Article Text

Download PDFPDF
Personalised risk prediction following emergency department assessment for syncope
  1. Venkatesh Thiruganasambandamoorthy1,2,3,
  2. Justin W Yan4,
  3. Brian H Rowe5,
  4. Éric Mercier6,7,
  5. Natalie Le Sage6,7,
  6. Mona Hegdekar8,
  7. Anne Finlayson8,
  8. Paul Huang9,
  9. Hassan Mohammad10,
  10. Muhammad Mukarram2,
  11. Phuong Anh (Iris) Nguyen2,
  12. Shahbaz Syed1,
  13. Andrew D McRae11,
  14. Marie-Joe Nemnom2,
  15. Monica Taljaard2,3,
  16. Marco LA Silviotti12
  1. 1 Department of Emergency Medicine, University of Ottawa, Ottawa, Ontario, Canada
  2. 2 Clinical Epidemiology Program, Ottawa Hospital Research Institute, Ottawa, Ontario, Canada
  3. 3 School of Epidemiology and Public Health, University of Ottawa, Ottawa, Ontario, Canada
  4. 4 Division of Emergency Medicine, Western University, London, Ontario, Canada
  5. 5 Department of Emergency Medicine and School of Public Health, University of Alberta, Edmonton, Alberta, Canada
  6. 6 Department of Family Medicine and Emergency Medicine, Universite Laval Faculte de Medecine, Quebec, Quebec, Canada
  7. 7 CHU de Québec-Université Laval Research Centre, CHU de Quebec-Universite Laval, Quebec City, Quebec, Canada
  8. 8 Department of Emergency Medicine, University of Manitoba, Winnipeg, Manitoba, Canada
  9. 9 Department of Emergency Medicine, The University of British Columbia, Vancouver, British Columbia, Canada
  10. 10 Faculty of Technology and Trades, Algonquin College, Ottawa, Ontario, Canada
  11. 11 Department of Emergency Medicine, and Department of Community Health Sciences, University of Calgary, Calgary, Alberta, Canada
  12. 12 Departments of Emergency Medicine and Biomedical, and Molecular Sciences, Queen's University, Kingston, Ontario, Canada
  1. Correspondence to Dr Venkatesh Thiruganasambandamoorthy, Department of Emergency Medicine, University of Ottawa Faculty of Medicine, Ottawa, ON K1N 6N5, Canada; vthirug{at}ohri.ca

Abstract

Background Published risk tools do not provide possible management options for syncope in the emergency department (ED). Using the 30-day observed risk estimates based on the Canadian Syncope Risk Score (CSRS), we developed personalised risk prediction to guide management decisions.

Methods We pooled previously reported data from two large cohort studies, the CSRS derivation and validation cohorts, that prospectively enrolled adults (≥16 years) with syncope at 11 Canadian EDs between 2010 and 2018. Using this larger cohort, we calculated the CSRS calibration and discrimination, and determined with greater precision than in previous studies the 30-day risk of adjudicated serious outcomes not identified during the index ED evaluation depending on the CSRS and the risk category. Based on these findings, we developed an on-line calculator and pictorial decision aids.

Results 8233 patients were included of whom 295 (3.6%, 95% CI 3.2% to 4.0%) experienced 30-day serious outcomes. The calibration slope was 1.0, and the area under the curve was 0.88 (95% CI 0.87 to 0.91). The observed risk increased from 0.3% (95% CI 0.2% to 0.5%) in the very-low-risk group (CSRS −3 to –2) to 42.7% (95% CI 35.0% to 50.7%), in the very-high-risk (CSRS≥+6) group (Cochrane-Armitage trend test p<0.001). Among the very-low and low-risk patients (score −3 to 0), ≤1.0% had any serious outcome, there was one death due to sepsis and none suffered a ventricular arrhythmia. Among the medium-risk patients (score +1 to+3), 7.8% had serious outcomes, with <1% death, and a serious outcome was present in >20% of high/very-high-risk patients (score +4 to+11) including 4%–6% deaths. The online calculator and the pictorial aids can be found at: https://teamvenk.com/csrs

Conclusions 30-day observed risk estimates from a large cohort of patients can be obtained for management decision-making. Our work suggests very-low-risk and low-risk patients may be discharged, discussion with patients regarding investigations and disposition are needed for medium-risk patients, and high-risk patients should be hospitalised. The online calculator, accompanied by pictorial decision aids for the CSRS, may assist in discussion with patients.

