来源：DOI 10.1016/j.im.2014.08.008

    知识的积累是一个领域科学和发展的基本条件[4]。更确切地说，进行有效的文献综述，对于推进知识的发展，了解某个感兴趣主题的研究广度，综合经验证据，发展理论或为后续研究提供概念背景，以及确定需要更多调查的主题或研究领域，是至关重要的[5-8]。

The accumulation of knowledge is an essential condition for a field to ‘‘be scientific’’ and to develop [4]. More precisely, conducting effective literature reviews is essential to advance the knowledge and understand the breadth of the research on a topic of interest, synthesize the empirical evidence, develop theories or provide a conceptual background for subsequent research, and identify the topics or research domains that require more investigation [5–8].
最普遍的评论类型通常被称为经验性文章中的 "文献回顾 "或 "理论背景"。论文的这一部分通常提供研究问题的理论基础和背景，并帮助研究问题成为焦点[14]。根据 Baker [15] 的说法，它代表了 "开展研究项目时必不可少的第一步和基础"（第 219 页）。更确切地说，文献综述部分有助于研究者理解现有的知识体系，为拟议的实证研究提供理论基础，证实研究问题的存在，证明拟议的研究对累积的知识有新的贡献，并且/或者为拟议的研究确定有效的研究方法、方法、目标和研究问题[12]。
The most prevalent type of review is commonly labeled the ‘‘literature review’’ or ‘‘theoretical background’’ within an empirical article. This section of a paper usually provides the theoretical foundations and context of the research question and helps bring the research question into focus [14]. According to Baker [15], it represents an ‘‘essential first step and foundation when undertaking a research project’’ (p. 219). More precisely, the literature review section helps the researcher understand the existing body of knowledge, provides a theoretical foundation for the proposed empirical study, substantiates the presence of the research problem, justifies the proposed study as one that contributes something new to the cumulated knowledge, and/or frames the valid research methodologies, approaches, goals and research questions for the proposed study [12].
还存在另一种类型的文献综述，它本身就构成了一种原创的、有价值的研究工作。与其说它为研究者自己的努力提供了基础，不如说它为所有对某一特定主题感兴趣的学术界其他成员创造了一个坚实的起点[9,16]。所谓的 "评论文章 "是一种期刊长度的文章，其首要目的是总结或综合某一领域的文献，而不收集或分析任何原始数据。撰写评论文章有几个原因，例如，分析特定研究流的进展、汇总研究结果或调和先前研究的含糊结果、回顾某个理论模型或方法学方法的应用、发展新的理论或研究模型，以及对先前关于特定主题或方法的研究进行批评性说明[5]。
There exists another type of literature review that constitutes an original and valuable work of research in and of itself. Rather than providing a basis for the researcher’s own endeavors, it creates a solid starting point for all other members of the academic community that are interested in a particular topic [9,16]. The so-called ‘‘review article’’ is a journal-length article that has an overarching purpose of summarizing or synthesizing the literature in a field without collecting or analyzing any primary data. Review articles can be undertaken for several reasons, such as analyzing the progress of a specific stream of research, aggregating findings or reconciling equivocal results of prior studies, reviewing the application of a theoretical model or a methodological approach, developing a new theory or research model and providing a critical account of prior research on a particular topic or method [5].
分类是最核心和通用的概念性工作之一。贝利[19]和史密斯[20]明确区分了两种形式的分类，即类型学和分类学。类型学是以演绎的方式得出的，而分类学通常是以经验或归纳的方式用聚类分析或其他统计方法得出。鉴于知识综合并不是一个新的概念，而且领先的方法论者已经提出了几种审查文献的方式和方法，显然，类型学似乎更符合我们最初的目标。
Classification is one of the most central and generic conceptual exercises. Bailey [19] and Smith [20] make a clear distinction between two forms of classification, namely, typologies and taxonomies. While a typology is derived in a deductive manner, a taxonomy is usually derived empirically or inductively using cluster analysis or other statistical methods. Given that knowledge synthesis is not a new concept and that leading methodologists have proposed several approaches and methods to review the literature, it clearly appears that a typology is more aligned with our initial objective.
