Concurrent Proximal Fractures Are Rare in Distal Forearm Fractures: A National Cross-sectional Study

Introduction Distal forearm fractures (DFF) account for 1.5% of emergency department (ED) visits in the United States. Clinicians frequently obtain imaging above/below the location of injury to rule out additional injuries. We sought to determine the incidence of associated proximal fractures (APF) in the setting of DFF and to evaluate the imaging practices in a nationally representative sample of EDs. Methods We queried the 2013 National Emergency Department Sample using International Classification of Diseases, 9th edition, diagnostic codes for DFF and APF. Current Procedural Technology codes identified associated imaging studies. We calculated national estimates using a weighted analysis of patient and hospital-level characteristics associated with APF and imaging practices. An analysis of costs estimated the financial impact of additional imaging in patients with DFF using Medicare reimbursement to approximate costs according to the 2018 Medicare Physician Fee Schedule. Results In 2013, an estimated 297,755 ED visits (weighted) were associated with a DFF, of which 1.6% (4836 cases) had an APF. The incidence of APF was lower among females (odds ratio [OR] (0.76); 95% confidence interval [CI], 0.64–0.91) but higher in metropolitan teaching hospitals compared to metropolitan non-teaching hospitals (OR [2.39]; 95% CI, 1.43–3.99) and Level 1 trauma centers (OR [3.9]; 95%, 1.91–7.96) compared to non-trauma centers. Approximately 40% (n = 117,948) of those with only DFF received non-wrist radiographs and 19% (n = 55,236) underwent non-wrist/non-forearm imaging. Factors independently associated with additional imaging included gender, payer, patient and hospital rurality, hospital region, teaching status, ownership, and trauma center level. Nearly $3.6 million (2018 U.S. dollars) was spent on the aforementioned additional imaging. Conclusion Despite the frequency of proximal imaging in patients with DFF, the incidence of APF was low. Further study to identify risk factors for APF based on mechanism and physical examination factors may result in reduced imaging and decreased avoidable healthcare spending.


INTRODUCTION
Distal forearm fractures (DFF) are some of the most common fractures evaluated and treated in the United States, and this incidence has been increasing over the last 50 years. [1][2][3][4][5] DFFs account for roughly 1.5% of emergency department (ED) visits annually 3 with complications including chronic pain, osteoarthritis, median nerve compression, loss of motion, and complex regional pain syndrome. 6,7 Most injuries are due to minor trauma such as accidental falls, especially in the geriatric population. 1,3,8 With an aging population, the Medicare costs for treating these fractures are also increasing. In 2007, $170 million (United States dollars) in payments were made by Medicare for distal radius fractures alone. 9 Many clinicians have been taught that elbow imaging should be a component of the evaluation of DFF to avoid missing corresponding injuries; however, there is a lack of primary literature to support this practice. 10 Excessive imaging continues to lead to additional expense and radiation risk, and the Choosing Wisely Campaign has targeted low-value imaging as one of its priorities in reducing unnecessary healthcare spending. 11 Describing the epidemiology and fracture patterns of DFF and associated proximal fractures (APF) could better target imaging to those most likely to benefit, and clinical decision rules could be developed to target imaging practices toward high-risk groups. The objectives of this study were the following: 1) to determine the proportion of concurrent APF in the setting of DFF; 2) to better understand the current imaging practice used in EDs to evaluate patients with DFF; 3) to perform a cost analysis on current imaging practices; and 4) to identify factors associated with APF among those with DFF.

