Association of State Share of Nonphysician Practitioners With Diagnostic Imaging Ordering Among Emergency Department Visits for Medicare Beneficiaries

Key Points

Question  Are emergency department (ED) nonphysician practitioner (NPP) encounters associated with more imaging studies than physician encounters?

Findings  In this cross-sectional study of 16 922 274 ED visits by Medicare fee-for-service beneficiaries in 2005-2020, the presence of NPPs in the ED compared with no NPPs was associated with 5.3% more imaging studies per ED visit.

Meaning  Use of NPPs in the ED setting appears to be associated with higher imaging use.

Importance  The use of nonphysician practitioners (NPPs) in the emergency department (ED) continues to expand, yet little is known about associations between NPPs and ED imaging use.

Objective  To investigate whether the state share of ED visits for which an NPP was the clinician of record is associated with imaging studies ordered, given that state NPP share is associated with state-level NPP scopes of practice.

Design, Setting, and Participants  This cross-sectional study compared diagnostic imaging ordering patterns associated with ED visits based on 2005-2020 Medicare claims for a nationally representative 5% sample of fee-for-service beneficiaries. For all 50 states and the District of Columbia, the state NPP share of ED visits by year was used to represent state-specific practice patterns for NPPs and physicians and how those patterns have evolved over time. The analysis controlled for patient demographic characteristics, Charlson Comorbidity Index scores, ED visit severity, year, and principal diagnosis.

Exposures  The share of ED visits in each state in each year (state share) for which an NPP was the evaluation and management clinician.

Main Outcomes and Measures  The main outcomes were the number and modality of imaging studies associated with ED visits. Analyses were by logistic regression and generalized linear model with γ-distribution and log-link function.

Results  Among 16 922 274 ED visits, 60.0% involved women, and patients’ mean (SD) age was 70.3 (16.1) years. The share of all ED visits with an NPP as the clinician increased from 6.1% in 2005 to 16.6% in 2020. Compared with no NPPs, the presence of NPPs in the ED was associated with 5.3% (95% CI, 5.1%-5.5%) more imaging studies per ED visit, including a 3.4% (95% CI, 3.2%-3.5%) greater likelihood of any imaging order per ED visit and 2.2% (95% CI, 2.0%-2.3%) more imaging studies ordered per visit involving imaging.

Conclusions and Relevance  In this study, use of NPPs in the ED was associated with higher imaging use compared with the use of only physicians in the ED. Although expanded use of NPPs in the ED may improve patient access, the costs and radiation exposure associated with more imaging warrants additional study.

In 2019, there were 151 million emergency department (ED) visits in the US.¹ Although ED visits per 1000 people have been relatively stable in recent years (420 in 2009 to 439 in 2018),² this rate has increased over time (eg, 353 in 1997 and 391 in 2007).³ Multiple studies have also shown increased use of diagnostic imaging in the ED over time. Of note, between 2004 and 2014, imaging studies per ED visit increased 29%.⁴ Overall, the use of ED imaging has increased for all modalities but more for computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound than radiography; in terms of volume, much of the increase in ED imaging has been due to CT.^4-14

Like other practice settings, the ED has seen substantial growth in nonphysician practitioners (NPPs), including nurse practitioners and physician assistants. In 1997, 5.2 million ED visits involved an NPP vs 15.2 million in 2006.¹⁵ More recent data have shown a continuation of this trend, with 22.5 million and 35.3 million ED visits involving an NPP in 2010 and 2018, respectively.^16,17 In the outpatient office setting, NPP encounters have been associated with higher use of imaging than physician encounters.¹⁸ One study that compared imaging use in EDs staffed with NPPs with those without NPPs found that EDs with NPPs had 11% more imaging per ED visit.¹⁹ However, the study was not able to control for patient- and visit-level factors, such as demographic characteristics, comorbidities, diagnoses, and visit severity, given that the unit of analysis was the ED rather than the ED visit.

