Rates, costs and determinants of lumbar spine imaging in population-based women born in 1973–1978: Data from the Australian Longitudinal Study on Women’s Health

Objective There are concerns that lumbar spine imaging represents low value care. Our aim was to examine the use of lumbar spine imaging [radiography, computed tomography (CT), magnetic resonance imaging (MRI)] over 20 years, and costs and person-level characteristics of imaging in a large cohort of Australian women. Methods The Australian Longitudinal Study on Women’s Health (ALSWH) is a longitudinal population-based survey of women randomly selected from national health insurance scheme (Medicare) database. This study examined 13458 women born in 1973–1978 who consented to link their ALSWH and Medical Benefits Scheme records. Self-reported data on demographics, body mass index, depression, physical and mental health, and back pain were collected in each survey performed in 1996, 2000, 2003, 2006, 2009, 2012, and 2015. Data on lumbar spine imaging from 1996 to 2015 were obtained from the Medical Benefits Scheme database. Results 38.9% of women underwent some form of lumbar spine imaging over 20 years. While radiography increased from 1996 to 2011 and decreased thereafter, CT and MRI continued to increase from 1996 to 2015. In women with self-reported back pain, depression and poorer physical health were associated with imaging, with no significant differences in types of imaging. Based on imaging rates in ALSWH, the estimated costs for Australian women aged 30–39 years were AU$51,735,649 over 2011–2015. Conclusions Lumbar spine imaging was common in population-based Australian women, with rates increasing over 20 years. Depression and poor physical health were associated with lumbar spine imaging. Raising awareness of this in clinicians is likely to result in significant cost savings if clinical guidelines are followed, with the potential of freeing resources for high value care and health outcomes.


Introduction
Low back pain (LBP), the leading cause of disability worldwide, is a common reason for seeking medical care and associated with substantial direct and indirect costs [1]. Non-specific LBP, defined as LBP not attributable to a known cause, represents 90-95% of the LBP cases [2]. Of concern is the overuse of lumbar spine imaging, including radiography, computed tomography (CT) and magnetic resonance imaging (MRI). A meta-analysis of over 4 million imaging requests across 21 years found one in four patients presenting to primary care with LBP received imaging, and complex imaging increased by over 50% from 1995 to 2015 [3]. This is despite evidence-based clinical guidelines [4][5][6][7] and Choosing Wisely campaigns [8][9][10] recommending against the use of routine diagnostic imaging for patients with non-specific LBP in the absence of clinical red flags (i.e. severe or progressive neurologic deficits or features suggesting a serious or specific underlying condition such as fracture, malignancy, cauda equine syndrome, or infection), thus representing low value care. Imaging for LBP has been listed in the evidence-based "Top 5" activities in primary care practice where change in practice could improve the quality of care and use of clinical resources [11]. Routine diagnostic lumbar spine imaging, in the absence of indications of serious underlying conditions, provides no benefit to clinical outcomes [12,13], while exposing patients to radiation, unnecessary additional tests and ineffective treatments, and increasing healthcare costs [14,15]. Various interventions have been tested to reduce the rate of lumbar spine imaging, but with limited effect [16].
There is little information about imaging utilization and its associated costs in communitybased populations, with most work having been performed in clinical populations with LBP [17,18]. Understanding this will be important for improving clinical decision making, optimising health resources, and providing valuable information about low value care in the general population. As women have more back pain and utilise more healthcare resources than men [19][20][21], we examined the use of lumbar spine imaging in population-based Australian women born in 1973-1978 over 20 years, costs and person-level characteristics associated with imaging, and estimated imaging costs for Australian women of the same age group. The Australian Longitudinal Study on Women's Health (ALSWH) was chosen because it is a largely representative sample of the general population of Australian women [22][23][24].

Participants
The ALSWH (www.alswh.org.au) is a longitudinal population-based survey of over 44,000 women randomly selected from the national health insurance scheme (Medicare) database which includes most permanent residents of Australia, with intentional oversampling from rural and remote areas [22,25]. The study first collected mailed survey data from three age cohorts in 1996: 1973-1978 cohort, 1946-1951 cohort, and 1921-1926

Lumbar spine imaging
All Australian citizens and permanent residents, including those without private health insurance, have access to quality health care service under Medicare. MBS items for lumbar spine imaging (radiography, CT, and MRI) were identified from the MBS Book from 1996 to 2015 which provides information on payment of Medicare benefits for professional services rendered by registered medical practitioners [26].

Demographics
Education levels were categorised as no formal qualifications to higher school certificate, and trade/apprenticeship/certificate/diploma/university degree or higher [27]. Working status was classified as full-time paid work, part-time or casual paid work, and no paid work [27]. Depression was assessed by the question "Have you ever been told by a doctor that you have depression (not postnatal)?" [27].

Body mass index
Body mass index (BMI) was calculated from self-reported height and weight [27] and obesity defined by BMI �30.0 kg/m 2 .

Physical and mental health
Physical and mental health were assessed using the 36-item Short Form Health Survey (SF-36) [28]. Physical component summary (PCS) and mental component summary (MCS) scores were calculated (0-100), with higher scores indicating better health.