Study Design, Setting, and Population
We conducted a cross-sectional study of data from the 2013 National Emergency Department Sample (NEDS), a dataset of a representative sample of U.S. ED visits developed by the Healthcare Cost and Utilization Project. 12 NEDS is a sample comprised of discharge data for ED visits across more than 900 hospitals located in 33 states and the District of Columbia. The data approximate a 20% stratified sample of U.S. hospital-based EDs with over 30 million ED visits annually, with a weighted estimate of 135 million ED visits. We included all records with DFF, defined by the International Classification of Diseases, 9 th edition, (ICD-9) codes 813.4-813.47, 813.5-813.54, 833.01. We excluded records with a discharge diagnosis consistent with DFF but without any imaging recorded, and we excluded visits requiring inpatient admissions.
This study was determined not to qualify as human subjects research by the local institutional review board and is reported in accordance with the Strengthening Observational Studies in Epidemiology (STROBE) publication guideline. 13

Definitions
DFF was defined through a series of ICD-9 codes (Supplemental File, Appendix A). Three independent experts in the management of DFFs identified ICD-9 codes that were "definitely" DFFs, codes that "could include" DFFs, and codes that were "not" DFFs. We used the most conservative "definite" definition of DFF (ie, the specific ICD-9 codes categorized as DFF obviously entailed a fracture in the distal part of the extremity), and other definitions were used for sensitivity analyses (Figure 1). We defined APFs as all other non-DFFs of the upper extremity. Other fractures of the upper extremity (humerus and elbow), as well as unspecified portions of the forearm, were categorized as APF in this conservative "definite" definition of DFF. We defined imaging as having a claim for a procedure code for imaging of the upper extremity, identified through Current Procedural Terminology (CPT)-4 codes (Supplemental Content, Appendix B). When evaluating for a DFF, we considered standard imaging to be of the wrist or forearm, while non-standard imaging was defined as imaging procedures performed at nonwrist and non-forearm sites (ie, elbow and humerus).

Cost Analysis
We estimated healthcare costs from a societal perspective of healthcare spending alone. The societal cost of the additional imaging procedures was approximated by the Medicare reimbursement rate. For the cost analysis, additional imaging was defined as a three-view elbow radiograph in We used a decision analysis model incorporating estimated base parameters (ie, prevalence of DFFs) and probability of APF, given DFF was used to estimate the population healthcare cost of imaging DFFs without APF. Finally, to account for potential variation in the actual cost of the additional imaging by facility, state, and payer, we performed a sensitivity analysis varying the cost by 75% and 150%. These differences were determined by the reported magnitude of differences in commercial insurance and Medicaid reimbursement compared to Medicare reimbursement. 14,15

Outcomes of Interest
The primary outcome of interest was the incidence of APF among patients with DFF. The secondary outcome was the proportion of patients with DFF who had non-standard imaging performed.

Statistical Analysis
To identify factors associated with APF we compared patient and hospital-level characteristics between DFF patients with and without APF, using weighted estimates. We conducted a bivariate analysis using variables in NEDS for primary sampling units, weights, and clustering to account for the sampling strategy and frame for this dataset. To ensure limiting this dataset would not introduce any bias, we evaluated the DFF subset with and without imaging across several patient and facility characteristics ( Table 1). We then assessed differences in patient or hospital-level characteristics of visits vs those with standard vs non-standard imaging (univariate logistic regression, OR [odds ratio], 95% confidence interval [CI]). We included all patient and hospitallevel characteristics in the final multivariate logistic model. Collinear variables were removed individually with those removed being ones of lower priority. As part of a sensitivity analysis, we evaluated whether a change in the definition of DFF and APF would influence the model estimates for each individual-and facility-level covariate.
We performed data management and statistical analysis using SAS v.9.4 (SAS Institute, Cary, NC), on a Unix-based institutional distributed computing cluster (High-Performance Computing, Information Technology Services, University of Iowa, Iowa City, IA).

Demographics
There were 464,597 visits indicating DFF, of which 166,842 (36%) were excluded for incomplete reporting with no CPTcoded imaging (eg, may have been transferred and had imaging performed elsewhere or had CPT codes incompletely reported) ( Table 1). The final sample analyzed included 297,755 visits with DFF identified. Demographic characteristics for excluded records   were similar to the included records. The majority of patients with DFF were <18 years (44.2%), female (55.4%), and had private insurance (41.2%) ( Table 2).

