Why Some Doctors Order Fewer Useless Back Pain Scans Than Others

Acute low back pain is one of the most common reasons people visit a primary care clinician, and it is also one of the most common reasons they walk away with an imaging order they do not need. Professional guidelines have been unambiguous for years: for patients with acute low back pain who show no red flag symptoms, immediate diagnostic imaging is considered low-value care. It is inconsistent with clinical recommendations, unlikely to improve outcomes, and carries a cascade of potential harms, from incidental findings that trigger further testing to unnecessary radiation exposure and cost. Yet the practice persists, and a new study published in BMC Health Services Research set out to answer a deceptively simple question: what actually distinguishes the clinicians who resist ordering these scans from those who do not?

The research, led by Michelle S. Rockwell of the Department of Family and Community Medicine at Virginia Tech Carilion School of Medicine together with colleagues across Virginia Tech, Carilion Clinic, the Virginia Center for Health Innovation, the University of Michigan, UCLA, and the RAND Corporation, took an unusual comparative approach. Rather than surveying a random sample of primary care clinicians, the team deliberately sought out the extremes. Within a large health system in the southeastern United States, they ranked 48 primary care practices according to their historical rate of low-value acute low back pain imaging, determined using insurance claims data. From that ranking, they identified the six highest-performing practices, meaning those with the least low-value imaging, and the six lowest-performing practices, meaning those with the most.

From those twelve practices, the researchers invited 93 clinicians, including both physicians and advanced practice providers, to complete a survey. The response rate was 66 percent, a solid figure for clinician survey research. The questionnaire probed several distinct domains: attitudes and beliefs about low-value imaging for back pain, perceptions of the clinicians’ own performance relative to peers, views on what drives low-value imaging and how to fix it, and personal characteristics, including whether the clinician had personally experienced low back pain. The design’s key strength was the linkage of these self-reported attitudes and beliefs to objective, claims-based utilization data, allowing the researchers to see which subjective factors actually tracked with measured performance.

To analyze the results, the team used LASSO regression, a statistical technique well suited to situations where many candidate predictor variables must be winnowed down to a parsimonious set. The method applies a penalty that shrinks the coefficients of less informative variables toward zero, effectively selecting the responses most strongly associated with membership in a highest- versus lowest-performing practice, while controlling for clinician demographics. This approach matters because clinician surveys generate dozens of potential correlates, and naive comparisons can easily produce spurious associations. LASSO’s built-in variable selection provides a more disciplined filter.

The baseline findings were, in some ways, reassuring. Across all respondents, clinicians reported strong agreement with low back pain imaging guidelines, scoring 9.1 out of 10 on average with a standard deviation of 1.7. In other words, almost nobody in the sample believed the guidelines were wrong. Trust in the claims-based performance data used to rank practices was more lukewarm, averaging 4.1 out of 10 with a standard deviation of 1.9. That moderate skepticism is itself informative, because any de-implementation strategy that relies on clinicians accepting feedback from administrative data will have to contend with the fact that many of them do not fully trust that data in the first place.

The statistically significant differences between the groups were more surprising, and arguably more consequential. Two clinician characteristics were associated with lower odds of belonging to a highest-performing practice. The first was a personal history of low back pain, with an odds ratio of 0.70 and a 95 percent confidence interval of 0.64 to 0.76. The second was a greater number of years in practice, with an odds ratio of 0.87 and a 95 percent confidence interval of 0.79 to 0.96. Both confidence intervals exclude one, indicating associations unlikely to be due to chance. The direction of these effects is striking: clinicians who had personally suffered back pain, and clinicians with more experience, were more likely to work in practices that ordered more low-value imaging, not less.

The interpretation of these associations is not settled by the study’s cross-sectional design, which captures a snapshot rather than tracking change over time. One plausible reading is that clinicians who have endured back pain themselves develop a stronger intuitive sense of their patients’ distress and a greater desire to rule out structural causes, even when guidelines say imaging is unnecessary. Similarly, longer-tenured clinicians trained in an era when routine imaging was more accepted may carry ingrained habits that resist guideline updates. Alternatively, the associations could reflect sorting effects, in which clinicians with particular styles gravitate toward particular practices. The study cannot disentangle these mechanisms, but it does establish that experience and personal history are not protective factors against low-value care, and may even be risk factors.

Perhaps the most humbling result concerned self-assessment. Clinicians from the lowest-performing practices were significantly more likely to rate their own performance as better than that of other clinicians in their practice, compared with clinicians from the highest-performing practices, at 57 percent versus 40 percent, a difference the authors report as statistically significant with a p-value of 0.031. This is a textbook illustration of a well-documented cognitive bias: clinicians who order the most low-value care tend to believe they order less than their peers. The finding suggests that simply telling clinicians their raw performance numbers may not be enough, because many will assume the numbers are wrong or that their cases were exceptional. Notably, the overall trust in claims data was only moderate, which compounds the problem of getting accurate self-perceptions to stick.

When asked what drives low-value imaging in the first place, clinicians across both groups converged on the same answer: patient demand was the most frequently identified driver. This attribution is common in the literature on medical overuse, and it frames the clinician as a gatekeeper responding to external pressure rather than an independent decision-maker. Interestingly, the two groups diverged on solutions. Clinicians from the highest-performing practices more frequently recommended health system-focused strategies, such as changes to workflows, decision support, or institutional policies, whereas clinicians from the lowest-performing practices more frequently recommended patient education. That split may reflect a self-serving logic, with lower performers locating the fix outside themselves, but it also carries practical weight: the strategies clinicians are willing to endorse are the strategies most likely to be implemented successfully in their own practices.

The study was conducted as part of the Virginia Center for Health Innovation’s Smarter Care Virginia initiative and a participating health system’s intervention to reduce low-value back pain imaging, both registered on ClinicalTrials.gov, with funding support in part from Arnold Ventures, which had no role in the study’s design, data collection, analysis, or manuscript preparation. The authors conclude that clinician characteristics, perceptions of performance, and preferred de-implementation strategies differ systematically by practice performance, and that these differences can help inform the selection and targeting of strategies to reduce low-value imaging. For health systems, the practical implication is that a one-size-fits-all campaign is unlikely to work. High-performing practices may respond best to system-level nudges, while low-performing practices may need interventions that confront the overconfidence gap directly, build trust in performance data, and address the patient-demand dynamic that clinicians themselves identify as the central pressure. As health systems worldwide grapple with the challenge of de-implementing low-value care, this study offers a reminder that the barriers are not ignorance of guidelines, which clinicians overwhelmingly endorse, but the subtler terrain of personal experience, habit, and self-perception.

Subject of Research: Clinician-level determinants of low-value acute back pain imaging in primary care

Article Title: Clinician-Level determinants of low-value acute back pain imaging in primary care

Article References: Rockwell, M. S., King, M., Mercogliano, E. H., Bortz, B. A., Karanjeet, R., Stewart, J., Fendrick, A. M., Mafi, J. N., & Epling, J. W. (2026). Clinician-Level determinants of low-value acute back pain imaging in primary care. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15609-5

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15609-5

Keywords: low-value care, acute low back pain, primary care, medical imaging, de-implementation, clinician survey, claims data, LASSO regression, health services research, patient demand, guideline adherence, overuse