Syncope: Emergency Department Evaluation and DispositionRichard A. Harrigan, MD
Introduction
Syncope is relatively common in the emergency department (ED) -- frequency is approximately 1% to 2% of all ED visits, and syncope accounts for roughly the same percentage of hospital admissions.[1-4] A syndrome rather than a diagnosis or distinct disease entity, syncope creates several diagnostic and disposition challenges for the emergency physician (EP).
The presentation of syncope evokes a wide differential diagnosis, including such potentially life-threatening etiologies as acute coronary syndrome, malignant dysrhythmia, aortic dissection, and pulmonary embolism, as well as neurologic emergencies, such as subarachnoid hemorrhage. The differential diagnosis also includes more benign entities, such as vasovagal events, orthostasis due to volume loss, autonomic disease, or medications; and situational syncope (eg, cough syncope, micturition syncope, defecation syncope). In some circumstances, syncope may mimic other disorders, such as seizure.[5]
To further add to the diagnostic challenge for the EP, syncope, which usually occurs outside of the hospital, can be difficult to elicit as part of the history of the present illness, as patients frequently appear normal during the ED evaluation. The event may also have been unwitnessed, and the patient may have little recall of prodromal symptoms or time to recovery, thus further complicating the diagnostic process.
Perhaps the biggest challenge is an appropriate ED evaluation and disposition. Although a patient with ongoing symptoms is easier to evaluate, there is a temptation to use a "shotgun" approach, including laboratory studies, radiographic imaging, and electrocardiography, in the well-appearing patient who presents to the ED after a syncopal event. When these tests are nondiagnostic, determining which patients may be discharged and which patients should be admitted for further evaluation can be difficult. The following is a discussion of the ED evaluation and disposition of patients with syncope, focusing on the current evidence-based recommendations and related literature.
Syncope: Guidelines and Clinical Policies
A wealth of papers has recently been published on syncope, both in the European and US literature. The American College of Emergency Physicians (ACEP) published a clinical policy on the topic in 2001,[6] which was updated in April 2007.[7] Specialty societies, such as the European Society of Cardiology (ESC)[8,9] and the American Heart Association (AHA),[10] have also issued specific guidelines on the syndrome. The reader is referred to these expansive documents for details. However, several points can be gleaned from each resource that reflect the current state of understanding of the syndrome, including an appropriate diagnostic approach and clinical decision rules for the disposition of patients with syncope.
The ACEP uses the standard process of offering recommendations based on classes of literature (class I, II, and III) and levels of evidence (A, B, and C). Generally speaking, the lower the class of literature or level of evidence (eg, class I or level A), the better the evidence and the stronger the recommendation. Class I evidence is derived from randomized clinical studies by using prospective data (or meta-analysis of the same); class II data come from nonrandomized trials, or retrospective or case-control data; and class III evidence is based on expert opinion or consensus and/or data from smaller, less well-controlled studies or case series. Analysis of the data leads to generation of level A, B, or C recommendations. Level A reflects a "high degree of clinical certainty;" level B a "moderate degree"; and level C means the recommendation is guarded -- based on "preliminary, inconclusive, or conflicting evidence, or in the absence of any published literature, [it is] based on panel consensus".[7]
A review of the ACEP document shows only 2 risk stratification issues that received level A support, and there is no level A support regarding admission decision.[7] Use of history and physical examination to detect congestive heart failure (CHF) (higher risk for adverse outcome) was level A, and a 12-lead electrocardiogram (ECG) was the only test with level A support for all patients with syncope; there were no diagnostic tests that received even level B support. The ACEP document contains 2 important caveats: Associated symptoms suggestive of an underlying disease process should be actively pursued (eg, syncope with headache or a new neurologic finding suggesting subarachnoid hemorrhage), and the guidelines exclude patients with obvious signs of illness for which syncope is part of the presenting symptom complex.