Laws governing the autonomy or physician oversight of NPPs vary by state and, over time, have generally increased scope of practice and/or autonomy for NPPs.²⁰ In states with an increased scope of practice, there has been more growth in the number of NPPs as well as increased access to and delivery of health care services by NPPs.^21-23

Given both the increased use of diagnostic imaging studies in the ED and the increased delivery of care in EDs by NPPs, we sought to understand how ED imaging ordering patterns differ on the basis of the share of ED visits managed by NPPs. Hypothesizing that a higher NPP share would be associated with both a greater likelihood of imaging during the ED visit and a greater number of imaging studies when a visit has imaging, we evaluated differences in ED imaging ordering patterns based on the share of ED visits in each state by year for which the clinician of record was an NPP rather than a physician.

This cross-sectional study used a Medicare 5% Research Identifiable File from 2005 to 2020 from the Centers for Medicare & Medicaid Services as the primary data source. These data include patient-level claims information and are a nationally representative sample of Medicare fee-for-service beneficiaries. On the basis of beneficiary zip code, we classified a beneficiary’s residence as metropolitan or not using rural-urban commuting area codes.

Our analysis included all ED visits from 2005 to 2020 at EDs in the 50 states and the District of Columbia (DC) for patients residing in the 50 states or DC. We restricted our analysis to ED visits for which the clinician submitting the evaluation and management claim (Current Procedural Terminology codes 99281-99285) was an emergency medicine, family practice, or internal medicine physician or nurse practitioner or physician assistant as designated in the Medicare data. We excluded ED visits for which the diagnosis on the evaluation and management claim was related to pregnancy, childbirth, perinatal conditions, congenital malformations, and supplemental codes (International Statistical Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] codes O00-Q99 and U00-Z99 and International Classification of Diseases, Ninth Revision, Clinical Modification codes 630-679, 740-779, V01-V91, and E000-E999). The Advarra institutional review board deemed this study exempt from oversight as it posed no or minimal risk to human participants. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Our primary outcome was the number of imaging studies ordered during the ED visit. We identified imaging studies using Neiman Imaging Types of Service codes, which classify Current Procedural Terminology and Healthcare Common Procedure Coding System codes by imaging type.²⁴ For each ED visit, we included all imaging claims with an ED place of service occurring on the day of the evaluation and management claim. In addition, because an ED visit might span midnight, we included the patient’s imaging the day before and day after the evaluation and management claim if the place of service was the ED. If the charge on the imaging claim was $0, we excluded it. Our secondary outcome was the number of imaging studies by modality, including CT, radiography/fluoroscopy (radiography), and other (MRI, ultrasound, etc, which were grouped together because of their small share of ED imaging). Note that the focus was on the imaging performed based on whether the evaluation and management clinician requesting the study was an NPP or a physician and not on the specialty of the clinician who interpreted the imaging.

The key variable of interest was the share of ED visits in each state in each year for which the evaluation and management clinician was an NPP, which we call state-year NPP share. We defined NPP share at the state level rather than for each specific visit because of state-specific scopes of practice and other variations in how NPPs practice in the ED. We considered the state-year NPP share an instrument or proxy for the likelihood that an individual ED visit was managed by an NPP. Furthermore, because claims data lack the granularity in patient and clinical data required to fully risk adjust by ED visit, use of the state-year NPP share avoids bias due to the endogeneity inherent in case severity and selection of clinician type for a specific ED visit.

Other covariates included the following patient-level variables: sex, age (<65, 65-69, 70-74, 75-79, 80-84, 85-89, and ≥90 years), Medicare-reported race and ethnicity as a control variable (Black, White, other [Asian, Hispanic, North American Native, other, and unknown]), Charlson Comorbidity Index score (1, 2, and ≥3), and metropolitan residence (vs nonmetropolitan, including micropolitan, small town, or rural). Emergency department visit–level variables included year, severity (on a 5-point scale of increasing severity of 1-3, 4, and 5), and diagnosis group by ICD-10-CM code (infections, A00-B99; neoplasms, C00-D49; hematologic, D50-D89; endocrine, metabolic, E00-E89; mental/behavioral disorder, F01-F99; nervous system including eye and ear, G00-H95; circulatory, I00-I99; respiratory, J00-J99; digestive, K00-K95; dermatologic, L00-L99; musculoskeletal, M00-M99; genitourinary, N00-N99; symptoms, signs, and abnormal findings, R00-R99; injury, poisoning, S00-T88). In addition, we controlled for patients with cancer, who died the day of the ED visit, who were admitted as an inpatient immediately after the ED encounter, or who had any claim in the preceding 30 days.

