Comparison of different risk stratication rules to predict short-term adverse outcomes after syncope in older Chinese adults

Background: Older adults with syncope are commonly treated in the emergency department. Clinical decision rules have been developed to assess syncope patients, but there have been no application or comparative studies in older Chinese cohorts until now. This study aimed to compare the values of ve existing rules in predicting the short-term adverse outcomes of older patients. Methods: From September 2018 to February 2021, older Chinese patients ( ≥ 60 yr) with syncope admitted to our hospital were investigated and evaluated by the Risk Stratication of Syncope in the Emergency Department (ROSE) rule, the San Francisco Syncope Rule (SFSR), the FAINT rule, the Canadian Syncope Risk Score (CSRS) and the Boston Syncope Criteria (BSC). After a one-month follow-up, the sensitivity, specicity, accuracy, positive predictive values (PPV), negative predictive values (NPV), positive likelihood ratios (PLR), and negative likelihood ratios (NLR) of each aforementioned rule were calculated and compared. Results: A total of 171 patients, with a mean age of 75.65±8.26 years and 48.54% male, were analysed in the study. Fifty-eight patients were reported to have experienced short-term adverse incidents during the month. The neurally mediated syncope group showed a signicant sex-specic difference in adverse incidences but the cardiac syncope group did not. There were some factors associated with signicant differences in adverse incidences, such as a history of hypertension, congestive heart failure, and chronic obstructive pulmonary disorder, as well as the levels of SpO2, B-type natriuretic peptide (BNP) and troponin T (TnT), while the levels of haemoglobin and creatinine suggested potential signicance. In order of the ROSE, SFSR, FAINT, CSRS and BSC rules in the analysis, the sensitivities were 81.03%, 77.59%, 93.10%, 74.14% and 94.83%, the specicities were 86.73%, 84.96%, 38.94%, 60.18% and 56.64%, the NPVs were 89.91%, 88.07%, 91.67%, 81.93%


Background
Syncope is a symptom of cerebral hypoperfusion and is de ned as a short, sudden, self-termination episode of transient loss of consciousness (T-LOC) with a brief period of unresponsiveness and a failure to maintain postural tone, ultimately resulting in spontaneous recovery that requires no resuscitation measures. Symptoms have different aetiologies and are a common clinical problem affecting millions of patients worldwide [1]. Although the incidence rate of syncope varies in different areas of the world, it accounts for approximately 1-5% of all emergency department (ED) visits and has shown an increasing trend in recent years [2][3]. Approximately 30-40% of syncope patients were admitted to the hospital for further diagnosis and treatment. For patients over 60 years old, the proportion of hospitalized syncope patients was as high as 50%. At the same time, body reactions and complications after syncope are even worse for elderly patients than for younger patients, which makes clinical evaluation challenging for physicians [4][5][6].
Early risk strati cation during the initial evaluation is crucial to guide clinical decisions and prevent short-term adverse events. The management of syncope patients is a di cult dilemma since its pathophysiological processes and causative factors vary and are complex.
Some patients appear well when they arrive at the hospital, and there are no witnesses to the event, while some will require emergency hospitalization for the treatment of potentially life-threatening causes [7][8]. In this regard, clinical decision rules (CDRs) can be helpful for clinicians. Several clinical models used in emergency departments are designed to assist physicians in making decisions at the bedside and performing risk strati cation of patients, such as the Risk Strati cation of Syncope in the Emergency Department (ROSE) rule, the San Francisco Syncope Rule (SFSR), the FAINT rule, the Canadian Syncope Risk Score (CSRS) rule and the Boston Syncope Criteria (BSC) rule.
Considering that these rules were derived in a few centres and some models are newly developed, there are limited external validations and even different ndings from some clinical studies [9][10][11][12][13][14][15][16]. Therefore, the present study aimed to explore and compare the values of the ROSE, SFSR, FAINT, CSRS and BSC rules in recognizing syncope patients at risk for short-term adverse events. To the best of our knowledge, there are no similar studies on older Chinese syncope patients regarding the application of these ve CDRs in a one-month follow-up investigation.

