The Value of the CHA2DS2VASc Score and the Essen Stroke Risk Score in a Nationwide Stroke Cohort
Background and Purpose—The CHA2DS2VASc score and the Essen Stroke Risk Score are respectively used for risk stratification in patients with atrial fibrillation and in patients with cerebrovascular incidents. We aimed to test the ability of the 2 scores to predict stroke recurrence, death, and cardiovascular events (stroke, transient ischemic attack, myocardial infarction, or arterial thromboembolism) in a nationwide Danish cohort study, among patients with incident ischemic stroke and no atrial fibrillation.
Methods—We conducted a registry-based study in patients with incident ischemic stroke and no atrial fibrillation. Patients were stratified according to the CHA2DS2VASc score and the Essen Stroke Risk Score and were followed up until stroke recurrence or death. We estimated stratified incidence rates and hazard ratios and calculated the cumulative risks.
Results—42 182 patients with incident ischemic stroke with median age 70.1 years were included. The overall 1-year incidence rates of recurrent stroke, death, and cardiovascular events were 3.6%, 10.5%, and 6.7%, respectively. The incidence rates, the hazard ratios, and the cumulative risk of all outcomes increased with increasing risk scores. C-statistics for both risk scores were around 0.55 for 1-year stroke recurrence and cardiovascular events and correspondingly for death around 0.67 for both scores.
Conclusions—In this cohort of non–atrial fibrillation patients with incident ischemic stroke, increasing CHA2DS2VASc score and Essen Stroke Risk Score was associated with increasing risk of recurrent stroke, death, and cardiovascular events. Their discriminatory performance was modest and further refinements are required for clinical application.
Various clinical scoring systems have been developed for the risk stratification in cardiovascular and stroke medicine. The CHADS21 and, more recently, the CHA2DS2VASc2 scores are based on well-recognized risk factors for stroke in atrial fibrillation (AF), including congestive heart failure, hypertension, age ≥75 years (2 points), diabetes mellitus, previous stroke or transient ischemic attack (2 points), vascular disease, age 65 to 74 years, and sex class (female). The scores have been validated in various cohorts3,4 and have become widely used for ischemic stroke risk stratification in patients with AF because of their simplicity.
To extend the possible application of these scores, there is some interest in validating the scores in non-AF patient populations. The CHADS2 and the CHA2DS2VASc scores have been shown to be associated with the risk of stroke in non-AF patients with ischemic heart disease,5,6 in patients undergoing coronary artery bypass grafting,7,8 in patients with acute coronary syndrome,9 and in addition, the scores hold promise for stroke and thromboembolism risk assessment in the general population.10
Risk stratification is of particular interest in patients with ischemic stroke as they are at high risk of a stroke recurrence (18% over a 5-year period11,12), and the recurrence risk is nonuniform across the population. A study based on the Athens Stroke Registry13 reported that the CHADS2 and the CHA2DS2VASc scores predict recurrent stroke and death in non-AF stroke patients. How these findings replicate in other patient populations is unknown. In addition, it is of interest to assess how the scores compare to risk scores developed specifically for the population of stroke patients. The Essen Stroke Risk Score is one such risk score, which is conceptually similar to the CHA2DS2VASc score, in that it adds points according to presence of risk factors—1 point for each of age 65 to 75 years, age above 75 years (2 points however), hypertension, diabetes mellitus, myocardial infarction, other cardiovascular disease (except myocardial infarction and AF), peripheral arterial disease, smoking, and previous stroke or transient ischemic attack, resulting in maximum score of 9 points.14,15
To evaluate how the CHA2DS2VASc score predicts recurrent stroke, death, and cardiovascular events, and in a first-ever stroke population, and how this general score compares to the more specific Essen Stroke Risk Score, we conducted a large-scale observational cohort study using nationwide Danish registry data. We hypothesized that both risk scores would exhibit comparable associations with the risk of stroke recurrence, death, and cardiovascular events.