Because many ED visits do not involve imaging, we used a 2-stage model to analyze the association between NPP share and imaging ordering patterns given the heavy weight on 0 (ie, no imaging).²⁵ Stage 1 used logistic regression to evaluate the association of NPP share with the likelihood of a patient receiving any imaging vs no imaging. For those with imaging, stage 2 used a generalized linear model with a γ-distribution and a log-link function to estimate the amount of imaging.²⁶ We selected the γ-distribution on the basis of the Akaike information criterion. The total number of imaging studies is the product of stages 1 and 2.

To estimate the association of NPPs with imaging performed, we used the model results to predict the amount of imaging if the state-year NPP share of ED visits was 0. In this way, the imaging volume comparison is between the current NPP-physician mix and what it would be if the ED visits included only physicians. It is not a comparison between only NPPs vs only physicians.

We conducted these analyses for imaging overall and separately for CT, radiography, and other modalities. A sensitivity analysis was conducted to confirm robustness of results to potential biasing factors. Specifically, we limited the analysis to ED visits for which the patient did not have cancer, was not admitted to the hospital, did not die the same day, and did not have any claim in the preceding 30 days (sensitivity sample). Finally, we repeated this analysis for subsets of ED visits based on diagnosis group. For statistical significance, we used an α of .05 and 2-sided tests. We constructed the analytical file using SAS, version 9.4 (SAS Institute Inc) and performed all statistical analyses using Stata, version 17.0 (StataCorp LLC).

There were 16 922 274 ED visits that met inclusion criteria. Sixty percent of patients were women and 40.0% were men; the mean (SD) age was 70.3 (16.1) years; 15.3% of patients were Black, 78.5% were White, and 6.2% were of another race or ethnicity (Table 1). A total of 1 911 255 visits (11.3%) were attributed to an NPP. Patients whose ED clinician was an NPP rather than a physician were more likely to be younger, White, and living outside a metropolitan area and had fewer comorbidities and a lower visit severity. The share of all ED visits for which the evaluation and management clinician was an NPP was 6.1% in 2005; this share steadily rose each year to 16.6% in 2020 (Figure).

Analysis of unadjusted means showed that 50% of all ED visits involved imaging and that for those visits with imaging, the mean (SD) number of imaging studies was 1.68 (1.13). Across all ED visits, the mean (SD) number of imaging studies was 0.84 (1.16) (Table 2). Specifically, the unadjusted mean number of imaging studies per ED visit was −0.04 less per visit with NPPs vs physicians (mean [SD] number of imaging studies, 0.80 [1.11] vs 0.84 [1.16]; P < .001).

Controlling for severity and other covariates, Table 3 shows the adjusted net increase in imaging studies using the 2-stage model to estimate the number of imaging studies if NPPs were not present in the ED (ie, 0% NPP share) compared with the observed mean NPP share of 11.3% over the study period. Across all ED visits with NPPs, the observed mean (SD) percentage of visits with imaging was 20.2% (0.4%) for CT, 39.3% (0.5%) for radiography, and 4.4% (0.2%) for other modalities. For imaging overall, state NPP share was associated with a 3.4% (95% CI, 3.2%-3.5%) higher likelihood of patients undergoing at least 1 imaging study. In addition, for those who underwent imaging, there was a 2.2% (95% CI, 2.0%-2.3%) increase in the number of studies associated with the presence of NPPs. Combined, there was a 5.3% (95% CI, 5.1%-5.5%) increase in imaging studies per ED visit given the presence of NPPs in the ED. By modality, the net increase on mean imaging studies was 7.3% (95% CI, 6.9%-7.7%) for CT, 3.2% (95% CI, 3.0%-3.5%) for radiography, and 14.2% (95% CI, 13.2%-15.2%) for other modalities. For the sensitivity sample, the changes overall (5.2%; 95% CI, 4.6%-5.7%) and by modality (CT, 7.6% [95% CI, 6.6%-8.6%]; radiography, 3.0% [95% CI, 2.4%-3.6%]; and other, 16.8% [95% CI, 14.3%-19.5%]) were consistent with the results for all ED visits. For the associations of patient and visit characteristics with imaging use, see eTables 1 and 2 in the Supplement. Specifically, patients who were older, were women, had comorbidities, or had higher visit severity were more likely to have imaging, whereas patients who were Black or of other race and ethnicity, died, or had other recent care were less likely to have imaging.

