Design

This study is a cross sectional survey of chiropractors in Australia.

Sample size and participants

The sample size was calculated to address the primary aim of the survey which was to identify factors (e.g. knowledge) that were predictive of intention to perform a particular behaviour, and then to use this information to develop the ALIGN implementation strategy. The sample size was calculated to detect a 0.5 difference in intention to perform a particular behaviour, measured on a 7-point Likert scale, between dichotomised factors (assuming an equal distribution of participants in each dichotomy), with 90% power. To detect this difference, a sample of 440 chiropractors was required, assuming a 5% significance level and standard deviation of 1.6 (estimated from a similar survey our study group had undertaken with general medical practitioners [9]). We allowed for a 50% response rate, and therefore approached 880 chiropractors to participate.

Chiropractors were randomly sampled from three strata, defined by States in Australia, with the same proportion of chiropractors approached in each state. The entire sample frame of 1760 was obtained from the Chiropractic Registration Boards. The numbers approached in Victoria, South Australia, and Western Australia were, respectively, 553, 165, and 162. These States were chosen as other Australian States' Registration Boards would not release contact details of their registrants for research purposes.

Survey instrument

Five patient vignettes (Additional File 1) were adapted from another study in a general medical practice setting [10]. The vignettes were designed to reflect people with acute LBP who would typically present to chiropractors. Elements of each vignette were drawn from the acute LBP guideline, from another study in Victoria that evaluated a media campaign for LBP [11] and from the North-East X-ray Utilisation (NEXUS) study, which evaluated the effectiveness of audit and feedback and educational reminders on United Kingdom general practitioners' ordering of lumbar spine and knee x-rays [12]. The chiropractor's role and patient elements that were included in each vignette were designed to contextualise the vignettes. These elements were chosen because they have been recognised in previous research as potential barriers to practitioners making decisions consistent with guideline recommendations [12].

Four of the five vignettes represented people who would not require a plain lumbar x-ray and would benefit from being advised to stay active as per the recommendations from the Australian guideline [4], and one vignette represented someone with a suspected vertebral fracture who would require a plain x-ray. From a selection of response options respondents were asked to indicate which investigation(s) they would order for the patient described in each vignette, and which intervention(s) they would recommend or undertake (see Additional File 2 for options given). The investigations or treatments offered as options in the responses were developed by two experienced chiropractors who were investigators on the project team. They were chosen by the chiropractic investigators (BW and SF) as those investigations or treatments most likely to be used or recommended by chiropractors. Chiropractor respondents also had three "other" response options and could write free text to describe their intention. The five patient vignettes were pre-tested in a pilot study using six (of eight randomly sampled) Victorian chiropractors. These chiropractors had previously been interviewed about LBP management in a separate phase of the ALIGN study [8]. The pilot results led to minor changes.

Survey administration

The Modified Total Design Method [13] was used for chiropractor recruitment. The sample initially received an invitation letter to participate in the study. This letter was accompanied by a plain language statement, consent form, and reply-paid envelope (for returning an individual's decision to opt-in [or opt-out] by mail). A reminder letter was sent to non-respondents every two weeks for up to a total of eight weeks after the initial letter was sent. Both the fourth and eighth week reminder letters were accompanied by a plain language statement, consent form, survey document and reply-paid envelope.

The Total Design Method usually includes the use of a return mail card to signify that the respondent has opted-in to the study. The return mail card can also be used by individuals to indicate that they do not wish to participate in the study. In this study we replaced the return mail card with an option on the consent form for individuals to indicate they did not wish to participate (i.e. opt-out) in the study (and not be contacted further about this study). It was therefore considered that a return mail card, as per the Total Design Method, was not necessary in this study.

Analysis

Double data entry was performed independently by two researchers (MP and KB) and discrepancies were resolved via discussion with a third researcher.

