Deriving the East Riding Elbow Rule (ER2): a maximally sensitive decision tool for elbow injury
- Correspondence to Will Townend, Emergency Department, Hull Royal Infirmary, Hull HU3 2JZ, UK;
- Received 13 November 2012
- Revised 9 January 2013
- Accepted 17 January 2013
- Published Online First 8 February 2013
Objective To derive a maximally sensitive decision rule for clinical practice to rule out the need for x-ray examination after elbow injury in adults and children.
Methods Emergency department patients with acute elbow injury were recruited. Practitioners used their usual judgement to assess whether x-ray examination was required. Radiographs were reported on by radiologists blind to clinical assessment. Patients not x-rayed were followed-up at 7 days by telephone interview, and those with ongoing pain were recalled for assessment. Recursive partitioning was used to derive a maximally sensitive decision tool. Inter-rater variability for significant discriminators was subsequently evaluated by a cohort of 20 emergency department clinicians.
Results 492 patients were recruited (May 2006–November 2008): 50.4% were male; 26.8% were children; 444 (90.2%) had an x-ray; 167 (37.6%) showed abnormality. A follow-up telephone interview was conducted with 28; none were recalled. Thirteen could not be contacted, none of whom returned within 3 months. Sixteen patients with fractures were able to fully extend their elbow. The sensitivity of elbow extension alone was 84% (95% CI 77% to 88%), with specificity of 54% (95% CI 53% to 58%). A 100% sensitive (95% CI 97% to 100%) decision rule for adults (n=348) was derived based on (1) inability to fully extend the elbow, (2) tenderness over radial head, olecranon or medial epicondyle, and (3) presence of bruising (specificity 24% (95% CI 19% to 30%)). A similar rule for children could not be derived.
Conclusions A simple and highly sensitive clinical decision rule for adult elbow fracture was derived in our cohort. A validation study in a second population is now required. At present, we are unable to recommend a rule-out strategy for elbow injuries in children.
- imaging, x-ray
- musculo-skeletal, fractures and dislocations
- Trauma, extremity
- clinical assessment
Elbow injury accounts for ∼1% of attendances at our inner city teaching hospital emergency department (ED) per annum (internal data review). The majority are assessed by junior doctors and emergency care nurse practitioners. Variation in clinician experience has been shown to lead to variation in ordering of imaging to determine bony injury.1 A relatively high proportion of radiographs may be performed for elbow trauma, but fractures are still missed.2 Clinical decision tools are widely available and accepted for standardising clinical assessment after limb injury,2–4 but no consistently sensitive prediction tool currently exists for elbow examination. Recent studies have shown high sensitivity and specificity for the range of movement of the elbow as a discriminator; however, this is not 100% sensitive and so a small number of fractures might be missed if used alone.2 ,5
We therefore used the recursive partitioning method to identify and combine useful elements of clinical examination in an attempt to derive a 100% sensitive decision rule for bony elbow injury.
Local research ethics committee approval was granted to include patients presenting with elbow injury to the ED of Hull Royal Infirmary, a hospital with an annual ED attendance of 125 000. All patients with elbow injury were eligible. We did not exclude those with previous elbow injury or upper limb pathology as we wanted a rule for general use. Written informed consent was taken from all adult patients, and parental consent taken for children. Patients were only excluded if unable to consent or if it was not given. ED clinicians (nurse practitioners and doctors) recorded their clinical findings on a data collection pro-forma, but based their decision to perform radiography on their own usual clinical judgement. Information about mechanism of injury, previous history and examination findings were recorded. Clinicians were asked specifically to record whether the elbow could be fully extended. How this was determined was left to the clinician's judgement. Areas of local tenderness were marked on a diagram of the elbow within the pro-forma. This diagram was retrospectively divided into eight quadrants (four anterior, four posterior), as seen in figure 1, for analysis. Any patient considered not to need imaging underwent follow-up by telephone using a standardised structured interview at 1 week. Any such patients still experiencing symptoms were invited back for review and x-ray.
All radiographs were initially reviewed by the practitioner and subsequently reported on by a radiologist or senior radiographer trained in reporting and blind to clinical assessment.
SPSS statistical analysis software was used for sensitivities and specificities for discriminators. Sub-group analysis of adults and children (age <16) was performed. CART software (Salford Systems V.1.S) was used for recursive partitioning and the derivation of a decision tool.
Inter-rater variability for clinical parameters used in the decision rule was assessed on 20 ED clinicians (nurse practitioners and doctors) in our department after completion of data collection for the elbow rule. The nurse practitioners were the same as in the original study, but the doctors were not. This investigation was conducted before data analysis, so the clinicians were unaware of the discriminators used in the rule derivation. They were asked to complete a copy of the data collection form on a standardised test patient. A member of the research team was used to act as the patient with an olecranon fracture. Bruising was simulated, although swelling could not be reproduced. Range of motion from 90° to 30° was used and standardised using a goniometer. The test patient was required to demonstrate bony tenderness only to palpation of the olecranon. Neurovascular testing was indicated as normal once examined by the rater. The same member of the team observed assessment for consistency. As raters were unaware of the results of the original study, they did not apply the rule in assessing the patient.
κ values were then calculated using an online calculator (Randolph J.J. 2008, http://www.justusrandolph.net/kappa).
