Explaining variation in GP referral rates for x-rays for back pain

Background Despite the availability of clinical guidelines for the management of low back pain (LBP), there continues to be wide variation in general practitioners' (GPs') referral rates for lumbar spine x-ray (LSX). This study aims to explain variation in GPs' referral rates for LSX from their accounts of the management of patients with low back pain. Methods Qualitative, semi-structured interviews with 29 GPs with high and low referral rates for LSX in North East England. Thematic analysis used constant comparative techniques. Results Common and divergent themes were identified among high- and low-users of LSX. Themes that were similar in both groups included an awareness of current guidelines for the use of LSX for patients with LBP and the pressure from patients and institutional factors to order a LSX. Differentiating themes for the high-user group included: a belief that LSX provides reassurance to patients that can outweigh risks, pessimism about the management options for LBP, and a belief that denying LSX would adversely affect doctor-patient relationships. Two specific differentiating themes are considered in more depth: GPs' awareness of their use of lumbar spine radiology relative to others, and the perceived risks associated with LSX radiation. Conclusion Several key factors differentiate the accounts of GPs who have high and low rates of referral for LSX, even though they are aware of clinical guideline recommendations. Intervention studies that aim to increase adherence to guideline recommendations on the use of LSX by changing the ordering behaviour of practitioners in primary care should focus on these factors.


Background
Low back pain (LBP) is a global and increasing problem [1]. Estimates of point prevalence range between 12% and 35% and lifetime prevalence between 49% and 80% [2]. The cost of LBP in the United Kingdom is high, both to the NHS (National Health Service) and in terms of the wider societal costs [2,3]. While there are a number of serious conditions that cause LBP, most LBP is non-specific, benign, and self limiting, although it may become recurrent [4,5]. Non-specific LBP is classified by the duration of symptoms into acute (less than 3-6 weeks), sub-acute (less than three months) and chronic (more than three months) [6][7][8]. One distinction between the acute and chronic forms of non-specific LBP is that the prognosis for the former is reported to be generally good [9]. For most people with acute LBP, symptoms rapidly improve within one month and continue to do so for up to three months, but from then on any residual symptoms of pain and disability remain constant [6,10]. The majority will experience at least one recurrent episode in the subsequent 12 months, and around 5% of those with acute lower back pain will develop chronic LBP [7].
The relationship between x-ray findings and non-specific LBP is unclear [11,12], and despite a link between radiological findings of degenerative disorders and LBP [11][12][13] such findings have little implication for the management of LBP. Recent trials have shown that radiography for primary health care patients with LBP has no effect on health outcomes, although patient satisfaction is higher [14,15]. Lumbar spine x-ray is associated with a dose of ionising radiation equivalent to approximately 65 chest x-rays [16]. Therefore, unnecessary examinations should be avoided.
The use of lumbar spine x-ray (LSX) in cases of LBP is not routinely indicated. Degenerative changes commonly detected by LSX are non-specific, and the main value of LSX, according to guidelines from the Royal College of Radiologists, is for young people (<20) where spondylolisthesis or ankylosing spondylitis are a concern, or for those older than 55. In the presence of specific 'red flag' symptoms, such as sphincter or gait disturbance, or widespread neurological deficit, Magnetic Resonance Imaging (MRI) is the preferred investigation [16]. Despite clinical guidelines [16] there is wide variation in the use of plain film x-rays for patients with LBP from primary care [17,18], and many requests do not conform to guideline recommendations [19,20].
Studies aiming to find reasons for GPs requesting LSX for patients with LBP have relied on quantitative methods (e.g., [21][22][23]) and have shown that GP requests for LSX derive from both health and non-health factors. Maintaining patient satisfaction with their care and offering reassurance to patients are often important. These studies, whilst informative, have not explored in-depth the rationales for referral behaviour. Two qualitative studies have been identified, both using focus group methods. The first, undertaken in the United States [24]., sought to explain the negative findings of an intervention study to improve referral practice. The second, a Norwegian study, identified factors affecting decisions to order spine radiography focussing on the barriers to guideline adherence [25]. No UK studies that seek to understand GP referral behaviour and explain different LSX referral practices for LBP have been found. This study aims to investigate reasons for GP referral for LSX for patients presenting with LBP and explain observed differences in referral rates. It is based on the premise that GPs with high and low referral rates for LSX will give different accounts of their perceptions, experience and management of LBP and LSX.