The ACEP guidelines offer several level B admission recommendations for patients with syncope of an unclear cause, including patients with a history, or evidence, of CHF; structural or coronary arterial disease; older age and associated comorbid conditions (exact parameters for "older age" and details of the associated comorbidities are unspecified); abnormal ECG (ie, ischemia, dysrhythmia, or significant conduction abnormalities); and a hematocrit < 30% (if obtained). Interestingly, obtaining a hematocrit is not recommended; however, recent strong literature from the San Francisco Syncope Research Group revealed that hematocrit < 30% is a predictor of short-term adverse outcome[11,12] and is the basis for this recommendation.
With the possible exception of a low hematocrit, these are common sense recommendations that may not meaningfully add to the EP's intrinsic clinical decision rubric. This observation is not a criticism of the ACEP guidelines, but instead is an indication of the dearth of good evidence on syncope in the medical literature. A look at the key literature behind these recommendations reveals the difficulties intrinsic to decision rules for syncope and suggests research trends for the future.
Syncope: Evidence-based Literature
Studies that seek to stratify risk for patients with syncope usually use a large dataset comprising many parameters that have been statistically analyzed to reveal effective diagnostic strategies. Although similar in design, key components of these studies differ, making direct comparison difficult. Each study usually defines a set of adverse outcome "predictors"; however, careful examination of how such predictors are operationally defined, as well as a focus on the follow-up period of the study, is recommended. These issues will be highlighted in the following discussion.
Perhaps the earliest paper of major significance on risk stratification and disposition of syncope patients in the ED is that by Martin and colleagues.[13] A total of 252 patients were prospectively evaluated in the derivation cohort (374 in the validation cohort, performed at the same center) Outcome was arrhythmia or death up to 1 year after evaluation. Four predictors emerged: (1) age > 45 years, (2) history of ventricular arrhythmia, (3) history of CHF, and (4) abnormal ECG.
One-year mortality was 2% in patients with no risk factors (1%, validation cohort), 5% in those with 1 risk factor (7%, validation group), and 37% in patients with 3 or 4 risk factors (27%, validation cohort). Defining "history of ventricular arrhythmia" was cumbersome in this study and hinders bedside application of these criteria. "Abnormal ECG" logically excluded persons with sinus bradycardia, sinus tachycardia, and nonspecific ST/T changes, but included potentially more benign disorders, such as left axis deviation. Also, the 1-year outcome period seems logical at first glance, but subsequent research has shown that a period of that length may be longer than necessary in syncope outcomes research.[11,12,14] Nonetheless, Martin and associates' work proved to be a valuable cornerstone of syncope outcomes research.
An Italian study derived and validated in a community sample also yielded 4 risk factors predictive of 1-year mortality[15]: (1) age > 65 years; (2) history of cardiovascular disease (including coronary artery, valvular heart, cerebrovascular, primary myocardial, and peripheral vascular disease and CHF; (3) syncope without prodrome; (4) abnormal ECG (not including "nonspecific repolarization abnormalities"). Each was given a score of 1, and the total was known as the "OESIL [based on an Italian acronym] risk score." The study sample was large (270 derivation and 328 validation cohorts, different centers). As with the study by Martin and colleagues,[13] age, underlying disease, and an abnormal ECG emerged as important predictors; however, the exact parameters differed.
A total score of 0 equaled 0% mortality in both groups, and a score of 1 was associated with < 1% mortality at 1 year in both cohorts. Scores of 3 (mortality of 35% derivation/29% validation sets) and 4 (mortality of 57% and 53% derivation and validation sets, respectively) carried significantly greater risk. Notably, no deaths were reported for patients with an OESIL score of 0 or 1 for more than 6 months, again raising the question of the optimal interval for a risk stratification tool used for ED disposition; indeed, the total 30-day mortality for all patients (derivation and validation sets) was only about 1% to 2%.[15]
A third study worth noting was included in the 2006 Life-long Learning and Self-assessment readings offered by the American Board of Emergency Medicine. The study, published by Sarasin and colleagues in 2003,[16] reflects a shift in focus in syncope outcomes research: The authors sought to determine risk factors to predict arrhythmia, not death, as the end point in patients with unexplained syncope after ED evaluation. Three factors emerged, mirroring those found in earlier studies: (1) age > 65 years; (2) a history of CHF; (3) an abnormal ECG.