Study design and setting
This was a single-centre retrospective observational cohort study performed in the ED of a training and research general hospital. All patients were observed at their rst consultation in our hospital's ED between September 2018 and February 2021. There were no xed decision tools for syncope management. The one-month follow-up study was designed to evaluate the predictive values of the ROSE, SFSR, FAINT, CSRS and BSC rules in recognizing adverse events. As the study was retrospective and observational, no identi able patient data were collected, and the involved data were stored on a dedicated secure computer. The study protocol was approved by the Ethics Committee of Beijing Tiantan Hospital, Capital Medical University. During the course of the study, the researchers adhered to the principles of the Declaration of Helsinki.

Study population and data collection
Patients aged 60 years or older presenting to the ED with syncope (de ned as T-LOC with or without prodromal symptoms characterized by a short duration and spontaneous full recovery) were eligible for the study. The exclusion criteria used to determine the target population were as follows: 1. Patients aged below 60 years; 2. Patients with con rmed nonsyncope syndromes (such as seizure, vertigo, coma, shock, sustained unconsciousness, head injury preceding the loss of consciousness, and stroke); 3. Patients with comorbidities with a low survival rate or a new serious diagnosis identi ed in the ED (such as advanced malignant neoplasm, dyscrasia, circulatory failure, respiratory failure, aortic dissection, pulmonary embolism, myocardial infarction, subarachnoid haemorrhage, major traumatic injury, and haemorrhage requiring blood transfusion); and 4. Patients with a history of drug or alcohol abuse. The nal decision of the patients who were enrolled was made after the patient evaluation forms were completed, and the form contained dozens of variables derived from existing syncope rules. The attending physicians, including an emergency medicine specialist, a cardiologist, and a neurologist, evaluated patients independently to assess subjective variables requiring interpretation. All the patients were treated in the usual manner by a physician blinded to the aims of the study.
A total of 230 older patients who were referred to the ED with complaints of syncope were screened, and 50 patients were excluded from this analysis due to a new serious diagnosis in the ED or a lack of su cient data to calculate CDR scores, so 180 patients were enrolled. Since 9 patients were lost to follow-up, 171 patients were eligible for the study, as shown in Figure 1.
Clinical and laboratory data were extracted for analysis and stored in a predesigned database. We used the dedicated database to calculate the scores together according to the ROSE, SFSR, FAINT, CSRS, and BSC rules to reduce personal interpretation bias. The ROSE rule considers patients positive and high risk, if any of the following are present: a B-type natriuretic peptide (BNP) level ≥300 pg/ml or bradycardia ≤50 in the ED/pre-hospital period, a rectal examination showing faecal occult blood, anaemia with a haemoglobin level ≤90 g/l, chest pain associated with syncope, an abnormal electrocardiogram (ECG) showing Q wave but not in lead III, or an SpO2 ≤94% on room air [9]. The SFSR deems patients positive and high risk in the presence of any of the following: a history of congestive heart failure, a haematocrit level < 30%, an abnormal ECG, shortness of breath, and a systolic blood pressure <90 mmHg [10]. In the FAINT rule, the presence of any sign of history of heart failure, history of cardiac arrhythmia, an initial abnormal ECG result, elevated N-terminal pro-BNP, or elevated high-sensitivity troponin T (TnT) puts the patients in the non-low-risk group for adverse outcomes. Abnormal ECG interpretations were similar to previous research [11]. Here, "FAINT2" stands for the cut-off value "2" used for evaluation by the FAINT rule, which was set up according to the latest study [12]. The CSRS rule was developed by assigning points to each of the factors based on the relative magnitude of the coe cient, with a score ranging from -3 to 11, and includes nine elements in the categories of clinical data, investigations, and presumptive ED diagnosis. A score of 1 or higher was judged as high risk [13][14]. In the BSC rule, the presence of any of the factors, such as acute coronary syndrome, cardiac history, a family history of sudden death, valvular heart disease, signs of cardiac conduction disease, circulating volume depletion, persistent abnormal vital signs in the ED, and brain lesions, put the patient in the high-risk group [15].
All patients were followed up by telephone call or by searching the electronic medical record system to identify adverse outcomes within one month of the rst referral to the hospital with complaints of syncope. Evaluated short-term adverse outcomes were de ned a priori as one of the following events: mortality, cardiac pulmonary resuscitation (CPR), acute myocardial infarction (AMI), cardiac arrest, severe arrhythmia, pacemaker implantation or implanted-cardiac de brillator (ICD), cardiac stent insertion, stroke, cerebrovascular accidents (CVAs), haemorrhage requiring urgent treatment, nontraumatic intracranial bleeding or pulmonary embolism (PE).