Descriptive statistics were used to describe demographic details of the chiropractic respondents. The percentage of chiropractic respondents indicating which investigation(s) and which intervention(s) they would undertake for each vignette was calculated. Confidence intervals were calculated adjusting for the stratification variable. Exact binomial confidence intervals were calculated when the normal approximation was not valid [14]. Confidence intervals calculated for the percentage of respondents indicating which investigation(s) and which intervention(s) they would undertake across multiple vignettes (e.g. recommending advice to stay active across vignettes 1 to 4) were adjusted to allow for the correlation of responses within chiropractor. Analyses were undertaken using the svy commands in the statistical package Stata (StataCorp. 2007. Stata Statistical Software: Release 10.1. College Station, TX: StataCorp LP) with Taylor series approximations for the estimation of standard errors.

Human research ethics approval for this study was granted by Monash University (CF07/1060 - 2007000274).

X-ray usage

Where an x-ray was not recommended according to the guideline 68% of chiropractors responded that they would order or take an x-ray, indicating an overall x-ray guideline adherence of 32%. It has been reported previously that chiropractors use plain x-rays at a high frequency for people with acute non-specific LBP, with rates varying from 12% to 63% [15–22]. Our findings are consistent with these studies and suggest that a high proportion of chiropractors who responded to this survey are not complying with the guideline on the use of plain x-ray for acute LBP. Some chiropractors order or take plain x-rays to rule out pathology, screen for contraindications to manipulation, assist in the selection and delivery of treatment, and monitor patient progress [23]. A post-hoc review of the literature found no substantive evidence of harm from chiropractic manipulation because plain x-ray had not been used nor enhanced efficacy.

Across all five vignettes approximately 25% (95% CI: 20%, 29%) of chiropractors would take, or order, a full spine plain x-ray. This was despite the lack of information in the vignettes about any symptoms or examination findings at the other regions of the spine. There is no evidence that full spine x-rays are warranted in these vignettes and this finding is troubling considering the radiation dose that the patient would receive [24, 25].

Around one third of chiropractors surveyed (35%; 95% CI: 29%, 40%) would perform spinal manipulation of the spine in a patient with a likely fracture. We did not investigate the reasons for this, however it may be that a) the potential for fracture diagnosis in the vignette was not considered, or b) the chiropractor intended to apply manipulation to regions above or below the fracture site.

The findings above have implications for patients' safety and health resource use. First, a full spine x-ray will expose the patient to needless ionising radiation, and manipulation of a spinal fracture may worsen the fracture with potentially serious consequences. Second, the findings have implications for third party payments for these extensive and expensive diagnostic imaging.

Giving advice to stay active

Over half the chiropractors (51%; 95% CI: 47%, 56%) indicated that in the first four vignettes they would advise the patient to stay active, while 60% (95% CI: 57%, 64%) responded they would give exercises (back specific or general) to people seen who were similar to the first four vignettes. In this sample of chiropractors, there appears to be a need to encourage the giving of advice to patients with acute LBP to stay active, regardless of any exercises they would recommend to the patient in order to bring chiropractic management in line with evidence-based practice.

Study limitations

The response rate was low (37%) and therefore the results have an increased potential for selection bias where those who choose to participate differ in important characteristics with those who do not choose to participate [26]. It may be speculated that those who responded were more likely to comply with the guideline, in which case we have calculated underestimates of guideline compliance. If this is the case the problem may be greater than our results demonstrate. It also may be speculated that those who responded were less likely to comply with the guideline, in which case we have calculated overestimates of guideline compliance. Regardless, we can only conclude for the respondents to this survey and are not able to generalise to the broader population of chiropractors in these three States of Australia. No data were available for the broader chiropractic population at the time we undertook this study, so we were unable to compare the characteristics of the respondents to the characteristics of the broader Australian chiropractic population to investigate non-response bias. The response rate is lower than those seen in other surveys of health professionals where a mean of 58% has been reported, however it has also been demonstrated that response rates of health professionals are decreasing over time [26]. Recently published surveys of chiropractors have reported response rates ranging from 38% to 88% [27–31].

Patient vignettes are limited by a lack of detail that may be gleaned during a patient consultation, including visual cues relating to pain during an examination. Further, patient vignettes may not be sensitive enough to pick up the fine nuances of practice that would better guide the practitioner [32]. Therefore, our conclusions are based on a measure of proxy behaviour rather than real behaviour. More research is needed to establish the extent to which proxy measures of behaviour can predict actual behaviour.