Back pain
For the question "In the last 12 months have you had back pain?" women were asked to circle one response for back pain frequency: 'never', 'rarely', 'sometimes' or 'often'. Those who responded 'never' or 'rarely' were categorized as 'no back pain', those who responded 'sometimes' or 'often' were categorized as 'back pain' [27].

Statistical analysis
Descriptive characteristics of imaging use were presented, for the whole study period  [29], as the patterns of imaging use over these two time periods better reflected the change in availability of imaging facilities for lumbar spine, particularly MRI. Participant characteristics were compared between women with self-reported back pain with and without lumbar spine imaging using independent samples t tests and chi-square tests, and binary logistic regression was used for adjustment for these characteristics. All analyses were performed using Stata SE version 14.0 (StataCorp).

Results
The number of all imaging procedures per year increased from 1996 to 2010 (2.4 times), stabilizing between 2010 and 2015 ( Fig 1A). Radiography increased by 1.7-fold from 1996 to 2011, reducing thereafter ( Fig 1B). CT increased from 1996 to 2015 (24.4 times) (Fig 1C), and MRI increased from 1999 (no MRI prior to 1999) to 2015 (16.3 times) ( Fig 1D). Table 1 presents the number of imaging procedures (original data for Fig 1) and number of women receiving imaging per year from 1996 to 2015. The trend of change in the annual percentage of women who underwent at least one imaging procedure (any imaging and by type) was similar to the trend of change in imaging procedures over the study period. There is also a trend for increasing numbers of women who received multiple imaging.
During 1996-2015, 5237 (38.9%) women had some form of lumbar spine imaging ( Table 2). While the numbers of women having CT and MRI increased from 1996 to 2015, the numbers of women having radiography increased from 1996 to 2010 then reduced in 2011-2015. Radiography only was most common (n = 3524, 67.3%), followed by radiography and CT (n = 752, 14.4%), CT only (n = 405, 7.7%), radiography, CT and MRI (n = 205, 3.9%), radiography and MRI (n = 167, 3.2%), MRI only (n = 133, 2.5%), CT and MRI (n = 51, 1.0%). The number of women receiving multiple types of imaging increased over the study period. S1 Table presents the numbers of women receiving different numbers of imaging over the study period.  (Fig 2B and 2C). Rates of MRI increased over all time periods (Fig 2D). These data indicated a time period effect of increasing imaging. At each time period from 2006-2010, older women had higher rates of imaging (Fig 2A-2D), showing an ageing effect on increasing imaging.
Costs of lumbar spine imaging for the study population are shown in Table 3.

PLOS ONE
Rates, costs and determinants of lumbar spine imaging in population-based women education, or working status) between the women who attended survey 6 (2012, n = 7510) and those who did not. In those who reported back pain in the last 12 months in survey 6 (n = 3739), we compared the characteristics of women with and without lumbar spine imaging

PLOS ONE
Rates, costs and determinants of lumbar spine imaging in population-based women

PLOS ONE
Rates, costs and determinants of lumbar spine imaging in population-based women performed in the preceding year (Table 4). Women receiving imaging were more likely to have depression and had lower PCS than those without imaging, with no significant differences between those receiving different imaging modalities. Depression and lower PCS score were associated with imaging in multivariable analysis.