This study defined "normal" to include sinus tachycardia, first-degree atrioventricular block, nonspecific ST/T changes, and premature atrial contractions; however, the ability to generalize the study data directly to the ED setting was questionable. Patients in the study with unexplained syncope had testing that typically would not be available on demand in the ED, such as echocardiography, 24+-hour Holter or telemetric monitoring, and electrophysiologic testing. Upright tilt testing and continuous-loop event recording were performed more frequently in the derivation than in the validation cohort. Again, the main outcome increased directly according to the number of risk factors. An important component of the study, however, is the concept that risk criteria for a bad outcome (death or arrhythmia) apply to patients after ED evaluation targeted not only to syncope but to the clinical presentation; this concept is carried forth by the researchers behind the San Francisco syncope studies.
The San Francisco Syncope Rule
The San Francisco Syncope Rule initially appeared in full form in 2004[11] and was validated with slight modifications by the same group in 2006.[12] Both papers are notable for design twists that are arguably better tailored to the EP in the practice environment. First, the main outcome measure was a "serious outcomes rate," and second, it was measured over a shorter period -- 7 days in the initial study,[11] and perhaps a more realistic 30 days in the validation study.[12] Serious outcomes were defined a priori and appear in
Table 1 . The concept that serious events include but are not restricted -- as in earlier studies -- to death,[15] dysrhythmia,[16] or even to death and dysrhythmia,[13] is attractive to the EP contemplating discharge of a patient with syncope not only to avoid premature death but also serious morbidity across organ systems.
The derivation set in the initial study[11] included 684 patients with syncope or near-syncope (another ED-appropriate modification, as risk stratification for and disposition of patients with near-syncope can be difficult). Of these, 79 (12%) developed a serious outcome by day 7. Statistical analysis yielded 5 criteria that were molded into the algorithm known as the San Francisco Syncope Rule. These criteria are easily recalled by using the mnemonic "CHESS," as outlined in
Table 2 .
The criteria demonstrated 96% sensitivity (95% confidence interval [CI], 92% to 100%) and 62% specificity (95% CI, 58% to 66%) for serious outcomes at 7 days.[11] The authors offer caution that although decision rules are ideally 100% sensitive, theirs was not. If the authors had added "age > 75 years," they would have increased the sensitivity to 100% (picking up 3 patients missed by the rule), but decreased the specificity to 44% and thus approximated the existing admission rate of the physicians whose patients comprised the study sample . The 3 missed patients did not have serious complications; 2 had minor increases in serum troponin levels (one with a negative cardiac catheterization), and the third patient was readmitted within the week with no cause of syncope found. The authors also caution that decision rules are more easily applied to black-and-white clinical entities, such as clinically significant ankle-and-foot fractures seen on radiographs, but are more difficult to use when the rule is applied to a syndrome (syncope or near syncope) that is actually a conglomerate of diagnoses. Understandably, some rare but catastrophic diagnoses are underrepresented in their study, despite its size, and thus a clinical decision rule might "miss" that entity when applied to someone outside the study sample[11]
Validating the San Francisco Syncope Rule
Subsequently, others have been unable to successfully validate the San Francisco Syncope Rule in their study samples, with sensitivities ranging from 52% to 77%[17-19] for the same serious outcomes. Although this research has appeared in abstract form and is thus difficult to fully evaluate, it appears that these studies have a common serious flaw: blanket application of the rule to all ED patients with syncope, rather than just to patients with unexplained syncope after ED evaluation. However, when the original San Francisco study group sought to validate their data in a new sample (albeit at the same single university center as the derivation cohort), their results were convincing. In a dataset of 791 visits, the rule was 98% sensitive (95% CI, 89% to 100%) and 56% specific (95% CI, 52% to 60%).[12] The one patient missed by the rule was a middle-aged diabetic with a cardiac history who had a negative cardiac evaluation after admission. This patient was found to have extensive posterior-circulation cerebrovascular disease that was thought to be the cause of his syncope. The overall serious outcome rate in this study was similar to the other dissenting studies (14% in the San Francisco group's validation study vs 14% to 17% in the other studies).[12, 17-19] However, the San Francisco group's validation study ultimately featured only a 7% serious outcome rate when the key statistic -- serious outcomes that were undeclared during the ED visit -- was considered. Notably, Quinn and colleagues expanded their serious outcome interval to 30 days in the validation study,[12] the time interval used in other recent studies.[17, 20, 21]
