Low-value imaging impacts approximately 10% of Medicare beneficiaries and costs nearly $300 million per year.¹ These services often fail to change the course of treatment and associated outcomes², and contribute to wasteful spending and potential harm to patients.³ These may include radiation exposure, false (positive or negative) test results, incidental findings which may lead to further testing and unnecessary treatment, and negative side effects.⁴ Current strategies for reducing low-value imaging – including education campaigns (eg, Imaging Wisely⁵; Choosing Wisely⁶), increased availability of alternative tests, greater specialist involvement, and electronic health record-based clinical decision support⁷ – are blunt, and have had limited success, with substantial variation across health systems and geographic regions. Developing a better understanding of how the use of these services varies across geographic regions and healthcare settings can provide insight into patterns of care for these services and potential opportunities to reduce them.

Prior studies demonstrate significant regional variation in utilization of low-value imaging. Recent work has focused on the international setting, showing that even in countries with relatively homogeneous populations (eg, Norway), low-value imaging rates varied by more than 50% across regions and hospital catchment areas.⁸ Older studies examining the U.S. healthcare system found the average use of imaging services varied threefold across hospital referral regions (HRRs) for Medicare beneficiaries between 1998 and 2007.⁹ Follow-on research showed regional variation in imaging utilization varied twofold from 2007 and 2011.¹⁰ While useful, trends from these studies are nearly two decades old, and therefore, lack relevance to the current healthcare landscape. In recent years, hospital consolidation has risen^11,12, a trend that is associated with lower hospital competition^13,14, higher prices^12,15–19, and potential reduction in quality²⁰ and access²¹ to care. Further, the COVID-19 pandemic upended the healthcare system and prompted a shift in the provision of discretionary healthcare services.²² An updated snapshot of systemic variation in low-value imaging in this context is crucial for developing successful containment strategies.

Meanwhile, reducing inappropriate use of imaging in the emergency department (ED) has become a priority issue among campaigns such as the American College of Emergency Physicians and the Society of Academic Emergency Medicine^23–25, with several studies documenting a dramatic rise in the use of these services^7,26–30 and variations across individual health systems.^31,32 To date, few studies have directly compared utilization rates of low-value imaging in the ED to other clinical settings.³³ The objective of this study was to examine HRR-level variation in the use of a common low-value imaging service – head imaging in the evaluation of syncope – among Medicare Advantage (MA) enrollees in the ED versus the ambulatory setting. We selected head imaging for syncope as a use case due to the clear clinical process for diagnosis and risk stratification, which does not incorporate head imaging.³⁴ In contrast, other low-value imaging services (eg, for an uncomplicated headache), pose greater uncertainty for clinicians due to fear of missing significant pathology, especially in the ED.³⁵ By examining geographic variation in low-value imaging through the lens of two distinct settings, we could explore the relative influence of underlying practice patterns on regional low-value imaging provision and develop evidence-based insights towards refining solutions.

The main outcome variable was the receipt of low-value head imaging for syncope services per 100 patients within an HRR, adapted from a measure developed by Schwartz et al. (2014). This measure was selected because it is commonly used when evaluating data generated from patients enrolled in Medicare¹ Further, there is minimal ambiguity regarding the provision of this service; imaging is not required for diagnosis and risk stratification,³⁴ and there is low yield of acutely abnormal findings on head computed tomography (CT) scans of patients presenting with syncope.³⁶ Diagnosis codes used to determine inclusion and exclusion in the denominator population are detailed in Table 1.

We separately examined receipt of low-value head imaging for syncope in the ED and ambulatory setting. We determined clinical setting using place or service (POS), revenue, or Current Procedural Terminology (CPT) codes. ED visits were identified by revenue code 045X, CPT for emergency evaluation and management (99281–99285), or POS 23. Ambulatory services were identified by one of the following POS codes: 02, 05, 06, 10, 11, 12,13, 14, 19, 20, 22, 24, 32, 33, 34, 49, 50, 71, 72. It is possible for a patient to qualify for the denominator in both the ED and ambulatory setting, if they meet the eligibility criteria in each clinical setting.