Discussion
In population-based Australian women born in 1973-1978, 38.9% underwent some form of lumbar spine imaging over 20 years. Although radiography remained the most common procedure, its use decreased after 2011, and the use of CT and MRI increased over the 20 years. Data on participant characteristics were from the 6th survey (2012) and presented as mean (standard deviation) or n (%). � For difference between women with and without lumbar spine imaging. The increased imaging rates were not simply due to ageing, but also time period related factors. Costs of lumbar spine imaging were estimated over 51 million Australian dollars for Australian female population aged 30-39 years over 5 years. Depression and poor physical health were associated with imaging in women with back pain, with no significant difference in types of imaging.
In primary care settings about 25% of patients with a new episode of LBP undergo imaging [3,30,31]. In our study 38.9% women underwent some form of imaging from 1996 to 2015, with annual rate increasing from 1.9% [250/13458] in 1996 to 4.1% [556/13458] in 2014 (Table 1). Overall lumbar spine imaging increased from 1996 to 2010, stabilising between 2010 and 2015. This was due to a slight reduction in radiography since 2011, but a continued increase in CT and MRI and more women receiving multiple imaging. The overall increase in imaging in our community-based study was consistent with findings from studies of patients with LBP [17,18,30], despite clinical guidelines discouraging routine imaging. Data from our study and others [3,17,18] demonstrated a significant rapid increase in expensive, advanced imaging, discordant with current guidelines [4][5][6][7].
Clinical guidelines recommend imaging where red flags are present [4][5][6][7]. In primary care settings, the rates of serious underlying pathology identified by diagnostic imaging are very low: fracture (4%), malignancy (0.7%), infection (0.01%), axial spondyloarthritis (0.1-1.4%), and cauda equine syndrome (0.04%) [32,33]. In those aged under 50 years, fracture and malignancy are even less common. In an Australian study of 1172 consecutive patients presenting to primary care clinics with acute LBP, although 80% had at least one red flag, only 0.9% had a serious underlying pathology [34]. This study showed that combination of positive red flag features yielded a sensitivity of 38% and a specificity of 100% to identify serious pathology [34], with similar findings observed in studies from other countries [35,36]. Thus based on these data, if clinical guidelines for lumbar spine imaging are followed, it is likely that approximately three LBP patients with red flags would be imaged to detect one case of serious pathology in primary care settings. In the ALSWH cohort, the proportions of women who reported "often" back pain and seeking help for back pain (i.e. those presenting to primary care and other health professionals) was 9.1% at survey 6 (2012) and 9.3% at survey 7 (2015). Therefore, the average, 9.2% per annum, was taken as the estimate of the prevalence during 2011-2015. Thus 34 (13458 � 9.2% � 0.9% � 3) women in our study would have received imaging each year as recommended by clinical guidelines [4][5][6][7]. Our estimate is likely to be overestimated as the ALSWH is a community-based population of younger age. If these women had one MRI each year, the estimated imaging costs over 5 years (2011-2015) would be AU$68,000 ($400 � 34 � 5), corresponding to AU$6,848,173 for Australian women of the same age over this 5 year period. Thus, there would be a potential cost saving of over 44 million Australian dollars in lumbar spine imaging over 5 years (2011)(2012)(2013)(2014)(2015) in this population alone if clinical guidelines are followed.
Clinical guidelines recommend radiography only for initial evaluation of LBP in patients with a history of low-velocity trauma, osteoporosis, or chronic steroid use [14,37]. If a red flag is present, MRI is preferred over CT, but CT can be performed if MRI is contraindicated or unavailable [14,37]. Consideration of unnecessary radiation is needed when assessing the cost benefit of imaging. We estimated 34 women in our study would receive lumbar spine imaging per year during 2011-2015, but 530 women had lumbar spine imaging each year (radiography 364, CT 139, MRI 79), suggesting significant low value care. We estimated that if clinical guidelines were followed, the cost saving would be over 44 million Australian dollars in Australian women aged 30-39 years over 5 years. This needs to be taken in context of the Australian population where women of this age represent 11.9% of Australian adults aged 20-65 years. The costs of lumbar spine imaging for Australian men and women aged 20-65 years in 2011-2015 would be estimated to be over 434.7 million Australian dollars, which was approximately 0.5% of expenditure on Medicare during 2011-2015 in Australia [38]. This is important information that extends current knowledge, providing policy makers with high quality information on the potential for cost savings to free resources for high value care. Patient expectations, practitioner beliefs, financial incentives, defensive medicine, and time constraint of clinicians may contribute to the discordance between current practice and guidelines and will need to be explored [15].
In our study, depression and poorer physical health were associated with imaging utilization in women with back pain, with no difference in types of imaging modality. Better characterising women at risk of excessive imaging will enable practitioners' awareness in their care to reduce imaging and offer the potential to develop targeted approaches to reduce low value care. Both patients and clinicians should be targeted for intervention to reduce unnecessary imaging for LBP, as there is evidence of erroneous beliefs from patients [39][40][41] and clinicians [42,43] that imaging is necessary in the management of LBP, contributing to the overutilization of imaging investigations.
Our study has limitations. As general practitioners cannot order MBS-funded lumbar spine MRI, all MRIs recorded were requested from non-general practice settings. Imaging procedures, such as MRIs performed privately, inpatient imaging procedures, and MRIs performed on machines that cannot claim a Medicare rebate, may not be recorded in the MBS database, resulting in an underestimation of imaging utilization and associated costs. Women who consented to data linkage (94.5% of the 1973-1978 cohort) had higher education levels and better self-rated health than non-consenters [44]. This might introduce a potential selection bias towards women with higher socioeconomic status and better health, which would have underestimated imaging utilization and associated costs. While our study identified patient characteristics associated with imaging utilization, clinical characteristics, such as the frequency and intensity of back pain and the presence of red flags, could not be assessed due to the unavailability of the data, and data were not available to investigate clinician characteristics of those requesting imaging. This is an area for future research, as clinicians should be targeted for education to reduce low value care. The costs of imaging were calculated over 20 years at 2015 prices, which is likely to overestimate the costs from previous years. Strengths of our study are linking data from a well-established cohort with comprehensive data collected at 3-year intervals over 20 years to the MBS database, and examining imaging use, associated costs and person-level factors over long term.
Lumbar spine imaging was common in population-based Australian women born in 1973-1978 with 38.9% having some form of imaging over 20 years and rates increasing from 1996 to 2015, explained by effect of ageing and time period. Depression and poor physical health were associated with imaging. Targeting women with these conditions and raising awareness of this in clinicians are likely to result in significant cost savings if clinical guidelines are followed, with the potential of freeing resources for high value care and health outcomes.
Supporting information S1 Table. Numbers of women with different numbers of lumbar spine imaging procedures from 1996 to 2015.
(DOCX) S1 File. STROBE statement-checklist of items that should be included in reports of cohort studies. (DOCX)