analysts as they work to leverage patient safety reports to mitigate risk and improve patient care.
Semistructured interviews were conducted with 21 analysts, from 11 hospitals across 3 healthcare systems. Data were parsed into utterances and coded to extract major themes.
From 21 interviews, 516 unique utterances were identified and categorized into the following 4 stages of data analysis: input (15.1% of utterances), transformation (14.1%), extrapolation (30%), and output (14%). Input utterances centered on the source (35.9% of inputs) and preprocessing of data. Transformation utterances centered on recategorizing patient safety events (57.5% of transformations) or integrating external data sources (42.5% of transformations). The focus of interviews was on extrapolation and trending data (56.1% of extrapolations); alarmingly, 16.1% of trend utterances explicitly mentioned a reliance on memory. The output was either a report (56.9% of outputs) or an action (43.1% of outputs).
Major gaps in the analysis of patient safety report data were identified. Despite software to support reporting, many reports come from other sources. Transforming data are burdensome because of recategorization of events and integration with other data sources, processes that can be automated. Surprisingly, trend identification was mostly based on patient analyst memory, highlighting a need for new tools that better support analysts.
随着患者安全报告系统的增加，有效分析这些数据以识别和减轻安全隐患的挑战也随之而来。管理报告的患者安全分析师可能缺乏理解报告数据的能力。我们试图了解患者安全分析师在利用患者安全报告来降低风险和改善患者护理方面的认知需求。
对来自 3 个医疗系统的 11 家医院的 21 名分析师进行了半结构化访谈。数据被解析成语句并进行编码，以提取主要主题。
从 21 次访谈中，共确定了 516 个独特的语句，并分为数据分析的以下 4 个阶段：输入（占语句的 15.1%）、转换（占 14.1%）、推断（占 30%）和输出（占 14%）。输入语句主要集中在来源（占输入的 35.9%）和数据预处理上。转换语句主要集中在重新分类患者安全事件（占转换的 57.5%）或整合外部数据源（占转换的 42.5%）上。访谈的重点是推断和趋势数据（占推断的 56.1%）；令人惊讶的是，16.1%的趋势语句明确提到依赖记忆。输出结果要么是报告（占输出的 56.9%），要么是行动（占输出的 43.1%）。
发现患者安全报告数据分析存在重大差距。尽管有软件支持报告，但许多报告来自其他来源。由于事件的重新分类和与其他数据源的集成，数据转换过程繁琐，这些过程可以实现自动化。令人惊讶的是，趋势识别主要基于患者分析师的记忆，这凸显了需要新工具来更好地支持分析师的需求。