Statistical analysis
A descriptive statistical analysis was carried out to characterize the pro le of the study population. The patients were divided into three age groups: 60 to 69, 70 to 79, and ≥80 years of age. Quantitative data are reported as the mean ± standard deviation or standard error, and qualitative data are reported as the frequency and percentage. Differences among variables were evaluated by the chi-square test and Fisher's exact test when appropriate.
To evaluate each CDR in predicting outcomes, the sensitivity, speci city, accuracy, positive predictive values (PPV), negative predictive values (NPV), positive likelihood ratios (PLR), and negative likelihood ratios (NLR) were calculated to identify patients who experienced adverse outcomes during the follow-up period. Receiver operating characteristic (ROC) curves were generated, and the areas under the ROC curve (AUCs) were compared using a previously proposed method to assess the CDRs' predictive values [17][18]. To determine the concordance of the CDRs' assessments with the presence of any adverse events, kappa coe cients for each criterion were calculated. A two-tailed p value <0.05 was considered to be statistically signi cant for all analyses. The statistical analysis was carried out using IBM SPSS version 20 (IBM, New York, USA).

Characteristics of the study subjects
In the present study, 171 patients with complete follow-up data at one month were analysed. The characteristics of the study patients are shown in Table 1. The mean age of the study sample was 75.65±8.26 years, and the median age was 75 years. Among the cohort, 48.54% were men and 51.46% were women. Compared with the men, the women were older (78.00±7.99 years vs. 73.16±7.83 years, respectively, p<0.05), and male sex was less common among the very elderly patients with syncope (15/59) but more frequent in the other age groups (68/112). The median concentration of BNP was 122.00 pg/ml, the mean concentration of haemoglobin was 129.82±24.35 g/L, the mean indoor oxygen saturation without oxygen inhalation was 95.46±2.81%, the mean systolic blood pressure was 126.14±16.05 mmHg, and the mean haematocrit level was 38.31±6.73%. There were signi cant differences among the three age groups in terms of haemoglobin, haematocrit, BNP, TnT, and BUN levels, and the ratio of BUN/creatinine. For CHF and creatinine levels, there were potential differences among the three age groups, with a wider statistical signi cance of a p value <0.1.    Figure 2).

