Abstract
Introduction:
Recent evidence suggests that the CHA
2
DS
2
-VASc score could be used to predict thromboembolic events (TE) (both CVA and noncerebral thromboembolism) in heart failure (HF) patients without atrial fibrillation (AF). This non-American cohort study had several limitations. We performed a retrospective analysis using the national Veteran Affairs database to externally validate the findings.
Hypothesis:
The CHA
2
DS
2
-VASc score can be used equally to predict TE in HF patients with or without AF.
Methods:
A retrospective cohort of the national Veteran’s Affairs database was used to identify HF patients discharged between 2002-2010. Rates for TE were calculated at both 1 and 5 years. C-statistics were calculated to test the performance of CHA
2
DS
2
-VASc in predicting the rate of TE in HF patients with and without AF. Patients who developed AF during the follow-up period (crossovers) and those on anticoagulation where excluded. Negative predictive value was calculated using cutoff value of 1.
Results:
A total of 77,601 patients were included in this analysis: 51,667 without AF and 15,630 with AF. 10,294 crossover patients were excluded. Baseline characteristics and medications are listed in Table 1. Event rates are reported in Table 2—there is a small trend towards higher incidence with increasing score in both groups. However, C-statistics were overall poor but similar for predicting TE in both groups (Table 3)
Conclusion:
Our analysis suggests CHA
2
DS
2
-VASc is a poor predictive model for TE in HF with or without AF. We found that previous conclusions regarding the predictive ability of CHA
2
DS
2
-VASc are not generalizable to our large American cohort. Like previously published studies, we show that the predictive ability of CHA
2
DS
2
-VASc was similar in both HF patients with or without AF, despite excluding patient receiving anticoagulation. This suggests that AF may not be the main predictor of TE in HF patients. There is a need for better predictive models for TE in HF.