Registry Data Sources
The study was based on The Danish Stroke Registry, The Danish National Patient Registry, The National Prescription Registry, and The Danish Civil Registration System. All 4 registries cover the entire Danish population, and linkage was facilitated via the unique personal identification number assigned to all Danish residents.
The Danish Stroke Registry was established in 2003. It is mandatory for all Danish hospital departments treating stroke patients to report a standardized set of data to the registry, including diagnosis and various clinical variables, such as selected comorbidities and smoking habits.16 The Danish National Patient Registry17 has registered all hospital admissions with corresponding discharge diagnoses since 1977. Up to 1993, all diagnoses were coded according to the 8th revision of the International Classification of Diseases (ICD-8), and since 1994, all diagnoses are coded according to the ICD-10. The National Prescription Registry18 holds data on all prescriptions dispensed from Danish pharmacies since 1994, coded according to the Anatomic Therapeutic Chemical Classification System. Finally, The Danish Civil Registration System contains information on date of birth, sex, migration, and vital status of all citizens.19
The study population consisted of all patients aged 18 years or above, registered in The Danish Stroke Registry with an incident ischemic stroke (ICD-10 codes I63 or I64) in the period January 1, 2003, and December 31, 2012 (index stroke). The diagnosis of stroke in The Danish Stroke Registry is based on the judgment of the treating physician(s). This judgment is based on clinical and neuroimaging information gathered during the admission. The sensitivity of a stroke diagnosis has previously been shown to be 91% to 97%.20 To include only patients with incident stroke, we identified the first occurrence of registration of ischemic stroke for each patient in the registry. If this record stated that the patient had experienced a previous stroke (judged by the treating physician based on the available information: medical files, previous neuroimaging, patients, or proxy self-report), the patient was excluded. Patients with AF and patients who died <14 days after the index stroke were excluded. To capture AF patients not diagnosed at a hospital, patients prescribed oral anticoagulants (warfarin, phenprocoumon, dabigatran, rivaroxaban, and apixaban) within 2 years before the index stroke were excluded.
To investigate potential differences between patients on secondary stroke prevention medications and the whole cohort, we identified a subgroup of patients treated with antiplatelet drugs (aspirin, clopidogrel, and dipyridamole) and statins. Treatment with these drugs was defined as a claimed prescription of both drugs in the period from 90 days before until 90 days after the index event.
Based on information in the registries, the CHA2DS2VASc score and the Essen Stroke Risk Score were calculated for all patients at the time of admission for the incident stroke. We followed registry-based algorithms previously decribed3 and added information from The Danish Stroke Registry. Definitions of comorbidities applied in the risk scores can be found in the online-only Data Supplement.
Patients were followed from the date of the index stroke and until an event of interest or death. The events of interest were recurrent ischemic stroke or a cardiovascular event. Cardiovascular events were defined as the composite of ischemic stroke, transient ischemic attack, myocardial infarction, or arterial thromboembolism. Follow-up was censored at time of emigration or end of study (December 31, 2012). Information on emigration or death was available from the Danish Civil Registration System. Cardiovascular events were identified in The Danish Stroke Registry and The Danish National Patient Registry. Recurrence of ischemic stroke was ascertained from The Danish Stroke Registry, defining recurrent stroke as admission with a new stroke after discharge for the index event >14 days after the index stroke. This restriction was used to reduce the risk of capturing double registrations of the index stroke.
Continuous data were summarized as mean values and standard deviations and categorical data as proportions. Event rates of recurrent ischemic stroke, death, and the composite cardiovascular end point were calculated for all patients, stratified by the CHA2DS2VASc score and the Essen Stroke Risk Score, respectively. Hazard ratios based on the Cox proportional hazards model were calculated for increasing values of the CHA2DS2VASc score and the Essen Stroke Risk Score, setting as a reference a CHA2DS2VASc score of 2 and an Essen Stroke Risk Score of 0, respectively. Patients with CHA2DS2VASc score ≥7 were merged into one stratum, and similarly patients with Essen Stroke Risk Score ≥5 were merged into one stratum. Further, we calculated estimates of Cox proportional hazards for each of the (binary) components comprised in the 2 risk scores.
