Narrative summary typically involves the selection, chronicling and ordering of evidence to produce an account of the evidence. Its form may vary from the simple recounting and description of findings through to more interpretive and explicitly reflexive accounts that include commentary and higher levels of abstraction. Narratives of the latter type can account for complex dynamic processes, offering explanations that emphasise the sequential and contingent character of phenomena (Abbot 1990). Narrative summary is often used in systematic reviews alongside systematic searching and appraisal techniques, as exemplified by Fairbank et al (2000). It can ‘integrate’ qualitative and quantitative evidence through narrative juxtaposition. Under the UK ESRC Methods Programme, methodological guidance on the conduct of narrative summaries has been developed, which informs good practice in this area (Popay et al 2006).

Thematic analysis, clearly sharing some overlaps with narrative summary and content analysis, involves the identification of prominent or recurrent themes in the literature, and summarising the findings of different studies under thematic headings. Summary tables, providing descriptions of the key points, can then be produced (Mays et al 2001). Several recent attempts at providing structured or systematic overviews of diverse areas of evidence have adopted this kind of approach (e.g. Garcia et al 2002).

Thematic analysis allows clear identification of prominent themes. It is flexible, allowing considerable latitude to reviewers and a means of integrating qualitative and quantitative evidence. It can be either data driven – driven by the themes identified in the literature itself – or theory driven – oriented to evaluation of particular themes through interrogation of the literature. However, there is frequently a lack of explicitness about procedures and aims in this area, including failure to specify the extent to which them-atic analyses are descriptive or interpretive. Questions remain about whether the structure of the analysis should reflect the frequency with which particular themes are reported, or should be weighted towards themes that appear to have a high level of explanatory value. If thematic analysis is limited to summarising themes reported in primary studies, it offers little by way of theoretical structure within which to develop higher-order thematic categories beyond those identified from the literature.

Grounded theory, originally formulated by Glaser and Strauss (1967), describes methods for qualitative sampling, data collection and data analysis. It sees the overriding concern of qualitative research as the generation of theory (generalisable explanations for social phenomena). The constant comparative method, the most widely used element of grounded theory, has the most obvious potential for application for systematic review in part (especially in later formulations) because it offers a set of procedures by which data may be analysed (see Strauss & Corbin 1998).

One of the most robust and theoretically sophisticated examples of the use of grounded theory for synthesis is Kearney’s (2001) grounded theory analysis of 15 qualitative papers on women’s experience of domestic violence. This study shows how grounded theory can deal with sampling issues and allow a synthesis of studies by treating study reports as a form of data on which analysis can be conducted using the constant comparative method. The generation of higher-order themes as a means of synthesis encourages reflexivity on the part of the reviewer while preserving the interpretive properties of the underlying data. Grounded theory, in the notions of theoretical saturation and theoretical sampling, also offers a means of limiting the number of papers that need be reviewed, especially where the emphasis is on conceptual robustness rather than on completeness of data.

Grounded theory does, however, have several disadvantages as a method for review. Even in its more proceduralised forms, it inherently lacks full transparency because it is an interpretive method. It also offers no advice on how to appraise studies for inclusion in a review. There are several important epistemological issues to be resolved, including the status of the accounts offered in the studies and how to deal with the varying credibility of these accounts. Moreover, the methodological anarchy that characterises the area, with ‘grounded theory’ being used to label many different types of analysis, should not be underestimated as a barrier to the development of this approach as a means of syn-thesising primary studies (Dixon-Woods et al 2004).

Meta-ethnography is a set of techniques specifically developed for synthesising qualitative studies. First proposed by Noblit and Hare (1988), it involves three major strategies:

Paterson 1998a and Paterson 1998b use the term ‘meta-study’ to encompass the overview of theory, method and data. They distinguish between meta-data synthesis, meta-method synthesis and meta-theory synthesis. Meta-data synthesis refers to the syn-thesis of data presented in reports; they suggest that the choice of analytic approach is up to the reviewers, with possible choices including grounded theory, meta-ethnography, thematic analysis and interpretive descriptive analysis (by which they seem to mean a narrative critical review).

Paterson et al propose that the key concern in meta-method is with identifying how the methods applied to an area of study shape understandings of it (e.g. interview-based studies compared with ethnographies). Meta-theory, on the other hand, in-volves a critical exploration of the theoretical frameworks that have provided direction to research (e.g. psychological and sociological approaches to understanding people’s experiences of chronic illness). They use the term ‘meta-synthesis’ to describe bringing together the ideas that have been deconstructed in the three meta-study processes. The primary goal of such an analysis is to develop mid-range theory (i.e. theories that are moderately abstract and have direct applications for particular defined areas of practice). However, the term ‘meta-synthesis’ is often deployed as a way of distinguishing reviews of qualitative studies from systematic reviews or meta-analyses (Walsh & Downe 2005). In this latter sense, ‘meta-synthesis’ has become widely used to describe an approach to synthesising qualitative research that draws broadly on the meta-ethnography tradition, and does not necessarily involve the strategies specified by Paterson et al. Though meta-study provides a very useful framework it also suffers from the limitations that it does not explicitly cope with quantitative evidence, and its processes are laborious and time-consuming.

Realist synthesis is an explicitly theory-driven approach to the synthesis of evidence (Pawson 2002a). Beginning with the theory that (apparently) underlies a particular programme or intervention, it seeks evidence in many forms, including formal study reports (both qualitative and quantitative) as well as case studies, media reports, and other diverse sources, and integrates them by using them as forms of proof or refutation of theory.

Examples of the approach have been provided in relation to the provision of smoke alarms as a means of improving safety, and public disclosure (Pawson 2002b). Key problems include the tendency to treat all forms of evidence as equally authoritative, the contingency of the chains of evidence, the vul-nerability to the robustness of the theory being evaluated rather than the evidence being offered, and the lack of explicit guidance on how to deal with contradictory evidence.

Miles and Huberman’s cross-case techniques (1994) offer a number of strategies for conducting cross-case analyses, which might also be suitable for synthesising across different studies. These include meta-matrices for partitioning and clustering data in various ways, sometimes involving summary tables based on content analysis, case-ordered displays, or time-ordered displays. Though Miles and Huberman appear to be discussing the analysis of primary data, their techniques are readily transferable to the synthesis of study reports. McNaughton (2000) describes using these techniques to conduct a synthesis of 14 qualitative reports of home visiting research.

Miles and Huberman’s approaches are highly systematic. The emphasis on data display assists in ensuring transparency, and the results of the synthesis are likely to be capable of being readily converted to quantitative variables. Software is available that can cope with this approach. However, Miles and Huberman’s emphasis on highly disciplined procedures is seen by some as unnecessarily and inappropriately stifling. Miles and Huberman offer no advice on sampling or appraisal of the primary papers.