Methods
The study was conducted among GPs in the North East of England during 2000. The number of LSXs requested by individual GPs in the preceding year was obtained from radiology departments in three hospitals. Absolute fre-quencies of LSX requests were adjusted to take account of working hours of GPs and list sizes, and the sample was rank-ordered according to the adjusted referral frequency. GPs were sampled sequentially from the high and low ends of the distribution. They were contacted by letter and subsequently by telephone inviting them to take part in a single interview lasting between 60 and 90 minutes and were reimbursed for their time at locum rates. Recruitment continued until categories were saturated. In addition, five GPs in research practices were interviewed first and provided comments on the interview and topic guide. No significant revisions were made as a result, and they are included in the analysis. Of 55 GPs who were invited to take part, 29 (53%) agreed and their characteristics are described in Table 1. Twenty six chose to be interviewed at their practice premises, two chose to be interviewed at home, and one at the university.
Interviews were conducted by RB (researcher), who remained blind during the interview to whether the GPs were categorised as high-or low-users of LSX (from radiology records). Interviews were informed by a topic guide, and were tape recorded and transcribed verbatim. Field notes [26] were dictated to tape after interviews and were transcribed. Emerging concepts and themes were recorded in a research diary [27]. The topic guide comprised four main sections: Section 1 was concerned with GP practice and locality characteristics; Section 2 aimed to elicit infor- mation about GPs' perceptions of patients presenting with LBP; and Section 3 focussed on how GPs handled specific cases of LBP, and they were asked in advance to retain the case notes of recent patients to discuss their histories, consultations, and decisions made. This case-based approach allowed probing of actual decisions to refer for LSX or choosing other courses of action. In Section 4, GPs were asked specifically about their beliefs and attitudes towards the use of LSXs. The order of topics in the interviews varied depending on spontaneous discussion and the emphasis accorded to topics by respondents. The topic guide was amended throughout the study as new themes were identified using constant comparative analysis.
A sub-sample of five transcripts was fully coded by two researchers (RB, SB) to develop an initial coding frame and to identify concepts and themes. Subsequent themes were debated and agreed upon by all three authors. Drawing on principles of constant comparison [28], the development of the coding frame and emergent themes were subject to deviant case analysis. NVivo qualitative analysis software [29] was used to index and interrogate the data. Themes that were consensual across high-and low-users of LSX were classified as 'convergent themes' and views which differentiated the two groups were classified as 'divergent themes.' Ethical approval to conduct the study was obtained from local research ethics committees in Newcastle and North Tyneside, Gateshead and North Tees. All interviewees were assured anonymity in any reports or publications of the findings.

Convergent themes
A number of convergent themes that related to the decision to request an LSX and showed no pattern of association with high-or low-users of LSX were identified. These were broadly categorised into three groups: 'clinical,' 'psycho-social,' and 'institutional' factors. GPs were knowledgeable about the existence and general thrust of clinical guidelines for the management of LBP, could articulate their main messages, and did not challenge their content. Some discussed general issues around the problems with adherence to guidelines in a clinical context, but these GPs did not take issue with the specific recommendations of either the Royal College of General Practitioners back pain guidelines [30] or the Royal College of Radiologist guidelines [16]. An x-ray which was deemed to be 'negative' (in terms of excluding certain diagnoses) was seen by some respondents as providing reassurance to patients, although this was qualified by some respondents who acknowledged that clinically insignificant findings on x-ray may raise anxiety due to continued uncertainty, rather than reduce it. Patient anxiety over LBP as a symptom of something more serious was an important influence, and LSXs were used to allay fears, particularly in relation to cancer. GPs were also influenced by the social and economic issues of importance to people with LBP: In summary, both high-and low-users of LSX were aware of the guidelines with respect to LBP and LSX, spoke about the pressures created by waiting times for secondary care or MRI, experienced patient pressure for something to be done and their anxieties (especially about LBP as heralding cancer), and were aware of social and socio-economic factors and the diagnostic limitations of LSX.

Divergent themes
Divergent themes were those more prevalent or given much higher degree of emphasis in either the high-or low-referral group.
Obtaining a 'negative' LSX result as a legitimate means to reassure patients that their LBP did not stem from serious pathology was a view more strongly associated with the accounts of high-user GPs.
We are well aware of the College  The need to preserve the doctor-patient relationship was an important factor for some GPs and was at times influential in their decisions to investigate and manage LBP. In the high-user group there was greater emphasis on the fragility of this relationship and concern over the detrimental consequences of it breaking down. LSX was used, at times, to meet patient demand in decisions that were inappropriate in strict clinical terms, in order to preserve relationships with patients.