The cumulative risk of stroke and cardiovascular events were calculated using the Aalen–Johansen estimator, taking into account the competing risk of death. The cumulative risk of death was calculated using the standard Kaplan–Meier estimator. The discriminatory performance of the scores was assessed with C-statistics, taking into account competing risks of death.21 Negative predictive values (the proportion of patients with a score below the cut-off who remained event-free and alive during follow up) were calculated for the end points of stroke and cardiovascular events, with cutoff values of 2 (CHA2DS2VASc) and 0 (Essen Stroke Risk Score), respectively.
We performed similar analyses for the subgroup of patients treated with antiplatelet drugs and statins.
Analyses were performed using Stata version 13 (Stata Corporation, College Station, TX). The study was performed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations.
The study was approved by the Danish Data Protection Agency (File No. 2012-41-0633). In Denmark, no ethical approval is required for anonymous registry studies.
In The Danish Stroke Registry, we identified 56 682 patients ≥18 years with incident ischemic stroke in the period January 1, 2003, to December 31, 2012. We excluded 3791 patients who died within 14 days after the index event, 9981 patients with AF, and 728 patients prescribed oral anticoagulants within 2 years before the date of the index stroke. Hence, the study cohort consisted of 42 182 patients with first-ever ischemic stroke and no AF.
Baseline characteristics of the cohort are shown in Table 1. Patients were followed up for an average of 3.5 (standard deviation [SD] 2.7) years. Based on the CHA2DS2VASc score, all patients were assigned 2 points for the index stroke, resulting in a mean score of 4.3 (SD 1.4). The mean Essen Stroke Risk Score was 2.4 (SD 1.4).
The overall incidence rates of recurrent stroke, death, and cardiovascular events during the first year of follow up were 3.6%, 10.5%, and 6.7% respectively. The corresponding overall incidence rates during the first 5 years of follow-up were 2.4%, 7.1%, and 4.3% per year, respectively. The stratified incidence rates are shown in Table 2. In the Cox proportional hazard analyses, the hazard ratios increased with increasing risk scores as shown in Table 3. Correspondingly, the cumulative risk of the outcomes increased with increasing risk score as shown in Figures 1 and 2.
Cox proportional hazard analyses for the single components of the scores are found in Table I in the online-only Data Supplement. All components, except for female sex, were positively associated with the outcome of stroke recurrence. The strongest predictor of stroke recurrence was peripheral arterial disease. All factors were positively associated with the risk of death, except for smoking.
The Essen Stroke Risk Score had a marginally better discriminatory performance in relation to stroke recurrence and cardiovascular event prediction than the CHA2DS2VASc score (Table 4), whereas the opposite was true regarding the prediction of death. Similarly, the negative predictive values for stroke recurrence and cardiovascular events with a score of 2 as cutoff value were marginally higher for the Essen Stroke Risk Score as shown in Table II in the online-only Data Supplement.
The subgroup treated with antiplatelet drugs and statins comprised 24 654 patients with a median age of 68.7 years (interquartile range 17.4 years), mean CHA2DS2VASc score 4.3 (SD 1.4), and mean Essen Stroke Risk Score 2.5 (SD 1.4). The overall incidence rates of recurrent stroke, death, and cardiovascular events during the first year of follow-up were 3.6%, 9.2%, and 5.5% respectively. The corresponding overall incidence rates during the first 5 years of follow-up were 2.3%, 7.1%, and 4.5% per year, respectively. The stratified analyses are shown in Tables III and IV in the online-only Data Supplement. C-statistics for the subgroup (not shown) and the whole cohort were comparable.
