缩短从证据到实践的旅程：混合实施效果试验的好处和复杂性

Full text

Translating evidence into practice is a long and arduous journey: 17 years, as the refrain goes (1). A traditional view sees this journey as linear; first, efficacy studies prove benefit, then effectiveness research creates knowledge in real-world settings, and finally, implementation science focuses on understanding and promoting uptake. With any one study in any phase potentially taking years to complete, the long period between the start of this journey and its end is unsurprising but, unfortunately, infeasible in the modern clinical care environment that needs treatment answers today. To address this evidence-to-implementation gap, hybrid effectiveness–implementation study designs have emerged (2), blending elements of effectiveness and implementation design to improve research efficiency and speed up knowledge translation. In this issue of AnnalsATS, Peltan and colleagues (pp. 424–432) give us an opportunity to consider the value and challenges of implementation research and hybrid implementation–effectiveness study designs. They report the results of a pilot hybrid trial of a complex and multifaceted program of several implementation strategies that, together, aimed to increase the use of lung protective ventilation directly (through audit and feedback, education, and championing of lung protective ventilation) and indirectly (by promoting increased use of clinical decision support tools in place to facilitate clinical decision-making) (3). Not only was the implementation itself multimodal, but the intended mechanism of practice change was also multipronged. They found that adherence to low tidal volume ventilation and other implementation outcomes indeed improved after intervention; however, no improvement was noted in several effectiveness outcomes, including mortality and ventilator-free days.

The study’s strengths reveal the promise of implementation studies to shorten that journey from evidence generation to improved healthcare delivery. The investigators increased the efficiency of a complex, costly study by leveraging a hybrid trial design to create knowledge about implementation outcomes and clinical outcomes in tandem. Furthermore, this design acknowledges that the phases from proving efficacy to clinical practice change need not be entirely sequential. Intentionally combining elements of effectiveness and implementation recognizes that these parts cannot, and indeed should not, be separated and that evaluating one without the other would be a missed opportunity. For example, capturing data on implementation factors could and perhaps should be standard for an effectiveness study to enable the exportation of successful strategies to other settings. Another strength of this study is the investigators’ approach to incorporating the possibility of secular trends in the analyses. By adding a secondary analysis using segmented regression, the investigators uncover that the findings of their primary analysis (that the period after intervention was associated with higher rates of adherence to lung protective ventilation) may have an alternative explanation and thereby potentially avoid incorrect conclusions about the impact of the specified implementation strategies. That finding also prompted them to recognize the silent or spillover implementation strategies that crept into the study, such as training local implementation teams before the study started.

Implementation science embraces complexity and messiness. But a tradeoff of encouraging complex and multifaceted implementation efforts is the difficulty of defining, studying, and reporting on all the parts that could drive change. For example, in this study (and to their credit), the authors acknowledge the possibility that the institution’s pretrial prioritization of adherence to lung protective ventilation was possibly an implementation strategy in and of itself, even before the deployment of the key strategies under study. Therefore, identifying and naming all implementation strategies a priori is critical. To guide researchers in doing so, implementation scientists have generated long lists of implementation strategies (4), many of which significantly overlap with each other and are almost by definition present concurrently in any implementation effort. Additional tools that can support clinicians and researchers in the identification and reporting of implementation strategies (5, 6) should be a minimum standard for the publication of implementation studies.

The study by Peltan and colleagues also reveals another important challenge of implementation research: the difficulty of operationalizing definitions of implementation outcomes. Lung protective ventilation is an effective intervention that has been shown to reduce mortality among patients with acute respiratory distress syndrome and likely has beneficial effects for many patients, even in the absence of acute respiratory distress syndrome. Yet, the intervention itself is not a simple treatment with a binary definition. The most cited efficacy study for lung protective ventilation tested a complex intervention involving 1) administration of low tidal volumes; 2) limiting inspiratory pressure; and 3) providing parameters for titration of inspired oxygen and positive end-expiratory pressure (7). For over a decade after this publication, numerous studies have tried to parse out what specific components of the multifaceted intervention were the drivers of the mortality benefit. Studies have explored the roles of initial tidal volume, duration and percentage of time with low tidal volumes, duration of time of exposure to injurious tidal volumes, and driving pressure as potential mechanisms by which lung protective ventilation improves mortality, to name just a few (8–10). It is infeasible to create an operational definition of adherence to the original intervention overall, encompassing all elements (tidal volume, inspiratory pressures, inspired oxygen, and positive end-expiratory pressure) and even more so to incorporate all these over a period of days of mechanical ventilation. Of the many possibilities, Peltan and colleagues selected a primary outcome of the percentage of time on mechanical ventilation during which low tidal volumes were administered and secondary outcomes of other operational definitions of adherence to low tidal volume ventilation as well as to other elements of the original intervention. Nonetheless, many elements of the original intervention went unmeasured.

