ACADEMIC EMERGENCY MEDICINE 2011; 18:555–558 © 2011 by the Society for Academic Emergency Medicine
Objectives: This study assessed the utilization and clinical yield of radiographs ordered to assist in the diagnosis of acute ankle injuries presenting to two emergency departments (EDs) in Kingston, Ontario, Canada, over a 7-year period.
Methods: A large case series was assembled. Records of ankle injuries from the Kingston sites of the Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP) were linked to hospital discharge records containing procedure codes. Utilization of radiographs and the clinical yield of these assessments were analyzed over time.
Results: Following exclusions, 7,706 acute ankle injuries were identified for analysis. Utilization of radiography increased modestly over time, to a high of 70.3% (95% confidence interval [CI] = 67.3% to 72.9%) in 2007. The percentage of cases positive for fracture remained consistent (18.3% to 21.9% annually).
Conclusions: Explanations for the observed increase in utilization of radiographs remain uncertain. Increased use of clinical decision tools such as the Ottawa Ankle Rules (OAR) appear to be required to reduce procedure costs within the ED and to limit patient exposure to radiography.
The Ottawa Ankle Rules (OAR) are a clinical decision aid designed to assist physicians in safely reducing the number of ankle radiographs ordered in the emergency department (ED).1 The rules state that ankle radiography is only necessary for adult patients presenting with malleolar pain and either specific bone tenderness or the inability to bear weight.2
When formally evaluated, implementation of the OAR consistently reduced rates of ankle radiography and associated ED wait times and procedural costs across a variety of hospital settings.1,3,4 Since the mid-1990s, adherence to these rules has not been monitored consistently in our country. It is possible that adherence to these rules and related health care benefits and cost savings have decreased, consistent with the early findings of others.5
Using available injury surveillance and hospital procedure records for radiography, we examined the utilization and clinical yield of radiography for acute ankle injuries in Kingston, Ontario, Canada, between 2001 and 2007. We did this to explore whether observed patterns of utilization were consistent with adherence to the OAR in our tertiary hospital ED.
This was a retrospective case series analysis. The study protocol was approved by the Queen’s University Health Sciences Research Ethics Board.
Data Sources and Linkage
Cases of acute ankle injuries were identified from the two Kingston sites of the Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP), a surveillance program that provides full (100%) capture of injuries in all age groups requiring assessment by radiography in our community.6 CHIRPP records contain information on patient demographics, the nature and anatomical site of injuries sustained, and external causes of injury. Patient identifiers and dates of presentation to the ED were used to link individual CHIRPP files to hospital discharge records containing procedure codes. We used SAS 9.1.3 (SAS Institute, Inc., Cary, NC) to perform this linkage.
Identification of Cases. All records coded as “550 Ankle” in the CHIRPP “Body part injured” field from January 1, 2001, to December 31, 2007 (n = 11,014), were identified. Exclusions were as follows: 1) children younger than 18 years (n = 2,733; as the original OAR were developed for adult patients)1,4 and 2) injuries where radiographic assessment is seldom indicated (n = 210; included superficial injuries, open wounds, injuries to nerves, burns/corrosion, animal/insect bites, and foreign bodies in soft tissue). Following exclusions, 7,706 adult cases were available.
Procedure and Diagnostic Codes. From January 2001 to March 2002, utilization of ankle radiographs was specified using code 268 (skeletal x-ray of ankle and foot) in any hospital prescription code category. In April 2002, our procedure codes were updated and prescription code 3WA10VA (x-ray; ankle joint without contrast [e.g. plain film]) was used. Proportions of cases that were positive for fracture were determined based upon presence or absence of ankle fractures, as indicated by CHIRPP injury codes. The latter was based upon final physician diagnoses.
Proportions of eligible cases that were referred for radiography and yielded positive results for fracture were analyzed over time. We calculated 95% confidence intervals (CIs) using standard methods. Chi-square and then multivariable Poisson regression analyses were used to explore whether the utilization of radiographs varied significantly across the 7 study years and whether such variation was associated with the severity of cases as inferred by the yields of radiographs that were positive for fracture.
Between 2001 and 2007, annual numbers of ankle injury cases that presented to the Kingston EDs declined (p < 0.05), while proportions of these cases where radiography was used in their assessment increased on average by approximately 2% per year from 2001 to 2007 (p = 0.02), from a minimum of 56.8% in 2002 to a maximum of 70.3% in 2007. After controlling for the proportion of total cases positive for fracture, the proportion of cases with radiographs increased on average by approximately 3% per year (p = 0.008; Table 1). This finding remained following removal of the 2002 year from the analysis in one sensitivity analysis.
|Year||Total Cases||Utilization: Referred for radiograph||Yield: Radiographs positive for fracture|
|n||%||95% CI||n||% Total Cases||(95% CI)||% Imaged Cases||(95% CI)|
|Χ2,6 df||58.2 (p < 0.0001)||5.5 (p = 0.48)||3.3 (p = 0.77)|
|Χ2trend, 1 df||31.6 (p < 0.0001)||0.40 (p = 0.53)||0.89 (p = 0.35)|
|Beta*||0.031 (0.008)||0.002 (p = 0.23)|
Percentages of total ankle injury cases that were positive for fracture (and hence considered serious in nature) remained stable across these 7 years, as did the percentages positive for fracture among cases where radiographs were ordered (Table 1). Percentages of cases that were minor in nature (i.e., no procedures, admission, or follow-up care required) also remained consistent by study year (data not shown).