Patient-level explanatory variables included the following characteristics: age, race and ethnicity, original reason for Medicare entitlement, dual eligibility for Medicare and Medicaid or receipt of a Part D low-income subsidy (LIS), plan type, and clinical risk. Age was included as a continuous variable. Race was assessed according to the Centers for Medicare and Medicaid Services (CMS) beneficiary race code and categorized as Black, White, underrepresented (Asian, Hispanic, North American Native, and Other), and unknown. Original reason for Medicare entitlement was included as a categorical variable and classified by age, disability only, end-stage renal disease only, and both disability and end-stage renal disease. Plan type was classified as HMO or PPO enrollees. Clinical risk was a count variable calculated using the Elixhauser comorbidity score.

HRR variables included in descriptive analyses only included the following characteristics: density of primary care physicians (PCPs), percent of hospitals classified as for-profit, percent of hospitals classified as non-profit, percent of hospitals classified as government, and MA penetration rate. Density of PCPs was defined as the number of people in the population per PCP. Data for this measure was obtained from the County Health Rankings and Roadmaps Database. Hospital classifications were determined using the CMS Hospital Enrollment files, and MA penetration rate was calculated using the CMS MA County Penetration Rate Monthly files. All CMS data were publicly available. Measures were aggregated to the HRR-level.

Among the final cohort of 114,542 MA beneficiaries with an encounter for syncope, 27.9% received a low-value imaging service. Eligible patients in the ED were less likely to be female compared to the ambulatory setting (54.9% vs. 56.5%, respectively), more likely to be Black (20.1% vs. 15.5%, respectively), and more likely to have become eligible for Medicare due to disability (33.2% vs. 31.2%, respectively). A slightly higher share of patients in the ambulatory setting lived in a rural zip code compared to those in the ED (22.2% vs. 21.0%, respectively). On average, the patient population in the ED had more clinical conditions, with a mean Elixhauser comorbidity score of 2.5, compared to 2.3 in the ambulatory setting. HRR-level characteristics were similar between settings, including density of primary care physicians (1,458.1 people per PCP for ED vs. 1,467.4 for ambulatory settings), percent of hospitals classified as for-profit (49.8% vs. 48.4%), percent of hospitals classified as non-profit (23.9% vs. 25.1%), percent of hospitals classified as government (18.1% vs. 18.2%) and MA penetration (49.9% vs. 50.1%) (Table 2).

Across clinical settings, 60.5% of eligible patients sought care in the ED and 57.6% sought care in the ambulatory setting, reflecting that some patients had encounters in both settings. Low-value imaging occurred more commonly for patients who sought care in the ED (41.2%) compared to the ambulatory setting (5.7%). Further, at the individual level, the unadjusted average count of low-value imaging services per 100 patients was more than seven times higher among the population in the ED compared to the population in the ambulatory setting (44.2 vs. 5.9 services per 100 patients, respectively) (Table 2).

Figure 2 shows the area-level variation in adjusted low-value imaging services per 100 in the ED and ambulatory setting. Consistent with unadjusted results, the mean adjusted HRR-level low-value imaging rate was more than seven times higher in the ED compared to the ambulatory setting (40.1 vs. 5.7, respectively). Further, geographic variation in low-value imaging was greater among patients receiving care in an ambulatory setting, in which HRRs in the top decile of low-value imaging had a rate 150% higher than HRRs in the bottom decile. In contrast, the ED had a 46% difference in the low-value imaging rate between HRRs in the top versus bottom deciles. As demonstrated in the maps, greater geographic variation among patients receiving care in the ambulatory setting is also evidenced by increased variation of colors, from light to dark. In contrast, the ED has a higher concentration of dark colors, indicating consistently high rates across HRRs.

Figure 2. Area-Level Variation in Adjusted Low-Value Imaging Services per 100 in 2023: A Comparison Between the ED and Ambulatory Setting

ED, Emergency department; HHR, Healthcare referral region

In this evaluation of over 100,000 MA beneficiaries across 187 HRRs, we examined geographic variation in the use of low-value head imaging for syncope in the ED versus the ambulatory setting. We found that the mean HRR-level rate of low-value imaging for syncope was more than seven times higher in the ED compared to the ambulatory setting (40.1 vs. 5.7 per 100 patients, respectively). However, while the imaging rates were higher in the ED setting, geographic variation in imaging rates was much greater in the ambulatory setting. Rates of imaging for syncope were 150% higher among patients receiving care in an ambulatory setting when comparing the ratio of rates in HRRs in the top decile of low-value imaging versus rates in HRRs in the bottom decile, relative to 46% higher in the ED.