在我们的类型学中，第一种类型的研究综合被称为叙述性评论。在其最简单的形式中，叙述性综述试图确定关于某个主题或话题的文字内容[10]。通常情况下，并不试图从所回顾的内容中寻求归纳或积累知识[37]。如表2所示，叙述性综述在许多其他方面与其他综述类型不同。叙述性综述通常是选择性的，因为它们不涉及对所有相关文献的系统和全面搜索。相反，叙述性综述通常是机会主义的，因为它们只调查研究人员容易获得的文献和证据[37]。重要的是，叙述性综述通常不提供任何关于如何进行综述过程的解释[2]。由于这个原因，它们容易受到主观因素的影响[10,57]。正如Dijkers[32]所认为的，"即使不存在偏见，传统的（叙述性）综述中缺乏关于如何搜索、选择和合并主要研究的信息，这使得其他人无法复制"（第425页）。这种缺乏明确的、可复制的方法被认为是叙述性综述的一个关键弱点[65]。关于数据分析，叙述性总结是指用于综合先前研究结果的非正式技术，通常包括某种类型的评论或解释[30]。
The first type of research synthesis in our typology is referred to as the narrative review. In its simplest form, the narrative review attempts to identify what has been written on a subject or topic [10]. Often, there is no attempt to seek generalization or cumulative knowledge from what is reviewed [37]. As shown in Table 2, narrative reviews differ from other review types in many other aspects. Narrative reviews are usually selective in that they do not involve a systematic and comprehensive search of all of the relevant literature. Instead, narrative reviews are often opportunistic in that they survey only that literature and evidence that are readily available to the researchers [37]. Importantly, narrative reviews usually do not provide any explanations of how the review process was conducted [2]. For this reason, they are vulnerable on the grounds of subjectivity [10,57]. As Dijkers [32] maintains, ‘‘even if no bias exists, the lack of information in the traditional (narrative) review on how primary studies were searched, selected, and combined makes replication by others impossible’’ (p. 425). This lack of explicit and reproducible methods has been identified as a key weakness of narrative reviews [65]. With regard to data analysis, narrative summary refers to the informal techniques used to synthesize prior study findings, often including some type of commentary or interpretation [30].
其次，描述性综述旨在确定某一特定研究领域的实证研究在多大程度上支持或揭示了与预先存在的命题、理论、方法或结论有关的任何可解释的模式或趋势[2]。为了实现这一目标并确保其结果的普遍性，描述性综述收集、编纂和分析反映现存文献中的主题、作者或方法的频率的数字数据[2,44]。它们通常采用结构化的搜索方法，在与特定调查领域相关的较大的已发表作品群中形成一个代表性样本。描述性综述的作者从每项研究中提取某些感兴趣的特征，如出版年份、研究方法、数据收集技术，以及最终结果的方向或强度（如阳性、阴性或不显著），以频率分析的形式产生定量结果。从本质上讲，描述性综述中包括的每项研究都被视为一个分析单位，而已发表的文献作为一个数据库，综述作者试图从中找出任何可解释的趋势和模式，或就现有概念、主张、方法或结果的优点得出总体结论。这样一来，描述性综述就可以宣称其结论代表了某个研究领域的技术水平[2]。
Second, descriptive reviews seek to determine the extent to which a body of empirical studies in a specific research area supports or reveals any interpretable patterns or trends with respect to pre-existing propositions, theories, methodologies or findings [2]. In fulfilling this objective and assuring the generalizability of their results, descriptive reviews collect, codify, and analyze numeric data that reflect the frequency of the topics, authors or methods found in the extant literature [2,44]. They usually employ structured search methods to form a representative sample of a larger group of published works that are related to a particular area of investigation. The authors of descriptive reviews extract certain characteristics of interest from each study, such as the publication year, research methods, data collection techniques, and direction or strength of the final outcomes (e.g., positive, negative, or non-significant) in the form of frequency analysis to produce quantitative results. In essence, each study included in a descriptive review is treated as a unit of analysis, and the published literature as a whole provides a database from which the review authors attempt to identify any interpretable trends and patterns or draw overall conclusions about the merits of existing conceptualizations, propositions, methods or findings. In doing so, a descriptive review may claim its findings to represent the state of the art in a research domain [2].