Discussion
Despite some limitations of our study, to the best of our knowledge, this is the rst reported cohort study on the comparisons of ve key CDRs for syncope patients in an older Chinese population. Since the number of patients suffering from syncope-related adverse outcomes increases sharply after 60 years [19], older patients mostly have multiple comorbid conditions, and risk strati cation is based on probable short-term adverse outcomes, it is crucial to distinguish non-low-risk patients for further timely precise therapies. We conducted this study to compare our clinical applications of the CDRs for the assessment of older adults with syncope. A total of 171 enrolled patients were followed up for one month and practical adverse outcomes were compared against the ROSE, SFSR, FAINT, CSRS, and BSC CDRs.
Overall, 33.92% of our study cohort had an adverse outcome at one month. The sex was roughly the same (males 83 vs. females 88), but the women were signi cantly older than the men (p<0.05). Women live longer and pay much more attention to their health care than men [20]. In our exploration, the adverse incidences were affected by the comorbidities of hypertension, CHF, and COPD, as well as the levels of SpO2, BNP, and TnT. In addition, our study also suggested that the levels of haemoglobin and creatinine have potential in uences on adverse incidences. SpO2, BNP and haemoglobin are three variables in the ROSE rule, CHF is a variable in the SFSR, FAINT, and BSC rules, and BNP and TnT are two variables in the FAINT rule. In addition, the SFSR, CSRS and BSC rules all emphasize the measurement of blood pressure [21]. However, creatinine has not been involved in any CDRs until now, while the level of creatinine is relevant to inner body uid metabolism and may affect the body's circulatory perfusion before syncope.
For the four diagnostic categories for syncope, the number of patients and the adverse incidences were greater in cardiac syncope patients than other patients, which is in accordance with previous studies on advanced age [22]. It seemed that NMS patients had sex-speci c and age-related differences in the incidence of adverse events, but CS patients suggested no similar results. Compared with CS, NMS is a kind of benign syncope, and neural regulation has differences between sexes and ages, while CS is a serious symptom with high mortality, so the syncope type itself has its own characteristics and clinical prognosis [23][24][25]. In the research, more than half of the adverse incidences in one month were relevant to AMI and arrhythmia. Cardiac syncope is associated with higher mortality, irrespective of age and sex. Thus, the ve CDRs in our investigation all included cardiac elements in their assessments and highlighted the signi cance of cardiogenic parameters, which re ects the importance of cardiac syncope in the elderly population.
The ROSE and SFSR rules showed higher speci cities and NPVs, and AUC areas were greater than 0.8, which meant that they had signi cance in identifying and screening non-high-risk syncope patients, which was roughly consistent with the previous clinical studies of each CDR [9][10]. In addition, the kappa coe cients for the concordance of these two CDRs with short-term adverse events were greater than 0.7, which suggests more reliable consistency. The AUC areas of the FAINT, FAINT2, CSRS, and BSC rules were all greater than or approximately 0.7. The FAINT and BSC rules had higher sensitivities and NPVs than the other rules, while the CSRS rule had moderate results in our clinical applications. It seemed that the FAINT and BSC rules had advantages in identifying high-risk syncope patients. The BSC rule has included the most comprehensive elements for syncope assessment until now, indicating high risk if any one item is positive. The kappa coe cient of the BSC rule was greater than 0.6, which suggests substantial consistency. The CSRS rule assigns various points to each of the factors based on the relative magnitude of the coe cient, and its external validation in an Australian cohort of 283 patients was completed in 2020, so larger and more widespread clinical applications are needed in the future. The kappa coe cient of the CSRS rule was close to 0.5, which suggests moderate consistency.
The FAINT rule was set up for elderly patients in 2020, and the preliminary results of the recent multicentre external validation in Europe and the US suggest that it is safe to evaluate elderly patients with scores of 0 or 1 [12]. The cut-off value of "1" was used in the original study for patient evaluation analysis, which means that 24 patients were "misjudged" as positive in our study if the cut-off score was 1, which may cause a higher proportion of low-risk patients to spend more money on further clinical examinations. Therefore, we additionally set "2" as another cut-off score for evaluation by the FAINT rule (FAINT2) to make the comparison with short-term prediction for elderly patients. Our FAINT results indicated that the sensitivity and NPV were higher and the speci city was lower, which was consistent with the original studies by Probst [11]. On the other hand, the results of the FAINT2 rule indicated that the speci city and