What Is the Specificity of the Aortic Dissection Detection Risk Score in a Low‐prevalence Population?

Abstract

Background Acute aortic syndrome (AAS) is a time‐sensitive and difficult‐to‐diagnose aortic emergency. The American Heart Association (AHA) proposed the acute aortic dissection detection risk score (ADD‐RS) as a means to reduce miss rate and improve time to diagnosis. Previous validation studies were performed in a high prevalence population of patients. We do not know how the rule will perform in a lower‐prevalence population. This is important because application of a rule with low specificity would increase imaging rates and complications. Our goal was to assess if the diagnostic accuracy of the score would be maintained in a low‐prevalence population that we are attempting to risk stratify in the emergency department (ED). Methods Retrospective cohort of patients age 18 years old and older who presented to two tertiary care EDs from January 1, 2015, to December 31, 2015, and underwent a computed tomographic angiography to rule out AAS. Two trained reviewers extracted data using a standardized data collection form. AAS was defined according to accepted radiologic standards. The components of the AHA risk score were defined a priori. Agreement was measured using kappa statistic. Sensitivity, specificity, and positive and negative likelihood ratios with 95% confidence intervals (CIs) were calculated. Analysis was performed using SAS 9.4 University Edition. Results A total 370 patients underwent computed tomography for suspected AAS. Chief presenting symptoms were chest pain (207, 58%), back pain (26, 7%), abdominal pain (32, 8.6%), syncope (7, 2.6%), and symptoms of stroke (6, 1.6%). AAS was finally diagnosed in 12 (3.2%) patients: five (1.4%) type A aortic dissection, four (1%) type B aortic dissection, two (0.5%) an aortic intramural hematoma, no penetrating aortic ulcer, and one a ruptured abdominal aortic aneurysm. The presence of one or more ADD risk markers (ADD‐RS ≥ 1) was associated with a sensitivity of 100% (95% CI = 73.5%–100%) and a specificity of 12.3% (95% CI = 9.1%–16.2%) for the diagnosis of AAS. The negative likelihood ratio was 0 and the positive likelihood ratio was 1.14 (95% CI = 1.1–1.2). Conclusions Our study confirms that in North America the prevalence of AAS in those undergoing advanced imaging is low. The ADD‐RS in this population has a low specificity. A lack of defined inclusion criteria and a low specificity limits the application of this rule in practice.

Systematic Review of Aortic Dissection Detection Risk Score Plus D‐dimer for Diagnostic Rule‐out Of Suspected Acute Aortic Syndromes

“…In North American retrospective series of patients undergoing CTA for suspected AAS, the prevalence of AASs was~3%. 8,37 In a vast out-of-hospital study evaluating the ADD-RS in nontraumatic emergencies, the prevalence of AASs was 0.9%. 38 In the studies reviewed herein, the prevalence of AASs was 13% to 23%.…”

“…First, ADD-RS, a decision rule derived from a retrospective register of AASs, has low specificity. 37 In addition, ADD-RS derivation methods have not been published, and it is currently unknown whether use of the ADD-RS provides any advantage in terms of diagnostic accuracy and of CTA ordering, compared to clinical gestalt. 39,40 In the future, focused ED-centered studies may provide alternative and more specific probability assessment tools.…”

Aortic Dissection Detection Risk Score: A Clinical Decision Rule That Needs Some Parenting

“…Even in the only prospective study to date, which did not require the ordering of diagnostic imaging, the prevalence of AAS was 13% . Ohle and colleagues, in this issue of Academic Emergency Medicine, find a somewhat expected result of applying a risk score validated in a high‐prevalence population to a lower‐prevalence population. With a prevalence of AAS of 3.2% in their cohort, they found that the ADD‐RS = 0 had a sensitivity of 100%.…”

“…Ohle and colleagues rightly bring up a number of concerns about applying this risk score to a low‐prevalence patient population. The low specificity and high sensitivity will result in unacceptably high false‐positive rates, resulting in increased rates of low‐yield, high‐cost diagnostic imaging.…”

“…Furthermore, the true sensitivity of ADD‐RS = 0 in a North American patient population is unknown. Ohle and colleagues, with a small retrospective patient population, report a sensitivity of 73.5% to 100%. In the higher prevalence cohort of the European validation studies sensitivity of ADD‐RS = 0 ranged from 91.1% to 98.9% …”