This last challenge feeds into a key dilemma of implementation research: how do we interpret a hybrid implementation– effectiveness study when the improvement in implementation outcomes does not correspond to an improvement in clinical outcomes? The debate over whether to prioritize process or outcomes is age-old in health services research, and this study provides a case study for further consideration. There are three key points we wish to make. First, the relationship between the operational definitions of the implementation outcomes and the clinical outcomes must be well understood to inform the study design. In this example, as mentioned above, the operational definition of adherence to the proven-effective intervention is challenging, so one explanation for the mismatch between implementation and effectiveness outcomes could be that the mechanisms for clinical effectiveness are not captured by these implementation definitions. Second, the study design helps us with the interpretation of the results. The hybrid type 3 design delineates that implementation outcomes are primary and clinical outcomes are secondary. Such a trial may, appropriately, not have sufficient power to detect differences in clinical outcomes. The nature of hybrid implementation–effectiveness studies requires that investigators be clear and upfront about the state of the evidence and carefully specify how outcomes are prioritized. Finally, we must remember that in the real-world settings of implementation and effectiveness studies, populations are more heterogeneous than in efficacy studies, and cointerventions are the norm. This study promoted low tidal volume ventilation among all mechanically ventilated patients; we would not expect an effect like that seen in a clinical trial of carefully selected patients with acute respiratory distress syndrome. And as the authors themselves point out, simply preparing for the implementation may have served as an implementation strategy itself. All of these issues will push the results of effectiveness analyses toward the null hypothesis. Therefore, the lack of evidence for effectiveness in a hybrid implementation-effectiveness study should not cause us to throw out the implementation strategies or the clinical intervention altogether.

The well-executed study by Peltan and colleagues highlights the benefits of hybrid implementation–effectiveness trials, as well as their complexities. Understanding their use and appropriately deploying these trials through thoughtful study design, careful statistical analyses, and a thorough understanding of the clinical environment is essential to moving from evidence to practice at a faster pace demanded by critical care in the modern day.

全文翻译（仅供参考）

将证据转化为实践是一个漫长而艰巨的旅程：17年了，就像叠句一样（1）。传统的观点认为这段旅程是线性的;首先，功效研究证明了益处，然后，功效研究在现实世界中创造了知识，最后，实施科学侧重于理解和促进吸收。由于任何阶段的任何一项研究都可能需要数年时间才能完成，因此从开始到结束的漫长时间并不令人惊讶，但不幸的是，在今天需要治疗答案的现代临床护理环境中是不可行的。为了解决这一证据与实施之间的差距，出现了混合有效性-实施研究设计（2），将有效性和实施设计的要素融合在一起，以提高研究效率并加快知识转化。在这一期AnnalsATS中，Peltan及其同事（pp. 424-432）给予了我们一个机会来考虑实施研究和混合实施效果研究设计的价值和挑战。他们报告了一项复杂和多方面计划的试点混合试验的结果，该试验包括几种实施策略，旨在直接（通过审计和反馈，教育和支持肺保护性通气）和间接（通过促进增加使用临床决策支持工具以促进临床决策）增加肺保护性通气的使用（3）。不仅实施本身是多模式的，而且预期的做法改变机制也是多管齐下的。他们发现，在干预后，坚持低潮气量通气和其他实施结果确实有所改善;然而，在几个有效性结果方面没有改善，包括死亡率和无呼吸机天数。

该研究的优势揭示了实施研究的承诺，以缩短从证据生成到改善医疗保健服务的旅程。研究人员通过利用混合试验设计来创建关于实施结果和临床结果的知识，从而提高了复杂且昂贵的研究的效率。此外，该设计承认从证明疗效到临床实践改变的阶段不需要完全顺序。有意将效力和执行的要素结合起来，是因为认识到这两个部分不能、实际上也不应该分开，如果只评价其中一个而不评价另一个，就会错失机会。例如，收集关于执行因素的数据可以而且也许应该成为有效性研究的标准，以便能够将成功的战略输出到其他环境。这项研究的另一个优势是研究者在分析中纳入长期趋势的可能性的方法。通过使用分段回归添加二次分析，研究者发现他们的主要分析结果（即干预后的时间段与肺保护性通气的依从性较高相关）可能有另一种解释，从而可能避免关于指定实施策略的影响的错误结论。这一发现也促使他们认识到悄悄进入研究的沉默或溢出执行战略，例如在研究开始前培训当地执行团队。