  • syncope
  • emergency department
  • hospitalisations

Data availability statement

No data are available.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

What is already known on this subject

  • Published risk tools do not provide possible management options for syncope in the emergency department (ED). We previously derived and validated the Canadian Syncope Risk Score (CSRS) to predict 30-day serious outcomes not identified during the index ED evaluation.

What this study adds

  • We pooled the previously published derivation and validation cohort data and used this larger cohort to develop more precise 30-day observed risk estimates based on the CSRS. These risk estimates provide personalised risk prediction including the overall risk and the likelihood of each type of serious condition such as death, arrhythmia or non-arrhythmia to guide management decisions. We also developed a freely available digital tool for risk calculation and corresponding infographics to assist with decision-making between healthcare providers and patients at the bedside.

  • Implications: The more precise 30-day observed risk estimates have the potential to standardise ED management decisions, improve identification of those at risk and perhaps reduce unnecessary hospitalisation.

Introduction

Syncope, a sudden transient loss of consciousness followed by spontaneous complete recovery, accounts for ~1% of all emergency department (ED) visits.1–3 Approximately 10% will harbour a serious underlying condition (eg, arrhythmia, pulmonary embolus) which caused the syncope: with 2% obvious even before ED arrival; 5% identified during the index ED evaluation; leaving ~3% of patients with occult conditions identified only in the 30 days following ED disposition (either during hospital admission or following discharge).1 4–6 These patients with occult serious conditions are challenging for emergency physicians, as fear of missing them and imprecise risk-stratification manifests as wide variations in ED management decisions, investigations and disposition.2

Previously published risk-stratification tools do not provide clear clinical management options.7–10 We previously derived and validated the Canadian Syncope Risk Score (CSRS; figure 1) that comprises of nine predictors with the total score for a given patient ranging from −3 to +11. The score provides risk-stratification into five distinct risk levels based on the total score: very low (−3 to –2), low (−1 to 0), medium (+1 to +3), high (+4 to +5) and very high risk (≥+6), with the 30-day risk ranging from <1% to >50%. The tool focuses on the identification of serious conditions that become evident after the index ED visit. Management implications based on the 30-day risk have not been previously reported.

Figure 1

The Canadian Syncope Risk Score. *Triggered by being in a warm crowded place, prolonged standing, fear, emotion or pain. †lncludes coronary or valvular heart disease, cardiomyopathy, congestive heart failure and non-sinus rhythm (ECG evidence during index visit or documented history of ventricular or atrial arrhythmias, or device implantation). ‡Includes blood pressure values from triage until disposition from the emergency department. § Shrinkage-adjusted expected risk.

We previously reported on the CSRS performance characteristics, discrimination and calibration, in the validation cohort (n=3819). With a large cohort of patients (n=8233) enrolled during the two study phases, derivation and validation, robust estimates on model performance, observed 30-day risk of serious outcomes, including the types of serious outcomes, is possible. Such robust prognostic information can provide vital information to the clinicians for disposition decision-making. If such estimates can be readily calculated on-line and displayed using easy-to-understand infographics, it will aid in personalised risk prediction and patient-centred disposition decision-making.

The primary objective of this analysis was to pool the derivation and validation cohort data to provide more precise 30-day serious outcome risk estimates observed after ED disposition for each CSRS/stratum in preparation for an on-line calculator and pictorial decision aids for shared management decision-making.

