Barriers to following imaging guidelines for the treatment and management of patients with low-back pain in primary care: a qualitative assessment guided by the Theoretical Domains Framework

This was an exploratory, qualitative study describing family doctors’ experiences and practices related to diagnostic imaging for LBP. We used the Atkins et al.[26] guide for applying the TDF to assess barriers and enablers to behaviour change. A description of how we applied the TDF in our data collection and analysis is described below.

Eligible participants included family doctors practicing in NL who were treating patients with LBP. Purposive sampling was used to identify study participants. We identified family doctors through an established practice-based network (Atlantic Practice Based Research Network) – a group of clinicians who have agreed to be contacted to participate in primary care research. To help ensure a diverse range of perspectives, we sought participants practicing in both urban and rural environments, as well as both academic and community settings. We planned to recruit and interview 10–15 family doctors or until we reached data saturation – the point at which no novel information was being contributed by additional participants [33].

Following previously established methods outlined in the TDF Guide [26] and other studies using the TDF for barriers assessments, [34] data were collected using in-depth, semi-structured interviews. Potential participants were emailed an invitation to participant in a study investigating family physicians views on imaging for LBP by a local researcher (AH). If interested, they were provided with additional information about the study (including the reasons for the study and our interest in the topic) and an interview time was scheduled.

The interview guide (please see Additional file 1) included 1–4 questions per domain for a total of 31 questions. Prompts were provided in the interview guide to assist the interviewer in clarifying participants’ responses if needed.

Interviews were conducted by two female healthcare researchers (PhD and Master’s – prepared) with experience and training in qualitative interviewing (AH, AP). An undergraduate student (RL) also attended some interviews to take field notes. Otherwise, notes were taken by the interviewers. Interviews were conducted in-person (at the participant’s or interviewer’s place of work) or via telephone, whichever was private and most convenient for the participant. They took approximately 1h to complete, and participants were provided a CAD $100 gift card honorarium (paid using funds from the CIHR grant supporting this research). All interviews were audio-recorded and transcribed verbatim; participants were not provided opportunity to review their transcripts.

To begin, coders read and reread the transcripts to become familiar with the data; they began coding after we had completed and transcribed three interviews. Using the TDF to generate a framework for content analysis, researchers analyzed the data deductively (assigning text to one or more domains) and inductively (generating themes at each of the domains) [26].

Nine family doctors (four males; five females) participated in this study. At the time of the interviews these participants were working in community (n = 4) and academic (n = 5) clinics in both rural (n = 6) and urban (n = 3) settings. Seven of nine participants had no previous interactions with the interviewers; two had a previous working relationship with them. None of the participants we contacted refused participation or dropped out. We initially planned to recruit and interview up to 15 family doctors. However, after completing and coding eight interviews we felt that we had reached saturation. To test our assumption, we interviewed one additional participant (adding some additional prompts in an attempt to elicit new information). Since no novel information was added after this interview, we did not pursue additional participants. The team did not feel it necessary to complete any repeat interviews.

Interrater reliability at each domain ranged from a low of κ = 0.67 (SD = 0.32) to a high of κ = 0.92 (SD = 0.21) (with 9 of 14 domains reaching κ = 0. 75 or above) thereby demonstrating substantial to almost perfect agreement [35, 36]. However, while interrater reliability was calculated to help ensure consistency between the coders, all interviews were coded to consensus (100% agreement).

Our analysis revealed a number of barriers related to the following domains: 1) beliefs about consequences, 2) beliefs about capabilities, 3) emotion, 4) reinforcement, 5) environmental context and resources, 6) social influences, and 7) behavioural regulation. We generated 49 themes, phrased as belief statements, from the data coded at these domains. Table 1 presents the specific beliefs together with illustrative quotes for each of the relevant domains. Overall, five main barriers to reducing imaging for patients with NSLBP were evident in the data: 1) negative consequences, 2) patient demand 3) health system organization, 4) time, and 5) access to resources.

The remaining seven TDF domains were not considered relevant and included: 8) knowledge, 9) skills, 10) social professional role and identity, 11) optimism, 12) intentions, 13) goals, and 14) memory, attention and decision-making. The data coded at each of these domains revealed 34 specific beliefs, presented in Table 2 with illustrative quotes. Below we provide a brief summary of these beliefs which did not indicate the presence of any barriers to reducing imaging for patients with LBP.

All participants were aware of the behaviours specified by the guidelines for the appropriate management of NSLBP and most understood them to be evidence-based (knowledge). However, a few were not sure of the quality of the evidence supporting the guidelines. Likewise, most of our participants believed that family physicians require good clinical skills (e.g., history-taking, symptom recognition) – acquired over the course of their educational experiences – to manage low back pain without the use of imaging (skills). There was no evidence of disagreement among our participants and no conflicting or incorrect beliefs noted.