荟萃分析使用特定的数据提取技术和统计方法，从两个或多个功能相似的研究中，以标准效应措施（如风险比、赔率、平均差异、相关系数）的形式汇总定量数据，同时考虑到每个研究的相对样本量[2,69]。他们通常有四个主要目标。(1)评估纳入综述的主要研究之间结果的一致性/可变性（即研究之间的异质性）；(2)调查并解释（如果可行）任何观察到的异质性的原因（例如。(2) 调查并解释（如果可行）任何观察到的异质性的原因（例如，通过亚组或元回归分析），以提高科学认识；(3) 计算汇总效应大小以及置信区间；(4) 通过敏感性分析和正式评估研究偏倚的潜在来源（包括发表偏倚）来评估累积效应大小的稳健性，这些偏倚来源于主要研究并可能对计算的汇总效应产生影响。
Meta-analyses use specific data extraction techniques and statistical methods to aggregate quantitative data in the form of standard effect measures (e.g., risk ratios, odds ratios, mean differences, correlation coefficients) from two or more functionally similar studies, taking into account the relative sample size of each study [2,69]. They usually have four main goals: (1) evaluate the consistency/variability of the results between the primary studies included in the review (i.e., the between-study heterogeneity); (2) investigate and explain (if feasible) the causes of any observed heterogeneity (e.g., through subgroup or meta-regression analyses) to improve scientific understanding; (3) calculate a summary effect size along with a confidence interval; and (4) assess the robustness of the cumulative effect size through sensitivity analyses and formal evaluations of the potential sources of study bias, including publication bias, that stem from the primary studies and might have an impact on the calculated summary effect.
通过将统计学上有意义的研究结果与纳入综述的所有相关研究的非有意义的研究结果汇集在一起，元分析能够计算出对所调查现象的影响的更精确的估计，而不是将这些研究单独作为离散的信息来源来研究时的估计[2,57]。荟萃分析被认为是一种强大的综合研究方法，它能使研究者通过解决由相互冲突的经验研究引起的现有争议，得出有意义的推论。与以叙述方式整合现有证据的定性系统回顾（见下文）相比，元分析具有许多潜在的优势，包括增加力量、提高精确度，以及发现模式、调节器、调解器或甚至分散研究结果之间关系的能力[57,69]。
By pooling statistically significant findings with non-significant findings from all relevant studies included in the review, metaanalyses are able to calculate more precise estimates of the effects of the phenomenon under investigation than those derived from the individual studies alone when these are examined individually as discrete sources of information [2,57]. Meta-analyses are considered to be a powerful research method of synthesis that allows researchers draw meaningful inferences by settling existing controversies that arise from conflicting empirical studies.
定性系统回顾试图从定量实证研究中搜索、识别、选择、评估和抽象数据，以回答以下主要问题：（1）效果的方向是什么？ 2）效果的大小是什么？ 3）效果在纳入的研究中是否一致？(4) 效果的证据强度如何？他们采用了典型的系统综述过程，但与元分析方法相反，他们使用叙述和更主观的方法（而不是统计方法）来汇集所纳入的研究结果，并对问题1至4进行跟踪[51,70]。尽管从实证研究中提取统计数据（如 P 值、比率或相关性）和数字分析可能同时发生，但定性系统回顾的决定性因素是在分析和综合过程中采用文本方法[71]。评审员通常是某一领域的专家，他们可能会使用各种内容分析方法，如分组、聚类、框架、分类方案和特征表，以总结所选研究的结果，叙述性地整合累积的证据，并得出结论和/或建议。此外，他们还可能使用一种称为计票的准定量技术[2]，以确定纳入综述的研究之间的效果或关系的方向和一致性。计票，也称为箱体评分，是一种利用各个研究中报告的概率和P值等统计数据来比较具有统计学意义或不具有统计学意义的研究的数量，以支持或不支持某一假设[2,72]。例如，它可以用来计算和比较在综述中，有多少实证研究表明采用某个特定的信息系统与某个组织特征（如规模）呈正相关，又有多少研究表明它与之无关。这被认为是一种简单明了、易于解释的方法。然而，计票有一个臭名昭著的记录，那就是误导性，特别是当被纳入综述的研究规模较小，而且没有足够的力量来达到统计学上的显著结果时[33,72]。
Compared to qualitative systematic reviews (see below) that integrate the available evidence narratively, meta-analyses hold many potential advantages, including an increase in power, an improvement in precision, and the ability to discover patterns, moderators, mediators, or even relationships between studies with dispersed findings [57,69].