A total of 312 500 patients attended the ED between May 2006 and November 2008; 3618 attended with elbow injury (adults 2343, children 1275), accounting for 1.2% of ED presentations. Of the latter, 492 were recruited to the study: 248 (50.4%) were male, and 132 (26.8%) were children under 16 years old. The youngest patient was 2 years old, and the oldest was 96. A total of 444 (90.2%) patients underwent radiography, and 167 (37.6%) of these showed abnormality as outlined in table 1.
Of the 48 participants not undergoing radiography, 28 had a telephone follow-up. None returned. Another 13 were not contactable by telephone, none of whom returned to our department within a 3-month period. One patient returned early for an unrelated complaint, and x-ray of the elbow was not required. The remaining six patients had no follow-up recorded. Case notes were reviewed, and either no follow-up was deemed necessary at the time of assessment by the clinician or no telephone number was given at presentation. None of these patients returned to the ED within 3 months.
The sensitivity of elbow extension alone for diagnosing elbow fracture was 84% (95% CI 77% to 88%), and specificity was 54% (95% CI 53% to 58%). In adults, the sensitivity was 86% (95% CI 80% to 90%) and specificity 55% (95% CI 52% to 58%). In children, the sensitivity was 78% (95% CI 66% to 87%) and specificity was 58% (95% CI 51% to 62%). The x-ray abnormalities of those able to straighten their elbow are listed in table 2.
Recursive partitioning of data from adult patients (n=348) identified a model with 100% sensitivity (95% CI 97% to 100%). The clinical parameters that identified all those with x-ray abnormality, and how they divided the patients are shown in figure 2. The specificity of the rule was 24% (95% CI 19% to 30%). Similar analysis of the data from children (n=132) did not yield a decision rule with 100% sensitivity that would have been useable.
An estimate of interobserver variability for variables collected on the study pro-forma in a simulated standardised patient was made. High κ scores were found for the clinical discriminators in the decision tool (table 3).
We were able to derive a 100% sensitive decision rule for adult patients in our cohort, but not for children. Our findings would need validation in a second cohort of appropriate size before they could be implemented. If the rule had been applied to the study sample from which it was derived, 56 fewer x-rays would have been ordered (15% reduction on current practice).
We found that elbow extension alone was less sensitive for x-ray abnormality than in a large previous study.2 The latter study of Appelboam and colleagues was larger than ours and multicentre. They also highlight that the olecranon must be reviewed when applying extension as a rule-out for fracture. The fractures in patients able to straighten their elbow in our study would be considered significant in some cases. We did not direct clinicians how to determine full extension—it was left to their discretion. This might account for the difference in our study compared with that of Appelboam and colleagues, as they gave a clear description of how elbow straightening was to be determined before patient evaluation. We did, however, find reasonable inter-rater reliability among our staff for evaluation of the ability of our test patient to straighten the elbow.
Strengths and limitations
We recruited a relatively small proportion of patients with elbow injury (492/3618; 14%), and our sample might not be representative. Our study was conducted without additional funding, and we had no research nurse in the department at the time to support recruitment. We therefore had to rely on the ED clinicians to support the study at a time of heavy service pressure and the Department of Health's 4 h ED access guarantee. This meant it was a challenging time to ask busy clinicians to spend more time with patients explaining the study and taking consent. We believe this, at least in part, explains the low recruitment to the study. Our recruitment rate raises questions about the generalisability of our findings. In comparison with the largest UK studies of elbow injury in the published literature, our study population was similar by gender: 51% male in our study, 51% in the South West Elbow Extension (SWEET) Study,2 and 54% in the study of Lennon and colleagues.5 Children made up 26.3% of our patients, compared with 45% in the SWEET Study. X-ray examination was performed on 90.2% of our recruited patients, compared with 73% in the SWEET study and 81% in the study of Lennon and colleagues. Our study may well be skewed to those perceived to have more serious injury. If this is so, we are likely to have included a relatively high proportion of those with fractures, and hence our estimate of sensitivity is likely to be reasonably robust, although specificity might be an overestimate.
Our design was, however, pragmatic and our rule is likely to be applicable to all ED settings. We found our clinicians to be safe at picking up elbow fracture; there were no missed fractures in those not undergoing x-ray based on clinical gestalt. We would, however, have been able to reduce the number of x-rays performed without missing a fracture with the application of our rule. ED clinicians have become accustomed to using decision tools with a number of steps to support decision making about the need for imaging. The systematic addition of a number of additional elements of evaluation to that of whether the patient can straighten their arm, at least in adult patients, seems to be one way of reducing the chance of a missed injury while reducing the overall use of radiography.
A simple and highly sensitive clinical decision rule for adult elbow fracture was derived in our cohort. A validation study in a second population is now required. At present, we are unable to recommend a rule-out strategy for elbow injuries in children.
We thank all the staff in the radiography department of Hull Royal Infirmary for their support, David Graham for his efforts with attendance data, all nursing and medical staff in the ED at Hull Royal Infirmary for their assistance and support of this study, and Sarah Midgley for refining the pro-forma.
Contributors DA and PW wrote the manuscript, overseen by WT. WT designed the study, undertook ethics approval, designed data collection pro-forma, and oversaw data collection. DA and PW collated and analysed data.
Competing interests None.
Ethics approval Hull and East Yorkshire Local Research Ethics Committee.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.