But at the end of the day you give in, cause it's not worth it, it's not worth losing a patient doctor relationship unless you want to. GP 24 High.
Low-users of LSX were similarly aware of the threat to the doctor-patient relationship by refusing to x-ray. They knew that patients who were not satisfied might 'shop around' to find a GP willing to comply. However, they were prepared to face these eventualities as part of the give and take of what they regarded as 'being a good GP.'

Concern about radiation
Concern about exposure to radiation was emphasised, often in strong terms, by most of the GPs in the low-referral group.

I am very anti X-ray. I see that X-ray has a dose of radiation associated [with it] and with back X-rays gives out a significant amount of radiation, and really I think a lot of people, particularly in casualty departments and GPs, don't take that into account and it's really important. GP06, low.
GPs who referred more patients for LSX were far less concerned about exposure to radiation. They explained their lack of concern in three ways. Firstly, the absolute dose of radiation was perceived to be minimal compared with other procedures such as computed tomography (CT) or barium studies. Secondly, the perceived benefits of the xray, including patient and GP reassurance, greatly outweighed the perceived radiation risks, and thirdly, relative to other treatment decision risks such as drug prescribing, the risk of exposing a patient to LSX radiation was regarded as small.
Such reasoning was couched in terms of their own experience and the absence of visible ill effects of x-rays on patients or NHS staff.

Knowledge of own use of x-ray
GPs from the low group were aware of and judged their relative use of LSX more accurately than those in the high group. Low-users had developed other ways of managing patients with LBP, and this was likely to be a feature of the practice as well as of individual GPs.
I rarely do lumbar x-rays but we don't, probably haven't done two in the last year, and one of those was as the request of the physio. GP16, low.
Nine of the 14 GPs who were relatively high-users of LSX stated that they were either unsure or were low-users of LSX. Despite their relatively frequent use of LSXs, they often perceived their use as infrequent.
I don't x-ray people's backs much at all really because they're not much use are they really -x-rays. GP27, high.

Discussion
Variation in the use of LSX in cases of LBP in primary care remains a problem. Before interventions can be designed to change behaviour and reduce variation there needs to be an in-depth understanding of current behaviour [31]. This study identified a number of themes which are common to, and which distinguish the accounts of, GPs with high and low referral rates for LSX, enabling a better understanding of GPs' decision making in this context. In this analysis we have created a picture of an archetypal high LSX user and have identified different exemplary characteristics of a low LSX user through GPs' own descriptions of their practices and beliefs.
There are, however, some limitations to this study that should be acknowledged. Since these interviews were carried out there have been changes in the organisation of UK primary health care, e.g., Practice-Based Commissioning (PBC), which may or may not have an impact on these findings. Whilst it is unlikely that initiatives such as PBC will have an effect on GPs' perceptions of LBP, clinical guidelines, LSX and their patients, such as are reported in this study, institutional and structural changes in the NHS may affect GPs' views of the treatment options available to them. This study cannot offer any insight into the possible effects of these recent changes. In addition, a purposive sampling strategy was used and respondents were selected on the basis of their levels of LSX use. As such, the sample is not representative and nothing can be said about the distribution of these themes in the population.
In terms of the application of our findings, two principal divergent themes have been identified that could be targeted in specific and simple interventions: firstly, GPs in the high-use group perceived their own use of LSX to be relatively low, and secondly, they had low levels of concern about the risks of radiation associated with LSX. That GPs are unaware of their relatively high use of x-ray suggests that informing them of their use relative to others may reduce the number of LSXs amongst this high group. However, intervention studies investigating the effect of feedback on referral rates for LSX have shown this strategy to be ineffective in changing behaviour [18,32].
The differing concerns between high-and low-users of LSX surrounding radiation risk may be of greater relevance to the design of future intervention studies. The importance of this theme also is substantiated by a recent qualitative study of the prescription of new drugs by GPs [33]. Using a similar comparative design the authors report 'attitudes to risk perception and benefit' amongst the key dimensions that classify high-and low-prescribers of new drugs and state that, "High-prescribers were more inclined to underplay the risks of new drugs, so freeing them to prescribe a new drug they believed offered therapeutic effectiveness.." p590. This study suggests that GPs who request relatively frequent x-rays for LBP differ from their colleagues in their assessment of radiation risk. It is possible, therefore, that these GPs may change their behaviour in response to information about the radiation risks associated with LSX, and revisit their assessment of the costs and benefits of requesting such examinations. This hypothesis provides an avenue for future research.