实施科学包含复杂性和混乱性。但是，鼓励复杂和多方面的实施工作的一个代价是难以定义、研究和报告可能推动变革的所有部分。例如，在这项研究中（值得称赞的是），作者承认，该机构在试验前对肺保护性通气依从性的优先排序可能是一种实施策略本身，甚至在部署研究中的关键策略之前。因此，事先确定和命名所有实施战略至关重要。为了指导研究人员这样做，实施科学家已经生成了一长串实施策略（4），其中许多策略彼此明显重叠，并且几乎可以定义为同时存在于任何实施工作中。可以支持临床医生和研究人员识别和报告实施策略（5，6）的其他工具应该是实施研究出版的最低标准。

Peltan及其同事的研究还揭示了实施研究的另一个重要挑战：实施成果的定义难以操作。肺保护性通气是一种有效的干预措施，已被证明可降低急性呼吸窘迫综合征患者的死亡率，即使在没有急性呼吸窘迫综合征的情况下，也可能对许多患者产生有益影响。然而，干预本身并不是一个简单的治疗与二元定义。最常被引用的肺保护性通气疗效研究测试了一种复杂的干预措施，包括1）给予低潮气量; 2）限制吸气压力;和3）提供用于吸入氧气的滴定和呼气末正压的参数（7）。在这篇文章发表后的十多年里，许多研究试图解析出多方面干预的哪些具体组成部分是死亡率获益的驱动因素。研究已经探索了初始潮气量、低潮气量的持续时间和时间百分比、暴露于有害潮气量的持续时间和驱动压力作为肺保护性通气改善死亡率的潜在机制的作用，仅举几例（8 - 10）。创建对原始干预的总体依从性的操作定义是不可行的，包括所有元素（潮气量、吸气压力、吸入氧气和呼气末正压），并且甚至更不可行的是在机械通气的几天的时间段内合并所有这些。在许多可能性中，Peltan及其同事选择了一个主要结局，即给予低潮气量的机械通气时间百分比，以及其他操作定义的次要结局，即坚持低潮气量通气以及原始干预的其他要素。尽管如此，最初干预的许多要素都没有得到衡量。

这最后一个挑战导致了实施研究的一个关键困境：当实施结果的改善与临床结果的改善不一致时，我们如何解释混合实施-有效性研究？在卫生服务研究中，是否优先考虑过程或结果的争论由来已久，本研究提供了一个案例研究供进一步考虑。我们希望提出三个要点。首先，必须很好地理解实施结果和临床结果的操作定义之间的关系，以告知研究设计。在该示例中，如上所述，遵守经证明有效的干预措施的操作定义具有挑战性，因此实施和有效性结果之间不匹配的一种解释可能是这些实施定义未捕获临床有效性的机制。其次，研究设计有助于我们解释结果。混合3型设计描述了实施结局是主要结局，临床结局是次要结局。这样的试验可能没有足够的能力来检测临床结果的差异。混合实施有效性研究的性质要求研究人员明确和预先了解证据的状态，并仔细说明结果的优先顺序。最后，我们必须记住，在实施和有效性研究的现实世界中，人群比有效性研究中的人群更具异质性，并且联合干预是常态。本研究在所有机械通气患者中推广低潮气量通气;我们不会期望像在精心挑选的急性呼吸窘迫综合征患者的临床试验中看到的效果。而且正如作者自己所指出的，简单地为实现做准备可能本身就充当了一种实现策略。所有这些问题都会将有效性分析的结果推向零假设。因此，在混合实施-有效性研究中缺乏有效性证据不应导致我们完全抛弃实施策略或临床干预。

Peltan及其同事的这项执行良好的研究强调了混合实施有效性试验的好处及其复杂性。通过深思熟虑的研究设计、仔细的统计分析和对临床环境的透彻理解，了解它们的使用并适当地部署这些试验，对于以现代重症监护所要求的更快速度从证据转向实践至关重要。

THE END