For the most part, the family doctors in our sample don’t feel that their ordering decisions are automatic but they also don’t struggle with those decisions. Most reported considering the patient’s history and physical exam findings when deciding whether or not an image is warranted. Other considerations include assessments for red flag conditions and surgical candidacy as well as response to previous treatments and resource stewardship (memory, attention, and decision-making).

All participants felt that managing patients with NSLBP without the use of imaging was a part of doing their job and meant they were practicing “good medicine” (social professional role and identity). Further, most reported that stewardship of health care resources was a part of their role as responsible practitioners and that avoiding low value care (such as imaging the low back) means they are practicing evidence-based medicine (social professional role and identity, goals). There was no evidence of discord or other salient information that leads us to suspect the existence of barriers related to this domain.

The majority of participants felt that, generally speaking, managing patients with NSLBP without imaging is a good idea (optimism). Further, most also reported wanting to manage their patients with NSLBP without imaging (goals) and their intention to manage their patients with NSLBP without imaging (intentions). There were no conflicts in these messages, however, two participants stressed that their priority is to ensure their patients’ well-being and not resource stewardship (goals).

Our findings are largely supported by previous studies that have investigated barriers and enablers to following guideline-recommended treatment and management of LBP in primary care. Two systematic reviews have assessed the literature focused specifically on physician-reported barriers to guideline-recommended imaging. In 2016, Slade et al., [30] performed a systematic review and meta-synthesis of qualitative studies investigating primary care physicians’ perspectives on clinical practice guidelines for LBP including barriers and enablers to their adherence. Building on this work, Hall et al., [31] completed a systematic review that included much of the same literature (plus two additional studies) and used a theoretical framework (the TDF) to guide the analysis. Several of our findings are in line with the results of these previous reviews. Briefly, social influence in the form of patient demand was an important factor in the decision to image for NSLBP, [30, 31] physicians felt imaging would ease patient anxiety and increase their satisfaction with care, [31] time is a barrier to reducing unnecessary imaging, [30, 31] and physicians imaged patients with NSLBP in part because they felt that it could act as a sort of “fail-safe” to protect them against missing a serious underlying condition [30].

There were, however, some important differences between the current investigation and existing literature. For example, unlike the Slade review [30], the participants in our investigation did not lack knowledge of guideline content and did not take issue with the credibility of clinical practice guidelines for LBP. Overall, they were also confident in their clinical skills to manage patients with NSLBP without the use of imaging. That being said, similar to findings noted in Hall’s review [31], the physicians we interviewed all reported some degree of struggle related to avoiding imaging for LBP. In many cases, they believe it is easier to order an image than negotiate with patients in order to avoid imaging. All physicians reported that they either a) don’t have the time to explain to patients that an image isn’t needed and/or b) struggle to convince patients who insist on imaging and/or c) believe that because some patients are reassured by imaging, it would damage the therapeutic relationship to deny it. Interestingly, a review including patient perspectives on imaging for low back pain found that some patients are, in fact, frustrated by imaging – particularly when the results are are inconclusive, don’t provide a clear reason for their back pain, or indicate degenerative or other issues perceived to be permanent or irreversible [37].

Our investigation also found an important barrier related to physician remuneration that has not been noted elsewhere in the literature on barriers to reducing imaging for LBP among family physicians. While physicians in NL are not remunerated for generating a referral, those we interviewed reported that the way they are remunerated (under a fee-for-service model) encourages imaging patients with NSLBP. They explained that under the current fee-for-service model in this province, they lose income by taking the time to explain why imaging is not necessary and counsel patients on alternative therapies for treatment. This finding underscores the importance of completing a context-specific barriers assessment before intervening to change behaviour.

A number of studies have now assessed barriers to following guidelines in the treatment and management of low back pain in a variety of settings in 10 countries and have identified similar barriers in these different contexts [31]. Our study revealed five important barriers to reducing unnecessary imaging, many of which align with the key barriers noted in previous reviews [30, 31]. Given the convergence of research in this area, it is unlikely that repeating barriers assessments in similar contexts will change these overall findings. Thus, any future research to examine barriers related to imaging guidelines for low back pain should consider focusing on areas that may add new and valuable information to the established knowledge base. For example, researchers may want to consider targeting particular contexts that have been under-studied in current literature such as resource-poor settings, or physicians with higher-than-average rates of imaging. Similarly, when an intervention to reduce imaging is being planned within a local context for which a full barriers assessment has not been conducted, we recommend using the existing literature as a reliable foundation to start selecting potential strategies and consider conducting 3–5 interviews with key contacts to confirm if there are any other issues that might be specific to the local context. These assessments would be undertaken not to generate new scientific knowledge but to optimize an intervention for a particular context.