Methods

Description of the CSRS studies

We performed a secondary analysis of pooled data from two previously completed, prospective cohort studies (the derivation and the validation phases of the CSRS), conducted to derive and to validate a risk tool for 30-day serious outcomes after the index ED visit in 11 Canadian EDs (online supplemental appendix 1). This study reports the results of the pooled data without any modifications to the methodology, CSRS predictors or the outcomes previously reported.5 11 The methods are described in the CSRS derivation and validation publications. In brief, both studies enrolled adults (≥16 years) with syncope and excluded patients who had a serious condition identified during the index ED visit. As the studies collected only routine clinical data and were non-interventional, the ethics committees approved the studies with the requirement of only verbal consent.

Supplemental material

Outcome

Our prespecified primary outcome for the current study was the identification of an occult serious condition (online supplemental appendix 2 ), the same as reported in the derivation and validation publications, and is subclassified as either arrhythmic (any serious arrhythmias; intervention to treat arrhythmias such as pacemaker/defibrillator insertion, or cardioversion; or any death due to an unknown cause) or non-arrhythmic (myocardial infarction, serious structural heart disease, aortic dissection, pulmonary embolism, severe pulmonary hypertension, significant haemorrhage, subarachnoid haemorrhage or any related serious condition) within 30 days of the index ED visit. For subjects who died, we reviewed all available medical records, and classified death due to an unknown cause as being due to arrhythmia. Our list of outcomes and the 30-day time frame were deemed most clinically relevant for ED risk prognostication by an international panel of experts.12 For patients who experienced any serious outcome, we collected the time and phase of care (ie, prehospital, in the ED, as inpatient, or after the index visit discharge) during which the serious outcome was first identified. Patients adjudicated to have a serious underlying condition identified in the prehospital setting and during the index ED evaluation were excluded from analysis. All serious outcomes including the time and the phase of care during which they were identified were adjudicated by a committee of two emergency physicians blinded to the predictors. Disagreements were resolved by a third physician. We did not readjudicate the serious outcomes previously adjudicated.

Data analysis

We used descriptive statistics and for each patient in the pooled cohort, we calculated their total CSRS, and their CSRS risk category: very low, low, medium, high and very high. For patients in whom either ECG or serum troponin testing was not performed, these variables were imputed as being normal to allow score calculation consistent with prior reporting.5 11 For the pooled cohort, using the CSRS predictors in a logistic regression model, we estimated the adjusted ORs with 95% CIs for each predictor. We report the area under the receiver operating characteristic curve (AUC) with 95% CI as a measure of discrimination and the calibration slope as a measure of calibration. Discrimination measures the ability of the model to distinguish patients with and without the study outcome, while calibration measures the numerical concordance between the expected and observed risk in the study cohort at different risk levels. A calibration slope measure ranges from 0 to 1 with one indicating excellent calibration across a range of risk groups.

Apart from the above outlined reporting measures for prediction models, for this study, we compared the observed and expected risk at each CSRS level in the pooled cohort and collapsed scores 6 or higher because of the small number of patients with higher scores. The sensitivity, specificity, and negative and positive predictive values of the CSRS dichotomised at each score level and risk category threshold were calculated. We reported the proportion of patients who suffered the serious outcomes (observed risk) for each score level and compared this risk among the CSRS risk categories using the Cochrane-Armitage trend test. We report the proportion of non-low-risk (medium or above) patients suffering non-arrhythmic serious outcomes each day in the 30 days following the index ED visit, to inform length of hospitalisation decisions. We used SAS (V.9.4) for data analysis. We used the observed serious outcome probabilities to develop the personalised online risk calculator and patient information materials.

We did not calculate a priori sample size as this was a secondary analysis of pre-existing data. The size of the pooled cohort was dictated by the sample size requirements for the derivation and validation phases of CSRS previously reported.5 11

Patient and public involvement

No patients or public were involved.