Qualitative systematic reviews attempt to search, identify, select, appraise, and abstract data from quantitative empirical studies to answer the following main questions: (1) What is the direction of the effect? (2) What is the size of effect? (3) Is the effect consistent across the included studies? (4) What is the strength of the evidence of the effect? They employ the typical systematic review process, but contrary to the meta-analytic approach, they use narrative and more subjective (rather than statistical) methods to bring together the findings of the included studies and follow through questions 1 to 4 [51,70]. Although the extraction of statistical data from the empirical studies (e.g., p values, ratios, or correlations) and numerical analyses may occur simultaneously, the defining element of qualitative systematic reviews is the adoption of a textual approach in the process of analysis and synthesis [71]. The reviewers, who are usually experts in a given field, might use various content analysis methods such as groupings, clusters, frameworks, classification schemes, and tabulations of characteristics to summarize the findings of the selected studies, narratively integrate the cumulative evidence, and arrive at conclusions and/or recommendations. Furthermore, they might use a quasi-quantitative technique called vote counting [2] to determine the direction and consistency of the effect or relationship between the studies included in the review. Vote counting, also known as box scoring, is a method that uses statistics such as probabilities and p values reported in the individual studies to compare the number of studies with statistically significant or non-significant results that favor or disconfirm a hypothesis [2,72]. For instance, it can be used to count and compare how many empirical studies within the review show the adoption of a particular information system to be positively associated with an organizational characteristic (e.g., size), and how many show that it is not. It is considered to be a straightforward and easily interpreted method. However, vote counting has a notorious record for being misleading, particularly when the studies included in the review are small and not sufficiently powered to reach statistically significant results [33,72].
伞状综述，也被称为综述概述，可被描述为一种三级研究，它将多个系统综述（定性或定量）的相关证据整合到一个可访问和可用的文件中，以解决一个狭窄的研究问题[47,60]。这是一种相对较新的证据综合方法，由于已发表的系统综述数量不断增加，在Cochrane合作组织内部和外部的健康科学领域都出现了这种方法[74] 。Bastian等人[75]最近的一项研究估计，每天有11篇新的系统综述和75项试验发表并被MEDLINE收录。这种增长导致针对相同或非常相似的研究问题的系统综述数量增加，同时，这些综述中不一致的发现或解释也随之增加[76]。这种不一致可能会引发激烈的争论，并对决策者（包括政策制定者、研究人员和其他依赖这些综述做出知情决定的利益相关者）构成重大挑战[76]。在这种情况下，综述是解决这些缺陷的下一个合理和适当的步骤。目前，对于这种类型的审查，还没有普遍接受的技术术语。因此，多年来，现存文献中出现了几个标签，其中最常见的是。伞式回顾"、"系统回顾概述"、"系统回顾的系统回顾 "和 "元回顾"[77]。这些术语都有其支持者，并在最近的出版物中被使用[例如，80,81]。
Umbrella reviews, also called overviews of reviews, may be described as a tertiary type of study that integrates relevant evidence from multiple systematic reviews (qualitative or quantitative) into one accessible and usable document to address a narrow research question [47,60]. It is a relatively new method of evidence synthesis that has emerged in the health sciences domain both within and outside the Cochrane Collaboration as a consequence of the constantly growing number of published systematic reviews [74]. A recent study by Bastian et al. [75] estimated that every day, 11 new systematic reviews and 75 trials are published and indexed in MEDLINE. This increase has led to a rise in the number of systematic reviews that address the same or a very similar set of research questions, with a concomitant increase in discordant findings or interpretations among them [76]. Such discordances may trigger passionate debates and pose a significant challenge for decision-makers, including policymakers, researchers, and other stakeholders who rely on these reviews to make informed decisions [76]. In this context, overviews of reviews represent the next logical and appropriate step to address these shortcomings. Currently, there is no universally accepted technical term for this type of review. As a result, over the years, several labels have appeared in the extant literature, of which the most common are: ‘‘umbrella review,’’ ‘‘overview of systematic reviews,’’ ‘‘systematic review of systematic reviews,’’ and ‘‘metareviews’’ [77]. Each of these terms has its proponents and has been used in recent publications [e.g., 80,81].