In this cohort study, we investigated the performance of the CHA2DS2VASc score and the Essen Stroke Risk Score for predicting stroke, death, and cardiovascular events in a nationwide cohort of non-AF patients with incident ischemic stroke. We found that increasing values of the CHA2DS2VASc score and the Essen Stroke Risk Score were associated with an increased risk of all 3 outcomes. The scores performed reasonably well in the prediction of survival free of stroke or cardiovascular events with negative predictive values for the lowest possible score around 0.95 and 0.85 (1-year and 5-year follow-up, respectively) for both scores. When taking into account the competing risks of death, the discriminatory performance in stroke recurrence and cardiovascular events was modest, with C-statistics around 0.55 for both scores.
In accordance with previous studies in non-AF cohorts,5,6 we demonstrated increasing stroke risk with increasing CHA2DS2VASc score, and we found similar results for the Essen Stroke Risk Score. However, in this cohort of patients with incident ischemic stroke, important findings differentiate it from the non-stroke cohorts. First, the baseline incidence rates of stroke are considerably higher in our cohort. With the lowest possible scores, we find incidence rates of recurrent stroke of 2.0% per year for the Essen Stroke Risk Score (score of 0) and of 3.1% per year for the CHA2DS2VASc (score of 2). In The Heart & Soul Cohort5 of patients with stable coronary heart disease, the annual incidence rate of stroke and transient ischemic attack was 0.9% for a CHADS2 score of 2. In their study on patients with acute coronary syndrome, Mitchell et al6 reported an annual incidence rate of stroke below 1% for all values of the CHA2DS2VASc score (stratified as 1, 2, 3, and ≥4), except for previous stroke or transient ischemic attack, which resulted in an annual stroke incidence rate of 1.54%, which is comparable to the baseline risk in our study. Second, the difference in stroke incidence rate between the lowest and the highest risk score is limited, which is reflected in the low hazard ratios, which only in the Essen Stroke Risk Score exceed a value of 2. That means, adding further risk factors does increase the stroke risk, but only modest, as seen from a 1-year stroke recurrence hazard ratio of 1.56 (1.20–2.02) for a CHA2DS2VASc score ≥7. Hence, previous stroke or thromboembolism confers a much higher baseline risk, which outweighs the impact of the other risk factors. However, in AF-patients as well, it seems that previous thromboembolism plays a special role, as reported in a recent Danish study.22 In that study, it was shown that previous thromboembolism as defined in the CHA2DS2VASc score adds more than the double to the total risk of stroke and other thromboembolic events than does the sum of any other 2 risk factors in the CHA2DS2VASc score.
We found the 5-year cumulative risks of stroke recurrence in the CHA2DS2VASc baseline group to be comparable to the results of the study from the Athens Stroke Registry (around 7% to 9%).13 By contrast, in the high risk groups, we found the cumulative risk of stroke recurrence to be somewhat lower: around 10% for a CHA2DS2VASc score ≥5 versus almost 20% for a CHA2DS2VASc score ≥2 in the Greek study (they did not count the incident stroke in the score). The Greek cohort was established over a wide time span (1993–2011), possibly resulting in greater inhomogeneity in treatment and recurrence risk, but also the influence of competing risks must be considered. From the total number of deaths, which in the present study exceeds 25%, it is clear that competing risks must be taken into account to avoid overestimating the risk of stroke.22,23
The Essen Stroke Risk Score is marginally superior to the CHA2DS2VASc score in stroke recurrence prediction when assessed with the C-statistics and estimates of the negative predictive values of the lowest possible score. In the Cox regression analysis of the single components of the scores, we find that female sex is significantly protective against recurrent stroke with a hazard ratio of 0.87 (95% confidence interval, 0.78–0.98) in the first year. This explains, at least partly, the difference between the 2 scores because sex is not comprised in the Essen Stroke Risk Score, and it emphasizes difficulties in applying the CHA2DS2VASc score in a non-AF cohort.
In the subgroup analyses of patients treated with antiplatelet drugs and statins, the rates of death and cardiovascular events after the first year of follow-up were marginally lower in the treatment group, but these differences disappeared after 5 years of follow up, not affecting the overall conclusion of the study.