Results

In the two original cohorts 25 617 patients were screened, 10 916 were deemed potentially eligible and 1938 patients (17.8%) were not enrolled. Of the 8978 patients enrolled, 465 patients (5.2%) had a serious underlying condition identified during the index ED visit and 280 patients (3.1%) were lost to 30-day follow-up leaving 8233 patients in the pooled cohort for analysis. Most were older (age 53.5±22.9 years), two-thirds arrived by ambulance and 739 (9.0%) were hospitalised despite no serious condition identified during index ED evaluation (table 1). Overall, 295 patients (3.6%; 95% CI 3.2% to 4.0%) had an occult serious condition identified within 30 days: 203 patients (2.5%) with an arrhythmic condition including 23 patients (0.3%) with death due to an unknown cause, and 92 patients (1.1%) with non-arrhythmic conditions (table 2). The proportion of patients who suffered 30-day serious outcomes during the derivation and validation phases are detailed in online supplemental appendix 3. The majority of serious conditions were cardiovascular, but gastrointestinal bleeding and pulmonary embolism accounted for most of the rest. A total of 34 patients (0.4%) died within 30 days of the index visit.

Table 1

Characteristics, emergency department management and outcomes of patients

Table 2

Thirty-day serious outcomes after emergency department disposition among patients with syncope

A total of 321 patients (3.9%) did not have an ECG performed and 3950 patients (48.0%) did not have troponin measured during the ED evaluation. These patients were generally younger and healthier, (online supplemental appendices 4 and 5) and hence, these missing predictors were imputed as normal. The adjusted ORs for the CSRS predictors in the pooled cohort were comparable to the original derivation model, but the larger sample allowed for more precise estimates (table 3). As in the derivation and validation phases, the CSRS maintained excellent discrimination with an AUC of 0.88 (95% CI 0.87 to 0.91). The AUC during the derivation phase was 0.87 (95% 0.84 to 0.89) and the validation phase was 0.91 (95% CI 0.88 to 0.93). The calibration slope was 1.0 for the observed risk against the expected risk, and the observed and predicted serious outcome probabilities for each CSRS level were similar (online supplemental appendix 6).

Table 3

Multivariable logistic regression model with Canadian Syncope Risk Score predictors using the derivation and pooled cohorts

In the pooled cohort, 0.3% (95% CI 0.2% to 0.5%) very-low-risk and 1.0% (95% CI 0.7% to 1.5%) of low-risk patients had an occult, 30-day serious outcome identified only after the initial ED visit; this proportion increased substantially to 42.7% (95% CI 35.0% to 50.7%) in the very-high-risk group (Cochrane-Armitage trend test p<0.001; table 4). The risk increased progressively at each score level and for each risk category, even when subdivided into arrhythmic and non-arrhythmic conditions (online supplemental appendix 7). The sensitivity, specificity and the negative predictive value for occult 30-day serious outcomes at CSRS low-risk level (score of ≤0, ie, patients with a score of 0 or lower discharged), were 87.5% (95% CI 83.1% to 91.0%), 78.1% (95% CI 77.1% to 79.0%) and 99.4% (95% CI 99.2% to 99.6%), respectively (online supplemental appendix 8).

Table 4

Thirty-day serious outcomes for each Canadian Syncope Risk Score category

Overall, 37 (2.5% (95% CI 1.8% to 3.4%)) medium-risk patients and 32 (6.2% (95% CI 4.4% to 8.7%)) high-risk and very-high-risk patients (table 4) experienced non-arrhythmic serious outcomes within 30 days after the index ED visit. Most of these occult conditions were identified within a few days of the index visit (online supplemental appendix 9). The proportion of patients hospitalised and those who had occult serious condition identified during hospitalisation for each CSRS risk strata are detailed in online supplemental appendix 10.

To facilitate bedside calculation of CSRS, we developed an online tool available at https://teamvenk.com/csrs for clinicians assessing ED patients with syncope (online supplemental appendix 11). Infographics corresponding to each risk stratum and each medium-risk score level were also developed to allow for individualised discussion before making disposition decisions. The infographic for the very-low and low-risk strata is shown in figure 2 and all other infographics are detailed in online supplemental appendix 12.

Figure 2

Fainting (syncope) patient information sheet.