PLR were higher and the sensitivity was lower. The areas of both AUCs were higher than 0.6, while the area and accuracy of the FAINT2 rule were slightly higher. The kappa coe cients of the FAINT and FAINT2 rules were both greater than 0.5, indicating moderate consistency. In that case, what is the signi cance of original research on the FAINT rule to improve the evaluation sensitivity at the expense of speci city? Considering that the proportion of cardiac syncope in elderly patients is higher, its complications are more dangerous when it occurs. The higher sensitivity of the FAINT rule may avoid the possibility of missing adverse events at best, but it requires more relevant medical resources, so the FAINT rule may be much more valuable for elderly people who have any history of cardiovascular diseases under the condition of su cient medical resources. Probst gave a deep explanation of the original exploration: The FAINT rule sets no serious adverse cardiac events' omission as the main point, so its high sensitivity and low NLR were the original intentions of the study. However, this rule is not suitable for screening low-risk patients at present since international multicentre validation has not been completed. As we know, the FAINT rule has been evaluated in a clinical study of elderly syncope patients with the largest sample size until now, but the population included patients aged 60 years and above with syncope or near syncope (the sensation of the impending loss of consciousness without the actual loss of it), and the patient refusal rate was as high as 53.2% when the study was implemented, which could be a kind of sample bias. The FAINT rule includes two biochemical elements relevant to cardiac function and myocardial damage. Compared with the ROSE rule, which includes only BNP levels, and the SFSR rule without biochemical elements, the FAINT rule certainly improves the laboratory work and medical cost, but it is helpful to reduce the diagnosis rate of "unexplained syncope" in clinics, so its net effect analysis in health economics is worth further exploration in the future. At the same time, the FAINT rule is much simpler than the CSRS and BSC rules to use in assessments. These ve rules all showed the assessment values for elderly syncope patients in practice. To date, there is still no syncope risk strati cation rule that can be used independently of physician judgement, and it only plays a role in assisting judgement and decreasing the cognitive load for physicians when making clinical decisions.
Overall, the proportion of one-month adverse incidences in our study was 33.92%. It was higher than that reported in the original studies of the ROSE, SFSR, FAINT, CSRS and BSC rules, but similar to that reported in other studies [26][27]. The discrepancy between the proportion of short-term adverse events was relevant to the different levels of public health services and disease preventive abilities in the cohort population. Additionally, it was related to the characteristics of the enrolled patients: the median age of the elderly patients in this study was 75 years, and 69.59% of them were over 70 years, whose incidence of adverse events was higher than that of patients aged 60-69 years (38.66% vs. 23.08%). The incidence of syncope in the elderly population generally increases with age, but the body is in a declining stage as it ages, so the physiological stress response after syncope may be partly damaged, and the complications could be more complex, which leads to the risks of adverse events being much higher [6, 28-31]. Moreover, the original FAINT study included "near syncope" patients whose clinical risk status was relatively low.

Study limitations
Our study was conducted at a single centre with limited resources. First, the international multicentre external validation of the FAINT and CSRS rules is still in progress. Larger sample sizes and multicentre data will certainly result in better assessments of the current risk rules. Second, the well-designed prospective study will be worthy of further exploration of its clinical signi cance based on previous retrospective work. Third, the confounders used in the risk strati cation may not have been inclusive of all causes or in uencing factors of syncope. For example, creatinine could possibly have in uenced the adverse incidences in our study, which needs further investigation in the future. Fourth, we did not collect information regarding secondary diagnoses when syncope patients arrived at the emergency department, and these data may be useful in further clarifying outcomes among patients. These limitations are expected to be improved in future research.

Conclusions
Our study could be helpful to understand the pros and cons of the risk strati cation rules, and the exible application of the ROSE, SFSR, FAINT, CSRS and BSC rules is bound to improve the e ciency of clinical evaluations. Physicians can judge patients' prognoses or short-term adverse outcomes more quickly and more scienti cally under the conditions of overloaded clinical work, considering the complexity of syncope aetiologies and most elderly patients with some chronic diseases.
Abbreviations ROC curves of the risk strati cation rules