Clearly, as shown in this study, stroke patients are at high risk of new cardiovascular events, and even at higher risk of death. The discriminatory performance of the scores in prediction of cardiovascular events is comparable to previous evaluations of The Essen Stroke Risk Score24 and is only modest. Overall, the performance in the prediction of death is somewhat better, which may be attributable to the strong influence of age on this outcome.
Strengths and Limitations
The main strength of this study is the nationwide design resulting in a large cohort with a large number of events in all strata. Our study is limited by its observational nature based on registry data. The diagnosis of stroke relies on correct diagnosis in The Danish Stroke Registry, which has previously been validated, showing a positive predictive value of 90% of a stroke diagnosis.20 We are not able to rule out that some cases of death could be caused by undiagnosed stroke as we do not have access to information of causes of death on all patients.
The frequency of risk factors in the cohort and the calculation of the risk scores may differ from the results of an individual assessment by a physician. However, we used a combination of previously validated algorithms3 and information from The Danish Stroke Registry to optimize the sensitivity of our comorbidity diagnoses.
The CHA2DS2VASc score and the Essen Stroke Risk Score predicted stroke, death, and cardiovascular events in this nationwide cohort of patients with incident ischemic stroke. Future refinements of clinical risk scores are required and decision-making based on these scores may not be warranted at present. Reliable risk stratification will enable the clinician to point out high risk patients who may benefit from a more intense follow-up and a more rigorous approach to the modifiable risk factors and have consequences for the choice(s) of antithrombotic treatment.
Sources of Funding
The study was supported by a grant from the Heinrich Kopp Foundation.
Dr Lip has served as a consultant for Bayer, Astellas, Merck, Sanofi, BMS/Pfizer, Daiichi-Sankyo, Biotronik, Portola, and Boehringer Ingelheim and has served as a speaker for Bayer, BMS/Pfizer, Boehringer Ingelheim, Daiichi-Sankyo, and Sanofi Aventis. Dr Larsen has been on the speaker bureaus for Bayer, BMS/Pfizer, Janssen Pharmaceuticals, Takeda, Roche Diagnostics, and Boehringer Ingelheim. The other authors report no conflicts.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.115.009912/-/DC1.
- Received April 29, 2015.
- Revision received June 13, 2015.
- Accepted June 18, 2015.
- © 2015 American Heart Association, Inc.
- Lip GY,
- Nieuwlaat R,
- Pisters R,
- Lane DA,
- Crijns HJ.
- Olesen JB,
- Lip GY,
- Hansen ML,
- Hansen PR,
- Tolstrup JS,
- Lindhardsen J,
- et al
- Friberg L,
- Rosenqvist M,
- Lip GY.
- Welles CC,
- Whooley MA,
- Na B,
- Ganz P,
- Schiller NB,
- Turakhia MP.
- Mitchell LB,
- Southern DA,
- Galbraith D,
- Ghali WA,
- Knudtson M,
- Wilton SB
- Hornero F,
- Martin E,
- Paredes F,
- Gil O,
- Cánovas S,
- García R,
- et al
- Lip GY,
- Lin HJ,
- Chien KL,
- Hsu HC,
- Su TC,
- Chen MF,
- et al
- Hillen T,
- Coshall C,
- Tilling K,
- Rudd AG,
- McGovern R,
- Wolfe CD
- Mainz J,
- Krog BR,
- Bjørnshave B,
- Bartels P.
- Lynge E,
- Sandegaard JL,
- Rebolj M.
- Kildemoes HW,
- Sørensen HT,
- Hallas J.
- Pedersen CB.
- Wildenschild C,
- Mehnert F,
- Thomsen RW,
- Iversen HK,
- Vestergaard K,
- Ingeman A,
- et al
- Olesen JB,
- Torp-Pedersen C.
- Weimar C,
- Diener HC,
- Alberts MJ,
- Steg PG,
- Bhatt DL,
- Wilson PW,
- et al