Imaging for Back and Joint Pain: Overuse Is an International Problem

Records in an Australian database indicated that primary care physicians very often order diagnostic imaging for patients with musculoskeletal pain shortly after first seeing them, despite evidence and guidelines that say these scans only rarely improve outcomes.

Among some 133,000 patients in Australia’s POLAR primary care database from 2014 to 2018 diagnosed with back, neck, or other joint pain, 36.2% were sent for some type of imaging (x-ray, CT, MRI, or ultrasound), according to Romi Haas, PhD, MPH, of Monash University in Melbourne, and colleagues.

About 30% of orders came within the 2 weeks prior to diagnosis and only a few were recorded in the weeks afterward. Only for MRI scans in patients with shoulder complaints was the median time substantially after diagnosis (42 days).

“[C]linical care standards and clinical practice guidelines discourage imaging for regional musculoskeletal complaints unless serious pathology is suspected, there is an unsatisfactory response to conservative care and/or imaging is likely to change management,” the researchers wrote in Arthritis Care & Research.

They cited recommendations from the “Choosing Widely” project that say, for example, “do not undertake imaging for low back pain for patients without indications of an underlying serious condition” and “do not request shoulder ultrasound to diagnose non-specific shoulder pain which on clinical evaluation is suggestive of rotator cuff pathology and in which surgery is not planned.”

Although campaigns like Choosing Wisely have been underway for more than 10 years, data suggest they have not had much impact. The National Committee for Quality Assurance recently posted U.S. statistics indicating that imaging rates for low back pain have actually increased for patients with commercial insurance. Even in the Veterans Affairs system, one recent study found substantial overuse, despite what the authors called “a setting largely unaffected by incentives of fee-for-service care.”

The VA study also confirmed that imaging is associated with negative “downstream consequences,” including increased rates of surgeries, painkiller use, and persistent pain, as well as higher treatment costs.

Australia has a universal, taxpayer-funded health insurance system with private supplemental policies also available. Thus, it may feature “incentives of fee-for-service care” that can drive up utilization, but that is likely offset to some degree by long wait times for non-urgent services.

For the new study, Haas and colleagues drew on the POLAR database, which takes electronic health records from 301 general medicine practices in the Melbourne area. Adult patients (de-identified) were included who had at least one in-person visit with a primary care doctor and who was diagnosed with pain in the neck, shoulder, low back, or knee. Among those with knee pain, only those 45 and older were included, because younger patients would be likely to have had an injury for which imaging would clearly be indicated. For other diagnoses, 18 was the lower age limit.

Overall imaging rates by pain location were as follows:

• Low back: 25.5%
• Neck: 33.6%
• Shoulder: 49.2%
• Knee: 43.1%

Of total imaging orders, 45% were for x-rays, 18% were for CT scans, 16% were for MRIs, and 21% were for ultrasound scans. MRI scans were most common for patients with neck and knee pain (37% and 25% of the MRI orders, respectively); CT scans were ordered most frequently for low back pain (50% of orders). Nearly 95% of imaging requests were for diagnosis rather than to guide injections or hydrodilatation.

Haas and colleagues also reported one important temporal trend: increases in the use of MRI scans for low back and neck pain (by 1.3% and 3.0% per year, respectively). These appeared to be replacing CT scans for low back pain and x-rays for neck pain, rates of which fell by the same degree as the increases in MRI orders. However, rates of overall imaging requests remained about the same over the study period.

Other studies had indicated that only about 1% to 6% of patients with low back pain have “serious pathology” that would warrant imaging, the group noted, versus the 25.5% imaging rate in the current study. “Based upon the small number of people attending general practice with serious pathology, it is likely our findings are at odds with recommended practice,” they concluded. “Qualitative research, GP [general practitioner] surveys and Australian Medicare Statistics suggest both clinicians and patients continue to have misconceptions about the value of diagnostic imaging for musculoskeletal complaints.” The investigators added that imaging ordered after diagnosis mainly consisted of MRI and CT scans for shoulder, knee, and low back pain, “suggesting perhaps that these are requested if initial management doesn’t help or if symptoms don’t subside.”

Haas and colleagues, in their discussion, addressed potential reasons for the apparently excessive use of imaging and how it could be reduced. An earlier analysis “identified social influence from patients, beliefs that a scan will reassure patients, and a lack of time to discuss why a scan is not needed, as the major barriers to reducing imaging for low back pain.” These are hard to overcome, Haas and colleagues suggested, but they pointed to a recent study in Australia in which clinicians with high imaging order volumes were approached for “individualized audit and feedback” about their practices, leading to significantly reduced rates of imaging orders in the months following.

Limitations to the study included the possibility that “our cohort received previous care for the same complaint or that imaging was performed only after a period of unsatisfactory improvement,” Haas and colleagues acknowledged. The data also did not capture clinicians’ rationales for requesting scans. Timing of imaging requests and diagnosis was also uncertain, the researchers indicated, because physicians “may record a diagnosis at the first presentation or at a later visit when the diagnosis is confirmed.” Generalizability to Australian practices not participating in POLAR and to other countries may be limited as well.