Our study provides an updated snapshot of earlier work demonstrating significant geographic variation in low-value imaging across U.S. HRRs from 1998 to 2011.^9,10 The present study indicates that high and variable use of low-value imaging continues to be an issue in today’s healthcare landscape, which is characterized by rising hospital consolidation^11,12 and changes in care patterns prompted by the COVID-19 pandemic.²² The consistency with which low-value imaging has contributed to unwarranted variation in healthcare service use over nearly three decades reflects the critical need for novel and effective utilization management strategies. Moreover, because the use of these services is sensitive to the supply of physician and hospital resources, rather than the health status of the population^9,10,37, a more refined, provider-centered approach may be needed.

Notably, this study builds on existing literature by demonstrating differences in the use of low-value imaging across clinical settings. The magnitude of our result is greater than that observed in an earlier systematic review comparing rates of diagnostic imaging in primary versus ambulatory care, which found that after pooling results from 45 different studies, one in four patients who presented in primary care received imaging compared to one in three in the ED.³³ To our knowledge, no prior work has directly compared systemic rates of low-value imaging between the ED and ambulatory setting. This distinction is critical because diagnostic imaging may be delivered differently across settings, given varying patient characteristics and physician practice patterns. Patients in the ED tend to have higher acuity levels and may experience reduced access to care based on factors like insurance coverage, distance to the hospital, and availability of primary care.³⁸ Meanwhile, physician behavior may differ due to the absence of a longitudinal patient relationship, incentives to provide defensive medicine, increased availability of imaging equipment, and limited ability of health plans to manage utilization.^35,39 Still, strategies for reducing low-value imaging are similar between settings. Results showing differences in the degree of geographic variation suggest that interventions to curb the use of these services should be tailored across settings.

Given the contrast in utilization and geographic variation across settings, a strength of this study is the examination of head imaging in the evaluation of syncope as a use case. Risk stratification for the diagnosis of syncope involves standardized guidelines that do not include head imaging as part of the assessment. Thus, we maximize opportunity for comparable provision of services in the ED and ambulatory setting.³⁴ Differences in care delivery across settings may increase as uncertainty of etiology, variation in guidelines by clinical scenario, and risk of long-term disability or death rises.³⁵ Syncope is a symptom that carries the possibility of very serious outcomes, whereby physical examination – not diagnostic imaging – provide the necessary information to determine an underlying diagnosis.³⁴ In contrast, while imaging for headache is considered low-value in some situations, diagnostic imaging may be appropriate with the presence of certain red flags. This nuance poses additional uncertainty for clinicians, especially in the ED.^35,40 Although this study only examines head imaging in the evaluation of syncope, future research exploring other low-value measures can provide additional information on how these findings generalize in other circumstances.

This study is not without limitations. First, head imaging in the evaluation of syncope is a single use case and may not represent care patterns in all low-value imaging services. Second, we used data from a single MA insurer, which may not be representative of the full MA population. Nonetheless, this insurer represents approximately one-fifth of all MA beneficiaries. Third, relying on claims data may fail to capture the qualitative reasoning behind physician decision making regarding the provision of low-value imaging. Clinicians may determine whether to provide an imaging service on a case-by-case basis, using lifestyle factors and symptoms that are not present in diagnosis codes. Finally, this research relies on the premise that geography is a useful unit of analysis in policy. While this work highlights meaningful variation in low-value imaging use among HRRs, it is possible that provider-level variation may be more appropriate for capturing the range in utilization of these services. Future analyses focusing on variation across clinicians could help to demonstrate the tradeoffs across the different levels of intervention.

Relatively higher rates of low-value imaging in the ED with lower geographic variation suggests a more intractable issue compared to the ambulatory setting, pervading across a broad swath of patients, providers, and regions. This result reinforces findings from earlier work highlighting the threefold rise in low-value imaging rates in the ED over the past three decades.^26,41 Although select studies have quantified moderate variation in the use of diagnostic imaging in the ED within individual health systems^31,32, this is the first to confirm the result by examining regional variation within the ED. By providing evidence of persistently high rates of low-value imaging with relatively limited variation, we show that systemic use of low-value imaging in the ED is a steadfast issue and contributes to a “culture of overuse.”⁴² Meanwhile, greater variation in the ambulatory setting may suggest more opportunity for improvement. These findings serve as a basis for policymakers to address potential differences in low-value imaging between clinical environments. Still, future work is needed to explore the individual and structural factors that contribute to low-value imaging across settings to develop comprehensive solutions.