Table 4 shows the net increase in the number of imaging studies per ED visit associated with the presence of NPPs. By ICD-10-CM diagnosis group, these counts ranged from a low of 2.7% (95% CI, 2.2%-3.1%) more for injury or poisoning to a high of 11.7% (95% CI, 6.7%-17.3%) more for neoplasms. For every diagnosis group, there was a significant increase in the mean number of imaging studies per ED visit. These increases were associated with increases in both the likelihood of imaging and the number of imaging studies when imaging occurred. These findings were consistent with the sensitivity analysis.

In this cross-sectional study of nearly 17 million Medicare ED encounters over 16 years, we found that the presence of NPPs in the ED compared with no NPPs was associated with 5.3% more imaging studies during ED visits. Such differential use of imaging is not unique to the ED, as NPPs have been found to be 1.34 times more likely to order imaging than physicians in the outpatient office setting.¹⁸ Another primary care study of low-value back imaging found that nurse practitioners had higher imaging rates than physicians, but the difference was not significant.²⁷

Our results are consistent with a previous study that investigated the association of NPPs in the ED with imaging use and found that EDs with NPPs had 11% more imaging per ED visit.¹⁹ Our results add credence to those findings by accounting for patient- and visit-level factors, such as demographic characteristics, comorbidities, diagnoses, and visit severity. In addition, incorporating the state-year NPP share of ED visits enabled us to distinguish and control for the level of NPP involvement given differences in the percentage of care delivered by NPPs across states and over time. Such differences may be associated with differences in NPP scopes of practice across states and over time along with other factors, such as ED team structure, incentive structure, and malpractice environment.

Three studies using the National Hospital Ambulatory Medical Care Survey showed fewer imaging studies for ED visits managed by NPPs vs physicians.^28-30 However, 2 of those studies^28,29 were descriptive only and did not control for patient and hospital characteristics. Although the third study³⁰ included these controls, all 3 studies lacked the clinical data necessary to fully risk adjust ED visits to make a fair comparison of ED visits with NPPs vs physicians. Although the current study also lacked clinical data, we minimized the potential endogeneity challenges in modeling imaging ordering patterns based on the use of an NPP or physician for a specific ED visit by using the state-year shares of NPPs as an instrument.

Because the Medicare Physician Fee Schedule pays for services rendered by NPPs at 85% of what it pays physicians, NPPs deliver care at a lower cost if their use of concurrent and downstream services is the same as that of physicians. However, such a cost advantage is offset to the degree that NPPs are associated with higher resource use. Thus, our findings are important because they show a higher use of imaging studies across all modalities. Although this study did not focus on the cost question, observation of increased imaging use confirms that there is at least some offset. Future research should explore the magnitude of this offset and whether it partially or completely offsets the savings associated with the 85% rule.

Beyond cost-per-episode comparisons between NPPs and physicians, the increased use of imaging over time has received attention because of increased aggregate imaging costs and population radiation exposure.³¹ Such concerns have led to efforts, such as Image Wisely, to limit injudicious imaging use that may be redundant, not clinically warranted, or wasteful.³² As such, appropriate imaging use is a broader policy issue. To the degree that NPPs in the ED use more imaging than physicians, their increased role in ED care may represent an opportunity for initiatives to ensure their judicious use of imaging. For example, we found that the 5.3% increase in imaging orders reflected both a 3.4% increase in the share of ED visits that had imaging and a 2.2% increase in the number of imaging studies for patients who had imaging. Accordingly, efforts to reduce variation in ordering practices in the ED need to address both when and what imaging is appropriate for specific clinical conditions. These efforts should be tailored to clinician type.³³

