Clinician, patient and general public beliefs about diagnostic imaging for low back pain: protocol for a qualitative evidence synthesis

Abstract

Introduction

Little is known about how to reduce unnecessary imaging for low back pain. Understanding clinician, patient and general public beliefs about imaging is critical to developing strategies to reduce overuse.

Objective

To synthesise qualitative research that has explored clinician, patient or general public beliefs about diagnostic imaging for low back pain.

Methods and analysis

We will perform a qualitative evidence synthesis of relevant qualitative research exploring clinician, patient and general public beliefs about diagnostic imaging for low back pain. Exclusions will be studies not using qualitative methods and studies not published in English. Studies will be identified using sensitive search strategies in MEDLINE, EMBASE, CINAHL, AMED and PsycINFO. Two reviewers will independently apply inclusion and exclusion criteria, extract data, and use the Critical Appraisal Skills Programme quality assessment tool to assess the quality of included studies. To synthesise the data we will use a narrative synthesis approach that involves developing a theoretical model, conducting a preliminary synthesis, exploring relations in the data, and providing a structured summary. We will code the data using NVivo. At least two reviewers will independently apply the thematic framework to extracted data. Confidence in synthesis findings will be evaluated using the GRADE Confidence in the Evidence from Reviews of Qualitative Research tool.

Ethics and dissemination

Ethical approval is not required to conduct this review. We will publish the results in a peer-reviewed journal.

PROSPERO registration number

CRD42017076047.

Strengths and limitations of this study.

• We will use systematic search strategies to identify relevant qualitative research.

• A priori methods will reduce bias related to searching, data extraction, study quality assessment and rating of confidence in synthesis findings.

• At least two authors will independently develop and refine a thematic framework to code data within included studies.

• Synthesis is limited to quotes and themes explored in the original reports.

Introduction

Low back pain is the leading cause of disability worldwide, and is responsible for an estimated 83 million years lived with disability per year.¹ In 2013, healthcare-related costs of low back pain were estimated to be $A5 billion in Australia² and US$87.6 billion ($A110.5 billion) in the USA.³ Much of this cost is related to unnecessary tests and treatments. For example, diagnostic imaging is an unnecessary test for the vast majority of patients with low back pain who present to primary care, yet in 2014, Australians spent $A214 million on lumbar radiographs.⁴ Internationally, rates of guideline-discordant diagnostic imaging across primary, secondary and emergency department settings have been reported to be as high as 55%.^5–12

Unnecessary diagnostic imaging is associated with substantial harm including the risk of overdiagnosis.¹³ Overdiagnosis occurs when diagnostic imaging detects incidental findings that are common in the asymptomatic population (eg, intervertebral disc degeneration¹⁴) and provides the patient with a diagnostic label that brings them no benefit or causes harm. Diagnostic labelling leads to medical overuse, a problem which is growing internationally.¹⁵ Unnecessary diagnostic imaging for low back pain drives flow-on effects such as overuse of advanced imaging, opioid prescriptions, spinal injections and surgery.^{16 17} Evidence from clinical guidelines suggests that most of these interventions have little to no benefit, and substantial risk for harms, in patients with non-specific low back pain.^18–20

Little is known about how to reduce unnecessary imaging. A recent systematic review found that of the five trials located on strategies to reduce imaging in primary care, only one was successful.²¹ That trial found evidence that distribution of radiological guidelines plus audit and feedback on practice imaging rates reduced spinal imaging by 20% (95% CI 3%–37%) compared with control.²² We are aware of one other successful trial in primary care. In 1986, Deyo and colleagues tested an approach that combined patient education with a delayed referral for imaging.²³ Though it was not the primary outcome for the trial, the authors found the delayed approach substantially reduced imaging rates (control imaging rate 83.9% vs intervention imaging rate 27.8%).²³ These results have yet to be replicated.

There is robust evidence that a delayed approach reduces unnecessary antibiotic prescriptions for acute respiratory infections.²⁴ Such an approach is promising for low back pain because, like many acute respiratory infections, natural recovery often occurs in the first 2 weeks.²⁵ A delayed referral for imaging would see many patients improve before they made the final decision to pursue imaging or not. Clinical guidelines endorse a delayed approach for most patients with low back pain with the exception of suspected cauda equina syndrome or vertebral infection.¹⁸ Survey data from the trial by Deyo et al suggested that delaying imaging was acceptable to patients.²³ However, such data provide little insight into the key, underlying beliefs about diagnostic imaging that could be gained, for example, from a qualitative study design.