接下来两种形式的研究综合旨在建立解释。
首先，理论综述借鉴现有的概念和经验研究，为识别、描述和转化为更高层次的理论结构和各种概念、构造或关系提供背景。它们的主要目标是建立一个概念框架或模型，并提出一系列研究命题或假设[11,82]。开展这些研究的目的是为了解决一个新出现的问题，该问题将受益于新的理论基础的发展，或者是一个成熟的课题，该课题已经有了一定的研究积累，但缺乏适当的理论，或者当前的理论不足以解决现有的研究问题[11]。理论综述汇集了不同的工作流，并使用各种结构化的方法，如分类系统、分类法和框架，以有效地组织先前的研究，检查它们的相互关系，并发现将促进新理论发展的模式或共同点[11]。因此，它不仅仅是对过去工作的集合和描述。这种审查的主要贡献和价值在于它能够通过识别和强调我们所知道的和我们需要知道的之间的知识差距来发展新的概念或扩展现有的概念[11]。
The next two forms of research synthesis aim at explanation building. First, a theoretical review draws on existing conceptual and empirical studies to provide a context for identifying, describing, and transforming into a higher order of theoretical structure and various concepts, constructs or relationships. Their primary goal is to develop a conceptual framework or model with a set of research propositions or hypotheses [11,82]. They can be conducted to tackle an emerging issue that would benefit from the development of new theoretical foundations or a mature topic for which an accumulated body of research exists but there is a lack of appropriate theories or current theories are inadequate in addressing existing research problems [11]. A theoretical review brings together diverse streams of work and uses various structured approaches such as classification systems, taxonomies and frameworks to organize prior research effectively, examine their interrelationships, and discover patterns or commonalities that will facilitate the development of new theories [11]. As such, it goes beyond merely assembling and describing past work. The primary contribution and value of this type of review lies in its ability to develop novel conceptualizations or extend current ones by identifying and highlighting knowledge gaps between what we know and what we need to know [11].
与聚合性综述相反，理论综述通常从一个广泛的综述问题开始，随着更多证据的收集和分析，这个问题通常会被完善[83]。在检索策略方面，Webster和Watson[11]建议采用系统化的方法，以确保积累相对完整的相关文献（包括理论和经验）的普查。一些人认为，正式的质量评估可能是不必要的[例如，86]，而另一些人则提出，弱点或有致命缺陷的论文（从概念或方法学的角度）应该完全排除在审查之外[例如，38]。理论综述中可以使用几种综合方法。这些方法可以根据评论所依据的认识论和本体论的假设进行分组。一方面，解释学方法包括基础理论、元民族志、元综合、元三角和元叙事[例如，54,59,83,87]。另一方面，也可以使用实证主义方法，包括内容分析和定性比较分析[例如，69]。
Contrary to aggregative reviews, theoretical reviews usually start with a broad review question that is often refined as more evidence is gathered and analyzed [83]. In terms of search strategy, Webster and Watson [11] recommend a systematic approach to ensure that a relatively complete census of the relevant literature (both theoretical and empirical) is accumulated. Some argue that formal appraisals of quality may not be necessary [e.g., 86], while others propose that weak or fatally flawed papers (from a conceptual or methodological standpoint) should be excluded from the review altogether [e.g., 38]. Several methods of synthesis can be used in theoretical reviews. These can be grouped according to the epistemological and ontological assumptions that underlie the review. On the one hand, interpretive methods include grounded theory, meta-ethnography, meta-synthesis, metatriangulation, and meta-narratives [e.g., 54,59,83,87]. On the other hand, positivist methods including content analysis and qualitative comparative analysis [e.g., 69] can also be used.