The most important finding from this study was that utilization of radiographs for the diagnosis of fracture in acute ankle injuries increased modestly over the 7 year study period, with no corresponding increase in the yield of such assessments in terms of cases positive for fracture. Utilization of radiography in 2007 was 70.3%. This and all other annual percentages were higher than post-OAR implementation (1993) levels of 57.6%4 and perhaps represented the culmination of an increasing trend in utilization observed during the study period. There was one deviation from this trend observed in 2002; speculatively this is attributable to a negative reporting or coding bias during the transition from ICD-9 to ICD-10 coding, and sensitivity analyses indicated that this had no bearing on the overall trend in annual utilization rates.
Annual proportions of radiographs that were positive for ankle fracture varied from 18% to 22% across the 7 study years, with no consistent trend in clinical yield. In post-OAR implementation evaluations conducted in multiple hospitals,1,4 15% to 17% of radiographs led to diagnoses of “clinically important” fractures. When compared with these historical norms, the higher levels of yield from radiographs in Kingston during our study period may indicate that the Kingston EDs had proportionately more serious injuries presenting to them now and this could partially account for higher rates of radiography utilization. The modest increase in utilization observed internally between 2001 and 2007 was not, however, tied to an analogous annual increase in clinical yield and hence severity of ankle injury cases.
Several explanations exist for the observed utilization patterns. First, there may have been a decrease in individual physician adherence to the OAR, as physicians in our setting are solely responsible for ordering radiographs. Compliance with clinical decision rules may decrease without ongoing reinforcement of their content. Second, the composition of our emergency physician group changed over time with a simultaneous reduction in clinical experience. Physician experience is an important determinant of resource use during patient encounters.7 Third, in the past decade the Kingston population has increased8 without substantial increases in ED resources. With busier waiting rooms, there is increased pressure to attend to more patients. While there have been no changes in the triage and patient management processes for ankle injuries, it is sometimes more efficient to order a radiograph than to perform a full clinical assessment to reach a definitive diagnosis. Finally, as Kingston was a site involved in the original evaluation of the OAR,4 it is possible that ordering behavior at that time was lowered by the knowledge that physician ordering practices were under study, as opposed to physician adherence to the rules themselves.
Given our findings, periodic initiatives to reinforce the content of the OAR to emergency physicians and the entire ED patient management system may be of benefit. Efforts to educate emergency medicine providers who are new to clinical practice may be efficacious. The presence of local advocates for clinical decision rules has been shown to be helpful,9 and the intentional incorporation of such advocacy in the management of EDs may be warranted. Finally, the incorporation of automated prompts to apply the rules from the electronic patient management system might be considered.
Strengths of this study include its large size and our use of a comprehensive data set that linked ED records with hospital procedure codes. All diagnoses were based on physician decisions informed by available radiography findings. The study also involved patients from general as opposed to specialized populations. Hence, findings may be cautiously applied to like EDs in tertiary general hospital settings.
The major limitation of our study is that we did not have access to physician ordering patterns to directly assess individual compliance with the rules. It is possible that utilization patterns vary widely between individual physicians, only some of whom account for any observed variations in utilization. Second, while the rules were also meant to be applied to some midfoot injuries that affect the ankle,1,2 these were not considered, as the vast majority of foot injuries do not involve the ankle complex and the Ottawa rules were meant to reduce the need for foot radiographs only. In Kingston, images are obtained of the entire foot as the standard of care. It was impractical for us to distinguish potentially eligible midfoot injury cases based upon existing CHIRPP diagnostic and radiography coding.
While the reasons for higher radiography rates remain unknown, study findings suggest that increased adherence to the Ottawa Ankle Rules may be needed within our setting. Evidence for this was our observed modest increase in utilization of radiographs with no concomitant increase in diagnostic yield. Attending ED staff generally view decision rules as a convenient source of advice, good educational tools, and a helpful adjunct for care.10 The Ottawa Ankle Rules are easy to learn and implement, and professional response to these rules is generally quite positive.1 Simple reminders to encourage their use can be incorporated into the patient management system. We have identified an opportunity to remind our colleagues about the use of these evidence-based decision rules in our clinical setting. This is required to reduce costs within the ED and to limit patient exposure to unnecessary radiographic procedures.
The authors thank David Barber, Kathy Bowes, Xiaoqun Sun, and members of the Clinical Research Centre of Kingston General Hospital.