Understanding clinician, patient and general public beliefs is an important step towards developing an acceptable delayed imaging approach. Although some clinician-level barriers to providing appropriate imaging for low back pain have been identified,²⁶ it is not known whether the same barriers would impact on a decision to delay a request for imaging. For example, while clinicians tend to believe patients expect imaging,²⁶ it is unclear how clinicians might perceive the benefits and harms of delaying, or avoiding, a referral. A previous qualitative synthesis found patients with chronic musculoskeletal pain value a diagnosis, but did not explore what participants believed about the role of diagnostic imaging.²⁷ Similarly, surveys suggest around 50% of patients and the general public expect diagnostic imaging to manage low back pain,^28–30 but it is unclear what people believe diagnostic imaging can actually provide; such beliefs could determine the acceptability of a delayed approach. For example, patients and the general public may be unwilling to accept a delay if they believe it will preclude effective diagnosis and treatment.³¹

We are not aware of any up-to-date qualitative evidence syntheses that could inform the use of the delayed imaging strategy for low back pain.

Aim

To synthesise qualitative research that has explored clinician, patient or general public beliefs about diagnostic imaging for low back pain.

Methods

We have registered this study on PROSPERO (CRD42017076047).

Selection criteria

We will include English language articles that fulfil the following criteria:

Types of studies

Used qualitative methods for data collection (eg, focus groups, interviews) and analysis. All qualitative designs will be included and studies that included mixed methods (qualitative and quantitative), where the qualitative data were collected and analysed independently to the quantitative data, will also be included.

Types of participants

Clinicians who treat low back pain (eg, general practitioner, rheumatologist, spine surgeon, physiotherapist, chiropractor, osteopath, and so on), patients with low back pain of any duration, or the general public interviewed about management of low back pain. Participants do not necessarily need to have received diagnostic imaging.

Types of settings

Any healthcare (primary, secondary or tertiary care) or non-healthcare setting in any country.

Types of imaging

Any diagnostic imaging (eg, X-ray, CT scan, MRI, and so on).

Types of outcome measures (findings on phenomena of interest)

Analysis describes beliefs about diagnostic imaging for low back pain. Beliefs are defined as ‘a proposition or a set of propositions held true.’³² We are interested in all beliefs about diagnostic imaging, not just those related to delaying or avoiding the test.

Search methods

We will develop a search strategy with the assistance of an Information Specialist based at the University of Sydney Library, and by using guidance from the Cochrane Qualitative and Implementation Methods Group (http://methods.cochrane.org/qi/supplemental-handbook-guidance). To locate relevant articles, we will search the following electronic databases: MEDLINE, EMBASE, CINAHL, AMED and PsycINFO. We will not use date limits, but will limit the search to articles in English. Sampling will be comprehensive rather than purposive; comprehensive searching will allow us to locate all relevant studies representing the phenomenon of interest (ie, beliefs about imaging for low back pain). We will perform the search in two parts as suggested by DeJean et al.³³ Part 1 aims to identify all studies on our topic of interest. The search strategy for part 1 will combine terms recommended by the Cochrane Back and Neck Review Group to identify studies of low back pain, terms describing the participants of interest (clinicians, patients, general public) and terms describing the phenomenon of interest (beliefs). Because the term ‘beliefs’ can be used interchangeably with others such as ‘views’, we will develop a list of terms that are conceptually related to beliefs. For example, we will include broad terms such as ‘perceptions,’ ‘feelings,’ ‘views’ and ‘understandings’ (online supplementary appendix). Part 2 will aim to identify all studies using qualitative methodology. Where available we will use qualitative filters to maximise sensitivity and specificity of searches to locate qualitative studies in part 2, and adapt these to the other electronic databases (EMBASE, CINAHL, PsycINFO).^33–35 To obtain final search results we will combine parts 1 and 2 using the syntax ‘AND’. An example search strategy for MEDLINE is provided in the online supplementary appendix. We will also search the reference lists and conduct citation tracking of included studies using Web of Science. To identify other potentially relevant studies we will contact experts in the field through existing professional networks of the review authors.

Selection of studies

One reviewer will download search results to reference manager software and delete duplicates. To detect duplicates we will use the Systematic Review Assistant-Deduplication Module which has been shown to improve efficiency of the deduplication process.³⁶ Two reviewers will screen titles and abstracts independently and perform full-text review as necessary. Disagreements about which studies to include will be resolved through discussion. If consensus cannot be reached an additional reviewer will make the final decision. The search and screening results will be presented in the form of a flow diagram as recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist.³⁷

Assessment of quality of included studies

There is currently no consensus on standard criteria to be used to assess the methodological quality of qualitative studies. For this qualitative evidence synthesis we will use the Critical Appraisal Skills Programme (CASP) quality assessment tool for qualitative studies.³⁸ Two reviewers will independently apply the CASP tool to the included studies. The tool evaluates rigour, credibility and relevance of qualitative research. Disagreements will be resolved by consensus or by discussion with an additional reviewer. Studies will not be excluded on the basis of quality but quality assessments will be used to determine confidence in the synthesis findings.