现实主义综述（也称为元叙事综述或定性证据综合综述）是理论驱动的解释性综述，其目的是通过对应用于不同环境的复杂干预措施的异质性证据进行理解，为政策决策提供信息、加强、扩展或替代性补充[86]。它们源于对传统系统回顾和非复杂干预措施的元分析的批评，这些批评集中在其简单化、实证主义的基本假设上[87]。如前所述，系统性回顾旨在确定因果关系。如果应用X，那么就会发生Y。这样的逻辑适合于医学和教育等科学领域，在这些领域中，随机对照试验和个别试验的结果可以结合起来，汇总起来，看一个新的治疗或干预措施是否能改善结果。然而，Pawson[88]认为，在社会政策等领域，不可能在干预和结果之间建立这种直接的因果关系，对于任何社会项目的干预，都不可能有一个常规的、一致的结果。相反，由于不同的动态环境，会出现各种结果模式，包括成功的、不成功的和 "两者都有"[88]。我们同意Oates[87]的观点，即Pawson对社会项目性质的描述与IS领域的情况相类似。事实上，人们早就认识到，任何涉及IT的干预都被插入到一个社会构建的环境中，而且干预可以被它所处的环境所改变[例如，91]。
为了规避这些限制，Pawson等人[55]开发了一种新的方法来综合先前的知识，试图解开 "复杂干预 "在特定背景或环境下如何发挥作用（或为何失败）的机制。这种形式的综合涉及批判现实主义方法。现实主义是一种方法论取向，它起源于哲学，应用于社会学、心理学和经济学等领域[90]。在现实主义下，"什么有效 "的基本评价问题变成了 "这种干预措施是什么，对谁有效，在什么情况下，在什么方面，为什么？这种方法以解释为重点，对定量或定性的证据没有特别的偏好。事实上，它认为多种方法都有好处，这样就可以对复杂的干预措施的过程和影响进行调查。现实主义评论从阐明可能的基本机制开始，然后分析可用的证据，以找出这些机制是否适用以及在何处适用[43,59]。文献中发现的任何相关初级研究都被视为案例研究，可以检验和修改最初的理论[43]。

现实主义评论的搜索过程通常是反复的、有目的的。初步搜索描绘了主题的范围，从而可以发展出一个初步的理论，随后的搜索可以寻找经验证据来检验理论，最后的搜索可以微调新出现的研究综合[90]。现实主义综述会评估每项主要研究的相关性，以及它为支持或反驳部分新兴理论所提供的证据的严谨性。它可能在某一领域有强有力的证据，而在另一领域有薄弱或没有证据；个别研究可能在整体上有缺陷，但它仍然可以在某些部分提供有用的证据[43]。如前所述，现实主义评论认为综合的任务是对理论的完善。现实主义回顾提出了相当大的概念和方法上的挑战，特别是与识别相关研究、评估纳入的研究以及综合证据的方法有关的挑战[63]。关于现实主义框架下的数据综合方法和出版标准的更详细讨论，可以在Wong等人的文章中找到。[62]、Pawson等人[55]、Shepperd等人[59]、Barnett-Page和Thomas[91]、Pearson[92]和Popay[93]对现实主义框架下的数据综合方法和出版标准进行了更详细的讨论。
Second, realist reviews (also called meta-narrative reviews or qualitative evidence synthesis reviews) are theory-driven interpretative reviews that were developed to inform, enhance, extend or alternatively supplement conventional systematic reviews by making sense of heterogeneous evidence about complex interventions applied in diverse contexts in a way that informs policy decision making [86]. They originated from criticisms of conventional systematic reviews and meta-analyses of noncomplex interventions that center on their simplistic, positivist underlying assumptions [87]. As explained earlier, systematic reviews seek to identify causation: If X is applied, then Y will occur. Such logic is appropriate for scientific fields such as medicine and education in which randomized controlled trials and the results of individual trials can be combined and aggregated to see whether a new treatment or intervention improves outcomes. However, Pawson [88] argues that it is not possible to establish such direct causal links between interventions and outcomes in fields such as social policy, where for any social program intervention, there is unlikely to be a regular, consistent outcome. Instead, there will be a pattern of outcomes, including successful, unsuccessful and ‘‘a bit of both’’ because of different and dynamic contexts [88]. We concur with Oates [87] that Pawson’s description of the nature of social programs is analogous to the situation in the IS field. Indeed, it has long been recognized that any intervention involving IT is inserted into a socially constructed context, and the intervention can be modified by the context in which it occurs [e.g., 91].