Several interventions have been implemented to address over-imaging for low back pain [38‐42]. However, most have focused on providing information to clinicians to increase their knowledge of the guidelines and have been largely unsuccessful in changing clinician ordering practices [42]. While lack of knowledge has been identified as a barrier to evidence-based ordering in some contexts, [30] the bulk of the evidence in this area shows us that addressing only knowledge via strategies like practitioner education or passive dissemination of guidelines is not enough [38‐42]. Rather, moving evidence into practice for this behaviour will require more comprehensive interventions that address the most relevant barriers using behaviour change techniques matched to as many of the implicated domains as possible. Following guidance from the Medical Research Council (MRC), theory-based interventions should be tested in randomized controlled trials (RCTs) that include a robust process evaluation [43]. A randomized design is important to consider wherever possible since we don’t have a good understanding of other confounders related to behaviour change, making it difficult to adjust for them in non-randomized designs. Process evaluations allow us to explain how complex interventions work by examining the processes through which an intervention generates outcomes. When testing behaviour change interventions (BCIs), a process evaluation should be carried out to assess (i) fidelity and quality of implementation of the intervention, (ii) clarify causal mechanisms and (iii) identify contextual factors associated with variation in outcomes [43, 44]. These evaluations are vital for understanding how interventions function in different settings.

Next steps will involve using what is known about the barriers to reducing imaging to develop a comprehensive intervention to improve uptake of imaging guidelines for the treatment and management of NSLBP using the resources developed by Michie et al. to complement the TDF [27, 45, 46]. These include the Behaviour Change Wheel—a systematic guide for designing BCIs and the Behaviour Change Technique Taxonomy – an extensive list of 93 techniques, divided into 16 categories that will form the active components of an intervention [27, 45, 46]. Using this method, an intervention to reduce imaging for low back pain should include behavior change techniques that target domains directly related to specific barriers. Each barrier may be related to a number of different TDF domains and each domain is, in turn, associated with several BCTs. The theory and techniques tool (https://​theoryandtechniq​uetool.​humanbehaviourch​ange.​org) has been developed to help researchers map behavior change techniques to identified TDF domains [47‐49]. The tool is essentially a heat map in which each cell represents a link between a BCT and a TDF domain with colour-coding used to provide indication of the strength of the evidence link between them. The strength of the links was determined through a literature synthesis study that extracted data from 277 behavior change intervention articles, [47] an expert consensus study including 105 international behavior change experts, [48] and a triangulation study (statistically assessing concordance between the first two studies supplemented by a consensus exercise to reconcile discrepancies) [49]. Tool developers have also established a repository where new behavior change intervention study data can be uploaded and synthesized in order to help keep this tool up-to-date. As such, it is susceptible to change as the evidence base grows. Using the theory and techniques tool, we have included an example of how BCTs could be used to build an intervention targeting barriers to reducing imaging for LBP in Table 3. For the purposes of this example, we have focused on only one barrier.

We will complement this process by also involving family physicians and other relevant stakeholders who can advise the research team on the acceptability and plausibility of successfully implementing the intervention. This intervention will be tested in a RCT that includes both a rigorous process evaluation (to identify causal mechanisms) and fidelity assessment (to determine the extent to which the intervention was implemented as intended).

As recommended by a host of national health and research organizations (e.g., the National Institute for Healthcare Excellence, the MRC, Health Canada, Canadian Institutes of Health Research, and the Quality Enhancement Research Initiative), [43, 50‐53] this study used a theory-informed approach to investigate the barriers and enablers of unnecessary imaging for NSLBP. While both physician and patient-related barriers to reducing imaging for NSLBP have been noted elsewhere in the literature, [30, 31] only one study that we are aware of used a theoretical framework to guide their assessment [54]. Our study was designed using the Atkins et al.guide [26] on how to apply the TDF to the development of data collection and analysis methods for assessing barriers and enablers to behaviour change. Using this approach allowed us to produce results that can be used to guide the development of an intervention that will appropriately tackle barriers related to each of the TDF domains.

Additionally, we used a rigorous approach to data analysis as recommended by Atkins et al. [26] in their guide to applying the TDF to barriers assessments which included double coding the transcripts, training coders extensively, oversight of analysis by a professional expert in the TDF, review of the results by a family physician team member in detail and again by the larger investigative team. The team also used the Consolidated Criteria for Reporting Qualitative Research (COREQ) 32-item checklist to guide our methods and reporting [55].

Despite these strengths, the results are limited in important ways. For example, our sample size was small and included only nine family physicians. While we did assess data saturation and found no new information after the eighth interview, it is possible that we reached data saturation prematurely by not interviewing participants with sufficient diversity to allow for more variety in responses. For example, although we sampled purposively to ensure that we interviewed rural and urban, community and academic, male and female physicians at varying stages of their professional practice, we were limited to a convenience sample of those who agreed to participate. We also didn’t assess the ordering rates of our sample to engage similar numbers of participants that order images at different rates or actively seek participants in equal numbers who had differing views on imaging conservatively.