Data extraction

Two reviewers will use a standard form to extract descriptive data from included studies (study details, design, sample details, and so on) and independently perform data extraction from 50% of the included studies. A third reviewer will extract descriptive data from all included studies and cross-check. Disagreements will be resolved by consensus or by discussion with a fourth reviewer. The descriptive data to extract will include:

• Study details (authors, year of publication).

• Research question.

• Design (phenomenological, thematic analysis, grounded theory, case study, ethnography, mixed methods).

• Participants (demographic characteristics, number, study location, history of low back pain and/or care seeking for low back pain, professional involved or professional background (eg, general practitioner, rheumatologist, spine surgeon, physiotherapist, chiropractor, osteopath, and so on)).

• Setting (country; urban/rural; primary, secondary, tertiary care).

• Diagnostic imaging type (eg, X-ray, CT, MRI, other, not specified).

• Method(s) (eg, focus group, interview) and timing in relation to seeking care (for patients with low back pain) of data collection and analysis, including if any theory or framework was used.

• Study findings. These will include themes, subthemes, discussion points and conclusions about imaging beliefs based on the original study authors’ interpretation of the first order constructs (ie, ‘second order constructs’).³⁹ This will be supported by specific beliefs expressed by study participants in the primary study in the form of quotes (ie, ‘first order constructs’).³⁹ To extract data on study findings, two reviewers will constantly update their own extraction sheets to include new headings (ie, themes) as they emerge. This process forms part of the thematic analysis described in more detail in the section below.

Data analysis and synthesis

We plan to synthesise findings using a narrative approach.⁴⁰ We chose a narrative synthesis method described by Popay et al⁴⁰ because of its suitability to the mixed methods expertise of our review team, the type of data we are seeking (information from a large number of studies, rather than a small number of conceptually rich studies), the potential relevance of aggregative synthesis results to policymakers, clinicians and commissioners of research, the comprehensive sampling technique, and the realist epistemological position.⁴¹ The synthesis will involve four steps conducted in an iterative process rather than sequentially⁴⁰:

1. Developing a theoretical model.

2. Developing a preliminary synthesis.

3. Exploring relationships in the data and emerging themes.

4. Providing a structured summary and assessing the robustness of review findings.

Developing a theoretical model

We plan to develop our theoretical model iteratively while performing open coding of the data, using constant comparison between existing themes and new data to incorporate new themes as they emerge. We will consider applying specific theories of behaviour change to contextualise our synthesis findings at the final stage of analysis, if they reflect the data adequately. However, we will not constrain the initial coding and synthesis procedures with an overarching existing theory.⁴⁰

Developing a preliminary synthesis

We will perform a preliminary synthesis of the findings by reading a diverse subset of included studies in detail and identifying the themes that emerge. To choose the subset of studies we will use purposive sampling based on key study features likely to influence beliefs, such as participants involved (clinicians, patients, general public), timing of interviews in relation to care seeking, healthcare setting, and duration of symptoms.

Two review authors will code findings from the same subset of studies and develop a preliminary list of themes. The review team will discuss the preliminary list of themes and revise the list in response to new themes that emerge in the remaining studies. We will tabulate this final list of themes and use it to organise findings from all studies. We will use NVivo software to facilitate coding processes. The final list of themes and accompanying data will be agreed on by discussion and consensus among the review team.

We will tabulate the full-text papers according to the presence or absence of key themes, with supporting data where available (eg, text extracts and participant quotes; table 1). We acknowledge that a theme occurring in more than one study does not provide evidence that the theme is important or common. However, we consider the number studies supporting a theme to contribute to the certainty that the theme is relevant to the phenomenon we are interested in.

Table 1.

Template for organising full-text studies into themes

	Theme 1	Theme 2	Theme 3
Study (date)	Supports theme/does not support theme/no data

Exploring relationships in the data

We will use two methods to explore relationships in the data: examining consistency of themes and idea webbing.⁴² In this step the ‘data’ will be second order constructs—themes, subthemes, discussion points, conclusions and selected participant quotes about imaging beliefs based on the original study authors’ interpretation of views expressed by participants. To examine consistency of themes identified in the preliminary analysis we will explore data extracted from studies with different characteristics such as healthcare setting (eg, primary, secondary, tertiary care), timing of interviews and duration of pain (ie, acute: <3 months, chronic: 3 months or longer). To further explore relationships in the data we will use idea webbing. Idea webbing maps major themes and subthemes using web diagrams to identify cross-cutting, contradictory or overlapping themes.⁴² Two reviewers will independently develop their own idea webs and decide on a final diagram after discussion with the review team.

