Clinical background
Syncope in the emergency department is common and usually benign, but a minority of patients have an underlying serious cause that is not evident at the initial assessment. The decision to admit or discharge rests in practice on an uncertain history, an often normal resting ECG and the absence of symptoms on arrival. The San Francisco Syncope Rule was constructed to standardise this assessment and to identify patients whose risk of a serious event within 7 days is low enough for discharge to be considered.
Applying the San Francisco Syncope Rule
The rule is a binary decision aid, not a scoring scale. A patient is classified as high risk if any of five criteria is present, following the CHESS acronym:
where CHF is a history of heart failure, is the haematocrit, an abnormal ECG means new changes or a non-sinus rhythm, and SBP is the systolic blood pressure at triage. If none of the five criteria is present, the patient is classified as low risk.
The derivation study was conducted at an American university hospital and included 684 visits for syncope or near-syncope [1]. Of these, 79 patients (12%) had a serious outcome within 7 days. A serious outcome was defined as death, myocardial infarction, arrhythmia, pulmonary embolism, stroke, subarachnoid haemorrhage, significant haemorrhage or any condition leading to a return visit and admission for a related event. Of 50 candidate variables, five were selected by recursive partitioning. In the derivation cohort, a sensitivity of 96% (95% CI 92 to 100%) and a specificity of 62% (95% CI 58 to 66%) were achieved. Applying the rule would potentially have reduced the admission rate by 10% in the cohort.
Interpretation in practice
The rule has only two outcomes and the guidance is unambiguous:
| Classification | Criterion | Clinical action |
|---|---|---|
| Low risk | None of the five CHESS criteria met | Discharge can be considered, provided no serious cause was identified at the initial emergency assessment. The patient should be given clear advice to seek urgent care if new symptoms occur. |
| High risk | At least one criterion met | Admission or observation for investigation and monitoring should be considered, however well the patient appears at the time of assessment. |
Since the rule is designed for high sensitivity at the expense of specificity, a positive result does not mean that the patient necessarily has a serious cause, but that the risk is raised enough to warrant further investigation.
Validation and performance
The first large external validation was carried out by Sun et al. at an American emergency department with 477 patients included, of whom 56 (12%) had a serious event within 7 days [2]. Sensitivity fell to 89% (95% CI 81 to 97%) and specificity to 42% (95% CI 37 to 48%). For events not identified during the initial emergency care, sensitivity was only 69% (95% CI 46 to 92%), illustrating that the rule is less good at capturing serious conditions that are not evident on arrival.
A systematic review and meta-analysis by Saccilotto et al. included 12 studies with a total of 5,316 patients, of whom 596 (11%) had a serious outcome [3]. The pooled sensitivity was 0.87 (95% CI 0.79 to 0.93) and the pooled specificity 0.52 (95% CI 0.43 to 0.62). Between-study variation was substantial, with a 95% prediction interval for sensitivity of 0.55 to 0.98. The probability of a serious outcome given a negative result was 5% or lower across the whole material, and 2% or lower when the rule was applied only to patients in whom no cause of the syncope had been identified after the initial emergency assessment. The commonest reason for a false negative classification was cardiac arrhythmia.
Serrano et al. carried out a separate meta-analysis and found a pooled sensitivity of 86% (95% CI 83 to 89%) and a specificity of 49% (95% CI 48 to 51%) for the San Francisco Syncope Rule [4]. Subgroup analyses showed that prospective studies performed better than retrospective ones (diagnostic odds ratio 8.82 versus 2.45) and that ECG interpretation by the treating physician gave better performance than when the ECG was interpreted by researchers or a cardiologist. This suggests that some of the variation between studies is due to differences in study design and in how the ECG variable is interpreted.
The most recent systematic review, from SAEM GRACE in 2025, identified 12 validation studies of the San Francisco Syncope Rule with a spread in positive likelihood ratio of 1.15 to 4.70 and in negative likelihood ratio of 0.03 to 0.64 [5]. The authors conclude that the quality of evidence for all syncope rules is low and that none of them reliably outperforms unstructured clinical judgement. An observational study comparing the rule with clinical judgement found that the San Francisco Syncope Rule had a sensitivity of 81% and a specificity of 63% (admission rate 40%), while clinical judgement had a sensitivity of 77% and a specificity of 69% (admission rate 34%). The rule would have required 29 additional admissions to avoid discharging one patient with a serious outcome, but clinical judgement missed two patients who died after discharge, both of whom the rule identified.
An individual patient data meta-analysis by Costantino et al. with 3,681 patients found that none of the three prediction rules tested (OESIL, the San Francisco Syncope Rule, EGSYS) performed better than clinical judgement in terms of sensitivity, specificity or prognostic value [6].
Limitations
The rule applies to patients presenting acutely with syncope or near-syncope in whom no obvious serious cause is identified at the initial assessment. It should not be applied to patients with an obvious cause of loss of consciousness, such as a seizure, hypoglycaemia, trauma or circulatory shock, nor to patients in whom a serious cause is already established on arrival.
The wide variation in sensitivity between validation studies, from approximately 74% to 98%, is the most important limitation. In populations with a low prevalence of serious outcomes, the negative predictive value may be high, but in cohorts with a higher prevalence it falls. False negatives are usually due to arrhythmias not captured by a single ECG on arrival, which argues for supplementing the rule with ECG monitoring or repeat ECGs.
Another problem is that the abnormal ECG variable is subjective. Serrano et al. showed that performance varied with who interpreted the ECG, with better results when the treating physician made the assessment [4]. This makes the rule hard to standardise between care settings.
The rule is furthermore not designed to distinguish between specific diagnoses, but only to stratify risk crudely. A positive result leads to no specific diagnostic action, only to consideration of admission. The low specificity means that a substantial proportion of patients without a serious cause will be classified as high risk, which can increase unnecessary admission if the rule is applied mechanically.
Finally, no randomised trial has shown that using the rule improves patient outcomes compared with clinical judgement. The recent systematic review from SAEM GRACE concludes that all the syncope rules have a low quality of evidence and that none reliably outperforms clinical judgement [5].
References
- Quinn JV et al. Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes. Ann Emerg Med 2004. PMID: 14747812
- Sun BC et al. External validation of the San Francisco Syncope Rule. Ann Emerg Med 2007. PMID: 17210201
- Saccilotto RT et al. San Francisco Syncope Rule to predict short-term serious outcomes: a systematic review. CMAJ 2011. PMID: 21948723
- Serrano LA et al. Accuracy and quality of clinical decision rules for syncope in the emergency department: a systematic review and meta-analysis. Ann Emerg Med 2010. PMID: 20868906
- Wakai A et al. Risk-stratification tools for emergency department patients with syncope: a systematic review and meta-analysis of direct evidence for SAEM GRACE. Acad Emerg Med 2025. PMID: 39496561
- Costantino G et al. Syncope risk stratification tools vs clinical judgment: an individual patient data meta-analysis. Am J Med 2014. PMID: 24862309