This study found that increases in imaging ordering patterns associated with the presence of NPPs varied by modality (7.3% for CT, 3.2% for radiography, and 14.2% for all other modalities combined). Correspondingly, studies of ED imaging trends have shown substantially more growth in CT, ultrasound, and MRI than radiography.^4,8,10,11 Given that in our study, 20.2%, 39.3%, and 4.4% of ED visits involved CT, radiography, and other imaging, respectively, it may be that higher variation is associated with less common but rapidly increasing use of CT, MRI, and ultrasound. Differences in imaging ordering practices by modality may inform efforts to reduce variation. Furthermore, greater imaging use was found across all diagnosis groups, ranging from 2.7% more for injury or poisoning to 11.7% more for neoplasms. Hence, differences in imaging ordering practices by diagnosis group may also inform variation reduction efforts.

The association of state NPP share with more imaging in the ED suggests variation in practice patterns of physicians compared with NPPs. To our knowledge, there is no literature on ED imaging ordering practice variation between physicians and NPPs. However, there are studies of ED imaging ordering practice variation among physicians. One single-site study found moderate variation,³⁴ and another single-site study estimated that 1% of variation was due to physicians.³⁵ More broadly, a variation study of chest CT in the diagnosis of suspected pulmonary embolism found that more-experienced physicians had lower imaging use and a higher diagnostic yield and that physicians who were board certified in emergency medicine had lower imaging use and the same diagnostic yield.³⁶ Hence, experience and expertise appear to be associated with less imaging.

This study has some limitations. First, some ED visits billed by a physician may have been jointly managed (ie, done in consultation) with an NPP; hence, there is some potential error in determining the complete role of NPPs in ED imaging ordering patterns. To the degree that NPPs are associated with more imaging ordering than physicians, these jointly managed visits would result in an underestimate of the increase on imaging ordering patterns associated with the presence of NPPs. Second, given the limitations of claims data, we cannot fully control for clinical differences across ED visits, such as case mix severity, which would enable a direct comparison of imaging ordering practices between NPPs and physicians on a visit-by-visit basis. For this reason, we used the state-year NPP share as an instrument. Differences across state NPP share may be partly due to differences in NPP scopes of practice across states and over time, as well as variation across states and over time in how NPPs are used in EDs, incentive structures, and malpractice environments, all of which may influence imaging ordering patterns.

In this cross-sectional study, we found that NPPs are associated with an increased likelihood of an ED visit involving imaging, and for ED visits with imaging, a greater number of imaging studies were performed per visit. Expanding the use of NPPs in the ED setting may improve patient access, but the associated increased use of imaging resources with associated cost and population and individual radiation dose must be considered. Furthermore, the presence of NPPs in the ED is associated with relatively more imaging for CT and other modalities (eg, MRI, ultrasound) than they are with radiography. Differences in imaging ordering practices by modality may inform efforts to reduce practice variation.

Accepted for Publication: September 26, 2022.

Published: November 10, 2022. doi:10.1001/jamanetworkopen.2022.41297

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Christensen EW et al. JAMA Network Open.

Corresponding Author: Eric W. Christensen, PhD, Harvey L. Neiman Health Policy Institute, 1891 Preston White Dr, Reston, VA 20191 (echristensen@neimanhpi.org).

Author Contributions: Dr Christensen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Christensen, Duszak, Hirsch, Rula.

Acquisition, analysis, or interpretation of data: Christensen, Liu, Duszak, Swan, Rula.

Drafting of the manuscript: Christensen, Liu, Duszak, Hirsch.

Critical revision of the manuscript for important intellectual content: Christensen, Duszak, Swan, Rula.

Statistical analysis: Christensen, Liu.

Administrative, technical, or material support: Duszak, Rula.

Supervision: Christensen, Duszak, Hirsch, Rula.

Conflict of Interest Disclosures: Dr Duszak reported receiving grants from the Harvey L. Neiman Health Policy Institute and serving as a medical advisor to and being a shareholder in Ethos Medical, Inc, outside the submitted work. Dr Hirsch reported receiving grants from the Harvey L. Neiman Health Policy Institute during the conduct of the study, consulting fees from Medtronic and Relievant, and personal fees for serving as chair of the data monitoring committee for Balt Therapeutics outside the submitted work. No other disclosures were reported.