Discussion

In this study using our original derivation and validation cohorts combined, we report with greater precision the risk of 30-day serious outcome after the index ED visit at each CSRS level and for each CSRS risk category. Collapsing the score into five distinct risk categories provides good separation into strata that are clinically relevant. The precise 30-day risk estimates show serious outcomes are low and specifically death and ventricular arrhythmia are rare among patients in the low-risk and very-low risk categories. The risk of death and ventricular arrhythmia was also very-low in the medium-risk group; however, a smaller but important proportion did suffer arrhythmic and non-arrhythmic serious outcomes. The overall risk of 30-day serious outcomes including death and ventricular arrhythmia is high for the CSRS high-risk and very-high-risk patients. We also reaffirm that the risk of an occult non-arrhythmic serious condition is front-end loaded in the first few days following syncope.

Our results are consistent with previous reports which estimate a very low overall short-term risk of serious morbidity (~0.3% ventricular arrhythmia) and mortality (<1%) following an ED visit for syncope when a cause is not identified in the ED.1 4 However, this risk is not equally partitioned across all patients. More challenging is to identify those who will benefit most from inpatient admission or expedited ambulatory cardiac rhythm monitoring. Three out of four in our cohort were designated as very-low risk or low risk and less than 1% experienced 30-day serious outcome with only one patient suffering death in the very-low-risk and low-risk categories. Our data suggest that these patients can be safely discharged following appropriate ED evaluation. The one death among more than 6000 patients classified as very-low risk or low-risk involved a young, immunocompetent female diagnosed with ‘vasovagal syncope secondary to groin pain’ at the index ED visit. Four days later she returned with sepsis, was admitted to the intensive care unit, developed septic shock and died 11 days after the index ED visit. Though ≤1% of the CSRS very-low-risk and low-risk patients suffered 30-day serious outcomes with only one death secondary to sepsis in this group, a sensitivity of 87.5% may be unacceptable to physicians. However, physicians in discussion with patients can make disposition decisions unconstrained based on the results reported in online supplemental appendix 8.

Another one in five patients in our cohort were classified as medium-risk. Of these, 7.8% suffered a 30-day serious outcome including 0.7% death, 0.6% ventricular arrhythmia and another 4.3% non-ventricular arrhythmia. Such patients may represent an appropriate target population for brief monitoring in the ED, followed by prolonged outpatient external cardiac rhythm monitoring to allow identification of most arrhythmias, which also tend to occur in the subsequent 15 days.6 Disposition for this group should be predicated on circumstances, values, preferences and goals, especially the ability to arrange outpatient monitoring.13 14 The patient infographics we developed (online supplemental appendix 13) can be used to aid in management and disposition decisions. In this study we neither attempt to develop or validate a patient decision aid by involving patients, or validate the tool, but rather present the outcomes for patients in the different CSRS risk strata to assist in the decision-making.15

The remaining patients, which represent approximately 1 in 20 of the enrolled cohort, were classified as either CSRS high-risk or very-high risk, and inpatient admission seemed appropriate. More than a quarter of these patients have a 30-day serious condition identified. Indeed, about half of the 30-day serious outcomes in our study were in these two strata, and most non-arrhythmic conditions occurred within 2–3 days suggesting that inpatient investigation and observation is a reasonable option.

Four other tools risk-stratify ED syncope patients for short-term serious outcomes.7–10 The San Francisco Syncope Rule (SFSR) has performed variably on external validation taking into account the methodological weaknesses in some of the validation studies.4 7 16 17 The Short-Term Prognosis of Syncope tool has not yet been validated.8 The Risk stratification Of Syncope in the Emergency department (ROSE) tool was derived and validated at a single centre in the UK.9 The FAINT Study (history of heart Failure or Arrhythmia, Initial abnormal ECG, elevated N-terminal pro-B-type natriuretic peptide, or elevated high-sensitivity troponin T) enrolled only older (≥60 years) adults with syncope, requires natriuretic peptide measurements, and also has not yet been validated.10 FAINT dichotomises patients into lower risk and higher risk with 97% sensitivity and only 22% specificity. The SFSR and ROSE derivation studies included patients with serious conditions clearly evident during ED evaluation, biasing these tools towards identification of the obvious.12 Consequently, an international panel recommended that ED syncope risk-stratification studies exclude patients with serious conditions identified in the ED.12 The CSRS addresses the above limitations and can aid in clinical decision-making regarding investigations and disposition.

