Trends and Disparities in Atrial Fibrillation and Stroke Related Mortality in the United States from 1999 to 2020

Atrial fibrillation (AF) is the most commonly encountered arrhythmia in clinical practice, and confers a 5-6-fold increased risk of stroke and a 2-fold increased risk of mortality [1]. It is associated with significant morbidity, mortality, and healthcare expenditure, with contemporary literature suggesting the annual direct and indirect cost of care being between $6 billion and $26 billion [2]. The prevalence of AF is expected to reach 12 million by 2030 in the United States, in conjunction with the rise in its risk factor prevalence [3]. Data from the early 21st century suggests a stagnation in stroke risk related to AF, driven in part by a plateau in anticoagulation use [4]. However, data regarding contemporary national mortality estimates related to stroke and AF are lacking. Updated epidemiological mortality trends are nonetheless important for targeted health policy interventions. This is especially important since recent data suggest that the overall AF-related mortality is on the rise [5]. The data that support the findings of this study are openly available at https://wonder.cdc.gov/mcd.html. We queried the Centers for Disease Control and Prevention Wide-Ranging OnLine Data for Epidemiologic Research (CDC WONDER) database to identify deaths in adults ≥25 years old where both atrial fibrillation/flutter and stroke (including ischemic, hemorrhagic, and nonspecified) were listed as an underlying or contributing cause of death between 1999 and 2020 (i.e., all these conditions were listed as an underlying or contributing cause of death on the death certificate) [6]. This database has previously been used to determine trends in mortality of cardiovascular diseases [7, 8]. AF/flutter cases were identified using ICD-10 codes I48.x, and stroke cases (including ischemic, hemorrhagic, and nonspecified) were identified using ICD-10 codes: I60.x, I61.x, I63.x, I64, I69.0, I69.1, I69.3, and I69.4). This study was exempt from local Institutional Review Board approval because the CDC WONDER database contains publicly available, anonymized data. AF + stroke-related deaths, population sizes, location of death, demographics (sex, race/ethnicity, and age), and regional information (urban-rural and state) were extracted. Race/ethnicities were defined as non-Hispanic (NH) White or NH Black based on race/ethnicity reported data on death certificates as previously described [6]. For urban-rural classifications, the 2013 National Center for Health Statistics Urban-Rural Classification Scheme was used to divide the counties into urban (large metropolitan area; population ≥1 million), medium/small metropolitan area (population 50,000–999,999), and rural (population <50,000) categories [9]. Crude and age-adjusted mortality rates per 100,000 population were determined. Crude mortality rates were determined by dividing the number of AF + stroke-related deaths by the corresponding US population of that year. Age-adjusted mortality rates (AAMR) were calculated by standardizing the AF + stroke-related deaths to the year 2000 US population. The Joinpoint Regression Program (Joinpoint V 4.9.0.0, National Cancer Institute) was used to determine trends in AAMR using annual percent change [10]. This method identifies significant changes in AAMR over time by fitting log-linear regression models where temporal variation occurred. Annual percent changes (APC) with 95% confidence intervals (CI) for the age-adjusted mortality rates were calculated using the Monte Carlo permutation test. Annual percent changes were considered increasing or decreasing if the slope describing the change in mortality significantly differed from zero using 2-tailed t-testing. Statistical significance was set at p < 0.05. Between 1999 and 2020, there were a total of 488,469 atrial fibrillation and stroke-related deaths. The absolute number of AF + stroke-related deaths stratified by sex and race/ethnicity are provided in Table S1. Between 1999 and 2020, there was a near doubling of AF + stroke-related deaths from 18,085 to 31,169, with a significantly higher number of deaths among women and NH Whites compared with their counterparts (Table S1). Of the deaths with information available regarding location of death, 45.3% occurred in medical facilities, 29.7% in nursing homes/long-term care facilities, 15.4% at home, and 5.8% occurred in hospice facilities (Table S2). The overall, sex and race/ethnicity stratified AAMR in AF + stroke-related deaths are listed in Table S3. The overall AAMR was 10.3 in 1999 and 11.6 in 2020, with a recent increase from 2008 to 2020 (APC: 1.16 [95% CI: 0.79–1.75]) after an initial period of decline between 1999 and 2008 (APC: −0.95 [95% CI: −1.88 to −0.36]) (Figure 1). While older (≥65 years) adults accounted for most of the deaths, young adults (25–64 years) had a steeper rise in mortality starting 2008 (APC: 6.28* [95% CI: 5.27–8.27]) (Table S4). White females had the highest mortality rates, whereas black females had the most rapid rise in mortality starting in 2018 (Figure 2). There were significant geographical disparities such that rural areas had significantly higher mortality rates compared with urban areas, and this gap widened in recent times with a steep rise in AAMR in rural areas starting 2009 (APC: 2.14 [95% CI: 1.66–2.92]) (Table S5 and Figure 1); and states in the 90th percentile of mortality (Minnesota, West Virginia, Oregon) had AAMR that were ∼3-fold higher than those in the bottom 10th percentile of mortality (Nevada, Florida, Louisiana) (Table S6 and Figure 3). These data demonstrate that, after an initial decline in atrial fibrillation and stroke-related mortality, there has been a reversal in trends with a recent rise in mortality rates starting in 2008. We also observed significant demographic disparities, such that women and NH White adults had higher mortality rates compared with their counterparts. There were also notable geographical differences in mortality burden, such that rural areas fared worse, and significant inter-state disparities were observed. These trends are consistent with the overall AF-related mortality trends over the same time period [5], and are likely driven by a variety of factors, including the overall increase in the prevalence of atrial fibrillation, as well as a possible underutilization of stroke prevention strategies including oral anticoagulation and left atrial appendage occlusion devices. While a decrease in case fatality rate could partly explain the rise in this trend, data from the Framingham study suggests that survival after AF diagnosis has not changed over the last two decades [11]. Consequently, the rise in AF + stroke-related mortality that we observed was likely secondary to an increase in the overall prevalence of AF as well as its risk factors including obesity, alcohol use, and obstructive sleep apnea, in recent years. This is particularly important when considering the sharp rise in mortality rates seen in young adults, indicating that the onset of AF in younger adults is rising and is translating into excess premature mortality. With recent data suggesting that early rhythm control may translate into mortality benefits [12], judicious use of anti-arrhythmic drugs and catheter-based ablation strategies should be considered when clinically appropriate and in a shared-decision making manner. Further, the use of newer smart wearables with built-in electrocardiogram (ECG) electrodes that can enhance our ability to detect AF early and intervene in a timely manner can help curb the excess preventable mortality associated with AF. Our findings are limited by the use of ICD codes and reliance on death certificate data which may be sources of misclassification bias. They also do not provide information about ejection fraction, ICD use, medication use, and socioeconomic factors. However, similar methodologies have been previously used to evaluate mortality estimates for AF and other cardiovascular diseases [3, 5]. Further, the positive predictive value ICD codes for identifying AF ranged from 70% to 96% in a recent systematic review [13]. Our study methodology also precluded us from capturing non-fatal strokes that occurred over the same time period. The data that support the findings of this study are openly available at 46 http://wonder.cdc.gov/mcd.html. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.