To circumvent these limitations, Pawson et al. [55] have developed a new approach to synthesizing prior knowledge that seeks to unpack the mechanism of how ‘‘complex interventions’’ work (or why they fail) in particular contexts or settings. This form of synthesis involves a critical realist approach. Realism is a methodological orientation that has its roots in philosophy and applications in fields such as sociology, psychology and economics [90]. Under realism, the basic evaluative question of what works changes to ‘‘what is it about this intervention that works, for whom, in what circumstances, in what respects and why?’’ This approach, which has an explanatory focus, has no particular preference of either quantitative or qualitative evidence. Indeed, it sees merit in multiple methods so that both processes and impacts of complex interventions may be investigated. A realist review begins by articulating the likely underlying mechanisms and then analyzes the available evidence to find out whether and where these mechanisms are applicable [43,59]. Any relevant primary studies found in the literature are viewed as case studies that can test and modify the initial theories [43].
The search process in realist reviews is often iterative and purposive. A preliminary search maps out the scope of the topic so that an initial theory can be developed, subsequent searches can look for the empirical evidence to test the theory and a final search can fine-tune the emerging research synthesis [90]. A realist review assesses each primary study for its relevance and the rigor of evidence it offers to support or contradict parts of the emerging theory. It might have strong evidence in one area and weak or no evidence in another; the individual study might be flawed overall, but it could still offer useful evidence in some sections [43]. As mentioned earlier, a realist review perceives the task of synthesis as one of refining theory. Realist reviews pose considerable conceptual and methodological challenges, particularly relating to the identification of relevant studies, the appraisal of included studies, and the methods of synthesizing evidence [63]. A more detailed discussion of data synthesis methods and publication standards within a realist framework can be found in Wong et al.[62], Pawson et al. [55], Shepperd et al. [59], Barnett-Page and Thomas [91], Pearson [92], and Popay [93].
The search process in realist reviews is often iterative and purposive. A preliminary search maps out the scope of the topic so that an initial theory can be developed, subsequent searches can look for the empirical evidence to test the theory and a final search can fine-tune the emerging research synthesis [90]. A realist review assesses each primary study for its relevance and the rigor of evidence it offers to support or contradict parts of the emerging theory. It might have strong evidence in one area and weak or no evidence in another; the individual study might be flawed overall, but it could still offer useful evidence in some sections [43]. As mentioned earlier, a realist review perceives the task of synthesis as one of refining theory. Realist reviews pose considerable conceptual and methodological challenges, particularly relating to the identification of relevant studies, the appraisal of included studies, and the methods of synthesizing evidence [63]. A more detailed discussion of data synthesis methods and publication standards within a realist framework can be found in Wong et al.[62], Pawson et al. [55], Shepperd et al. [59], Barnett-Page and Thomas [91], Pearson [92], and Popay [93].
例如，Levy和Ellis[12]提出了一个进行和撰写有效文献综述的通用框架。他们的框架遵循系统的数据处理方法，包括三个步骤，即：（1）文献收集和筛选（输入）；（2）分析证据（处理）；以及（3）撰写文献回顾（输出）。