The editorial[22] accompanying the validation study by Quinn and colleagues emphasizes the importance of not blindly applying this "rule" to all comers with syncope or near-syncope, but rather applying it after the ED evaluation (ie, after the history of the present illness, physical examination, evaluation of the ECG, and any other tests the presentation suggests). Thus, for example, normotensive patients with a history of hypertension who have syncope accompanied by ripping chest pain radiating to the back will not have the critical diagnosis of thoracic aortic dissection missed if reasonable clinical judgment prevails and the EP does not simply order an ECG and a hematocrit. The authors of the editorial also highlight the importance of confidence intervals; although the sensitivity of 98% in Quinn and colleagues' validation study[12] was strong, the lower end of the confidence interval (89%) suggests that up to 11% of patients discharged home with 0 out of 5 on the "CHESS" algorithm may still have a serious outcome as defined in the study within 30 days -- an unacceptably high risk.[22]
Syncope: Current Status and Future Directions
The available evidence, although seemingly inconclusive in helping to determine which patients presenting to the ED with syncope should be admitted to the hospital for further evaluation and observation, still offers much guidance to the practicing EP. In the first instance, the time-honored dogma that an appropriate history of the present illness, thorough physical examination, and an ECG are the key components of the ED evaluation of patients with syncope has been borne out in the literature. Key historical features include a history of CHF and/or dyspnea as a presenting sign . Quinn and coworkers have found low triage blood pressure to be an important risk stratification factor.[11, 12] An abnormal ECG, somewhat variably defined in the literature[11-16] but perhaps most consistently characterized as non-sinus rhythm or new changes compared with a prior tracing[11, 12], is another high risk factor. Age, by definition a continuous variable, is understandably difficult to force into the binary world of "yes-admit/no-discharge," and no consistent cut-off for a critical age can be gleaned from the literature for use as a criterion for disposition. Nonetheless, the EP should be aware of consistent red flags -- old age, CHF, and an abnormal ECG -- when risk stratifying the patient with syncope to determine whether the patient should be admitted to the hospital or discharged. The recent work by the San Francisco Syncope Group also has established a low hematocrit (< 30%) as a critical component of the disposition algorithm. One can speculate that this data point emerged from their research because they considered serious hemorrhage, as well as the more amorphous entity of adverse outcome necessitating readmission, as "serious outcomes," whereas prior studies looked only at death and/or dysrhythmia -- both of which are unlikely to be causatively tied to serum hematocrit.
Future directions include prospective validation of the San Francisco Syncope Rule by an outside group by applying the rule at the back end of the ED evaluation as determined by history and physical examination. New decision rules, such as the Boston Syncope Criteria, are expected in the near future.[21] The EP should be mindful that, unlike some clinical entities, syncope does not lend itself easily to hard-and-fast decision rules because of its categorization as a syndrome with a broad differential diagnosis rather than as a discrete medical entity.
Table 1. Serious Outcomes as Defined by the San Francisco Syncope Rule[11,12]Death
Myocardial infarction
Arrhythmia (on monitor and tied to the syncopal event)
Pulmonary embolism
Stroke
Subarachnoid hemorrhage
Significant hemorrhage (tied to syncope and requiring transfusion)
Any condition causing return to the ED and hospitalization for related event
Table 2. San Francisco Syncope Rule[11,12]Congestive heart failure history
Hematocrit < 30%
ECG abnormal (non-sinus rhythm, or new changes compared with old ECG)
Shortness of breath
Systolic blood pressure < 90 mm Hg at triage