Distal Radius and Associated Proximal Fractures
The number of DFF records with APF was 1.6% (n = 4836, 95% CI, 1.2-2.1%) with the majority of the APF being radial shaft fractures (15.2%), radial head fractures (14.9%), and supracondylar humerus fractures (12.9%) ( Table 3). Although these were the most common APF they were still exceedingly rare in those with DFF, with radial shaft fractures occurring in 0.56%, radial head fractures occurring in 0.55%, and supracondylar humerus fractures occurring in 0.48% of patients with DFF (Table 3) 3) had non-wrist/non-forearm imaging. Dedicated imaging of the humerus or elbow occurred less frequently at 1.4% (95% CI, 1.2-1.5), and 8.1% (95% CI, 6.9-9.2), respectively. There were differences in the cases with non-standard imaging (imaging at locations other than the wrist or forearm) performed by demographic-and facility-level characteristics ( Table 3). Among those with DFF only, the odds of non-standard imaging were approximately two times greater among those ≥18 years of age compared to those <18 years. Additional imaging occurred more frequently among females (uOR [1.09]; 95% CI, 1.01-1.17). Compared to those with private insurance, additional imaging that was non-standard occurred most frequently among

Multivariable Analysis
Among patient-level factors in the final multivariable model, age, sex, and payer were still independently associated with non-standard imaging. Compared to metropolitan non-teaching facilities, the unadjusted odds of non-standard imaging were 1.28 (95% CI, 1.02-1.62) and 0.73 (95% CI, 0.61-0.87) among metropolitan teaching facilities and non-metropolitan hospitals, respectively. This suggests patients presenting to teaching hospitals receive more radiographs than those at rural hospitals. The unadjusted odds of non-standard imaging was 2.16 (95% CI, 1.41-3.30) among Level 1 trauma centers compared to nontrauma centers.

Cost Analysis
If every DFF presenting to the ED received a radiograph (assumed to be a three-view elbow radiograph) to evaluate for APF, it would cost $21.2 million yearly and $4,455 at $71 per radiograph per APF identified. In our sample, 8.1% of those with DFF received this radiograph series costing $1.7 million. Using Medicare reimbursement as a proxy for health system cost, $3.95 million is spent annually for additional imaging of DFF who do not have APF. In sensitivity analyses varying the cost of a radiograph (to account for potential underestimation of the true cost of imaging using the Medicare reimbursement rate), the cost of identifying an APF through imaging of all DFF patients ranged from $3,341 to $6,683.