While the proportion of ED patients with syncope who subsequently experience a 30-day serious outcomes has consistently been reported at approximately 10%, wide variations (12% to 83%) in hospitalisation exist even among similar institutions.2 18 We believe that this is not only due to variations in provider/patient risk tolerance, but more fundamentally because of the poor precision of risk estimates following ED syncope evaluation. Given that the causes for syncope can be heterogeneous and complex, the CSRS is able to provide risk estimates as recommended by the professional societies.19 We believe that use of CSRS can help to standardise hospitalisation decisions and will provide support for a more informed reduction in some.

Limitations

Our study has limitations. We have previously reported the limitations pertaining to derivation and validation studies and include inability to enrol approximately 18% of potentially eligible patients, troponin values missing in near half of those enrolled, 3% of patients lost to follow-up (online supplemental appendix 13), and potential bias related to difficulty in differentiating between syncope and seizure.5 11 As previously reported, eligible patients were not enrolled as the physicians were too busy to complete the forms.5 However, we did not find any systematic bias for non-inclusion. Patients with missing troponin values were young with no comorbidities and those lost to follow-up did not die when crosschecked against regional or provincial health data platforms. The CSRS includes a subjective ED diagnostic impression predictor, vasovagal or cardiac syncope. As the data used were from the previously studied cohorts, the risk estimates reported are limited to the two cohorts and both studies were conducted at Canadian centres and hence, the findings may not be generalisable. Additionally, the availability of appropriate investigations to rule out underlying serious cause for syncope in the ED and system mandates such as the ‘4 hour target’ by NHS UK will impact the ability to complete the evaluation prior to CSRS application.

Conclusions

Using a large data set from two prior studies, the CSRS derivation and validation, we found that the score can be used to assign patients to one of five risk categories with good accuracy and precision to identify the 30-day risk for an occult serious condition after the index ED visit. This stratification can be used for management decisions. We provide an online tool for point-of-care risk calculation, as well as infographics for patient-centred discussion and decision-making. We believe these results will improve and personalise risk prediction; the online tool will facilitate accurate score calculation with precise definitions used in the development, and generate patient information materials to promote better risk communication with patients. Overall, our study results have the potential to standardise ED disposition decisions, improve identification of those at risk and perhaps reduce unnecessary hospitalisation.

Data availability statement

No data are available.

Ethics statements

Patient consent for publication

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Handling editor Edward Carlton

  • Twitter @TeamVenk?s=20

  • Contributors VT, JWY, BHR, EM, NLS, MH, PH, MM, ADM, MT and MLAS conceived the idea, contributed to the study design, developed the study protocol and applied for funding. VT, JY, BHR, EM, NLS, MH, AF, PH, MM, PAN, ADM, M-JN, MT and MLAS supervised the conduct of the studies including recruitment of patients, data collection, data management including quality control. MT provided statistical advice on study design. M-JN analysed the data under MT’s supervision. HM, PAN and SS interpreted the results, designed and developed the online calculator, pictograms and patient information materials. VT drafted the manuscript. All authors reviewed the manuscript and contributed substantially to its revision. VT takes responsibility for the paper as a whole.

  • Funding The two prospective studies from which these data were taken were funded by The Physicians’ Services Incorporated Foundation (09q4017), Canadian Institutes of Health Research (MOP-114927), Heart and Stroke Foundation Canada (G-15–0009006), and the Cardiac Arrhythmia Network of Canada (SRG-15-P10-001) as part of the Networks of Centres of Excellence (NCE).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.