Validating a data science methodology for patterns of mental health services use: The patient record of clinical experience sequence study (PROCESS)

背景：有效的学习型医疗系统需要帮助管理者的措施 确定如何促进系统级别的改进。一次影响系统级别的机会 改进是直接测量提供给患者的护理序列，这可能反映 效率降低，护理分散加剧。我们建议确定退伍军人管理局 管理数据可用于构建护理序列的可靠测量。 意义/影响：我们的创新之处在于采用数据科学序列分析 退伍军人事务部行政记录的方法论。这种方法论有可能突出护理 碎片化和一体化。可以说，碎片化是最重要的未被强调的 退伍军人事务部的进球。有关于护理和获取质量的绩效目标，并且有很强的 用于管理成本的基础设施。然而，对减少碎片化和 改善一体化，部分原因是缺乏适当的措施。退伍军人管理局的优先事项包括更多 高效地利用资源。提高效率的两个VHA策略是勤奋地寻找 浪费和纠正以产生节约，并通过以下方式改善保健服务的提供 确保所有护理环境中的护理协调。基于序列的分段和 一体化措施有可能直接为这些战略提供信息。 具体目标：这项为期两年的提案的具体目标是确定退伍军人管理局 管理数据可以用来可靠地衡量心理健康护理序列。作为一种 为了探索性目的，我们将确定序列是否可能代表CARE碎裂。 方法：我们将使用退伍军人事务部数据仓库收集内部证据 一致性和重测信度。大约46,000名退伍军人将在每个 54个医疗中心的6个年度队列(2013财年至2018财年)。我们将使用序列分析 识别具有共同共识特征的相似序列的簇 顺序模式。内部一致性将通过比较随机患者来确定 用于确定患者序列是否在共识序列中更相似的样本 而不是模式之间的差异。重新测试的可靠性将随着时间的推移比较模式。我们 期望一个阶跃函数，其中序列随着时间的推移将是相似的，具有如下周期性变化 能力得到提升。一般来说，近端序列将比远端序列更相似。 为了探索性的目的，我们将使用管理数据库计算患者级别的相关性 医疗服务碎片化的衡量标准和设施水平与退伍军人管理局绩效指标的相关性。 带有专家小组的Delphi过程将在多大程度上由 方法论可以衡量碎片化。这将为下一步的研究提供初步数据。 下一步/实施：这项为期两年的研究将确定序列分析是否 该方法可应用于退伍军人管理局的管理数据，以确定可靠的护理序列。输出 Delphi过程以及会聚和判别效度测试将允许团队 开发关于退伍军人抑郁症护理序列的特定假设，这些假设将在后续测试中进行- 向上学习。下一步将是确定抑郁症护理序列与 心理健康症状、功能、满意度和成本。我们的长期目标是开发一种 用于近实时测量护理序列并向管理者提供反馈的方法和 临床医生根据可能需要干预的护理顺序确定患者。

Background: An effective learning healthcare system needs measures that help managers identify how to promote system-level improvements. One opportunity to influence system level improvements is to directly measure the care sequences provided to patients that may reflect decreased efficiency and increased care fragmentation. We propose to determine whether a VA administrative data can be used to construct reliable measures of care sequences. Significance/Impact: Our innovation is in adapting a data science sequence analysis methodology to VA administrative records. This methodology has the potential to highlight care fragmentation and integration. Fragmentation is arguably the most important underemphasized goal in VA. Performance goals exist for quality of care and access, and there is a strong infrastructure for managing cost. However, there is limited focus on reducing fragmentation and improving integration, in part due to the lack of adequate measures. VA priorities include more efficient resource use. Two VHA strategies to increase efficiency are to diligently find areas of waste and correct to generate savings, and to improve the delivery of health care services by ensuring care coordination across all care settings. Sequence-based fragmentation and integration measures have the potential to directly inform these strategies. Specific Aim: The specific aim of this two-year proposal is to determine whether VA administrative data can be used to reliably measure mental health sequences of care. As an exploratory aim, we will determine whether sequences may represent care fragmentation. Methodology: We will use the VA Corporate Data Warehouse to collect evidence for internal consistency and test-retest reliability. Approximately 46,000 Veterans will be sampled in each of 6 annual cohorts (FY2013-FY2018) across 54 medical centers. We will use sequence analysis to identify clusters of similar sequences that are characterized by a common consensus sequential pattern. Internal consistency will be determined by comparing random patient samples to determine if patient sequences are more similar within consensus sequential patterns than between patterns. Test-retest reliability will compare patterns over time. We expect a step function where sequences will be similar over time, with periodic changes as capabilities improve. Generally, proximal sequences will be more similar than distal sequences. For the exploratory aim, we will calculate patient-level correlations with administrative database measures of care fragmentation and facility-level correlations with VA performance measures. A Delphi process with an expert panel will the degree to which each sequence generated by the methodology may measure fragmentation. This will provide preliminary data for the next study. Next Steps/Implementation: This two-year study will determine whether the sequence analysis method can be applied to VA administrative data to identify reliable care sequences. The output of the Delphi process and the convergent and discriminant validity tests will allow the team to develop specific hypotheses about VA depression care sequences that will be tested in a follow- up study. The next step will be to determine the association of depression care sequences with mental health symptoms, functioning, satisfaction, and cost. Our long-term goal is to develop a method for measuring care sequences in near-real time and provide feedback to managers and clinicians to identify patients regarding care sequences that may require intervention.