DISCUSSION
We report a low incidence (1.6%) of APF associated with the diagnosis of DFF. The low incidence of APF is likely a significant reason the previous literature on APF has been limited to case reports. 12,[17][18][19][20][21][22][23][24][25][26][27][28][29] In our series, the most common APFs were radial shaft fractures (15.2%), followed by radial head fractures (14.9%), and supracondylar humerus fractures (12.9%) ( Table  3). Forty percent of patients with an APF had fractures that could have been identified on elbow radiographs. Nearly half (45%) of those with an APF had elbow radiographs performed (Table 4). Although this fracture rate is 5% lower than the percentage of patients who had an APF and received an elbow radiograph, this may be an acceptable rate of potential imaging. However, combined with the 8.1% of those without an APF who received radiographs of the elbow, this may be an area where particular attention should be paid to the physical examination in identifying patients who are at risk for osseous injury.
The use of the physical examination to identify patients at very low risk for fractures of the knee and ankle has been used to reduce low-value imaging. 30,31 That being said, the use of physical examination to accurately assess who is at risk for osseous injury at the elbow has had mixed results. [32][33][34] The East Riding Elbow Rule, which combines elbow extension, osseous tenderness, and bruising, boasted 100% sensitivity for elbow fracture and would decrease elbow radiographs by an estimated 15%. 31 Subsequently, studies using similar methodology have not had as promising results in accurately identifying those at risk of elbow fracture through the use of physical exam; sensitivities for elbow extension alone ranged from 73-88% with the combination of elbow extension and osseous tenderness having sensitivities from 96-98%. 32,33 It is unclear whether routine imaging of the elbow is necessary or cost effective in those with DFF. However, the routine practice of obtaining imaging of the joint proximal to the known fracture site has been evaluated in patients with ankle fractures with nearly 64% of those patients receiving adjacent joint imaging and only 9.9% of patients having an APF, although it is unclear how these results would translate to the upper extremity. 35 Demographic considerations may also play a role in the need for additional imaging. The higher proportion of APFs in trauma centers is noteworthy, because it suggests that either 1) increased imaging identifies fractures that are missed in non-trauma centers; or (2) the patient population in trauma centers is different from those in non-trauma centers. Patients being treated at Level 1 trauma centers were 2.42 times more likely (95% CI, 1.62-3.61) to undergo imaging of the non-wrist or non-forearm in patients without an APF. They may also be more likely to have sustained a more significant mechanism of injury necessitating additional imaging. Furthermore, trainees at these institutions initially evaluate patients, and prior reports have associated junior trainees with increased diagnostic testing. Additionally, patients who receive care at academic institutions have a higher level of testing performed. 36 These findings have been consistent across a variety of hospital settings including EDs, intensive care units, general internal medicine wards, and units treating ischemic strokes. [36][37][38][39] Our analysis also showed that imaging of the non-wrist and non-forearm occurred more frequently among females who only had a DFF (unadjusted odds ratio [1.09]; 95% CI, 1.01-1.17). This could be related to previous work revealing that DFF is more common in females. 2 However, females were less likely to have an APF in our study.

LIMITATIONS
Our study has several limitations. First, our analysis was done retrospectively using the NEDS database to obtain a large, diverse, and generalizable data sample. However, there are several inherent limitations to a retrospective database analysis. The NEDS database is a collection of claims data, not medical records. This may be relevant given that only 64% of patients diagnosed with DFF had complete data in the NEDS database. We limited our analysis to records from all those with DFF who had recorded imaging in the database. Accordingly, all patients without imaging were eliminated from our analysis since the diagnosis of DFF was contingent upon imaging.
Second, when defining our cohort we used increasingly stricter ICD-9 definitions and ultimately ran an analysis on the strictest definition to minimize uncertainty regarding the precise anatomic location of the DFF. This may have excluded some DFFs that were coded using general codes, which could lead to an underestimate of concomitant fractures. We intentionally used this strategy to define an upper limit for the actual estimate, because the rate of APF in reality may be lower than the 1.6% we report. However, model estimates from our sensitivity analysis were similar across all three definitions of DFF, suggesting the APF rate of 1.6% may be accurate.
Third, our cohort was limited to patients who were discharged from the ED. One could argue patients admitted after sustaining a DFF were more likely to experience more significant trauma, which could put those patients at higher risk for APF.
Fourth, in our analysis APFs were seen more often in teaching hospitals. In this setting more radiographs were also performed. With that said, even those without APFs were more likely to receive non-standard imaging in teaching hospitals when compared to non-teaching hospitals (Table 4). One could contend that the direct correlation between the increased testing and the greater rate of APF identified justifies performing additional testing in all patients with DFF. We assert that there are other potential means to identify those at risk for APF in a more practical and cost-efficient manner (eg, the physical examination). However, this study cannot address which radiographs were clinically indicated.
Lastly, we assume that all APFs were identified. We were unable to determine whether a patient was subsequently diagnosed with an associated proximal fracture that was missed during the ED visit.

CONCLUSION
In patients with a DFF, the incidence of having an APF is low. Further study to identify risk factors for APF based on mechanism of injury, physical examination, and demographic factors may result in identifying patients at variable degrees of risk for APF.