A Systematic Review and Meta-synthesis of the Impact of Low Back Pain on People's Lives

Robert Froud; Sue Patterson; Sandra Eldridge; Clive Seale; Tamar Pincus; Dévan Rajendran; Christian Fossum; Martin Underwood


BMC Musculoskelet Disord. 2014;15(50) 

In This Article


Databases Searched and Inclusion and Exclusion Criteria

To identify qualitative studies of patients' experiences of chronic (i.e. ≥ 12 weeks) non-specific LBP, published in English peer-reviewed journals, from database inception until July, 2011, we searched CINAHL, Embase, PsycINFO, PEDro, and Medline.[21] Whilst PEDro primarily indexes clinical trials, systematic reviews, and guidelines, our scoping searches revealed it to index a nested qualitative study within a back pain trial, hence our inclusion of this database. We included all studies in which conclusions were based on qualitative data collected during face-to-face interviews or focus groups, reporting patients' experiences of chronic non-specific LBP. We excluded studies solely reporting on experiences of trial participation.

Search Strategy

We used terms that were developed from the Cochrane back review search strategy, scoping searches, and team discussion.[22] Our search strategy is detailed in full online (Additional file 1: Table S1). Briefly, it included variations on the following terms: 'low back pain, backache, lumbago, grounded theory, interview, focus group, phenomenology, action research, ethnographic, and epistemology'. Two reviewers (RF & SP) independently searched titles and abstracts, and full texts where necessary, to agree on included studies. Disagreements were resolved by an arbitrator (MU).

Appraisal of Included Papers

Two of four researchers (RF, SP, DR, or CF) independently assessed the reporting of each included study against the Consolidated Criteria for Reporting Qualitative Research (COREQ) framework.[23,24] Disagreements were resolved by arbitration by one of the remaining two researchers. RF and DR also abstracted data on described analytical approaches, individual study aims, and findings, for provision as an online resource.

Data Extraction and Analysis

We modified an approach described by Britten and colleagues, using Marston and King's approach to coding.[18,19] Britten developed a seven-step approach originally described by Noblit and Hare.[25] The first three steps encompass 1) 'getting started', i.e. identifying the research area of interest, what has previously been done, gaps in research, and forming a specific research question, e.g. What is the impact of non-specific low back pain on patients' lives? 2) 'deciding what is relevant', in terms of choosing material that is broad enough to enable the research question to be addressed, whilst not too disparate to raise questions regarding commensurability. In our case, we focused on any qualitative research featuring face-to-face discussions about the experience of non-specific low back pain; and then 3) 'careful reading and re-reading' of all included studies. The fourth step involves 'determining how the studies relate to each other', which was done by considering relationships between the concepts arising from the included papers; for example, work-based stigma being perceived by participants to be associated with the 'invisibility' of the condition, or a lack of credible diagnosis. A grid of concepts is formed during the fifth step and studies are 'translated into one another', which involves paraphrasing concepts in the grid cells – e.g. stigma, depression, trouble sleeping, etc. We modified the fourth and fifth steps using Marston and King's approach, which permits concepts and themes to be rigorously and systematically considered following coding, and the synthesis, once expressed, to be supported by typifying quotes and descriptions of individual study participants' responses (first-order constructs). For example, consider an original (first-order) quote presented in an included study under the theme of 'confrontation', stated "I remember the look in my managers eyes when I spoke of my back pain – I knew he didn't believe me". We might code this under a theme with the same name, and/or re-code it under different themes in-line with our research question, for example 'stigma', 'deligitimisation', or 'perceived disbelief'. Our coded themes were assigned to nodes, which facilitated their arrangement to form a framework; for example, a 'relationships' theme might comprise sub-themes of nodes containing coding for family relationships, social relationships, sexual relationships, and so on. We also separately coded authors' interpretations of the impact of LBP (second-order constructs), when these were reported.[26] We had access only to data presented by authors in the included study reports, and not to any raw data from these included studies. We did not seek validation of our coding from the original authors of our included studies.

RF thematically coded the first-order constructs within a thematic framework that was developed by RF, SP, and TP, following a preliminary analysis of selected studies. The coding framework was modified by RF as remaining studies were reviewed, with these modifications agreed with SP and TP. RF coded second-order constructs independently of the thematic framework. Nvivo 9 (QSR International, Victoria) was used to manage these data and their coding.

A second reviewer (SP) independently reviewed all included studies and analysed data using a meta-narrative approach described by Greenhalgh and colleagues, which involved identifying the key dimensions of the research problem, taking each dimension in turn and giving a narrative account of its contributions, before summarising the overall messages from the research literature.[20] The results of both processes were then compared and informed the sixth step of synthesis as described by Britten and colleagues, 'synthesising translations'.[18] This involved considering each of the concepts and second-order constructs in turn, the meta-narrative review, and whether these were refutational or reciprocal.[18] For example, if an interpretation in one study is that people with back pain strongly depend on close relationships for emotional support, yet an interpretation in another study is that people in pain seek solitude, distancing themselves from close relationships, one might consider these interpretations refutational. However, in the case a third study suggests that people with back pain depend on close relationships for emotional support generally, but isolate themselves from others during periods of intense pain, one might reason these interpretations are reciprocal, and go on to form an argument that back pain suffers generally depend on close relationships but seek solitude during episodes of relatively intense pain. RF, SP, MU, TP, CS, and SE all contributed to the synthesis following discussion and deliberation of coded constructs and the meta-narrative description. The seventh and final step is the expression of the synthesis, which we do in this paper. We describe first and second-order themes, as well as our third-order interpretation.