Significant progress has been made in analysing indexing systems of databases for qualitative studies. The Hedges Project at McMaster University has expanded its coverage of empirically-tested methodological filters to include qualitative research filters for MEDLINE (Wong 2004), CINAHL (Wilczynski 2007), PsycINFO (McKibbon 2006) and EMBASE (Walters 2006). Nevertheless evidence from qualitative studies collected and reported within randomized trials or as part of linked studies are difficult to retrieve (Evans 2002). MEDLINE introduced the MeSH term ‘qualitative research’ only in 2003. CINAHL introduced ‘Qualitative Studies’ in 1988, reflecting particular interest in qualitative studies for nursing researchers, with a corresponding focus on ‘quality of life’ issues (see Chapter 17, Section 17.3). However, locating qualitative studies remains problematic because of the varied use of the term ‘qualitative’ (Grant 2004).
In addition, current strategies for indexing terms related to qualitative study designs and protocol-driven search strategies are only of limited value (Evans 2002, Barroso 2003, Greenhalgh 2005). Review authors must be aware that limiting a search to well-known databases may result in missing much useful information. An audit of sources for a review of complex interventions (including qualitative evidence) found that only 30% were identified from databases and hand searches. About half of studies were identified by ‘snowballing’ and another 24% by personal knowledge or personal contact (Greenhalgh 2005). Search strategies to identify qualitative studies using a range of different qualitative methods need to be further developed.
While there is general agreement on the need for search strategies aiming to identify qualitative research to be systematic and explicit, there is recent debate on whether qualitative evidence syntheses share the need for comprehensive, exhaustive searches. It has been argued that a more purposive sampling approach, aiming to provide a holistic interpretation of a phenomenon, where the extent of searching is driven by the need to reach theoretical saturation and the identification of the ‘disconfirming case’ may be more appropriate (Dixon-Woods 2006). Nevertheless this places an even greater imperative to improve quality of reporting standards of search methods (Booth 2006).