Providing a structured summary

We will provide a structured summary of synthesis findings in the form of a table (table 2). The table will include a summary of the robustness of the synthesis findings according to the GRADE Confidence in the Evidence from Reviews of Qualitative Research (CERQual) approach.⁴³ The GRADE CERQual approach is described further below.

Table 2.

Template for summary of qualitative findings

Review finding	Studies contributing to review finding	Assessment of methodological limitations	Assessment of relevance	Assessment of coherence	Assessment of adequacy	Overall GRADE CERQual assessment of confidence	Explanation of judgement
Finding 1

CERQual, Confidence in the Evidence from Reviews of Qualitative Research.

Assessment of confidence in review findings

We will use the CERQual tool to assess confidence in our findings.⁴³ This tool considers four factors that influence the confidence in syntheses of qualitative studies: (1) methodological limitations of the studies contributing to the review finding; (2) the relevance of contributing studies to the review finding; (3) the coherence of data contributing to the review finding; and (4) the adequacy of data contributing to the review finding.

Methods on how to operationalise the CERQual items are still in development.⁴³ Below we describe how we plan to operationalise these items for this review. Criteria will be used as a guide only; at least two independent reviewers will make a subjective judgement about confidence in each review finding. Reviewers will judge overall confidence in a review finding as high, moderate, low, or very low, with a justification for this rating (box 1). Each finding will begin with a rating of high confidence; reviewers will downgrade if there are concerns with any of the four CERQual criteria. A final decision on confidence in review findings will be reached through discussion and consensus among the review team. To assist with the GRADE CERQual approach we will use draft worksheets that the GRADE CERQual Working Group have provided to us, on request, via http://www.cerqual.org.

Box 1. GRADE Confidence in the Evidence from Reviews of Qualitative Research (CERQual) approach⁴³ to confidence in findings of qualitative evidence synthesis.

High confidence: highly likely that the review finding is a reasonable representation of the phenomenon of interest.

Moderate confidence: likely that the review finding is a reasonable representation of the phenomenon of interest.

Low confidence: possible that the review finding is a reasonable representation of the phenomenon of interest.

Very low confidence: unclear whether the review finding is a reasonable representation of the phenomenon of interest.

Methodological limitations of the studies contributing to the review finding

To examine item 1 we will use the CASP tool to consider the quality of studies contributing to the finding. If many of the contributing studies have important problems with design or conduct we will consider that to raise ‘major concerns’ about methodological limitations.

Relevance of contributing studies to review finding

To examine item 2 we will consider factors such as healthcare setting, duration of low back pain (acute or chronic), history of care seeking, timing of interviews (before or after consultation in care seekers), professional background (for clinician participants), socioeconomic background, and reimbursement system for imaging (publicly subsidised or free, privately funded). If studies contributing to review finding are based on data that have contextual features that do not relate to our review question, or to our objective to develop strategies to reduce overuse of imaging, we will consider that to raise major concerns about relevance.

Coherence of data contributing to review finding

To examine item 3 we will create a table to depict the number of studies that contributed to a given finding. Rows will contain primary studies, columns will contain higher order themes and subthemes (table 1). If some of the data supporting a review finding are ambiguous, or contradict the review finding, we will consider that to raise major concerns about coherence.

Adequacy of data contributing to review finding

To examine item 4 we will consider the quantity and ‘richness’ of the data supporting a review finding. We consider rich descriptions to be those which include detailed narrative descriptions, such as quotes that allow the reader to visualise the participant–researcher interaction and provide sufficient detail to support or refute the phenomenon of interest. If most studies lack rich narrative descriptions, we will consider that to raise major concerns about adequacy of data.

Reporting

We will report this study in accordance with the Enhancing Transparency in Reporting the synthesis of Qualitative research (ENTREQ) statement.⁴⁴

Discussion

Here we present the design of a qualitative evidence synthesis to explore beliefs about diagnostic imaging for low back pain. We anticipate this review will provide a robust picture on how healthcare professionals, patients and society might view the utility of spinal imaging. Understanding such beliefs is critical to determining the feasibility of strategies aimed at reducing unnecessary use, including those that encourage a delayed referral.